Copper-dependent regulation of NMDA receptors by cellular prion protein: implications for neurodegenerative disorders

Abstract

Abstract

N-Methyl-d-aspartate (NMDA) receptors mediate a wide range of important nervous system functions. Conversely, excessive NMDA receptor activity leads to cytotoxic calcium overload and neuronal damage in a wide variety of CNS disorders. It is well established that NMDA receptors are tightly regulated by a number of cell signalling pathways. Recently, it has been shown that NMDA receptor activity is modulated by cellular prion protein (PrP^C) in a copper-dependent manner. Here we give an overview of the current state of knowledge concerning the novel concept of potent modulation of this receptor's kinetics by copper ions, and the interplay between NMDA receptors and PrP^C in the context of neurological diseases such as Alzheimer's disease, epilepsy, pain and depression.

Drs Gerald Zamponi and Peter Stys work at the University of Calgary's Hotchkiss Brain Institute. Dr Zamponi is the Head of the Department of Physiology and Pharmacology and a basic neuroscientist. His interest lies in the molecular physiology of voltage and ligand gated ion channels with an emphasis on pain and neurodegenerative disorders. Dr Stys is a neurologist and clinician scientist with expertise in multiphoton imaging techniques and a research interest in multiple sclerosis and neurodegeneration. Through a collaborative effort Drs Zamponi and Stys have discovered a new mechanism by which Aβ mediates neuronal degeneration via a copper dependent dysregulation of NMDA receptors.

NMDA receptor function

N-Methyl-d-aspartate (NMDA) receptors are one of the major classes of ionotropic glutamate receptors in the mammalian brain (McBain & Mayer, 1994). Their activation by glutamate or the synthetic agonist NMDA, together with the co-agonist glycine (or d-serine) (Shleper et al. 2005), results in the opening of a non-selective cation channel that mediates entry of calcium and sodium ions into the cytosol. At the synapse, glutamate release from presynaptic terminals diffuses across the synaptic cleft to activate these receptors, which in turn depolarize the postsynaptic membrane to induce an excitatory postsynaptic potential (EPSP). In addition, there are extrasynaptic NMDA receptors that are activated by glutamate spillover (Asztely et al. 1997; Lozovaya et al. 1999; Diamond, 2001). It is well established that NMDA receptors play a critical function in learning and memory (Herron et al. 1986; Collingridge, 1987; MacDonald et al. 2006). However, altered expression or function of these receptors is likely to play a role in the pathophysiology of a wide variety of CNS disorders including ischaemia, epilepsy, many neurodegenerative diseases and even neuropsychiatric disorders such as schizophrenia (Lipton & Rosenberg, 1994; Aarts et al. 2002; Loftis & Janowsky, 2003). The fact that NMDA receptors fulfil key physiological functions poses a dilemma with regard to designing therapeutics targeted towards these receptors, although a balance can be struck as evidenced by the Alzheimer's drug memantine, which is an NMDA receptor antagonist with relatively modest side effects (Lipton, 2006; Olivares et al. 2011). At negative voltages near the resting membrane potential, most NMDA receptors are tonically inhibited by extracellular magnesium ions (Nowak et al. 1984). Membrane depolarization dislodges magnesium from its binding site in the NMDA receptor pore, thus conferring a voltage dependence on receptor activity (Mayer et al. 1984). In addition to magnesium, zinc ions also bind to the receptors outside of the pore to modulate activity (Peters et al. 1987). As we will outline below, another metal ion that is emerging as an important NMDA receptor modulator is copper.

NMDA receptors activate (i.e. open) in response to ligand binding, and they deactivate (i.e. close) upon ligand unbinding (Vance et al. 2011). During prolonged application of agonist, these receptors undergo desensitization (Mayer et al. 1989), a process conceptually similar to voltage-dependent inactivation in voltage-gated calcium and sodium channels (Stotz et al. 2004), and designed to limit toxic calcium overload of cells during periods of prolonged glutamate elevations. Receptor desensitization is modulated by the aforementioned co-agonist glycine, such that increasing concentrations of this ligand produce a dramatic slowing of receptor desensitization kinetics (Mayer et al. 1989; Vyklicky et al. 1990). The physiological importance of NMDA receptor desensitization is underscored by the observation that application of high glycine concentrations is toxic to neurons, and blocking glycine reuptake increases NMDA receptor mediated neuronal excitability (Chen et al. 2003). Hence, receptor kinetics must be tightly controlled to maintain critical NMDA receptor functionality while preventing receptor hyperactivity and cellular damage.

