From bedside-to-bench: What disease-associated variants are teaching us about the NMDA receptor

Abstract

NMDA receptors (NMDARs) are glutamate-gated ion channels that contribute to nearly all brain processes. Not surprisingly then, genetic variations in the genes encoding NMDAR subunits can be associated with neurodevelopmental, neurological, and psychiatric disorders. These disease-associated variants (DAVs) present challenges, such as defining how DAV-induced alterations in receptor function contribute to disease progression and how to treat the affected individual clinically. As a starting point to overcome these challenges, we need to refine our understanding of the complexity of NMDAR structure-function. In this regard, DAVs have expanded our knowledge of NMDARs because they do not just target well-known structure-function motifs, but rather give an unbiased view of structural elements important to the biology of NMDARs. Indeed, established NMDAR structure-function motifs have been validated by the appearance of disorders in patients where these motifs have been altered, and DAVs have identified novel structural features in NMDARs such as gating triads and hinges in the gating machinery. Still, the majority of DAVs remain unexplored and occur at sites in the protein with unidentified function or alter receptor properties in multiple and unanticipated ways. Detailed mechanistic and structural investigations are required of both established and novel motifs to develop a highly refined pathomechanistic model that accounts for the complex machinery that regulates NMDARs. Such a model would represent a template for rational drug design and a starting point for personalized medicine.

Abstract figure legend

Full-length structure of the NMDA receptor (PDB: 4TLM) with known disease-associated variants (DAVs) in the GluN2 subunits highlighted in red. These DAVs highlight known as well as novel structure-function motifs in NMDA receptors, listed at right. In this review, we discuss the structure-function implications of DAVs domain by domain, focusing first on the core gating machinery (LBD, linkers, TMD) and then on the modulatory domains (ATD and CTD).

Introduction

The NMDA receptor (NMDAR) is a glutamate-gated ion channel that participates in nearly all brain functions. The central and diverse role of NMDARs is highlighted by the identification of numerous inherited and de novo nonsense and missense variations in the genes encoding NMDAR subunits. These variations – many of which are extremely subtle in terms of the mature protein – are associated with diverse brain disorders including autism spectrum disorder, epilepsy, schizophrenia, and severe intellectual disability, among many others (Burnashev & Szepetowski, 2015; Hardingham & Do, 2016; XiangWei et al., 2018; Xu & Luo, 2018). These disease-associated variants (DAV) present challenges clinically, such as how to treat the affected individual: how to ameliorate the determinantal effects of the variation, while minimizing negative side effects. DAVs, however, also present opportunities. DAVs do not just target well-known structure-function motifs, but rather give an unbiased view of structural elements important to the biology of NMDARs.

Targeting NMDARs in the clinic is a general challenge since they are ubiquitous and have diverse signaling properties. While tremendous advances have been made in terms of NMDAR pharmacology (Ogden & Traynelis, 2011; Hackos & Hanson, 2017; Wu & Tymianski, 2018), we still lack the tools to rationally target them. One promising approach in terms of DAVs has been to characterize their functional effects – how they alter ion channel function – and tailor the treatments accordingly (Pierson et al., 2014; Yuan et al., 2014; Soto et al., 2019). However, there are advantages of going beyond characterizing functional effects and defining in mechanistic and structural detail how DAVs alter NMDAR properties important to their physiology (Zhou & Wollmuth, 2017). Developing a detailed structure-function model of the NMDAR has the potential to form a template for rational drug design, permitting more refined treatments, and may permit predictions of functional effects of newly discovered DAVs without detailed characterization, accelerating personalized medicine.

NMDARs are obligate heterotetramers, composed of two GluN1 subunits, encoded by GRIN1, and typically some combination of GluN2(A-D) subunits, encoded by GRIN2A-D. DAVs have been identified in the obligatory GluN1 subunit (Lemke et al., 2016; Fry et al., 2018) as well as in GluN2A (Lemke et al., 2013; Xu & Luo, 2018), GluN2B (Hu et al., 2016), GluN2C (Yu et al., 2018), and GluN2D (XiangWei et al., 2019) subunits. Disease phenotype, progression, and severity are critical considerations of DAVs and are addressed in the literature noted above. In addition, an overview of the clinical phenotypes is shown in Table 1, and the specific disease associated with each DAV can be found in the citation associated with it (Table 2). Here, we will focus on structure-function issues: how those DAVs that have been characterized affect receptor function and how they highlight known structure-function motifs and have aided in the identification of novel mechanisms. We must emphasize, however, that the vast majority of DAVs remain unexplored and often reside in regions of unknown function. Thus, there remains much new information to be garnered.

Table 1.

Overview of the clinical phenotypes associated with DAVs.

Subunit	Domain	Sub-domain	Total DAVs	Most common phenotype	2nd most common phenotype	3rd most common phenotype
GRIN1	ATD		5	ID/DD (4/5)	Hpt (2/5)	multiple (1/5)
	LBD	S1	1	ID/DD (1/1)
		S2	5	Epi (5/5)	ID/DD (4/5)	multiple (3/5)
	Linkers	S1-M1	7	Epi (5/7)	ID/DD (5/7)	Hpt (4/7)
		M3-S2	2	ID/DD (2/2)	multiple (1/2)
		S2-M4	0
	TMD	M1	0
		M2	2	ID/DD (2/2)	Hpt (2/2)
		M3	9	ID/DD (7/9)	Epi (6/9)
		M4	6	ID/DD (5/6)	Epi (5/6)	multiple (4/6)
	CTD		1	multiple (1/1)
GRIN2A	ATD		13	Epi-Aph (8/13)	Epi (3/13)	ID/DD (2/13)
	LBD	S1	11	Epi-Aph (5/11)	Epi (3/11)	multiple (1/11)
		S2	14	Epi-Aph (11/14)	ID/DD (2/14)	multiple (1/14)
	Linkers	S1-M1	4	ID/DD (2/4)	Epi (2/4)	LD (2/4)
		M3-S2	0
		S2-M4	2	ID/DD (1/2)	Epi (1/2)	Epi-Aph (1/2)
	TMD	M1	2	ID/DD (1/2)	Epi (1/2)	ASD (1/2)
		M2	3	ID/DD (3/3)	Epi-Aph (2/3)	Hpt (2/3)
		M3	4	Epi (3/4)	ID/DD (2/4)	multiple (1/4)
		M4	2	Epi (2/2)	ID/DD (2/2)	ASD (1/2)
	CTD		18	Scz (8/18)	Epi-Aph (5/18)	Epi (3/18)
GRIN2B	ATD		5	Scz (4/5)	ID/DD (1/5)
	LBD	S1	11	ID/DD (11/11)	ASD (6/11)	Epi (4/11)
		S2	17	ID/DD (15/17)	Epi (7/17)	ASD (4/17)
	Linkers	S1-M1	4	ID/DD (3/4)	multiple (1/4)
		M3-S2	1	ID/DD (1/1)
		S2-M4	3	ID/DD (3/3)	Epi (2/3)	Ana (1/3)
	TMD	M1	2	ID/DD (2/2)	Epi (1/2)
		M2	9	ID/DD (9/9)	Epi (5/9)	Ana (3/9)
		M3	4	ID/DD (4/4)	Epi (3/4)	CVI (3/4)
		M4	9	ID/DD (8/9)	Epi (4/9)	multiple (2/9)
	CTD		16	Scz (6/16)	ASD (6/16)	ID/DD (3/16)

