nmda receptor physiological activators and inhibitors a three-fold molecular and kinetic story

19
NMDA Receptor Physiological Activators and Inhibitors: A Three-fold Molecular and Kinetic Story Laurensius E. Mainsiouw. ARTICLE INFO ABSTRACT Article history: Submitted 25 March 2015 NMDA receptors are part of the tetrameric ionotropic glutamate receptor family of which the function is mainly tied to the electro- physiological function known as long term potentiation (LTP). This important phenomena underpins our understanding of memory, plasticity and neuronal development. The receptor itself demonstrates slow kinetics, is highly permeable to Ca 2+ and is voltage activated by virtue of the intrinsic Mg 2+ block. Each subunit of the tetramer contains functionally important N-terminal, Ligand binding and transmembrane domains which co-ordinate receptor function to exhibit its properties. Furthermore the GluN2 subunit has four different isoforms which contributes to differences in selective allosteric modulators. Data from recent NMDA crystal structures, electrophysiological and mutational studies are used to draw conclusions and generate ideas of receptor function. The mechanisms of three events will be investigated: Activation, Allosteric modulation and Voltage dependence. Single channel recordings and other biophysical methods have given us insight into kinetic schemes of receptor interactions between the agonists: Glutamate/Glycine, Zn 2+ and Mg 2+ block. However, understanding of the proton and possible voltage sensor is limited and unified kinetic model still needs to be elucidated. Keywords: LTP Agonist binding domain N-Terminal domain Kinetic model Single Channel Recordings Introduction N-Methyl D-Aspartate (NMDA) receptors are one of the four glutamate receptor variants. Similar to AMPA and kainate receptors, they are ligand gated ion channels. In synaptic transmission it is known that these receptors mediate the slow component of the excitatory post- synaptic potential, whereas the AMPA glutamate receptors are responsible for the fast component. However, these ion channels conduct sodium, potassium and importantly calcium ions across the cell membrane. This calcium entry is important as NMDA receptors are typically considered as ‘coincidence detectors’ (Malenka & Bear, 2004). This coincidence detection is due to NMDA receptor activation having a voltage dependence. Although these glutamate receptors are ligand activated, under resting physiological conditions an Mg 2+ ion is situated within the pore. In membrane depolarised states the Mg 2+ block is surmounted and the receptor is permeable to cations. This process is thought to occur by two ways: one, summation of frequency stimulation from pre-synaptic cells and two, integrating dendritic inputs via back propagating action potentials both of which contribute to a physiological response. This consolidation results in an altered threshold potential, and is the basis of long term potentiation (LTP), which has been linked to plasticity events such as cytoskeleton remodelling (Schwechter & Tolias, 2013). As well as plasticity events, NMDA receptors have also been implicated in a phenomenon known as excitotoxcicty

Upload: laurensius-mainsiouw

Post on 17-Feb-2017

273 views

Category:

Documents


1 download

TRANSCRIPT

NMDA Receptor Physiological Activators and Inhibitors:

A Three-fold Molecular and Kinetic Story

Laurensius E. Mainsiouw.

ARTICLE INFO ABSTRACT

Article history:

Submitted 25 March 2015 NMDA receptors are part of the tetrameric ionotropic glutamate

receptor family of which the function is mainly tied to the electro-

physiological function known as long term potentiation (LTP). This

important phenomena underpins our understanding of memory,

plasticity and neuronal development. The receptor itself demonstrates

slow kinetics, is highly permeable to Ca2+ and is voltage activated by

virtue of the intrinsic Mg2+ block. Each subunit of the tetramer

contains functionally important N-terminal, Ligand binding and

transmembrane domains which co-ordinate receptor function to

exhibit its properties. Furthermore the GluN2 subunit has four

different isoforms which contributes to differences in selective

allosteric modulators. Data from recent NMDA crystal structures,

electrophysiological and mutational studies are used to draw

conclusions and generate ideas of receptor function. The mechanisms

of three events will be investigated: Activation, Allosteric modulation

and Voltage dependence. Single channel recordings and other

biophysical methods have given us insight into kinetic schemes of

receptor interactions between the agonists: Glutamate/Glycine, Zn2+

and Mg2+ block. However, understanding of the proton and possible

voltage sensor is limited and unified kinetic model still needs to be

elucidated.

Keywords:

LTP

Agonist binding domain

N-Terminal domain

Kinetic model

Single Channel Recordings

Introduction

N-Methyl D-Aspartate (NMDA) receptors

are one of the four glutamate receptor

variants. Similar to AMPA and kainate

receptors, they are ligand gated ion

channels. In synaptic transmission it is

known that these receptors mediate the

slow component of the excitatory post-

synaptic potential, whereas the AMPA

glutamate receptors are responsible for the

fast component. However, these ion

channels conduct sodium, potassium and

importantly calcium ions across the cell

membrane. This calcium entry is important

as NMDA receptors are typically

considered as ‘coincidence detectors’

(Malenka & Bear, 2004). This coincidence

detection is due to NMDA receptor

activation having a voltage dependence.

Although these glutamate receptors are

ligand activated, under resting

physiological conditions an Mg2+ ion is

situated within the pore. In membrane

depolarised states the Mg2+ block is

surmounted and the receptor is permeable

to cations. This process is thought to occur

by two ways: one, summation of frequency

stimulation from pre-synaptic cells and

two, integrating dendritic inputs via back

propagating action potentials both of which

contribute to a physiological response. This

consolidation results in an altered threshold

potential, and is the basis of long term

potentiation (LTP), which has been linked

to plasticity events such as cytoskeleton

remodelling (Schwechter & Tolias, 2013).

As well as plasticity events, NMDA

receptors have also been implicated in a

phenomenon known as excitotoxcicty

which too also has a calcium basis

(Hardingham & Bading, 2010).

Receptor function is closely tied to

structure. This is evident as the three

ionotropic glutamate receptors share

common structural homology. The receptor

itself contains a tetrameric quaternary form,

in which the individual subunits have

distinct domains. They contain an

extracellular N-terminus and agonist

binding domains (ABD). The

transmembrane region of the protein is

made up of three helices M1/3/4 and a re-

entrant p-loop (pore-lining), sometimes

referred to as the M2 region. Finally it

contains an intracellular C-terminal domain

which has been associated with protein

trafficking and auxiliary proteins (Traynelis

et al., 2010).

To further understand how NMDA

receptors mediate their role in neuronal

physiology, we must look into how

activation of this receptor is controlled. Of

course as a glutamate receptor its

endogenous ligand is glutamate, however

various other mechanisms and ligands

influence how this receptor is activated. For

a complete understanding of how these

molecular machines work, we must be able

to quantitatively describe their action. To

do this kinetic models have to be created.

Multiple studies have allowed the

construction of such models (Banke &

Traynelis, 2003; Lester & Jahr, 1992).