NMDA receptor structure and subunit composition

NMDA receptors are heterotetramers that contain two obligatory GluN1 subunits (formerly known as NR1; Collingridge et al. 2009) and various combinations of GluN2A–D (formerly NR2A–D) and GluN3A–B (formerly NR3A–B) subunits (Paoletti & Neyton, 2007). Each of these subunits share a common transmembrane topology with a large extracellular N-terminal domain, and four membrane helices, as well as a large cytoplasmic C-terminal region (see Fig. 1) (Paoletti, 2011). In mammals, there is only one gene encoding GluN1, although different isoforms can be generated through alternative splicing (Tingley et al. 1993). The GluN1 subunit contains the binding site for the co-agonist glycine and d-serine which interact at a site formed by the N-terminal domain and the extracellular linker connecting the third and fourth transmembrane domains (for review see Paoletti, 2011). There are four different types of GluN2 subunits (GluN2A, B, C and D) which can be incorporated with the two obligatory GluN1 subunits into the assembled NMDA receptor complex, or with GluN3 to form triheteromers (Low & Wee, 2010; Rauner & Kohr, 2011). The GluN2 subunit contains the glutamate binding site, which, similarly to the GluN1 glycine binding site, is formed by interactions between the N-terminal domain and the loop between M3 and M4. The GluN2 subunit also contains the zinc interaction site which is located within the N-terminal domain (Choi & Lipton, 1999). The GluN3 subunits (i.e. GluN3A and GluN3B) do not bind glutamate, but instead interact with glycine, thus giving rise to a ‘glycine-only NMDA receptor’ when assembled with GluN1 in the absence of GluN2 (Chatterton et al. 2002; Piña-Crespo et al. 2010). In the context of this article, these glutamate-independent receptors are not considered further. In a fully assembled receptor, the pore region is formed by residues within the M2 regions which function similarly to the p-loops found in other types of ion channels (Kupper et al. 1996). Specific amino acid residues in the pore region ensure selectivity for cations, and the ability to bind magnesium ions or blockers such as MK801 (Mori et al. 1992).

Figure 1. NMDA receptor structure and kinetics.

A, transmembrane topology of a basic NMDA receptor subunit (i.e. GluN1, GluN2 or GluN3). Note the large N-terminus region that is involved in ligand binding. B, schematic representation of an NMDA receptor. One GluN1 and one GluN2 subunit are shown, although the fully assembled receptor complex contains two GluN1 and two GluN2 (or GluN3) subunits. C, effect of glycine on NMDA receptor desensitization. Two current recordings obtained from a wild-type mouse hippocampal neuron are depicted and scaled to overlap at peak. Both traces were evoked by 500 μm NMDA, the traces shown in black and grey were obtained, respectively, in the presence of 300 nm and 10 μm glycine.

The different types of GluN2 subunits generate receptor complexes with distinct functional and pharmacological properties (Loftis & Janowsky, 2003; Cull-Candy & Leszkiewicz, 2004; Paoletti & Neyton, 2007; Paoletti, 2011). As demonstrated in heterologous expression systems, GluN2 subunits govern a number of important NMDA receptor properties, including deactivation and desensitization rates, glycine and glutamate affinity, maximal open probability and susceptibility to magnesium block (reviewed in Paoletti, 2011). Furthermore, in neurons, the subcellular distribution of the receptors is GluN2 subtype dependent with GluN2A-containing receptors being localized to synaptic sites, whereas GluN2B containing receptors are found extrasynaptically (Liu et al. 2007), although this arrangement is not absolute (Thomas et al. 2006). At least in rodents, the expression of GluN2 subunits is thought to change with age, thus adding further complexity and suggesting a specific role of GluN2 subunits at various stages of neuronal development (Loftis & Janowsky, 2003).

