Memory  >   The Extinction (Desensitized) State of NMDA Receptors

In the dentate gyrus of the hippocampus, more than 90% of granule cells are silent, namely, they do not respond to environmental stimuli, and no longer engage in acquiring new memory (Alme et al., 2010). Silent cells are also quite prevalent in the neocortex (Barth and Poulet, 2012), and other subregions of the hippocampus (Thompson and Best, 1989). What causes these cells to become silent?

For memory engram cells, neuronal firing depends largely on the NMDA plateau produced by a cluster of synapses. The action potential can be generated only when the NMDA plateau is produced. In silent cells, the NMDA receptors (NMDARs) may somehow be inhibited such that they are unable to produce NMDA plateau. Evidence will be presented in this and subsequent chapters for the CABT Hypothesis that the GluN2B-containing NMDARs could be blocked by the CABT complex, which consists of a CRMP2 monomer, alpha (α) and beta (β) tubulin. (Figure 12-1). Binding between GluN2B and CRMP2 has been demonstrated experimentally (Al-Hallaq et al., 2007).

The Structure of CRMP2

Collapsin response mediator protein-2 (CRMP2) was first discovered as a principal regulator of axonal extension (Goshima et al., 1995). Since then, its functions have expanded to include dendritic branching (Niisato et al., 2013), calcium channel regulation (Khanna et al., 2012), microtubule transport (Hensley and Kursula, 2016), and dendritic spine development (Zhang et al., 2018; Jin et al., 2016).

CRMP2 is encoded by the DPYSL2 gene which is implicated in bipolar disorder and schizophrenia (Fallin et al., 2005; Liu et al., 2014; Pham et al., 2016; Tobe et al., 2017). It has three subtypes: A, B and C. The CRMP2B protein contains 572 amino acids (aa) while CRMP2A comprises 677 aa. In the brain, CRMP2B is ~20 times more abundant than CRMP2A (Balastik et al., 2015). Perhaps for this reason, previous studies focused on CRMP2B. However, CRMP2A, being a larger protein than CRMP2B, could be more effective in forming the CABT complex to occlude GluN2B-NMDARs.


Figure 12-1. The three dimensional structure of a CRMP2 monomer. H19 is the last helix of CRMP2, which could bind to the cytoplasmic domain of GluN2B. PDB ID = 5X1A.

The three dimensional structure of CRMP2B has been determined (Figure 12-1). Biochemical studies have also revealed its phosphorylation sites (Figure 12-2). The phosphorylation state of CRMP2 is critical for its biological functions. There is compelling evidence indicating that the spine loss induced by beta amyloid oligomers (a hallmark of Alzheimer's disease) is caused primarily by CRMP2 hyperphosphorylation (more info).


Figure 12-2. The phosphorylation sites and binding domains of CRMP2. Single letter abbreviations are used to represent its amino acid sequence. T555 has been shown to bind both tubulin and kinesin (more info).

The CABT Complex

Tubulin is the canonical binding partner of CRMP2. It has two isoforms, α and β, that usually form heterodimers. The dendritic spine contains abundant tubulin heterodimers. In the postsynaptic density (PSD, a structure just beneath the postsynaptic membrane), the amount of α-tubulin molecules accounts for 8% of the PSD protein mass, far exceeding the amount of PSD-95 molecules, which account for only ~ 0.8% (Yun-Hong et al., 2011). CRMP2 forms a tetramer in solution. Interaction with tubulin heterodimers breaks the tetramer of CRMP2 into monomers to form CABT complexes (Niwa et al., 2017).


Figure 12-3. Structure of the CABT complex, consisting of a CRMP2 monomer, α and β tubulin. The red line represents the C-terminal domain (CTD) of the GluN2B subunit. DWED are the four residues within CTD, predicted to interact with the H19 helix of CRMP2. [Adapted from: Niwa et al., 2017]

In dendritic spines, the function of the CABT complex was largely unknown. This book proposes that it could play a central role in memory extinction and retrieval.

The Extinction State of NMDA Receptors

In humans, GluN2B contains a long CTD, including residues 867 - 1484. In rats and mice, their CTD is only slightly shorter, starting from residue 867 to 1482. Both CRMP2 and tubulin have been shown to interact with the GluN2B subunit of NMDARs. The tubulin heterodimer may bind to the CTD of GluN2B at the region 1243 - 1376 (van Rossum et al., 1999). CRMP2 binds preferentially to GluN2B over GluN2A (Al-Hallaq et al., 2007). By utilizing a peptide array, Brittain et al. (2012) identified two peptides in GluN2B that may bind to CRMP2: peptide KPGMVFSISRGIYSC (residues 857–871 of the rat GluN2B sequence) and DWEDRSGGNFCRSCP (residues 1205–1219 of the rat GluN2B sequence). Notably, the second peptide contains the negatively charged sequence, DWED, which could interact with the positively charged H19 in CRMP2. Therefore, the CABT complex may bind to the CTD of GluN2B, with CRMP2's H19 near DWED (1205 - 1208) and the tubulin heterodimer around the region 1243 - 1376 (Figure 12-3).

The association between CABT and GluN2B could block NMDAR currents (Figure 12-4), abolishing the NMDA plateau, consequently resulting in neuronal silence. With this function, CRMP2 is expected to play an important role in memory processing. Indeed, experiments have demonstrated that the antibody against CRMP2 causes amnesia (Mileusnic and Rose, 2011). Furthermore, memory extinction is impaired in schizophrenia (Holt et al., 2009; Holt et al., 2012).


Figure 12-4. Schematic drawing for the extinction (desensitized) state of NMDAR, where the ion-conducting pore is occluded by the binding of the CABT complex to the CTD of the GluN2B subunit.

The NMDAR blocked by CABT is a form of NMDAR desensitization, which is broadly defined as reduction of NMDAR currents in the continuous presence of glutamate. There are several forms of NMDAR desensitization: glycine-dependent (Cummings and Popescu, 2015), calcineurin-dependent (Krupp et al., 2002), Ca2+-dependent (Sibarov and Antonov, 2018), and PKA-dependent (Aman et al., 2014). The NMDAR desensitization resulting from CABT occlusion is identical to the PKA-dependent desensitization where NMDAR currents decrease when GluN2B is dephosphorylated at Ser-1166. Phosphorylation of Ser-1166 by PKA increases NMDAR currents (Aman et al., 2014). According to the CABT Hypothesis, dephosphorylation of Ser-1166 allows CABT to occlude the GluN2B-containing NMDARs, thereby reducing NMDAR currents (Chapter 13). The desensitization process also depends on the actin depolymerizing factor, cofilin, which can be activated by Ca2+ influx through NMDARs (Chapter 20).

To distinguish from other forms of desensitization, the NMDAR blocked by CABT will be called "NMDAR extinction", because it may give rise to the macroscopic memory extinction (inhibition of memory retrieval). This notion is supported by mounting evidence that PKA plays a key role in memory extinction and retrieval (Koh et al., 2002; Isiegas et al., 2006; Mueller et al., 2008; Nijholt et al., 2008; Menezes et al., 2015).


Author: Frank Lee
First published: October, 2017
Last updated: August, 2019