Total DAVs include unique positions, as used in Figure 1B, as well as different variations at any one position (Table 2). Patients with co-morbidities were placed in each appropriate category. References for specific mutations can be found in Table 2. ‘multiple’ indicates that different phenotypes were tied in number.

Abbreviations: Ana: Anatomical; ASD: Autism spectrum disorder; CVI: Cortical visual impairment; Epi: Epilepsy; Epi-Aph: Epilepsy with aphasia; Hpt: Hypotonia; ID/DD: Intellectual disability/developmental delay; LD: language disorder; Scz: Schizophrenia.

Table 2.

Disease-associated variants (DAV) covered in the present review.

Domain	GluN1		GluN2A		GluN2B
ATD	Val27Gly1		Pro31Thr13	Cys231Tyr/Arg14, 16	Ile50Asn4
	Arg217Trp2		Pro79Arg14	Ala243Val14	Val65Ile18
	Asp227His3		Thr143Ile4	Ala290Val14	Ile150Val26
	Arg306Gln4		Phe183Ile14	Gly295Ser15	Ala271Val4
	Arg397Ser5		Ile184Ser15	Pro336Ser4	Leu362Met4
			Thr189Asn4	Arg370Trp14
LBD	Pro532His1		Cys436Arg14	Met705Val14	Glu413Gly26,	Gly684Arg/Glu28
	Asn674Ile6		Asn447Lys17	Gly723Ser16	Cys436Arg26	Thr685Pro19
	Ser688Tyr7		Val452Met4	Glu714Lys14	Cys456Tyr26	Pro687Arg/Leu26
	Met706Val8		Cys455Tyr16	Ala716Thr15	Cys461Phe26	Gly689Ser26
	Asp789Asn6		Gly483Arg15	Ala727Thr14	Gly499Arg26	Ser690Asn28
	Arg794Gln6		Arg504Trp15	Asp731Asn15	Thr514Ala26	Arg693Ser26
			Val506Ala18	Val734Leu14	Asn516Ser26	Ile695Thr26
			Arg518His/Cys15, 1	Thr749Ile20	Asp524Asn27	Arg696His26
			Pro522Arg1	Lys772Glu14	Phe525Val26	Met706Val26, 27
			Thr531Met13	Asp776Tyr13	Gly533Asp26	Ala734Val26
			Lys669Asn15		Arg540His26	Ile751Thr26
			Val685Gly19		Gln662Pro26	Met789Lys18
			Ile694Thr15		Asp668Tyr26
			Pro699Ser14		Arg682Cys26
LBD-	Ser549Arg/Gly1, 9	Pro557Arg1	Ala555Thr21	Leu812Met13	Ser541Arg26	Glu807Lys26
TMD	Leu551Pro6	Gln559Arg11	Ser545Leu1	Ile814Thr14	Pro553Leu/Thr26	Ser810Arg/Asn26
Linkers	Asp552Glu1	Arg659Trp6	Ala548Thr15		Ser555Ile26
	Ser553Leu6	Glu662Lys1	Pro552Arg13		Glu657Gly26
TMD	Gly618Arg1	Phe817Leu1	Arg586Lys22		Val558Ile26, 27	Ala639Val26
	Gly620Arg1	Gly827Arg1	Lys590Asn23		Ala590Thr18	Ile655Phe26
	Gly638Val/Ala1	Leu830Pro1	Leu611Gln24		Trp607Cys26	Met818Leu/Thr26
	Met641Ile1		Asn614Ser18		Gly611Val26	Ala819Thr26
	Ala645Ser1		Asn615Lys24		Asn615Lys/Ile26	Gly820Ala/Glu/Val26
	Tyr647Ser/Cys1, 6		Ala643Asp18		Asn616Lys26	Met824Arg26
	Asn650Lys/Ile1, 6		Thr646Ala1		Val618Gly26	Leu825Val26, 4
	Ala653Gly6		Leu649Val13		Val620Met26	Gly826Glu26
	Ala814Asp12		Phe652Val15		Ser628Cys/Phe26
	Gly815Arg/Val1		Met817Val/Thr18		Ala636Pro/Val26
CTD	Arg844Cys1		Ile876Thr23	Asn1076Lys14	Phe1037Leu28	Met1339Val18
			Asp884Asn23	Arg1159Pro4	Gln1014Arg4	Met1342Arg4
			Ile904Phe14	Thr1229Ser4	Gly1026Ser4	Ser1415Leu4
			Ala922Val4	Asp1251Asn15	Arg1111His26	Leu1424Phe4
			Asp933Asn15	Ala1276Gly4	Thr1228Met18	Ser1446Thr28
			Ala968Thr4	Ile1295Thr21	Ala1267Ser18	Ser1452Phe4
			Asn976Ser14	Ile1379Val25	Thr1273Lys18
			Val998Met4	Asp1385Tyr16	Lys1292Arg29
			Val1000Met23		Lys1293Arg21
			Thr1064Ala4		Met1331Ile18

Listed DAVs include unique positions, as used in Figure 1B, as well as different variations at any one position. Signal peptide mutants not included. Earliest Review used as primary reference when applicable. DAVs were included only if a defined clinical phenotype was associated with it.

GluN1 (GRIN1):

¹Center for Functional Evaluation of Rare Variants (CFERV);

²(Lemke et al., 2016);

³(Bosch et al., 2016);

⁴(Tarabeux et al., 2011);

⁵(Halvardson et al., 2016);

⁶(Fry et al., 2018);

⁷(Zehavi et al., 2017);

⁸(Papa et al., 2018);

⁹(Helbig et al., 2016);

¹⁰removed;

¹¹(Tan et al., 2017);

¹²(Ortega-Moreno et al., 2017).