NMDA receptors do not operate in isolation. They are tightly regulated by second messengers such as src and fyn kinases and CamKII which act on the large cytoplasmic C-terminal domain of the receptor (Xu et al. 2008; Sanhueza et al. 2011; Trepanier et al. 2012). Furthermore, NMDA receptors interact with certain types of G-protein coupled receptors (such as dopamine receptors) to form macromolecular signalling complexes (Lee et al. 2002). Recent work from our laboratory has revealed that NMDA receptors interact with, and are regulated by, cellular prion protein (PrP^C) (Khosravani et al. 2008; You et al. 2012). Below, we will focus on the effect of PrP^C on NMDA receptor function, and its possible pathophysiological roles.

Structure and physiological function of PrP^C

Before normal cleavage, PrP^C is a 254 amino acid protein that is ubiquitously expressed and has highly conserved homologues even in simple organisms such as yeast. It contains a 22 amino acid N-terminal signal peptide, followed, depending on the species, by four to five octapeptide repeats contained within an unstructured N-terminal region (Riesner, 2003; Aguzzi & Heikenwalder, 2006). The C-terminal half of the protein contains three α-helices and a glycophosphatidylinositol (GPI) anchor at position 231 that mediates an association with the extracellular leaflet of the plasma membrane (Aguzzi & Heikenwalder, 2006). The octapeptide repeats contain up to five copper binding sites with affinities varying from the femtomolar to the micromolar range (Jackson et al. 2001; Burns et al. 2003). Conversion of PrP^C into the pathological β-sheet-rich scrapie conformation (i.e. PrP^Sc) has been associated with neurological disorders such as Creutzfeldt–Jakob disease (Kingsbury et al. 1983; Palmer et al. 1991; Mallucci et al. 1999), bovine spongiform encephalopathy (BSE) (Hill et al. 1997; Prusiner, 1997), and Kuru (Gajdusek, 1977). Once misfolded, PrP^Sc acts as a chaperone that triggers the misfolding of normal PrP^C, thus leading to a progressive accumulation of PrP^Sc (Aguzzi et al. 2008). This in turn gives rise to aggregates that may cause massive neuronal degeneration that ultimately culminate in death of the subject (Caughey et al. 2009). Therefore, much effort has focused on understanding the role and function of PrP^Sc, and the molecular and biochemical mechanisms that trigger misfolding of normal PrP^C. In contrast, the normal physiological function of PrP^C remains incompletely understood (Linden et al. 2008). Interestingly, copper, likely acting via the copper-binding octarepeats, induces significant conformational changes to PrP^C (Wong et al. 2003), which may have important implications for the regulation of NMDARs (see below). There is also accumulating evidence that PrP^C may not only play an important role in neuroprotection (Khosravani et al. 2008), but also act as a conduit for compromised neuronal function and ultimately degeneration in disorders not typically considered prionopathies, such as Alzheimer's disease (Lauren et al. 2009).

Mice lacking PrP^C display a surprisingly mild behavioural phenotype (Bueler et al. 1992; for review see Steele et al. 2007) unless subjected to an insult. Under normal circumstances, these mice reportedly show slight deficits in spatial learning (Nishida et al. 1997; Valenti et al. 2001; Criado et al. 2005) and at an advanced age, peripheral nerve demyelination (Bremer et al. 2010). Reports based on brain slice recordings suggest altered long term potentiation (Collinge et al. 1994; Johnston et al. 1998; Curtis et al. 2003; Rangel et al. 2009), consistent with these memory deficits. Notably, PrP^C null mice are resistant to infectious prions, because these mice lack the template for further conversion of PrP^C into PrP^Sc and accumulation of the latter (Sailer et al. 1994). Along these lines, knockdown of PrP^C prevents PrP^Sc toxicity in rodents (White et al. 2008; White & Mallucci, 2009). However, there is increasing evidence that the absence of PrP^C can in fact be harmful. For example, PrP^C null mice show greater mortality following chemically nduced (i.e. pentylenetetrazole and pilocarpine) seizures, and a greater incidence of seizures in the kainic acid model (Walz et al. 1999), although these findings have been disputed in a more recent study (Ratte et al. 2011). PrP^C null mice display enhanced and prolonged neuroinflammation in a model of experimental autoimmune encephalomyelitis (EAE), whereas overexpression of PrP^C was protective in the EAE model (Tsutsui et al. 2008). Moreover, silencing of PrP^C has been shown to accelerate CNS autoimmune disease in a T-cell-dependent manner (Hu et al. 2010). Finally, recent findings from our laboratory have revealed that PrP^C null mice display depressive-like behaviour (Gadotti et al. 2012), and exhibit heightened nociception and increased inflammatory pain, as well as central pain sensitization (Gadotti & Zamponi, 2011). Interestingly, both of these behaviours could be abrogated by treatment with the NMDA receptor antagonist MK-801, suggesting the possibility that NMDA receptor activity may be enhanced by the absence of PrP^C. Together, these findings indicate that PrP^C may have multiple neuroprotective roles, thus perhaps explaining why this protein is so widely expressed.