These studies have extensively used

excised membrane patch clamp techniques

to study the electrophysiology of the

NMDA receptors. Typically they

investigate outside-out patches of

membrane containing single channels

Figure 1. Typical data analysed in single channel

experiments. (A) Demonstrates the burst recordings

of a GluN1/N2B receptor, notice how the waveform

is not that of an EPSP but rectangular implying the

quantal nature of receptor response. (B&C) Shows

the frequency of open and close times of the burst

data represented as a histograms. The MLE technique

was used to approximate the exponent probability

density functions. Importantly this statistical method

allowed estimates the five time constants of each

exponential, as shown in the shut time distribution

(B). (Banke & Traynelis, 2003)

A

B

C

(Stern, Behe, Schoepfer, & Colquhoun,

1992). The time recordings of the open or

closed states are plotted as histograms and

maximum likelihood estimation (MLE) is

used to generate an averaged trace. The

traces usually have multiple exponential

components, in which the time constants

are determined by MLE. Figure one is an

example of this technique. Here Banke &

Traynelis have measured and analysed the

burst patterns of a patch of membrane

containing one channel over a certain time

period. The square root of the counts are

represented by the histograms (ordinate)

and the shut time in milliseconds is

represented on the abscissa. The time

constants of each exponential are important

as they provide insight into all the rate

constants for entering that specific

conformational change. These

conformational changes are affected by

various ligands and the specific subunit

composition of the NMDA receptor in

question. Therefore, often is the case that

these time constants are analysed to

determine the rate constants of a reaction

mechanism.

Molecular and Pharmacological Traits

The kinetic properties of the receptor are

influenced by a multitude of factors. One of

these factors is the heterologous expression

of the three different types of subunits.

These are encoded for by seven genes. One

gene generates the GluN1 subunit and its

eight splice variants. Four genes encode the

GluN2 A/B/C/D subunits. Finally two

genes encode the GluN3 A/B subunits

(Glasgow, Siegler Retchless, & Johnson,

2015). Typically these NMDA subunits

form a tetramer from obligate heterodimers

in which a GluN1 combines with a GluN2

or GluN3 monomer, a dimer of dimers if

you will (Paoletti, 2011; Traynelis et al.,

2010). These tetramers are referred to as

diheteromeric however there are some

cases in which two different obligate

dimers come together to form a

triheteromeric structure. Recently it was

found that this triheteromeric state occurs

naturally within hippocampal synapses

(Tovar, McGinley, & Westbrook, 2013).

These differential structures alter the

binding kinetics of various agonists,

allosteric modulators and antagonists. This

is due to each subunit containing an altered

sequence and motif composition within the

N-terminal, agonist binding and

transmembrane domains, the locations

being where the ligands bind to. This can be

seen in Table.1, which describes these

gating and ligand binding differences.

Table 1. GluN2-subunit specific gating and ligand

binding properties

GluN1-1a+

N2A N2B N2C N2D

Conductance (pS) Main 50 50 37 37

Sub 38 38 18 18

Mean open time (ms) 3-5 3-5 0.5-1 0.5-1

EC50 (Glycine), µM 1.7 0.4 0.3 0.1

EC50 (Glutamate), µM 4 2 1 0.4

τoff (Glycine), ms 140 680 τoff (Glutamate), ms 40 300 300 2000

Po,peak 0.4-0.5

0.1-0.2

~0.01 ~0.01

Pf (Ca2+), % 18 18 8* n.d.

IC50 (Mg2+), µM (VM

=-100 mV) 2 2 12 12

IC50 (Zn2+), µM 0.01 2 20 10

IC50 (ifenprodil), µM >20 0.15 >20 >20

EC50 (CIQ), µM >30 >30 2.7 2.8

IC50 (DQP), µM 206 164 7.0 2.5

IC50 (H+), pH 6.9 7.5 6.6 7.5

Table was modified from (Acker et al., 2011; Mullasseril

et al., 2010; Paoletti, 2011), included in this table is a list

of the electrophysiological constants including

pharmacological constants of known ligands. DQP IC50

were averaged between human and rat samples.

This data also demonstrates that the N2A

receptors have similar electrophysiological

characteristics to N2B receptors, similarly

the same is also found between N2C and

N2D variants (Paoletti, 2011; Siegler

Retchless et al., 2012). This phenomena is

thought to be due to the high sequence

similarity between each subunit pair.

Gating Sequence Mechanisms

The ligand requirements for NMDA

receptor gating has been elucidated and

requires the binding of both glutamate and

glycine (Kleckner & Dingledine, 1988;

Mayer, 2006; Traynelis et al., 2010), the

affinities of which are found in table.1. The

binding affinities of each agonist is

determined by the individual subunit types,

however glycine binds to GluN1/3 and

glutamate to GluN2, as a rule of thumb. If

you notice in table.1 that the EC50 of

glycine is altered by the different isoforms

of GluN2, which will be explained by

certain structural elements. The molecular

structure of the external N-Terminus and

ABD are both described as bilobate or

containing a Venus fly trap motif (Figure.2)

(Glasgow et al., 2015; Karakas &

Furukawa, 2014; Traynelis et al., 2010).

Both lobes (D1 top, D2 bottom) of the ABD

have been found to be dimerized, which

allow the obligate subunits to alter the

binding affinities of the other subunits

respective ligand, and is important in

channel opening (Mayer, 2006). It has also

been found using mutational studies that

exchange of GluN2A/B/C/D N-terminal

domain (NTD) and linkers changes the

receptor Popen via interactions to the top

LBD lobes (D1) of the ABD (Gielen,

Siegler Retchless, Mony, Johnson, &

Paoletti, 2009; Yuan, Hansen, Vance,

Ogden, & Traynelis, 2009). In other

mutational experiments NTD and linkers

were altered on different GluN2 subunits

which subsequently altered ligand

energetics (Figure.2) (Glasgow et al., 2015;

Yuan et al., 2009).

In context to activation, glutamate and

glycine are both found to bind to sites

within the open top lobe clamshell ABDs.

Structural and electrophysiological data

from experiments using various partial

agonists of increasing size, showed that

agonist size decreased the measured current

from the receptor. Using X-ray crystal

structures, a model was produced

describing the mechanistic closing of the

lobes lead to channel opening (Furukawa &

Gouaux, 2003; Jin, Banke, Mayer,

Traynelis, & Gouaux, 2003). This

Figure 2. Above is a schematic representation of a NMDA obligate hetero-dimer consisting of GluN1-1a and GluN2B

subunits. The cartoon illustrates the movement within bottom ABD lobes upon agonist binding, which in turn causes a

reorientation of the TMD allowing the channel to open (Mayer, 2006). The lines connecting the A2 and D1 protein

domains are the N-Terminus linkers which also convey conformational information from the NTD to the ABD, however

this topic is under contention. The bi lobed venus fly trap motifs of both the NTD and ABD can also be seen on the crystal

structures (Karakas & Furukawa, 2014). Notice how the NTD can also bind ligands, in the above example the allosteric

modulator ifenprodil site is shown, but this region is also responsible in binding Zn2+, another allosteric modulator

(Erreger & Traynelis, 2008). Diagram adapted from (Karakas & Furukawa, 2014; Mayer, 2006).