GluN2A (GRIN2A):

¹³(Strehlow et al., 2019);

¹⁴(Lemke et al., 2014);

¹⁵(Lesca et al., 2013);

¹⁶(Yang et al., 2018);

¹⁷(Xu et al., 2018);

¹⁸(Hardingham & Do, 2016);

¹⁹(Swanger et al., 2016);

²⁰(Yoo et al., 2017);

²¹(Yuan et al., 2015);

²²(Endele et al., 2010);

²³(Yu et al., 2018);

²⁴(Li et al., 2019);

²⁵(Lal et al., 2015)

GluN2B (GRIN2B):

²⁶(Platzer et al., 2017);

²⁷(Lelieveld et al., 2016);

²⁸(Grozeva et al., 2015);

²⁹(Takasaki et al., 2016).

Uneven distribution of DAVs in NMDAR domains

NMDARs are members of the ionotropic glutamate receptors (iGluR) superfamily (Traynelis et al., 2010). iGluRs function as tetramers and have a highly modular, layered structure (Figure 1A) (Sobolevsky et al., 2009; Karakas & Furukawa, 2014; Lee et al., 2014): The extracellular amino-terminal (ATD) and ligand-binding (LBD) domains, the membrane embedded transmembrane domain (TMD) forming the ion channel, and the intracellular C-terminal domain (CTD). All of these domains are critical to the physiology of NMDARs (Traynelis et al., 2010; Paoletti et al., 2013; Hansen et al., 2018). The LBD, the ion channel, and the short polypeptide chains connecting these domains, the LBD-TMD linkers, are directly involved in converting agonist binding into ion channel opening, which mediates much of the biology of NMDARs. These domains also participate in receptor assembly and trafficking (She et al., 2012; Gan et al., 2015). The ATD and CTD modulate the core gating function, contributing to the richness of NMDAR signaling (Paoletti et al., 2013; Hansen et al., 2018). The ATD and CTD also carry out a variety of cell biological functions, including roles in receptor assembly (Hansen et al., 2010; Sukumaran et al., 2012), trafficking to the membrane, and surface distribution (Gladding & Raymond, 2011; Ladepeche et al., 2014; Won et al., 2017), and downstream signaling (Chen & Roche, 2007; Sun et al., 2018).

Figure 1. Disease-associated variants (DAV) are prominent in the core gating machinery in NMDA receptors (NMDAR).

(A) Membrane topology of the tetrameric NMDAR, which are composed of two GluN1 (= A/C conformation = light orange) and two GluN2 (= B/D conformation = gray) subunits. NMDARs are highly modular proteins composed of 4 layered domains: extracellular amino-terminal (ATD) and ligand-binding (LBD) domains, a membrane embedded transmembrane domain (TMD) forming the ion channel, and an intracellular C-terminal domain (CTD). The LBD and TMD are connected by short polypeptide chains called the LBD-TMD linkers. Model structure based on 4TLM of GluN1/GluN2B (Amin et al., 2017).

(B) Percent and number (in parenthesis) of unique positions containing a DAV, relative to total number of residues, within each structural layer for GluN1 (GRIN1), GluN2A (GRIN2A), and GluN2B (GRIN2B). DAVs are also present in GluN2C (Yu et al., 2018) and GluN2D (XiangWei et al., 2019) subunits, but are too few in number to include in this analysis. Unique DAVs are missense mutations only, excluding terminations and deletions, and do not include different variations at the same position (Table 2).

(C) Topology of an individual subunit (GluN1) highlighting that the modular domains are intrinsic to subunits. The various structural levels are colored: ATD (blue), LBD (green), LBD-TMD linkers (red), TMD (orange), and CTD (light gray).

(D) Cartoon of the core gating machinery in NMDARs transitioning from the agonist unbound, ion channel closed (left) to the agonist bound, ion channel open (right) conformation. Binding of agonists to the LBD (green) causes the LBD clamshell, formed by D1 and D2, to close (transition from left to right). The movement of D2 away from the membrane pulls on the M3-S2 linker which in turn pulls on the activation gate formed at the apex of M3 leading to ion channel opening (transition from left to right). Open NMDARs permeates monovalent cations (Na⁺) and Ca²⁺. Below the cartoons are high resolution structures of the AMPAR M3 helices from a top down view as the channel transitions from closed (left) to open (right). PDBs 4WEK & 4WEO (Twomey et al., 2017). At present, open state structures are only available for AMPARs.

DAVs are distributed throughout NMDAR subunits, but are not distributed equally across these domains (Hardingham & Do, 2016; Hu et al., 2016; Lemke et al., 2016; Xu & Luo, 2018). One possibility for the uneven distribution is that a domain or a subdomain is so critical to function that it is intolerant to variations. However, regions that appear to be the most intolerant (Ogden et al., 2017) are also the regions having the greatest concentration of DAVs per domain (Figure 1B). Indeed, DAVs are most heavily concentrated in those regions directly involved in ion channel gating (Figure 1B).

The ATD and CTD, which show a low concentration of DAVs (Figure 1B), also show the highest number of non-disease associated missense variations (e.g., GRIN1 (Lemke et al., 2016); GRIN2A (Strehlow et al., 2019); and GRIN2B (Platzer et al., 2017)). Nevertheless, those DAVs identified in the ATD or CTD must carry out important roles in NMDAR function, although to date most remain unexplored.

Core gating mechanism and ion permeation

The modular structure of iGluRs is intrinsic to individual subunits (Figure 1C). The LBD consists of two lobes, D1 and the membrane proximal D2 (Figure 1C, green). The TMD consists of three transmembrane segments, M1, M3, and M4, and a reentrant M2 pore loop (orange). The transmembrane segments are connected to the LBD by the LBD-TMD linkers (red): S1-M1 for M1, M3-S2 for M3, and S2-M4 for M4. The ion channel is formed when the TMD of the four subunits comes together, with the M3 segment and the M2 pore loop being the major pore-lining structures (Figure 1D).