PrP^C regulation of NMDA receptors

PrP^C appears to regulate the function of a wide range of different ion channels and receptors. For example, L-type calcium channel amplitude is decreased when the channels are exposed to the copper binding domain of recombinant PrP^C (Korte et al. 2003). Knockout of PrP^C results in a reduction of calcium activated potassium currents in cerebellar Purkinje cells (Herms et al. 2001), although it is not clear if this is a direct action on the potassium channel, or an indirect effect mediated via altered calcium entry. Along these lines, application of a 20 amino acid central fragment of PrP^C (residues 106–126) triggers the activation of potassium currents in basal forebrain neurons (Alier et al. 2010). A subsequent report revealed that PrP^C interacts with the potassium channel tetramerization domain (Huang et al. 2012), together indicating that PrP^C is an important regulator of various types of potassium channels. A recent study revealed that PrP^C interacts with mGLuR1 and mGluR5 receptors to stimulate ERK1/2 activation (Beraldo et al. 2011) suggesting that PrP^C can regulate G protein-coupled receptor-mediated cell signalling. Ligand-gated channels are also regulated by PrP^C. For example, the α7-nicotinic receptor activity is regulated by stress-inducible-protein-1 via PrP^C (Beraldo et al. 2010), and PrP^C is known to regulate kainate receptor expression (Rangel et al. 2007). Altogether these examples illustrate that PrP^C is a promiscuous protein that has the propensity to regulate signalling by plasma membrane receptors and channels (for a more detailed review, see Linden et al. 2008).

It may thus not be surprising to note that NMDA receptor activity is subject to modulation by PrP^C and PrP^Sc. Indeed, in addition to the in vivo data mentioned in the previous section, regulation of NMDA receptor function by PrP^C is also supported by findings that MK801 protects from increased cell death in PrP null mice subjected to kainate excitotoxicity (Rangel et al. 2007), presumably because the kainate-induced membrane depolarization leads to a secondary activation of NMDA receptors. Furthermore, neuronal cultures infected with PrP^Sc show increased survival in the presence of MK801 (Muller et al. 1993).

A recent study from our group examined the cellular basis for these effects (Khosravani et al. 2008). In field potential recordings from hippocampal slices from PrP null mice there was an increase in the number of population spikes compared with those seen in WT slices. This effect was exacerbated upon removal of magnesium ions, and normalized in the presence of the NMDA receptor blocker amino-5-phosphonovaleric acid. This hyperexcitability seen with PrP null slices fits with the increased susceptibility of PrP^C null mice to seizure-like behaviour. Miniature synaptic NMDA currents were also enhanced both in amplitude and duration, and whole cell NMDA currents in cultured pyramidal neurons showed greater current amplitude and dramatically slowed deactivation kinetics. Consistent with NMDA receptor hyperfunction, direct injection of NMDA into PrP null mouse brains resulted in dramatically increased lesion size compared to WT animals (Khosravani et al. 2008). The slowed deactivation kinetics appear reminiscent of what is observed with transiently expressed NMDA receptors containing the GluN2D subunit (Cull-Candy & Leszkiewicz, 2004), perhaps suggesting that the absence of PrP^C may lead to altered NMDA receptor subunit composition. However, although GluN2D siRNA treatment resulted in smaller, faster deactivating currents, it is unlikely that PrP null mouse neurons contain pure GluN1/GluN2D receptors, because whole cell currents could still be blocked by a combination of GluN2A and GluN2B blockers (H. You and G. W. Zamponi, unpublished observations). Rather, we suspect that in the absence of PrP^C, there is an increased proportion of heteromeric GluN1/GluN2(A or B)/GluN2D receptors which may give rise to altered deactivation kinetics, thus potentially reconciling our findings. The observation that synaptic NMDA currents are upregulated may explain the increased pain phenotype in PrP null mice, as NMDA receptors expressed in the dorsal horn of the spinal cord are similarly upregulated during central pain sensitization (i.e. ‘wind-up’) (Woolf & Thompson, 1991; Salter, 2005; Liu et al. 2008).