Closed Open

hypothesised model proposed that the

binding of agonists induces a

conformational change whereby the bottom

lobes (D2) move distally away from the

TMD (Figure.2). The moving bottom D2

lobes are connected to the transmembrane

region of the receptor via linkers. Therefore

when these lobes undergo the proposed

conformational change, it is also transduced

into the pore region. This new conformation

of the p-loop and other transmembrane

helices allow for favoured cation entry

(Furukawa, Singh, Mancusso, & Gouaux,

2005; Mayer, 2006; Paoletti, 2011).

A caveat to the partial agonist experiments

is that they were conducted on AMPA

receptors, although similar work has been

done in NMDA systems. Using glycine

antagonists and X-ray crystal structures

another group confirmed a similar

mechanism exists for NMDA receptors. It

was found that the glycine antagonist

DCKA (5,7‐dichlorokynurenic acid)

stabilised GluN1 D1 and D2 domains in the

open confirmation at an angle of ∼21°,

whilst also observing no current flow. The

study also utilised site-directed

mutagenesis techniques and two critical

residues were found to be essential for

glycine binding at the GluN1 site. The first

residue is V689. Intriguingly the AMPA

glutamate binding GluA2 subunit has an

equivalent T655 residue in which the

hydroxyl group would hydrogen bond to

the glutamate γ‐carboxylate oxygen, a

chemical trait V689 lacks (Figure.3). The

second residue is W731 with the

homologous L704 in GluA2 subunits. The

molecular differences here come from the

indole ring found on the GluN1 W731

which would sterically hinder the carboxyl

group of glutamate upon binding

(Figure.3). It is important to understand the

similarities and evolutionary differences

that lead to a preferred glycine binding in

GluN1 subunits compared to GluA2, as the

glutamate binding residues in GluA2 are

conserved in the GluN2 subunit which

perhaps undergoes the same binding and

gating mechanism as AMPA receptors.

Therefore it is probable that NMDA and

AMPA receptors gate in a similar manner

(Furukawa & Gouaux, 2003; Furukawa et

al., 2005).

Crystal data obtained from a GluN1a/N2B

NMDA receptor (Figure.2) has shown that

it does share morphological similarity to the

AMPA receptor (Karakas & Furukawa,

2014). However, upon closer inspection it

is much more compact and the domains

have been pictured with high intimacy with

regards to intra-protein interactions,

especially in the NTD. It is clear that the

previous hypothesised model will most

likely require revision to include the

different sub-conductance states and pre-

gating steps (Banke & Traynelis, 2003;

Schorge et al., 2005). An implication in

regards to the compactness of the receptor

is that this may be the reason why it is a

target of allosteric modulation of which the

AMPA receptor is not. The high subunit

interfaces could be a factor in the subtype

dependent affinities as discussed

previously. A small aside to the crystal data

Figure 3. Superimposed image of a GluN1 subunit

binding glycine (Black sticks) and GluA2 subunit

binding glutamate (Grey sticks). The changes in

amino acid residues are shown by the green (GluN1)

and red (GluA2) colouring. The red boxes indicated

the two critical substitutions to achieve glycine

affinity. Adapted from (Furukawa & Gouaux, 2003).

is that it was only determined at a low

resolution of 4Å, due to the difficulties in

purifying membrane protein. This means

that the data is useful to identify critical

secondary structure, however it is not

precise enough to look at residue

interactions.

Overall however it must be remembered

that the Mg2+ pore block must still be

overcome after the gating procedures for

induction of the NMDA component of the

excitatory post synaptic potential. This

mechanism will be subsequently discussed

in a later chapter.

Receptor Desensitisation

Desensitisation events in this receptor is

poorly understood, nevertheless

preliminary ideas have been suggested. It is

thought that upon binding of the agonists

and activation of the NMDA receptor, two

possible conclusions can occur. One, the

agonists unbind and the receptor returns to

rest, or two, the agonist bound receptor

enters a desensitised state. The proposed

extension to the previous model is that the

top lobes (D1) of the ABD de-dimerize and

prevent the bottom lobes moving which in

effect stops the TMD reorganising into the

conducting state (Mayer, 2006). However,

as with previous NMDA receptor

hypothesises this was speculated from

AMPA crystal data. With the recently

determined structure however it is now

possible to test these hypothesises by trying

to image the receptor in such

conformations.

Kinetics of Activation/Desensitisation

To determine kinetic schemes patch clamp

data is analysed, as explained in the

introduction. The shut-time distributions of

a single NR1/2B receptor patch in

Figure.1C gave five exponential

components. The hypothesis was that of

these time constants, some belonged to the

binding of glutamate and glycine (Banke &

Traynelis, 2003). They reasoned that if

partial agonists were applied and there was

an observed change in a specific time

constant, then that constant must be

attributed to the binding site for the

substituted agonist. Upon use of the

glutamate partial agonists with glycine, the

third shut time constant τSHUT3 was found to

have doubled when either quinolinic acid

(QUIN) or NMDA was applied, 208±24

and 196±29 percent of the original value

respectively (table Figure.4). When glycine

partial agonists: HA-966 and cycloserine

(CS) were tested, the second time constant

τSHUT2 also increased by 186±28 and

167±20 percent of the original value,

respectively (table Figure.4). The first time

constant (τSHUT1) did not alter with the use

of any substituted partial agonists. It was

interpreted that this time constant defined

that gating conformation as only one

conductance state was observed. This

gating and co-agonist kinetic data was

thought to show a concerted process. It was

suggested that the kinetic scheme contained

a cycle, in which glutamate or glycine

would be required to bind and each subunit,

GluN1 and GluN2 must enter some active

state before channel opening occurs

(Figure.4). These determined time

constants reflect the molecular movements

involved in binding and hence provide a

temporal understanding of the molecular

mechanism.

The original ideas for the physico-chemical

basis of desensitisation was explored by

Lester and Jahr. They found that the affinity

of the agonist altered the rate of

desensitisation. From this finding they

suggested that desensitisation accounted for

the slow component of the NMDA EPSC,

however this is not the case as proven by

Banke & Traynelis. Their work did produce

an initial bi-agonist kinetic scheme

whereby the receptor could enter some

desensitised state as shown below, which

provided the groundwork for later models

(Lester & Jahr, 1992):

This scheme was further elaborated on by

Banke & Traynelis partial agonist study,

where their model of independent

activation events in GluN1 (fast) and

GluN2 (slow) subunits could have different

kinetics. They also postulated that this

could also be the case for desensitisation in

the receptor, where either subunit could

enter and exit this state:

This model was the first to explain the

kinetic contributions of both agonist

binding subunits. It also conceptually

explains the slower kinetics of the NMDA

receptor in the EPSC compared to the

AMPA component. This model also shows

some agreement to the X-ray crystal data in

which two lobes of both subunit LBDs must

close upon bound agonist to induce an

opening in the TMD, which leads to the

activation of the receptor. However the

study falls short by answering this question

by only testing GluN2B subunits, if all four

N2 subunit isoforms (and LBD chimeras)

were tested and changes in τ(SHUT3) were

observed it would be logical to accept the

hypothesis. Importantly if the GluN1

subunit was preserved in the above

experiment which resulted in an altered

τ(SHUT2); this observation gives further

evidence that the dimerization between the

LBD lobes may allow the lobes to co-

ordinate agonist binding conformational

changes. A final caveat to this data is that

the five time constants are minimised

estimates and are highly reliant on the

Figure 4. Cartoon Scheme for NMDA

receptor activation upon glutamate and

glycine binding. The top mechanism is

demonstrating the N1-N2-N1-N2

configuration (Karakas & Furukawa,

2014). The table below lists the shut time

constants (τ) when alternate glutamate and

glycine agonists were used. The first shut

time constant τ(SHUT1) was not changed

relative to the control in any of the agonist

substations. However, τ(SHUT2) appeared to

increased depending if different glycine

agonist was applied, whereas τ(SHUT3)

changed when glutamate agonist was

substituted Adapted (Banke & Traynelis,

2003).

resolution of the shut/open time

distribution. Therefore it is possible that

one single component could be a composite

of multiple components.

Further review of this kinetic model by

Schorge et al has shown that even Banke-

Traynelis model may still be an

oversimplification (Schorge et al., 2005).

This is because in the previous model only

one open time component was considered

and steps involving agonist binding were

not elaborated. Schorge et al found that

instead of one there was two conducting

states of the receptor, similar to nAChR and

GlyR. It was proposed that the multiple

conducting states were due to partially

bound states, however, when single channel

recording were conducted in four differing

solutions of alternating saturating

concentrations of glycine and glutamate, it

was observed that the correlation between

open and shut times was constant. It was

instead interpreted as two possible open

states could be reached after the pregating

sequence was complete (Figure.5). Their

mechanism also included a tetrameric

binding sequence in which all nine possible

NMDA receptor bound states were possible

with rate constants entering and leaving all

states (Figure.5).

NMDA NTD Allosteric modulation

Allosteric modulators can be defined as

molecules that bind to a receptor protein

and indirectly alter its function without

directly acting on the agonist orthosteric

binding site, but rather bind to its own

distinct site. Interestingly the NMDA

receptor is a target for many allosteric

modulators, in which there are multiple

binding sites. These allosteric sites are less

conserved between the GluN2 subunit

isoforms and therefore these ligands can be

subunit specific. The effect produced upon

binding can be either an up regulation of

function, and these molecules are known as

positive allosteric modulators (PAMs) or

molecules that cause a down regulation of

function, which are referred to as negative

allosteric modulators (NAMs). (Glasgow et

al., 2015; Paoletti, 2011; Traynelis et al.,

2010).

One of these allosteric modulators is Zn2+.

These endogenous small ions are found to

accumulate in vesicles of specific

glutamatergic forebrain neurons (Amico-

Ruvio, Murthy, Smith, & Popescu, 2011;

Traynelis et al., 2010). Zn2+ is known to act

as a NAM and modulate GluN2A NMDA

receptors with high specificity; EC50 of

10nM (Table.1). The mechanism of action

is not well understood, however it is

thought that these ions act in a bipartite

fashion: binding on the NTD reducing the

receptor Popen and cause a voltage

dependent block of the pore similar to

Mg2+. To understand the modulatory effect

of Zn2+, we must re-examine the molecular

structure of the receptor. As mentioned

earlier the NTD is bilobate similar to the

LBD (Figure.2), and it is within this domain

where the Zn2+ allosteric site lies. It was

determined that this site contained two

histidine residues (H42/H44) which co-

ordinate with the Zn2+ and underlie its high

Figure 5. Schorge et al NMDA receptor mechanism

with rate constants of each state being determined by fits

to the data. However, only 12 of the possible 33 (bi-

agonist action) were determined. The first 9 (from the

left) states describe the agonist binding stage. The next 4

states within the cycle incorporates Banke & Traynelis

pre-gating steps. Of the last 3 states the non-shaded

represents the desensitised receptor, whereas the shaded

versions are the two possible open states. Taken from

(Schorge, Elenes, & Colquhoun, 2005)

specificity and mutation of these residues

caused a 200-fold increase in EC50 (Choi &

Lipton, 1999). The effect mediated by Zn2+

differs depending if the receptor is exposed

to a saturating or sub-saturating Zn2+

concentration. It was found that in a sub

saturating concentration the glutamate

induced current rapidly declines. The

reason for this is an increase in Zn2+ affinity

upon glutamate binding in GluN2A.

Whereas in a saturating concentration the

NTD Zn2+ binding sites are already

occupied and therefore already have been

inhibited before glutamate is applied. This

phenomena was termed Zn2+ dependent

desensitisation (Amico-Ruvio et al., 2011;

Zheng et al., 2001). The Zn2+ inhibition

decreases the Popen of the receptor and this

was due to an increase in desensitised states

which was observed as a decrease in

openings within a burst with no change in

inter burst time. However a caveat with the

experiments is that the N/G NMDA

receptor mutants used to remove Mg2+

inhibition displayed different gating

kinetics compared to the native receptors. It

is important to note that this NTD allosteric

Zn2+ site is also the target of other

molecules, most notably drugs such as

ifenprodil acting on the GluN2B subtypes

Figure 6. Schematic diagram of what is thought to occur upon Zn2+ allosteric modulation. Experiments comparing inward

current between steady-state Zn2+ application and pulse (peak) Zn2+ application. The graph of % current change against

[Zn2+] is plotted and a bell shaped curve is found (Zheng et al., 2001). This is indicative of the following mechanism,

whereby binding of glutamate to the GluN2 subunit increases Zn2+ affinity to the bound receptor complex. Zn2+ binds to

the NTD and causes a conformational change resulting in LBD de-dimerization and the TMD-LBD linkers to be altered,

allowing for higher proton affinity.

(Paoletti, 2011; Traynelis et al., 2010).

However, this is not the only allosteric site

for which there are other molecules which

are selective for the GluN2C/D subunits.

Zn2+ and H+ Inhibition Mechanism

Interaction at the NTD appears to reduce

receptor activity, however the next question

is, how does the binding of these allosteric

modulators affect gating and what are the

molecular dynamics behind this

phenomenon? To answer this question we

must look at another NMDA receptor

allosteric modulator, protons. It has long

been known that NMDA receptors are

sensitive to protons (H+), and it was

observed that these molecules do inhibit

receptor activity but not in a competitive

manner (Traynelis & Cull-Candy, 1990;

Traynelis et al., 2010). Further work in the

phenomena has found that mutations in the

linker between the M3 transmembrane

helix and the D2 domain in the GluN1

subunit alter proton sensitivity and gating.

Similarly mutations within the D2-M4 helix

linker in the GluN2A/B subunit also

affected pH sensitivity (Figure.6) (Low et

al., 2003). Data produced from the Zn2+

inhibition experiments have also shown that

the Zn2+ modulation has a pH dependence

(Choi & Lipton, 1999; Zheng et al., 2001).