The basic principles of ion channel gating or pore opening in iGluRs are understood (Figure 1D) (Traynelis et al., 2010; Dai & Zhou, 2013; Plested, 2016; Hansen et al., 2018; Twomey & Sobolevsky, 2018). Two adjacent clamshell-like LBDs pair through their upper (D1) lobes forming what is called a dimer pair and in the absence of agonist are in the open clamshell configuration (Figure 1D, left). The M3 transmembrane segment contains elements of the activation gate, the structure that precludes the flux of ions in the closed state (Sobolevsky et al., 2009; Traynelis et al., 2010). In the absence of agonists, the M3 gate is closed (Figure 1D, left, upper & lower panels). Agonist binding induces clamshell closure which translates into the lower (D2) lobes moving away from the membrane (Figure 1D, right). The D2-attached M3-S2 linkers pull on the M3s, opening the ion channel pore (Kazi et al., 2014; Ladislav et al., 2018). The transition of the M3 transmembrane segments from the closed (Figure 1D, left bottom) to the open splayed (Figure 1D, right bottom) conformation permits the flux of ions, including Ca²⁺, across the membrane. Although the electrical signal mediated by NMDARs can impact signaling, it is the Ca²⁺ influx that has the most notable effects.

DAVs in the core machinery: Ligand-binding domains (LBD)

At the core of iGluR function are the LBDs, which bind glycine (GluN1) or glutamate (GluN2) (Figures 2A & 2B). Viewed simplistically, the agonist-induced displacement of the two lobes of LBD causes clam-shell closure (Figures 1D & 2A). However, high resolution structures, both of the isolated domain (Furukawa et al., 2005; Mayer, 2006) and of near full-length receptor (Tajima et al., 2016; Zhu et al., 2016; Twomey & Sobolevsky, 2018) have highlighted the more complex 3-dimensional movements and movements of local structural elements. In addition, the LBDs have numerous interfaces important to their function, including that between LBDs within a dimer pair (Borschel et al., 2011), the ATD-LBD interface (Sun et al., 2017; Regan et al., 2018), the interface between dimer pairs (Swanger et al., 2016), and the LBD-TMD linkers (Kazi et al., 2014). Although the general principles of how agonist-induced displacements lead to pore opening are understood (Figure 1D), many of the details such as how the LBD fully drives pore opening and the three-dimensional rearrangements of the LBD remain incompletely defined.

Figure 2. DAVs in the ligand-binding domains (LBD) have diverse functional effects.

(A) Cartoon of agonist-bound core gating machinery.

(B) Location of DAVs in the agonist-bound GluN1 (left), GluN2A (middle), and GluN2B (right) LBDs. The D1 and D2 lobes are highlighted and the ligand binding pockets are circled. DAVs are in red. PDBs 5I57 (Yi et al., 2016) & 4PE5 (Karakas & Furukawa, 2014).

(C) Zoomed in view of the ligand binding pocket for GluN1 (left), GluN2A (center), and GluN2B (right). The molecular structure of agonists (glycine, GluN1; glutamate, GluN2 subunits) are indicated. DAVs located in the vicinity of the binding pockets are indicated in green and are labeled.

(D) Venn diagrams showing the functional effects of DAVs tested from the GluN1 (left), GluN2A (middle), and GluN2B (right) LBDs (Fry et al., 2018).

DAVs are found throughout the LBDs (Figure 2B). The GluN1 LBD shows fewer DAVs than GluN2 subunits (Figure 1D), which may reflect in part that GluN1 has an extremely high affinity for glycine (Traynelis et al., 2010). Since ambient glycine is thought to saturate this site, small variation in binding may not impact function. Nevertheless, GluN1 is not inert (Zhu et al., 2013), and it is surprising that more DAVs have not been identified in this domain.

The primary function of the LBDs is to bind agonist. Still, only a limited number of DAVs directly interact with the agonist binding pocket (Figure 2C). In addition, when tested in detail, many of the DAVs do affect agonist binding, but they also have effects on other receptor properties including receptor gating and assembly/trafficking (Figure 2D) (Swanger et al., 2016; Fry et al., 2018). This is most notable with GluN2A where the most detailed characterization has occurred (Figure 2D, middle panel) (Swanger et al., 2016) and highlights the complex roles of the LBD in receptor function beyond just agonist binding.

One significant challenge are those DAVs that negatively impact assembly/trafficking. The majority of pharmacological agents targeting iGluRs are designed to enhance or suppress the activity of surface-expressed receptors, yet the receptors have to be on the membrane surface to be affected. Regulating the surface expression of receptors using small molecules might be one approach to overcome this limitation. Nevertheless, while a receptor’s ability to bind agonists can impact assembly/forward trafficking (Barria & Malinow, 2002; Kenny et al., 2009; She et al., 2012), how it does so mechanistically remains poorly understood. Given the diversity of DAVs in the LBD that affect assembly/trafficking (Swanger et al., 2016), more detailed investigations of their actions could contribute to a more detailed mechanistic understanding for the LBD’s involvement in this role.

Characterizing the effects of LBD DAVs on numerous properties is extremely valuable and provides guideposts for future experiments (Swanger et al., 2016; Fry et al., 2018). However, additional approaches need to compliment these functional experiments to provide a more global view of their action. Obviously high resolution structures would be helpful, but other approaches such as single-molecule FRET (Landes et al., 2011) and molecular dynamic (MD) simulations with enhanced sampling methods (Lau & Roux, 2007; Dai & Zhou, 2015; Yu & Lau, 2018) could identify how DAVs affect large-scale conformational changes and the energetics of agonist binding.

DAVs in the core machinery: The pore lining structures

The major structural elements lining the ion channel permeation pathway are the M3 transmembrane segments, which cover the extracellular half, and the M2 pore loops, which covers the intracellular half (Figure 3A) of the pore. The apex of the M3 segments form the activation gate (Figure 3A), which must be opened for ion flux to occur (Figure 1D). The M2 loop forms a narrow constriction in the pore that controls ion selectivity and channel block.

Figure 3. DAVs in the M3-S2 linkers and major pore lining structures.

(A) View of GluN1 (light orange) and GluN2B (gray) M3-S2 linkers and pore forming structures. The iGluR pore is formed by the four M3 transmembrane helices and on the intracellular half, by the four M2 pore loops. The M3 segments contain SYTANLAAF, the most highly conserved motif in iGluRs. At their apex, the M3 segments form an activation gate, which prevents the flux of ions in the closed state. The M2 pore loop forms the narrow constriction or selectivity filter.

(B) View of the GluN1 pore lining elements with DAVs highlighted in hot pink. The location of the SYTANLAAF motif is indicated in cyan.

(C) View of the GluN2 pore lining elements with DAVs in blue (GluN2A), red (GluN2B), or purple (both GluN2A & GluN2B).

(D) View of M2 pore loops with the nearest pore loop removed for clarity. The GluN1 N site and the GluN2 N+1 site form the selectivity filter (Wollmuth et al., 1996) and play key roles in Ca²⁺ permeability and Mg²⁺ block (Traynelis et al., 2010). PDB 5UN1 (Song et al., 2018).