As noted above, NMDA receptor inhibitors prevent cell death of cultured neurons exposed to PrP^Sc (Muller et al. 1993). In principle, this could be explained by one of three different scenarios. First, exposure to PrP^Sc may cause the misfolding of endogenous PrP^C, thus leading to a de facto PrP^C knockdown and increased NMDA receptor activity. Indeed, a reduction in PrP^C levels has been reported in mouse models of prion disease (Griffin et al. 2007). Alternatively, it is conceivable that PrP^Sc could directly stimulate NMDA receptor function. Finally, it is possible that PrP^Sc may be unable to inhibit NMDA receptor activity, thereby acting as a dominant negative inhibitor of PrP^C action on the receptor complex.

Copper modulates NMDA receptors via PrP^C

It is well established that PrP^C is a high affinity copper binding protein (Brown et al. 1997; Jackson et al. 2001; Brown & Sassoon, 2002); however the physiological role of the copper binding sites on PrP^C has been elusive. We have recently shown that chelation of copper ions potently regulates native NMDA receptors in rat and mouse hippocampal neurons (You et al. 2012). As noted earlier, glycine acts as a potent co-agonist at the GluN1 receptor and a negative regulator of NMDA receptor desensitization. Many investigators in the NMDA receptor field conduct recordings in high concentrations of glycine as an experimental manipulation to boost current responses. Interestingly, both glycine and d-serine chelate copper ions (Martin et al. 1971), therefore the balance between applied agonist concentration and ambient copper levels becomes critically important. For this reason, we conducted our experiments using a physiological glycine concentration that is thought to reflect the submicromolar levels normally present at the synapse (Supplisson & Roux, 2002; Yang & Svensson, 2008). When neurons were exposed to 300 nm glycine/500 μm NMDA, currents in cultured pyramidal cells desensitized almost completely (Fig. 1C, black trace). Exposure of neurons to selective copper chelators such as bathocuproine sulfonate (BCS) or cuprizone induced a persistent non-desensitizing current component that could be overcome by addition of excess copper ions. Interestingly, in 300 nm glycine (without copper chelation), neurons from PrP null mice behaved exactly like BCS treated WT neurons; moreover the effects of PrP^C knockout and BCS were not additive, suggesting that the effects of copper on NMDA receptor desensitization were mediated via PrP^C. Furthermore, PrP^C could be immunoprecipitated with GluN1, and the strength of this interaction was greatly weakened upon copper chelation. This, together with the observation that the glycine affinity for the receptor was enhanced in PrP^C null mice (or upon application of BCS) led us to propose a model in which copper-dependent allosteric interactions between PrP^C and the GluN1 subunit regulate the affinity of the receptor complex for the co-agonist glycine, in this manner leading to non-desensitizing currents at any given glycine concentration (Fig. 2). This then suggests that perhaps a key physiological role of the copper binding sites on PrP^C is to regulate NMDA receptors. However, even in the absence of PrP^C, micromolar concentrations of copper still speed NMDA receptor desensitization (You et al. 2012; see also Vlachova et al. 1996), indicating the presence of a second copper modulation mechanism that operates independently of PrP^C. One possibility may be an as-yet-unidentified copper binding site on the receptor, as was discovered for zinc (Rachline et al. 2005). The intrinsic mechanisms that give rise to persistent currents in NMDA receptors upon copper chelation/deletion of PrP^C are not known. One attractive possibility is a switch in modal gating of the receptor, as described recently in a study by Zhang et al. (2008). It is also interesting to note that GluN2D-containing receptors show little desensitization and display the highest glycine affinity of all NMDA receptor subtypes (Erreger et al. 2007; Chen et al. 2008). However, given that the effects of copper chelation occurred rapidly, a switch in NMDA receptor subunit composition is unlikely to account for the persistent currents.

Figure 2. Model depicting the effects of copper and PrP^C on NMDA receptor function.