The proposed mechanism for Zn2+

modulation is when glutamate binds to the

GluN2 subunit, this increases the affinity

for Zn2+ to the NTD binding site. Upon

binding this causes a conformational

change in the receptor and the linker

between LBD and TMD, resulting in a

higher proton affinity and thus a lowered

Popen. This conformational change is

perhaps transmitted via the de-dimerization

of the LBD domains, a phenomena/idea

previously described. This conclusion was

made from data recorded from outside-out

patch experiments where two types of Zn2+

application was used:

1. Steady-State (SS), where the Zn2+

concentrations were allowed to

equilibrate with the patch.

2. Peak (PK), where the Zn2+ is

applied in a short pulse.

It is also important to note that glutamate

was applied in a short pulse (400-500ms) to

prevent glycine induced desensitisation;

this was kept constant. The data was plotted

as % current change as the ordinate and

Zn2+ concentration as the abscissa. It was

found that relationship between the two

variables was that of a bell shaped curve.

This is in agreement with our model as at

low steady state Zn2+ concentrations it

would not saturate the binding sites and

when the Zn2+ pulse is applied the short

application time would most likely cause

less than average binding saturation to

occur. Therefore ISS>IPK and 1 −𝐼𝑆𝑆

𝐼𝑃𝐾 would

be small at low [Zn2+]. This would also be

the case for high concentrations as all the

receptors in the steady state conditions will

be equilibrated, whereas when a pulse

application of high Zn2+ would cause a

larger population of the receptor population

to be in the Zn2+ bound state (Zheng et al.,

2001).

Although we have discussed Zn2+ binding,

how do protons fit into these kinetic

schemes, the decreased Popen is supposed to

be due to a molecular change in the proton

sensor. Further studies have shown that

protons and Zn2+ inhibit in a similar manner

(Erreger & Traynelis, 2008). Here they

explored the possible kinetic schemes in

which the proton interaction is accounted

for. They first suggested that either

protons/Zn2+ bind to the receptor in a

sequential manner leading to an alteration

in channel opening, in which protonation

and channel closure are independent events.

The second scheme utilised ideas from the

previously described activation studies

(Banke & Traynelis, 2003), whereby the

protons/Zn2+ bind and alter the GluN1/2

subunits independently. They found this

hypothesis to marginally better fit the

experimental data leading to a second

scheme (Erreger & Traynelis, 2008):

This model is novel as it does explain Zn2+

role in proton sensitivity as a reduction in

the proton dissociation rate and the

decreased channel open time is portrayed as

the faster closing of protonated receptors

(data not shown) (Erreger & Traynelis,

2008). Although this model does fit the data

well, their other proposed models did so as

well but were rejected upon conceptual

discrepancies (independent protonation and

gating steps). It should also be noted that

Zn2+ has been exclusively considered as an

allosteric modulator that acts on the NTD.

However, evidence suggests that it acts as a

pore blocker similar to Mg2+ and as such

these previous models should be revised to

account for such a phenomenon. This

compounded with our understanding that

Zn2+ and H+ are highly concentrated in

glutamatergic vesicles (Salazar, Craige,

Love, Kalman, & Faundez, 2005; Smart,

Hosie, & Miller, 2004) and by extension

could play an important part in glutamate

signalling. Therefore it is imperative that

we determine a rigorous kinetic model. This

will most likely be the case once the

molecular determinants of the proton sensor

have been elucidated and higher resolution

patch clamping and X-ray crystal methods

have been developed.

GluN2C/D Allosteric Modulators

Allosteric modulators have also been found

to act on the GluN2C/D receptor subtypes.

However, compared to Zn2+ (2A) and

ifenprodil (2B) the two new modulators,

CIQ and DQP-1105, are both selective for

2C/D (Table.1) (Acker et al., 2011;

Mullasseril et al., 2010). Another difference

of CIQ and DQP is that they do not act on

the NTD, but rather on the TMD helices

(CIQ) or the ABD lower lobe (DQP)

(Figure.7).

CIQ is also a PAM whereas the other three

are NAMs. Using site directed mutagenesis

and measuring % current from maximum it

was determined that the allosteric site was

on the M1 transmembrane helix and

potentiation is mediated by GluN2 pre-M1

region and GluN1 M4 helix (Figure.7)

(Ogden & Traynelis, 2013).

As previously stated DQP does not act on

the NTD but rather the D2 lobe of the ABD

(Figure.7). It was postulated that the

measured decreased POpen was due to the

molecule increasing the energy barrier of

the Banke & Traynelis pre-gating step. This

allosteric site was found using site directed

mutagenesis whilst measuring the %

inhibition change attributed to that mutation

(Acker et al., 2011).

These findings are seminal, as the

development of N2 specific ligands

especially allosteric modulators will allow

us to better target pathologies associated

with NMDA receptors (excitotoxicity).

This is due to the N2 subunit being

differentially expressed in tissues, which

the disease may be based. Since the finding

of the crystal structure it may now be

possible to identify how the conformational

changes induced by these modulators are

propagated.

Figure 7. Cartoon Electron density map of a typical

NMDA receptor. Above are all the known allosteric

modulator binding sites. Taken (Ogden & Traynelis,

2013)

Mg2+ Selectivity of Physiological Block

Briefly described earlier was the

phenomena whereby at resting conditions

the NMDA receptor pore is blocked by

Mg2+. This has interesting consequences, as

the Mg2+ block has to be surmounted by the

membrane becoming depolarised before it

enters the conducting state (Figure.8)

(Khazipov, Ragozzino, & Bregestovski,

1995; Traynelis et al., 2010). This means

that although the NMDA receptor is a

ligand gated ion channel it also has an

associated voltage dependence (Figure.8).

Mutagenesis studies have allowed us to

observe the critical amino acid residues

within the structure that govern the Mg2+

block. It was found that on both GluN1/2B

subunits asparagine 589 on the P-loop (M2)

was connected to mediating the Mg2+ block

of the receptor. In this study a N589Q

mutation was used in both subunits and

these receptors were expressed in

Xenopus/HEK systems. This point

mutation (in either subunit) caused the

receptor to lose its affinity to Mg2+ where it

was found that two orders of magnitude

higher concentration of Mg2+ were required

to reach the same amount of block as wild

type receptors. The same study also

demonstrated that the block of the drug

MK-801 was similarly affected, implying

that both Mg2+ and MK-801 bind to the

same site and block the ionophore (Mori,

Masaki, Yamakura, & Mishina, 1992). A

caveat to their work was that only the

GluN2B isoforms were used. However,

another caveat to the study was that only a

glutamine mutation was used. Glutamine is

structurally and physicochemically similar

to asparagine, as the R group of glutamine

is only extended by CH2 group. Although

this small change did give rise to a large

difference in Mg2+ affinity, it would be

fascinating to assess the effects of other

substituted amino acids. For instance if an

aspartate mutation was used would the

extra negative charge cause an increase in

Mg2+ binding affinity? If that’s the case

then perhaps the 589 residue directly

interacts with the pore blocking agent.