(E) Sequence of GluN1, GluN2A, and GluN2B M2 pore loops highlighting secondary structures and positions of N and N+1 sites (Song et al., 2018). Location of DAVs are indicated in red.

The M3-S2 linkers

The pore-lining M3 segments are connected to the LBD (lobe D2) via the M3-S2 linkers (Figure 3A). Clamshell closure of the LBDs pull on these M3-S2 linkers causing the external M3 segments to splay away from the central axis of the pore (Figure 1D) (Kazi et al., 2014; Twomey & Sobolevsky, 2018).

The GluN1 M3-S2 also contains a structural element, the DRPEER motif, which is critical to the high Ca²⁺ permeability of NMDARs (Watanabe et al., 2002). DAVs are present in this motif (Arg659Trp, DRPEER and Glu662Lys, DRPEER) (Figure 3B), though the specific effect of these DAVs on receptor function including Ca²⁺ permeability is unknown. However, Glu662Lys, which is a negative to positive charge variant certainly would strongly reduce Ca²⁺ permeability (Watanabe et al., 2002), and this action may contribute to its clinical pathology.

The M3-S2 linkers take on distinct subunit-specific orientations (Sobolevsky et al., 2009) with GluN1 (A/C conformation) M3-S2s being perpendicular to the membrane (Figure 3B) whereas those in GluN2 (B/D conformation) being nearly parallel (Figure 3C) (Karakas & Furukawa, 2014; Lee et al., 2014). The only DAVs in the LBD proximal to the LBD-TMD linkers are those in the GluN2 subunit (Figure 3C). For pore opening to occur, the GluN2 M3 segments require more energy (Kazi et al., 2014), which may reflect in part these distinct M3-S2 orientations. Of those DAVs in S2 (LBD) proximal to the GluN2 M3-S2, only N2A-Lys669Asn has been tested and shows strong effects on receptor gating (Swanger et al., 2016), which may reflect the energy difference between GluN1 and GluN2 subunits, but more rigorous experiments will be needed to clarify.

The pore-lining M3 segment

The external M3 segment contains the SYTANLAAF motif, the most highly conserved motif in iGluR subunits (Wollmuth & Sobolevsky, 2004). SYTANLAAF contains elements of the activation gate (Chang & Kuo, 2008; Sobolevsky et al., 2009) including the lurcher site (SYTANLAAF), which initially identified the M3 segment as a gating element in iGluRs (Zuo et al., 1997). The majority of DAVs in either the GluN1 (Figure 3B) or GluN2 (Figure 3C) M3 segments appear in the extracellular half of M3 and largely encompass the SYTANLAAF motif (cyan). When tested, DAVs in or around the SYTANLAAF alter receptor gating (Fry et al., 2018; Vyklicky et al., 2018), though at present there has been no detailed mechanistic study of the effects of DAVs in the M3 segments on receptor function.

Although GluN2A and GluN2B have the most identified DAVs, DAVs have recently been identified and characterized in GluN2D (XiangWei et al., 2019). These DAVs are predominantly in the M3 segment and highlight an important concept: DAVs will ultimately appear in important structure-function motifs. For example, recent cryo-EM open state structures of AMPARs revealed an alanine hinge in SYTANLAAF that allows the channel gate to splay open (Twomey et al., 2017). DAVs are not present at this site in GluN2A or GluN2B (Figures 3B & 3C), but are present in GluN2D (XiangWei et al., 2019). The absence of DAVs at this site in GluN2A and GluN2B might reflect that it plays an insignificant role in these subunits, but we think this alternative is unlikely since perturbations of this alanine in GluN2A has strong effects on gating (Sobolevsky et al., 2007). Rather, the lack of DAVs at this site in GluN2A and GluN2B more likely reflects either that: (1) mutations at this site are embryonic lethal; or (2) they just have not yet been identified and will be in the future. We think alternative #2 is more likely and reflects that with time, as more DAVs become identified, they will further delineate key structure-function motifs.

The M2 pore loop

The M2 pore loop harbors the functionally critical N site, homologous to the Q/R-site in non-NMDARs (Figure 3D) (Wollmuth, 2018), that regulates Ca²⁺ permeability, Mg²⁺ block, and block by the clinically used ketamine and memantine (Traynelis et al., 2010; Huettner, 2015). The N site asparagine in GluN1 (Figure 3D, blue) and the N+1 site asparagine in GluN2 (Figure 3D, green) form the narrow constriction or selectivity filter in NMDARs (Wollmuth et al., 1996). A number of DAVs are present in this region (Figure 3E). Not surprisingly, these DAVs when tested have strong effects on receptor function including disrupting Ca²⁺ permeation, Mg²⁺ block, and receptor gating (Fedele et al., 2018; Li et al., 2019; Marwick et al., 2019). Indeed, many of the M2 loop DAVs nearly eliminate Mg²⁺ block. Structurally, these DAVs most likely disrupt the overall structure of the M2 loops, which in turn alter the positioning of side chains at the N site and N+1 site, which are critical in preventing Mg²⁺ yet allowing Ca²⁺ permeation (Mesbahi-Vasey et al., 2017; Fedele et al., 2018).

DAVs in the M2 loop present significant challenges to targeting them in the clinic. One alternative for gain-of-function DAVs are open channel pore blockers such as memantine (Pierson et al., 2014; Platzer et al., 2017), which is a low-affinity pore blocker that is in clinical use for moderate Alzheimer’s disease. Memantine in part acts by lodging in the channel pore at the M2 narrow constriction (Johnson & Kotermanski, 2006; Johnson et al., 2015; Song et al., 2018). However, DAVs in the M2 loop drastically alter the efficiency of memantine block, in some cases allowing for its permeation and a complete lack of channel block (Fedele et al., 2018). Because NMDARs are allosteric proteins, DAVs occurring at distant sites can also affect the M2 pore loop (Amin et al., 2018). Thus, understanding how NMDAR structural domains and subdomains coordinate with each other to affect changes at key active sites is critical for guiding both best clinical practice and rational drug design.

DAVs in the outer structures identify novel structure-function motifs

The splaying of the M3 segments, mediated by the M3-S2 linkers, is central to pore opening (Figures 1D). However, this inner M3/M3-S2 gating core is surrounded by the M1 and M4 transmembrane segments as well as the linkers connecting them to the LBD, S1-M1 and S2-M4 (Figures 4A) (Sobolevsky et al., 2009; Karakas & Furukawa, 2014; Lee et al., 2014). These outer structures affect gating (Schmid et al., 2007; Talukder et al., 2010; Ren et al., 2012), contain sites for allosteric modulators (Ogden & Traynelis, 2013; Perszyk et al., 2018; Shi et al., 2019), and must be displaced for efficient pore opening to occur (Figure 4B) (Kazi et al., 2013). Although these outer structures undergo agonist dependent conformational changes (Dolino et al., 2017; Twomey et al., 2017), how they mechanistically contribute to the gating process is unclear. It is here where DAVs have provided the most novel mechanistic insights into NMDAR structure-function.