A, under normal circumstances, PrP^C in its copper-loaded state exists in a complex with the NMDA receptor, reducing the affinity of the co-agonist glycine for the receptor. At physiological levels of glycine (1 μm or lower), this results in strong desensitization of the receptor (see current trace obtained from a cultured hippocampal neuron). B, knockout of the gene encoding PrP^C results in a receptor with higher affinity for glycine, thus leading to persistent non-desensitizing currents and overall greater accumulation of calcium in the cell. C, chelation of copper ions by the selective chelator bathocuproine sulfonate or by Aβmonomers causes a weakening of the association between PrP^C and the NMDA receptor, thus producing currents similar to those seen in PrP^C null mouse neurons. D, Aβoligomers bind directly to PrP^C to disrupt PrP^C regulation of NMDA receptor activity, producing the same phenotype as that shown in panels B and C.

The regulation of NMDA receptors by copper ions deserves additional discussion. First, BCS might be a selective chelator of Cu⁺ over Cu²⁺ whereas PrP^C is thought to specifically bind Cu²⁺. However, the interactions between BCS and the two oxidation states of copper are non-trivial, with BCS able to bind Cu²⁺ with high affinity to form bis µCu(BCS)₂½²⁻ complexes (Al-Shatti et al. 1981; Sayre, 1996) or the protonated form of BCS µH·BCS½⁻ associates with Cu²⁺ to form complex species such as µCu²⁺ (BCS) (H₂O)_x½ and a proton (Xiao et al. 2011); in a pH buffer, this proton will be buffered driving the reaction in favour of complete chelation of all Cu²⁺ by BCS (or more precisely, by its mono-protonated form). The fact that the canonical Cu²⁺-selective chelator cuprizone (Peterson & Bollier, 1955) yielded identical results supports the idea that BCS indeed acted as a high-affinity copper chelator independent of this metal's oxidation state.

There is a large body of literature on the copper concentrations in CSF, with published values ranging widely from 0.12 to 7 μm (Kanabrocki et al. 1964; Agarwal & Henkin, 1982). In the whole brain, levels are around 5 μg (g wet weight)^-1 (∼80 μm equivalent) (Warren et al. 1960). The problem with determining CSF or whole brain copper levels is that almost all of this metal is bound to proteins and amino acids; thus the copper available for binding to PrP^C is determined not so much by the measured total concentrations, but rather by the relative affinities of PrP^C vis-à-vis those of other copper binding species. Measured copper in our culture media was around 100 nm although other media range up to the low micromolar level. The resting copper concentration in the synaptic cleft where the NMDA receptors are located is also estimated to be ∼1 μm, though transient peak levels may approach 250 μm (reviewed in Millhauser, 2007). Given the ability of copper to potently modulate NMDA receptor kinetics and limit agonist-induced inward current and therefore Ca²⁺ loads (see above), it is plausible that ambient copper in culture media and brain extracellular space functions to limit NMDA receptor-dependent excitotoxicity. In line with this prediction, treatment of cultured neurons with copper chelators causes cell death that can be prevented by the NMDA receptor antagonist 5,7-dichlorokynurenic acid or by supplying excess copper (You et al. 2012), underscoring the potential key importance of trace copper in cell culture media and experimental perfusing solutions. Neurons have active copper transport mechanisms such as the copper ATPases (Niciu et al. 2006; Veldhuis et al. 2009) that, along with copper binding proteins such as PrP^C, ensure proper copper homeostasis in the brain. When cultured neurons or brain slices are placed in recording chambers filled with external solutions such as aCSF (which according to our measurements contain anywhere between 4 nm and 50 nm total copper) for prolonged periods of time, one might expect a time-dependent partial depletion of copper to occur. As noted earlier, PrP^C contains several copper binding sites whose affinities vary widely from the micromolar to the femtomolar range (Jackson et al. 2001). In media or perfusing solutions containing sub-physiological concentrations of copper (e.g. in the tens of nanomolar, compare with published CSF concentrations noted above), it is conceivable that the lower affinity copper-binding sites on PrP^C will be depleted of this metal, potentially significantly altering the physiology or even the survival of the neuron. On the other hand it is important to note that prolonged exposure of neurons to higher levels of copper is also toxic due to the generation of free radicals (Simpson et al. 1988). This may explain findings from Sassoon et al. (2004) showing that prolonged exposure of neurons to 10 μm copper exacerbated the effects of a toxic PrP^C fragment.