Another point of interest is that this residue

was determined due to homologous residue

differences defining Ca2+ permeability’s of

some AMPA receptors.

Mg2+ sensitivity has also been found to be

different depending on which GluN2

subunit is present in the tetramer. Two

groups emerged from observing the four

subtypes A/B/C/D, whereby A and B were

found to have a high affinity to Mg2+ block

and high permeability to Ca2+, but C and D

subunits displayed the opposite properties.

A critical residue substitution within the

transmembrane M3 helix causes this

difference in properties (Figure.9). It was

found that at position 632 a serine is present

in the A/B subunits, whereas a leucine is

present in the C/D variants. Site directed

mutagenesis exchanging the respective

amino acid to the other groups: GluN2A

S632L and GluN2D L632S; caused the

electrophysiological properties of the

channel to mimic the respective group

(Figure.9) (Siegler Retchless et al., 2012).

As this residue site was on the M3 helix and

not directly in the pore, it was hypothesised

that its action was mediated by an

Figure 8. I-V plot of a NMDA receptor in presence

and absence of Mg2+. In presence of Mg2+ not much

current is being conducted by the receptor until the

membrane is sufficiently depolarised. In contrast to

this the data without Mg2+ displays linear changes in

current with respect to membrane potential.

Interestingly the data also shows that the reversal

potential (Vrev) of Mg2+ is 0mV. (Data produced from

NEURON simulation plotted on OriginPro)

interaction with another amino acid. To test

this hypothesis bioinformatical methods

such as homology modelling, where known

crystal structures such as the bacterial

KcsA/NaK cation channels are used to

predict residue interactions in the NMDA

pore which at the time had not been

crystallised (Glasgow et al., 2015). These

methods found two possible tryptophan

residues (608 & 611) on the GluN1 subunit

which could mediate the effect of S/L 632

residue. Mutant Cycle analysis quantified

by the coupling coefficient showed that

W608 had an interaction with the subunit

determining 632 S/L amino acid (Siegler

Retchless et al., 2012).

Another study has revealed that the NMDA

receptor voltage dependence may not be

wholly due to the Mg2+ ion occluding the

pore, but rather an intrinsic sensor

positioned on the GluN2 subunit (Clarke &

Johnson, 2008). A previous paper showed

that the GluN2A/B subtypes regulate Mg2+

block in different kinetic manner, with two

steps where one is fast and the other slow

(Clarke & Johnson, 2006; Vargas-

Caballero & Robinson, 2003). In

experiments involving a Mg2+ free medium

it was observed that a repolarisation from

different voltage steps (95mV to -65mV

and 55mV to -65mV) causes a change in tail

current amplitude and relaxation kinetics.

In the larger step it was shown that current

A

C

B

D

Figure 9. Data from mutagenesis experiments of the 632 residue, comparing the effect of Mg2+ affinity between GluN2

A and D subtypes. (A) Is a cartoon schematic of the transmembrane and pore region of the NMDA receptor, showcasing

the 632 residue found on the M3 helix (Square). The two mutations are the opposite residues found on the GluN2A group

and GluN2D group. (B) Demonstrates the changes in Mg2+ affinity relative to membrane potential, notice how the 2D

subtype has a lower affinity at all voltages compared to the 2A subtype. (C&D) Are the same plots as B, however the

additional trace shows the effects of the mutation. The mutation of S to L in the 2A subtype clearly causes the Mg2+

voltage dependent affinity to closely resemble the unchanged 2D subtype. In contrast to this the reverse mutation on the

2D subunit causes its Mg2+ affinity to be similar to the unchanged 2A subtype. Adapted from (Siegler Retchless, Gao, &

Johnson, 2012)

amplitude was greater and relaxation was

slower compared to the smaller step,

implying the receptor has voltage

sensitivity even without Mg2+ present

(Figure.10)(Clarke & Johnson, 2008). A

caveat to this work is that the large voltage

step of 95mV is not a typical physiological

membrane potential and thus may not

accurately model the receptor, to counteract

this perhaps multiple voltage increments

could be used. This phenomenon may be

mediated by conformational changes within

the pore and TM regions previously

discussed (M3 helix), however no evidence

has been found for this hypothesis but may

be a point of interest in the future.

These data further reinforce how the

voltage control of the receptor is centralised

to the pore and transmembrane region. The

control of Mg2+ block has also shown to be

tightly regulated by various amino acids

and also is categorised by the GluN2

subunit isoform. It was also found that the

GluN2 subunit alters Mg2+ gating in a

kinetic fashion (Clarke & Johnson, 2006).

This may imply that differential expression

of NMDA receptor tetramers in brain

structures leads to altered kinetics of

plasticity utilised in different neuronal

outputs. The elucidation of the critical

tryptophan residue on the GluN1 subunit

may imply that conformational changes

occur within the tetramer upon

depolarisation leading to relief of the Mg2+

on the pore. Another hypothesis is that

these residues may be involved in some

intrinsic voltage sensor. However, it is clear

that understanding the amino acids

underpinning the Mg2+ and voltage

sensitivity could give us insights into LTP,

and may aid in designing new drugs which

could alter these plasticity events or block

the channel itself.

Kinetics Underlying Mg2+ Block

Since the nature of block is uncompetitive

initial kinetic del-Castillo-Katz models

conceptually added another term in after

receptor activation, indicative of two states

blocked and unblocked:

𝐴2𝑅𝐸⇔𝐴2𝑅

∗ +𝑀𝑔2+𝐾𝑀𝑔⇔ 𝐴2𝑅

∗𝑀𝑔

Where efficacy (E) is a dissociation

constant defining the receptor entering its

conductive state and KMg is the Mg2+

dissociation constant. The A2R term defines

the agonist bound state but the channel is

still closed, A2R* defines the conducting

state of the channel and A2R*Mg describes

the open but blocked form of the channel.

This interpretation was developed by

studying local anaesthetics on endplate Ach

receptors, but is still applicable as the local

anaesthetic blocks pore similarly to Mg2+

(Khazipov et al., 1995; Neher & Steinbach,

1978).

This model appears to be too simplistic in

describing the Mg2+ block, as it was found

that there were two kinetic components

involved the Mg2+ release: a fast and slow

phase (Vargas-Caballero & Robinson,

2003). This was determined when current

kinetics were fitted with the archetypical

exponential function (Figure.10). However,

the best fits to the trace were not made with

one exponential component but rather two:

𝐼(𝑡) = 𝐼𝑆 {𝐴1 [1 − 𝑒−𝑡

𝜏1] + 𝐴2 [1 − 𝑒−𝑡

𝜏2]} (1)

Where A1+A2=1 and IS is the stationary

current at each voltage step. Figure nine

showcases the time constants for both

components. It was also observed that as

the cell become more depolarised A1

linearly decreased whereas A2 increased.