Figure 4. DAVs in the outer structures, M1 and M4, identify key structure-function motifs.

(A) Top down view of the TMD. The inner structures, the M2 loops (not shown) and M3 helices, form the ion channel pore (solid circle). The outer structures, M1 and M4 helices and the linkers that connect them to the LBD, S1-M1 and S2-M4, surround the inner core. Together, the S1-M1 (notably the pre-M1 helix), extracellular M3, and extracellular M4, make up a structure-function motif critical to gating known as the gating triad (red circles and dashed lines). The gating triad is present in both the GluN1 and GluN2 subunits though it has most notable effects in GluN2 (Ogden et al., 2017).

(B) Cartoon of the inner and outer structures showing the general displacement of structures in the transition from the closed (left) to open (right) conformation.

(C) Topology of the outer structures, the M1 helices and their LBD-TMD linker (S1-M1) (left) and the M4 helices and their LBD-TMD linker (S2-M4) (right) for GluN1 and GluN2. The extracellular portion of the TMD and the lower portion of the LBD-TMD linkers are termed the gating collar (Yelshanskaya et al., 2017) or gating triad (Gibb et al., 2018). DAVs are highlighted in hot pink (GluN1), blue (GluN2A), red (GluN2B), or purple (both GluN2A & GluN2B). The helices are oriented such that the M3 facing portion is to the left.

(D) The three-dimensional arrangement of the gating triad in the closed state. The GluN2 pre-M1 helix, extracellular M4 and M3 make up the gating triad. With the exception of the M2 pore loop and pore lining portions of M3, most DAVs in the TMD appear in this region. DAVs are labeled as in (C).

DAVs in the outer structures are predominantly localized either to the LBD-TMD linkers or in the case of the M4 segments to the upper third of the helix (Figure 4C). High resolution structures have highlighted the close link between the M3 segments (SYTANLAAF motif), the S1-M1 linker (most notably the pre-M1 helix), and the extracellular M4 transmembrane segment (Sobolevsky et al., 2009; Tajima et al., 2016; Zhu et al., 2016; Twomey et al., 2017). These interacting motifs have been referred to as a “gating collar” in AMPARs (Figure 4C) (Yelshanskaya et al., 2017) and a “gating triad” in NMDARs (Figure 4D) (Gibb et al., 2018). The presence of numerous DAVs in the major elements of the gating triad, and their notable absence from surrounding structures like the M1 and lower M4 segments, highlight their functional importance (Figure 4D).

A conserved proline in the pre-M1 helix in S1-M1 mediates pre-gating steps

NMDARs presumably have multiple intermediate steps or conformational changes between agonist binding and ion channel opening (Howe et al., 1991; Banke & Traynelis, 2003; Popescu & Auerbach, 2003). Some of the structures mediating these pre-active or pre-gating movements have been speculated on with the S1-M1 being a prime candidate (Sobolevsky et al., 2009).

The S1-M1 contains a proline (P), P557 in GluN1, P552 in GluN2A, and P553 in GluN2B (Figure 4C), that is highly conserved across a variety of species (Alsaloum et al., 2016). This conserved P is located in a short helix in S1-M1 referred to as the pre-M1 helix. DAVs at the conserved P enhance agonist potency (Ogden et al., 2017). Surprisingly, when the conserved P in the GluN2 S1-M1 are mutated, channel activation – the rate at which the M3 segment presumably goes from closed to open – became dramatically slower and thus NMDARs lose their ability to respond rapidly to transient glutamate, suggesting that the conserved P mediates agonist-induced movements that occur before opening of the ion channel (Figure 5). Additionally, only one intact proline site in one of two GluN2 subunits is required to maintain fast NMDAR activation (Ogden et al., 2017). Hence, the displacement of a single pre-M1 helix is sufficient to allow fast pore opening to occur, suggesting that the S1-M1s act independently and that the core gating machinery (M3-S2/M3s) is energetically prepped for ion channel opening. Molecular dynamics (MD) simulations have further indicated that S1-M1 movements are independent of the actions of neighboring subunits, suggesting that this element mediates subunit specific features of NMDAR activity (Gibb et al., 2018)(McDaniel et al., in revision).

Figure 5. Agonist-induced displacements of the outer structures prime the ion channel for pore opening.

Top down views of the gating triad in AMPARs for the B/D (≈GluN2) subunit. In the transition from the closed (left) to the open (right) states, all three elements – the pre-M1 helix, and the tops of M4 and M3 – undergo extensive and coordinated movements. The splaying of the M3 segments, highlighted in red (right), is the final step in ion channel opening. It seems likely that the outer structures undergo displacements (pre-M1, blue & M4, green) (middle panel) prior to the final M3 displacement. These displacement between agonist binding and M3-mediated pore opening are called pre-active or pre-gating movements.

These movements suggest a temporal model for the gating triad (Figure 5) (Gibb et al., 2018). Agonist binding induces multiple conformational changes in the LBD. These are propagated to the TMD with an early step being displacement of at least one pre-M1 helix in the S1-M1 (Figure 5, middle, pre-active closed). Subsequently, M3 segments can then be splayed (Figure 5, right, bound, open). Hence, the displacement of S1-M1 (as well as M4) prime the channel for opening. This model is simplistic and will require detailed studies to further refine Nevertheless, targeting the S1-M1 region, or the gating triad in general, is a promising means to modulate NMDAR activity in the clinic. Indeed, promising drug candidates targeting this region have already begun to appear (Perszyk et al., 2018).

The M4 segments regulate the function of the inner pore-lining structures

A high density of DAVs occur in the upper third of the GluN1 and GluN2 M4 segments and the S2-M4 (Figure 4C), a region critical to receptor gating in NMDARs (Amin et al., 2017). Most notable is a conserved glycine (G) located at the juncture of the upper third and lower two-thirds of the M4 segment (GluN1 G815 & GluN2B G820). Numerous DAVs have been identified at this conserved G (Figure 4C)(Amin et al., 2018), including glycine-to-arginine, the most common mutation in the transmembrane region (Partridge et al., 2004; Molnar et al., 2016), but also subtle changes in side chains such as glycine-to-valine or glycine-to-alanine with equal effects (Amin et al., 2018).