In our experiments, appearance of persistent currents upon application of BCS occurred with a lag of approximately 2 min, consistent with a mechanism whereby PrP^C is initially loaded with copper, and BCS (which binds copper with attomolar affinity) slowly chelates copper ions that unbind from PrP^C. This suggests that copper ions are likely to be bound to PrP^C (at least at the higher affinity sites) even when bathed in aCSF, either from trace copper in the solutions or from copper ions released by the neurons themselves, or because copper ions may be leaching from the experimental apparatus. Altogether, these considerations suggest that maintenance of consistent copper homeostasis during experiments is critical. By extension, it is conceivable that some of the variability in the extensive published literature pertaining to NMDA receptor physiology could be attributed to variable trace copper levels in media and perfusing fluids. We would therefore argue that copper ions might have to be supplied exogenously to maintain consistent concentrations, similar to what is done routinely with magnesium.

It should be noted that copper ions not only modulate NMDA receptors, but also a variety of other types of ion channels. For example, AMPA receptors in rat cortical neurons are blocked with IC₅₀ values of around 5 μm, and at saturating levels of copper (i.e. 30 μm) agonist affinity for these receptors is reduced (Weiser & Wienrich, 1996). Copper ions also block T-type calcium channels and high voltage activated calcium channels with affinities ranging from ∼1 to 30 μm depending on calcium channel subtype (Jeong et al. 2003; Lu et al. 2009), and they modulate ENaC channels in the submicromolar range (Lu et al. 2009). These blocking affinities are similar to those observed in our own experiments on NMDA receptors in the absence of PrP^C, which fits with the idea that voltage gated cation channels may show common architectural features in their pore forming regions – the presumed site of action of copper. In the nanomolar range, copper ions have been shown to enhance the firing of olfactory epithelium neurons, whereas concentrations above 1 μm were inhibitory (Aedo et al. 2007). These potent effects of copper on ion channel physiology and neuronal activity further underscore the need to clamp copper concentrations at a physiological level during in vitro studies of the nervous system.

Aβ-mediated regulation of NMDA receptor function

NMDA receptors have been implicated in Alzheimer's disease (AD), and the NMDA receptor blocker memantine is Food and Drug Administration (FDA) approved for the treatment of this disorder (Kalia et al. 2008; Kotermanski & Johnson, 2009; Di Stefano et al. 2011). One of the key aspects of AD is an overproduction of toxic Aβ_1-42 peptides in the brain. Application of Aβ_1-42 oligomers to cultured neurons has been shown to trigger spontaneous NMDA currents and longer term exposure of neurons to Aβ_1-42 leads to NMDA receptor internalization (Snyder et al. 2005; Texido et al. 2011). It was recently reported by Lauren et al. (2009) that Aβ_1-42 oligomers can interact with PrP^C and that the toxic effect of Aβ_1-42 is dependent on these interactions. Furthermore, Aβ_1-42 is a high affinity copper binding peptide (Atwood et al. 2000), with values similar to those reported for BCS. Altogether, these considerations suggest a possible link between Aβ_1-42, copper, PrP^C and NMDA receptors. Indeed, as we have shown recently (You et al. 2012), application of nanomolar concentrations of Aβ_1-42 oligomers, or micromolar concentrations of Aβ_1-42 monomers, produced effects on NMDA receptors that were indistinguishable from those of BCS or removal of PrP^C (either acutely via GPI anchor cleavage or by gene knockout). Furthermore, recordings from neurons cultured from hemizygous 5XFAD mice, a mouse model of Alzheimer's disease where large amounts of Aβ_1-42 are produced within neurons (Oakley et al. 2006), revealed more slowly desensitizing NMDA currents compared to wild-type littermates, again with characteristics very similar to those observed with BCS treatment or PrP^C ablation. Like BCS, application of Aβ_1-42 to cultured neurons causes cell death that could be prevented either with NMDA receptor antagonists, or by addition of excess copper (You et al. 2012). Together these data suggest that an interplay between copper, Aβ_1-42, PrP^C and NMDA receptors may play a pivotal role in AD pathogenesis. We envision that high affinity binding of oligomers directly to PrP^C mediates an allosteric inhibition of NMDA receptor desensitization (similar to the absence of PrP^C), whereas higher concentrations of monomer might act simply by copper chelation (similar to BCS). Therefore, Aβ peptides are likely to mediate their deleterious effects by two related, but distinct mechanisms: one dependent on PrP^C where oligomers interfere with normal physiological function (i.e. regulation of NMDA receptor kinetics) of this ubiquitous cuproprotein, and the second by more directly interfering with the ability of copper to modulate these receptors. Such a dual mode of Aβ action would elegantly reconcile the ongoing controversy surrounding the requirement of PrP^C for the observed pathological effects of this peptide (Lauren et al. 2009; Calella et al. 2010; Gimbel et al. 2010; Kessels et al. 2010; Freir et al. 2011) (see below).