Caveats to this work is that NMDA was

used as the agonist. As the recordings

utilised cortical pyramidal cells, it was

necessary to selectively activate NMDA

receptors and no other glutamatergic

receptors for example metabotropic

receptors. This methods drawback is that

NMDA kinetics are slightly different

compared to glutamate (Banke & Traynelis,

2003) which may contribute to the current

kinetics data. i

Another group found a two exponential

equation also gave the best fit to Mg2+

block, yet they found that the slow second

component was in fact due to the receptor

having intrinsic voltage dependence,

similar to the nicotinic acetylcholine

receptor (Figure.10) (Clarke & Johnson,

2008; Magleby & Stevens, 1972). As

described earlier, they observed that the tail

currents produced by different steps of

repolarisation showed different time decays

in the peak current (Figure.10) (Clarke &

Johnson, 2008). Furthermore it was

suggested that this voltage dependence was

due to the S632L mutation on the M3 helix

(Siegler Retchless et al., 2012). With this

information they produced an alternate

kinetic model using earlier ideas from

Banke & Traynelis. The original model

utilised two different gating steps, RA2s and

RA2f states, where the receptor has to enter

both states in any order before an active

receptor conformation is attained. Clarke &

Johnson have postulated that the rate

constants that determine entry and

departure of these steps may carry some

form of voltage dependence. Simulations

showcased that if this voltage attribute was

modelled on the slow constant (Ks+/Ks-) it

generated better fits onto the experimental

waveforms compared to modelling the fast

state constants. When modelling, they also

considered that the Mg2+ may be bound

within every state in Banke & Traynelis’s

scheme, as the slow unblock and current

relaxation in 0 Mg2+ shared many features

(Clarke & Johnson, 2008). It was suggested

that a mechanism would contain a blocked

arm that describes the events preceding

Mg2+ relief:

These findings are useful as understanding

the timings and mechanism of the Mg2+

block underpins long term potentiation.

LTP underlies many neuro-biological

phenomena, such as memory/learning, pain

sensitisation and in some pathological

states such as Alzheimer’s disease.

Conclusion

The NMDA receptor is an immensely

complicated macromolecule. Part of the

complexity stems from the multiple ligands

that are able to bind to it and alter its

function. This is further compounded by the

multiple isoforms of the subunits of the

receptor which demonstrate different

affinities to the ligands, Popen and Mg2+

block. Although, this review has not

heavily considered that these subunits can

be differentially expressed in cortical

Figure 10. (A) The top four panels showcase current-

time plots generated from data at different voltage

clamps. The line that passes through the trace is the

fitted two exponential equation (1). The two time

constants generated by this model show that relief of

Mg2+ block is not as instantaneous as originally

thought. (B) The bottom fifth panel highlights the

voltage dependence of the fast unblocking fraction of

current (A1). (C) The sixth panel are repolarisation

tail currents measured at different voltage steps (95

and 55mV to -65mV). You can see the relaxation

kinetics are different for each step. Importantly this

data was obtained in a 0 Mg2+ solution. Adapted from

(Clarke & Johnson, 2008; Vargas-Caballero &

Robinson, 2003)

structures, it is important to identify these

differences and the contextual use within

these structures (Paoletti, 2011).

Furthermore identification of receptor

properties aids in understanding the role it

plays in physiology and pathology. For

example we know that the receptor has an

associated voltage dependence which is

integral to its physiological function, LTP.

Additionally, multiple studies have been

successful in elucidating the molecular

determinants and kinetics behind binding

and activation.

AMPA receptors and other ion channels

have previously been used to model the

NMDA receptor, which has successfully

led to many novel advancements. But

recently the crystal structure has been

determined which will now be used as the

basis in future studies. However, further

advancement in this field requires higher

resolution structures of the NMDA receptor

itself to be used to observe activation,

gating and desensitisation conformational

changes specific to this receptor. These

high resolution models could also be used

to identify key elements and residues of the

receptor that we do not understand yet, for

instance the placement and activity of the

proton and voltage sensors.

The ‘holy grail’ of biological understanding

so to speak is the structure function

relationship. To truly understand the

molecular mechanisms underlying function

of receptors, data from single channel

recordings interpreted into kinetic models

are used to provide a timeline of such

events. For the NMDA receptor such ideas

have been successful in describing certain

parts of the ‘timeline’, for instance

activation and desensitisation has been

explained by Banke & Traynelis, where

they suggested the pre-gating steps are

conformational changes in the GluN1/2

subunits. Similarly for Zn2+ and H+

attempts have been made by Low et al,

Erreger and Traynelis and finally Mg2+

block have been made by Clarke &

Johnson. However the final goal is to

describe all of these molecular activity’s in

one singular model. This will be extremely

difficult as there may be more ligands for

the NMDA receptor and the currently

known ligands may interact unexpectedly

with other parts of the receptor. However, a

complete understanding in this receptor

function could have large ramifications in

treating many neurological pathologies

such as Alzheimer’s, Parkinson’s disease

and phenomena such as excitotoxicity.

References

Acker, T. M., Yuan, H., Hansen, K. B., Vance, K. M., Ogden, K. K., Jensen, H. S., . . . Traynelis, S. F. (2011). Mechanism for noncompetitive inhibition by novel GluN2C/D N-methyl-D-aspartate receptor subunit-selective modulators. Mol Pharmacol, 80(5), 782-795. doi: 10.1124/mol.111.073239

Amico-Ruvio, S. A., Murthy, S. E., Smith, T. P., & Popescu, G. K. (2011). Zinc effects on NMDA receptor gating kinetics. Biophys J, 100(8), 1910-1918. doi: 10.1016/j.bpj.2011.02.042

Banke, T. G., & Traynelis, S. F. (2003). Activation of NR1/NR2B NMDA receptors. Nat Neurosci, 6(2), 144-152. doi: 10.1038/nn1000

Choi, Y. B., & Lipton, S. A. (1999). Identification and mechanism of action of two histidine residues underlying high-affinity Zn2+ inhibition of the NMDA receptor. Neuron, 23(1), 171-180.

Clarke, R. J., & Johnson, J. W. (2006). NMDA receptor NR2 subunit dependence of the slow component of magnesium unblock. J Neurosci, 26(21), 5825-5834. doi: 10.1523/JNEUROSCI.0577-06.2006

Clarke, R. J., & Johnson, J. W. (2008). Voltage-dependent gating of NR1/2B NMDA receptors. J Physiol, 586(Pt 23), 5727-5741. doi: 10.1113/jphysiol.2008.160622

Erreger, K., & Traynelis, S. F. (2008). Zinc inhibition of rat NR1/NR2A N-methyl-D-aspartate receptors. J Physiol,

586(3), 763-778. doi: 10.1113/jphysiol.2007.143941

Furukawa, H., & Gouaux, E. (2003). Mechanisms of activation, inhibition and specificity: crystal structures of the NMDA receptor NR1 ligand-binding core. EMBO J, 22(12), 2873-2885. doi: 10.1093/emboj/cdg303

Furukawa, H., Singh, S. K., Mancusso, R., & Gouaux, E. (2005). Subunit arrangement and function in NMDA receptors. Nature, 438(7065), 185-192. doi: 10.1038/nature04089