DAVs at the conserved G in either GluN1 or GluN2B subunits have strong effects on receptor function (Amin et al., 2018; Vyklicky et al., 2018). Notably, they significantly speed up the slow deactivation time, a key regulatory component of Ca²⁺ signaling, by dramatically restricting the ability of the pore-lining M3 segments to stay open, with these effects being strongest in GluN1 (Amin et al., 2018). Experiments using a combination of relative MD simulations (Figure 6A) and functional studies revealed that the conserved G in GluN1 acts as a hinge with the lower two-thirds of the GluN1 M4 existing in a ‘constrained’ (Figure 6B, left) or ‘expanded’ (Figure 6B, right) conformation (Amin et al., 2018). Highlighting the importance of integrating diverse techniques was that these conformations of the GluN1 M4 also affected the M2 pore loop conformation (Figure 6B). While the MD simulations predicted such an effect, functional experiments verified that DAVs at the GluN1 conserved G affected Ca²⁺ permeability (Amin et al., 2018).

Figure 6. DAVs identify critical gating hinges.

(A) Top down view of the NMDAR TMD in the closed state with M1 omitted for clarity (left) and a simulated open state (right) (Amin et al., 2018). There are multiple hinge points including an alanine hinge in the M3 SYTANLAAF motif (discussed with M3 segment) and a glycine hinge (conserved G) in the M4 segments (hinges labeled in cyan). These hinges facilitate the transition to the open state (red arrows).

(B) Constrained (left) and expanded (right) conformations of the GluN1 M4 helix and the M2 pore loop. The lateral expansion of the GluN1 M4 at the conserved G hinge allows the M2 loop to expand, permitting the transition from a low (left) to high (right) Ca²⁺ permeability state. DAVs at the GluN1 conserved G keep the M4 segments in the constrained conformation (left).

(C) Cartoon representation of the pore in two conformations: Unstable open (left) where the M3 gate is open but the GluN1 M4 is constrained; and Stable open (right) where both the M3 gate is open and GluN1 M4 is expanded. In the ‘unstable open’ conformation (left), NMDARs show brief ion channel openings at the single channel level (lower traces) and low Ca²⁺ permeability. In the ‘stable open’ conformation (right), NMDARs show more characteristic long-lived single channel openings (lower traces) and high Ca²⁺ permeability (Amin et al., 2018).

DAVs at the GluN1 conserved G revealed new features of NMDAR gating (Figure 6C). If the ion channel pore is opened by the M3 gate (with pre-M1 movements before this), but the GluN1 M4 is ‘closed’ (Figure 6C, left, unstable open), the channel only shows brief openings and a low Ca²⁺ permeability. In contrast, when the M3 gate is open and the GluN1 M4s are expanded (Figure 6C, right, stable open), the receptor enters into long-lived open states and high Ca²⁺ permeability (Amin et al., 2018). Thus, the conformation of the GluN1 M4 helix affects both receptor gating and Ca²⁺ permeability.

In contrast to the GluN1 M4, the lower two-thirds of the GluN2 M4 only show limited outward splaying, and therefore DAVs at the GluN2 conserved G do not affect the M2 pore loop or Ca²⁺ permeability (Amin et al., 2018). Nevertheless, DAVs at the GluN2 conserved G do have strong effects on receptor gating (Amin et al., 2018), but its mechanistic role remains unexplored. The upper third of the GluN2 M4/S2-M4 has extensive contacts with M3 and S1-M1. These contacts, which also show numerous DAVs (Figures 4C & 4D), presumably play key roles in regulating the triad (Figure 5).

The M4 segments directly attach to the highly modifiable intracellular CTD, which is known to affect receptor gating (Maki et al., 2012; Murphy et al., 2014) and Ca²⁺ permeability (Aman et al., 2014). The M4 segments are also exposed to lipids, which can alter NMDAR function (Korinek et al., 2015). Nevertheless, whether the CTD and/or lipids act via the M4 to alter receptor function remains unknown, yet this might represent a novel pathway to modulate NMDAR function.

NMDARs can signal via metabotropic pathways (Nabavi et al., 2013), which can impact synaptic plasticity. These signaling pathways are dependent on displacements of the GluN1 CTDs (Aow et al., 2015; Dore et al., 2015). Given that the GluN1 CTD is directly attached to the M4 lever, it is possible that the conserved glycine mediates the movements required to initiate the metabotropic action of NMDARs. Hence, DAVs at the conserved glycine, or the M4 segment in general, might not only disrupt the ionotropic but also the metabotropic function of NMDARs.

DAVs in the modulatory amino-terminal (ATD) and C-terminal (CTD) domains: lots of them but not lots of information.

The ATD and CTD carry out critical cell biological and gating functions that are fundamental to the role of NMDARs at synapses (Martel et al., 2012; Ryan et al., 2013). While there are numerous DAVs in these domains, the vast majority of them have an unknown effect and often reside in regions of the protein – especially in the CTD – without a currently known biological function. Hence, there is a tremendous need to expand detailed studies of these DAVs. Such information has the potential to provide novel insights into the role of these domains in NMDAR function.

The amino-terminal domain (ATD)

The ATD is a major hub for allosterically regulating NMDAR gating and participates in receptor assembly (Hansen et al., 2010; Paoletti et al., 2013). In terms of gating, three major ligands act via the ATD: Zn²⁺ and protons, both of which act as negative allosteric modulators, and polyamines such as spermine that act as positive allosteric modulators. The action of protons overlaps with that of Zn²⁺, though their sites of action are more diffuse. Structurally, the ATD is composed of two lobes: R1 and R2 (Figure 7A) (Karakas et al., 2011; Romero-Hernandez et al., 2016). These allosteric ligands act to modify gating by either occupying a Zn²⁺ binding site, which has the highest affinity in the GluN2A subunit, at interfaces either between R1 and R2, or between the ATD and the LBD (Mony et al., 2011; Romero-Hernandez et al., 2016; Jalali-Yazdi et al., 2018; Zhang et al., 2018).

Figure 7. DAVs in the modulatory amino-terminal (ATD) and C-terminal (CTD) domains.

(A) Location of DAVs in the GluN1 (left), GluN2A (middle), and GluN2B (right) ATDs. The R1 and R2 lobes are highlighted. Only GluN2A binds the inhibitory divalent zinc under physiological conditions. DAVs are in red. PDBs 5TQ2 & 5TPZ (Romero-Hernandez et al., 2016).