If elimination of PrP^C mediates similar electrophysiological consequences compared to those of Aβ_1-42, why do PrP null mice not exhibit synaptic and neuronal degeneration as in AD? There are at least two key differences between the absence of PrP^C and the accumulation of Aβ_1-42. First, a slowing of NMDA receptor desensitization only matters under conditions where there is prolonged exposure to (i.e. excess) glutamate. Indeed, Aβ_1-42 is known to reduce glutamate reuptake in hippocampal neurons (Li et al. 2009), which would lead to increased glutamate concentrations and toxic persistent currents that would not be observed in PrP null mice. Second, as noted earlier, NMDA receptors can be blocked by copper independently of PrP^C. Given that higher copper concentrations can compensate for the absence of PrP^C by restoring NMDA receptor kinetics to near normalcy, it is possible that in the PrP null brain, compensatory mechanisms are able to maintain free copper in key locations (e.g. in the synaptic cleft) at higher levels to mitigate the adverse influence of a PrP^C deficit on this key receptor.

It is important to note that the involvement of PrP^C in the toxic effects of Aβ_1-42 have become somewhat controversial. In support of the finding of Lauren and colleagues, Barry et al. (2011) reported that suppression of in vivo long-term potentiation by Aβ_1-42 oligomers was dependent on PrP^C. Freir et al. (2011) reported that Aβ_1-42 isolated from human AD brains could alter neuronal plasticity in a manner dependent on PrP^C. Finally, Gimbel et al. (2010) reported that deletion of PrP^C could abrogate the memory impairment seen in a mouse AD model. However, several other groups were unable to confirm a PrP^C dependence of the effects of Aβ_1-42 on synaptic physiology (Calella et al. 2010; Kessels et al. 2010). It is possible that different levels of copper and glycine used in these various studies lie at the root of these discordant findings, again stressing the need to maintain copper at a known fixed concentration in neurophysiological studies involving NMDA receptors.

Another perplexing question that arises is, why is PrP^Sc so toxic to neurons? If complete absence of PrP^C, by our arguments a very important NMDA receptor regulator, results in minimal neuronal toxicity (perhaps because ambient copper levels are adjusted accordingly), what is the mechanism of such extreme toxicity of some species of scrapie? The answer is not known, but it is conceivable that prion protein may assume many different conformation states, with PrP^Sc being the most nefarious. In this conformation, it is possible that the modulatory influence on NMDA receptors is so disadvantageous, potentially inducing high persistent currents that cannot be rescued by any compensatory mechanisms that are invoked in tissue merely lacking PrP. As a result of neurons being locked into a ‘chronic excitotoxic state’, massive degeneration occurs, which has been shown to be at least partially NMDA receptor dependent (Muller et al. 1993; Schroder et al. 1998). Clearly, this hypothesis will have to be tested experimentally.

Concluding remarks

Copper ions appear to mediate potent regulation of NMDA receptors akin to what has been described for zinc nearly two decades ago. Dysregulation of copper homeostasis near the NMDA receptor can trigger aberrant, slowly desensitizing currents that result in toxic calcium overload in the presence of sustained glutamate levels. This in turn may contribute to the pathology of Alzheimer's disease. Copper dysregulation of NMDA receptors may also play a role in other neurological disorders. For example, in Menkes disease, a mutation in the copper ATPase mediates NMDA receptor-dependent neurodegeneration (Schlief et al. 2005, 2006; Schlief & Gitlin, 2006). Furthermore, both α-synuclein and huntingtin are known to be copper binding proteins (Fox et al. 2007; Davies et al. 2010), and both have been associated with NMDA receptor mediated neuronal toxicity (Fan & Raymond, 2007; Adamczyk et al. 2009). This raises the possibility of copper modulation of NMDA receptors (perhaps via PrP^C) as a unifying theme in many neurodegenerative disorders.