Gielen, M., Siegler Retchless, B., Mony, L., Johnson, J. W., & Paoletti, P. (2009). Mechanism of differential control of NMDA receptor activity by NR2 subunits. Nature, 459(7247), 703-707. doi: 10.1038/nature07993

Glasgow, N. G., Siegler Retchless, B., & Johnson, J. W. (2015). Molecular bases of NMDA receptor subtype-dependent properties. J Physiol, 593(1), 83-95. doi: 10.1113/jphysiol.2014.273763

Hardingham, G. E., & Bading, H. (2010). Synaptic versus extrasynaptic NMDA receptor signalling: implications for neurodegenerative disorders. Nat Rev Neurosci, 11(10), 682-696. doi: 10.1038/nrn2911

Jin, R., Banke, T. G., Mayer, M. L., Traynelis, S. F., & Gouaux, E. (2003). Structural basis for partial agonist action at ionotropic glutamate receptors. Nat Neurosci, 6(8), 803-810. doi: 10.1038/nn1091

Karakas, E., & Furukawa, H. (2014). Crystal structure of a heterotetrameric NMDA receptor ion channel. Science, 344(6187), 992-997. doi: 10.1126/science.1251915

Khazipov, R., Ragozzino, D., & Bregestovski, P. (1995). Kinetics and Mg2+ block of N-methyl-D-aspartate receptor channels during postnatal development of hippocampal CA3 pyramidal neurons. Neuroscience, 69(4), 1057-1065.

Kleckner, N. W., & Dingledine, R. (1988). Requirement for glycine in activation of NMDA-receptors expressed in

Xenopus oocytes. Science, 241(4867), 835-837.

Lester, R. A., & Jahr, C. E. (1992). NMDA channel behavior depends on agonist affinity. J Neurosci, 12(2), 635-643.

Low, C. M., Lyuboslavsky, P., French, A., Le, P., Wyatte, K., Thiel, W. H., . . . Zheng, F. (2003). Molecular determinants of proton-sensitive N-methyl-D-aspartate receptor gating. Mol Pharmacol, 63(6), 1212-1222. doi: 10.1124/mol.63.6.1212

Magleby, K. L., & Stevens, C. F. (1972). A quantitative description of end-plate currents. J Physiol, 223(1), 173-197.

Malenka, R. C., & Bear, M. F. (2004). LTP and LTD: an embarrassment of riches. Neuron, 44(1), 5-21. doi: 10.1016/j.neuron.2004.09.012

Mayer, M. L. (2006). Glutamate receptors at atomic resolution. Nature, 440(7083), 456-462. doi: 10.1038/nature04709

Mori, H., Masaki, H., Yamakura, T., & Mishina, M. (1992). Identification by mutagenesis of a Mg(2+)-block site of the NMDA receptor channel. Nature, 358(6388), 673-675. doi: 10.1038/358673a0

Mullasseril, P., Hansen, K. B., Vance, K. M., Ogden, K. K., Yuan, H., Kurtkaya, N. L., . . . Traynelis, S. F. (2010). A subunit-selective potentiator of NR2C- and NR2D-containing NMDA receptors. Nat Commun, 1, 90. doi: 10.1038/ncomms1085

Neher, E., & Steinbach, J. H. (1978). Local anaesthetics transiently block currents through single acetylcholine-receptor channels. J Physiol, 277, 153-176.

Ogden, K. K., & Traynelis, S. F. (2013). Contribution of the M1 transmembrane helix and pre-M1 region to positive allosteric modulation and gating of N-methyl-D-aspartate receptors. Mol Pharmacol, 83(5), 1045-1056. doi: 10.1124/mol.113.085209

Paoletti, P. (2011). Molecular basis of NMDA receptor functional diversity. Eur J Neurosci, 33(8), 1351-1365. doi: 10.1111/j.1460-9568.2011.07628.x

Salazar, G., Craige, B., Love, R., Kalman, D., & Faundez, V. (2005). Vglut1 and ZnT3 co-targeting mechanisms regulate vesicular zinc stores in PC12 cells. J Cell Sci, 118(Pt 9), 1911-1921. doi: 10.1242/jcs.02319

Schorge, S., Elenes, S., & Colquhoun, D. (2005). Maximum likelihood fitting of single channel NMDA activity with a mechanism composed of independent dimers of subunits. J Physiol, 569(Pt 2), 395-418. doi: 10.1113/jphysiol.2005.095349

Schwechter, B., & Tolias, K. F. (2013). Cytoskeletal mechanisms for synaptic potentiation. Commun Integr Biol, 6(6), e27343. doi: 10.4161/cib.27343

Siegler Retchless, B., Gao, W., & Johnson, J. W. (2012). A single GluN2 subunit residue controls NMDA receptor channel properties via intersubunit interaction. Nat Neurosci, 15(3), 406-413, S401-402. doi: 10.1038/nn.3025

Smart, T. G., Hosie, A. M., & Miller, P. S. (2004). Zn2+ ions: modulators of excitatory and inhibitory synaptic activity. Neuroscientist, 10(5), 432-442. doi: 10.1177/1073858404263463

Stern, P., Behe, P., Schoepfer, R., & Colquhoun, D. (1992). Single-channel conductances of NMDA receptors expressed from cloned cDNAs: comparison with native receptors. Proc Biol Sci, 250(1329), 271-277. doi: 10.1098/rspb.1992.0159

Tovar, K. R., McGinley, M. J., & Westbrook, G. L. (2013). Triheteromeric NMDA receptors at hippocampal synapses. J

Neurosci, 33(21), 9150-9160. doi: 10.1523/JNEUROSCI.0829-13.2013

Traynelis, S. F., & Cull-Candy, S. G. (1990). Proton inhibition of N-methyl-D-aspartate receptors in cerebellar neurons. Nature, 345(6273), 347-350. doi: 10.1038/345347a0

Traynelis, S. F., Wollmuth, L. P., McBain, C. J., Menniti, F. S., Vance, K. M., Ogden, K. K., . . . Dingledine, R. (2010). Glutamate receptor ion channels: structure, regulation, and function. Pharmacol Rev, 62(3), 405-496. doi: 10.1124/pr.109.002451

Vargas-Caballero, M., & Robinson, H. P. (2003). A slow fraction of Mg2+ unblock of NMDA receptors limits their contribution to spike generation in cortical pyramidal neurons. J Neurophysiol, 89(5), 2778-2783. doi: 10.1152/jn.01038.2002

Yuan, H., Hansen, K. B., Vance, K. M., Ogden, K. K., & Traynelis, S. F. (2009). Control of NMDA receptor function by the NR2 subunit amino-terminal domain. J Neurosci, 29(39), 12045-12058. doi: 10.1523/JNEUROSCI.1365-09.2009

Zheng, F., Erreger, K., Low, C. M., Banke, T., Lee, C. J., Conn, P. J., & Traynelis, S. F. (2001). Allosteric interaction between the amino terminal domain and the ligand binding domain of NR2A. Nat Neurosci, 4(9), 894-901. doi: 10.1038/nn0901-894