(B) Linear representation of the GluN1, GluN2A, and GluN2B CTDs. At present, there are no high-resolution structures of the CTDs. Sites for known phosphorylation (Src kinase, Protein kinase C (PKC), Protein kinase A (PKA), palmitoylation, and putative phosphorylation (yellow) are indicated along with essential protein-protein interactions. DAVs are indicated by red bars. The vast majority of these DAVs reside at positions of unknown function.

A variety of DAVs have been identified in the GluN1, GluN2A, and GluN2B ATDs (Figure 7A). Given the more prominent role of the GluN2 subunit in mediating the allosteric role of the ATD, it is not surprising that the majority of DAVs are identified here. Nevertheless, despite the prevalence of the GluN2A subunit in mediating the Zn²⁺ action, only two DAVs in the GluN2A ATD (Pro97Arg & Arg370Trp) out of seven tested had an effect on Zn²⁺ with Arg370Trp actually enhancing its action (Serraz et al., 2016). These same DAVs had no notable effect on proton inhibition. The lack of changes to the ATD’s canonical modulatory function suggest that there are unknown critical functional features of the ATD. More detailed mechanistic studies of DAVs in the ATD may provide clues to its complex role.

Another function of the ATD is receptor assembly (Hansen et al., 2010). Several of the GluN2A DAVs do reduce surface expression (Serraz et al., 2016; Addis et al., 2017). In addition, the DAVs in GluN1 and GluN2B are not associated with elements involved in Zn²⁺ inhibition (Figure 2A). Hence, these DAVs may also impact receptor assembly, though they have not yet been explored. DAVs may thus represent a unique opportunity to clarify the role of the ATD in receptor assembly, which is poorly understood.

The C-terminal domain (CTD)

The NMDAR CTDs are a site of numerous post-translational modifications as well as protein-protein and protein-lipid interactions (Figure 1A). Most notably are binding sites for PSD-95 which anchors the receptor transiently, and the Ca²⁺-induced activation of CaMKII which drives plasticity and receptor turnover in the N2 subunits (Hell, 2014; Won et al., 2017). Additionally, the CTD is known for its role in trafficking, including harboring an ER retention motif (Petralia et al., 2009). Although binding and phosphorylation sites have been located across the CTD (Chen & Roche, 2007), the specifics of the structure are unknown due to a lack of high-resolution structures reflecting that many elements are disordered (Choi et al., 2011).

A single functional study of DAVs in the CTD has been published, linking their phenotype to impaired MAGUK binding (Liu et al., 2017). This leaves an opportunity for further investigation of DAVs, which at first glance do not overlap with major sites however are still able to confer a phenotype. What we have learned from DAVs is that protein binding in the CTD plays an important role in disease mechanisms. Many of the CTD DAVs are associated with schizophrenia (Tarabeux et al., 2011; Hardingham & Do, 2016). Specifically, in N2B schizophrenia and cognitive impairments has been connected to the CTD through phosphorylation sites, actin adaptor proteins, and MAGUKs (Liu et al., 2017). Although no known mutants disrupt phosphorylation sites directly, the impact of structural disruptions or steric interference reducing affinity and binding are not ruled out.

A DAV inspired model of NMDAR structure-function offers new avenues for drug design and personalized therapy

Tremendous advances have been made in understanding the structure-function of NMDARs. DAVs are inspiring new efforts to further refine our understanding of NMDAR function. Indeed, DAVs do not emerge at random but rather appear at positions critical to NMDAR biology and target not only our favorite structure-function motifs but also novel motifs – parts of the protein we did not know were important. Established structure-function motifs like the N site, DRPEER, SYTANLAAF, and more have been validated by the appearance of disorders in patients where these motifs have been disrupted. By taking advantage of these observations, our knowledge of NMDARs is greatly accelerated, using DAVs to drive our research. Already DAVs have allowed us to define new elements of Mg²⁺ block (Fedele et al., 2018), allosteric structures like the gating triad that mediate pre-gating steps in NMDAR activation (Ogden et al., 2017; Gibb et al., 2018), and gating hinges in outer structures that allosteric regulate the inner M2 pore loop (Amin et al., 2018). Nevertheless, DAVs at many points in the protein, such as the ATD and CTD, remain largely unexplored. A similar problem exists even in parts of the protein well-established such as the M3 segment. Such information is required to develop a highly refined pathomechanistic model with insights from DAVs that accounts for the complex machinery that regulates NMDARs. Such a model would represent a template for rational drug design and a starting point for personalized medicine.

Future directions and challenges

A full model of NMDAR structure-function, derived largely from in vitro studies as we have described in the present review, will be clinically invaluable. Still, it is only a starting point. Ultimately, to be truly of value to patients, this model must also take into account the complex regulation of NMDARs in vivo. We highlight just a few of the challenges here.

Most of the work done on DAVs has been to identify the properties of the channel for individual DAVs in vitro. However, less has been done to show how DAV-mediated changes in channel properties contribute to the clinical phenotype. For example, how DAVs that induce NMDAR hypofunction cause epilepsy, a disease associated with hyperfunction, cannot be understood without studying DAVs in a broader physiological context. The development of mouse models as well as other model organisms, such as zebrafish, using previously characterized DAVs will do much to address this shortcoming.

Another challenge is the cell biology of NMDARs. Clearly subunit composition, the number of receptors on the membrane, and their distribution, plays critical roles in NMDAR physiology. At present, most work to characterize NMDAR trafficking and surface expression has been done in vitro, typically in HEK293 cells. While this information is invaluable, it is only a starting point as real neurons handle these processes much differently. In addition, our traditional framework of the function of NMDARs is expanding (Dore et al., 2017). However, few models take these additional modes of action into account.

Current pathomechanistic models broadly distinguish DAVs by domain or by gain-of-function versus loss-of-function. However, the clinical challenge presented by patients is that DAVs have diverse and often unique effects, indicating that such broad characterization is insufficient. Hence, any model of NMDAR structure-function has to take into account the diversity of actions as well as the complex regulation of NMDARs in vivo. It will be a challenging task, but the reward – individuals who can survive and prosper with an NMDAR DAV – is worth the effort.

Abbreviations

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

AMPAR

AMPA receptors

ATD

amino-terminal domain

PDB

protein data base^*

CTD

carboxy-terminal domain

DAV

disease-associated variant

iGluR

ionotropic glutamate receptors

LBD

ligand-binding domain

MD

molecular dynamics

NMDA

N-methyl-D-aspartate

NMDAR

NMDA receptor

TMD

transmembrane domain