Calcium- and calmodulin-dependent inhibition of NMDA receptor currents

Abstract

Calcium ions (Ca²⁺) reduce NMDA receptor currents through several distinct mechanisms. Among these, calmodulin (CaM)-dependent inhibition (CDI) accomplishes rapid, reversible, and incomplete reduction of the NMDA receptor currents in response to elevations in intracellular Ca²⁺. Quantitative and mechanistic descriptions of CDI of NMDA receptor-mediated signals have been marred by variability originating, in part, from differences in the conditions and metrics used to evaluate this process across laboratories. Recent ratiometric approaches to measure the magnitude and kinetics of NMDA receptor CDI have facilitated rapid insights into this phenomenon. Notably, the kinetics and magnitude of NMDA receptor CDI depend on the degree of saturation of its CaM binding sites, which represent the bona fide calcium sensor for this type of inhibition, the kinetics and magnitude of the Ca²⁺ signal, which depends on the biophysical properties of the NMDA receptor or of adjacent Ca²⁺ sources, and on the relative distribution of Ca²⁺ sources and CaM molecules. Given that all these factors vary widely during development, across cell types, and with physiological and pathological states, it is important to understand how NMDA receptor CDI develops and how it contributes to signaling in the central nervous system. Here, we review briefly these recent advances and highlight remaining questions about the structural and kinetic mechanisms of NMDA receptor CDI. Given that pathologies can arise from several sources, including mutations in the NMDA receptor and in CaM, understanding how CaM responds to intracellular Ca²⁺ signals to initiate conformational changes in NMDA receptors, and mapping the structural domains responsible will help to envision novel therapeutic strategies to neuropsychiatric diseases, which presently have limited available treatments.

Significance

Ratiometric approaches to quantify NMDA receptor CDI have allowed rapid advancements in understanding the mechanism by which intracellular Ca²⁺ levels regulate the magnitude of NMDA receptor currents and to identify the many factors that can control it during physiological and pathological processes in which it is involved.

Introduction

Calcium ions are universal second messengers

Calcium ions help to translate extracellular information into powerful cellular instructions (1). Despite the long and arduous research history into the mechanisms and physiological roles of calcium signaling (2), many gaps remain in understanding how intracellular calcium elevations affect cellular functions and, conversely, how cells control calcium admission. The cellular repercussions of any calcium-signaling event depend on specific arrays of calcium-sensitive proteins expressed, which are cell specific and often compartmentalized. Moreover, the biophysical and pharmacologic properties of the calcium-sensing proteins present endow cells with the ability to detect the spatial and temporal dynamics of calcium elevations (3), such that intracellular calcium transients have specific cellular consequences depending on the differential and conditional activation of calcium-responsive targets (4). In this review, we focus on how NMDA receptors and calmodulin (CaM) collaborate to control calcium signaling in cells of the central nervous system (CNS).

NMDA receptors are calcium-permeable ion channels

In the CNS, calcium signaling has an oversized role in orchestrating fundamental processes across a broad timescale. Calcium influx triggers presynaptic neurotransmitter release and, thus, has immediate effects on synaptic transmission and excitability (5): it is required for synaptic development, plasticity, and pruning across the life span (6,7); and directs the fate of neurons and glial cells toward survival, senescence, or death (8). Accumulating evidence indicates that NMDA receptors participate in all these processes. However, the bulk of existing literature centers on their roles at postsynaptic sites of excitatory synapses. Early measurements estimated that a single NMDA receptor can raise the free calcium concentration in a dendritic spine to 0.5 mM within 1 ms (9). This high and rapid calcium flux coupled with high affinity for the neurotransmitter glutamate (EC₅₀, 1 μM), and magnesium-mediated voltage sensitivity (10), identified NMDA receptors as critical mediators of Hebbian plasticity (11) and stimulated intense interest in postsynaptic receptors. For this reason, although NMDA receptors are widely expressed on neuronal soma and axons, at extrasynaptic sites on dendrites, and at presynaptic terminals, as well as on glial cells, and on many cell types outside the CNS, the biological roles and regulatory mechanisms associated with non-neuronal NMDA receptors remain poorly understood (12). Moreover, due to considerable molecular diversity of endogenous receptors, much of the present knowledge about the biophysical and pharmacologic properties of NMDA receptors was obtained from experiments with recombinant receptors (13).

NMDA receptors are large transmembrane proteins with diverse molecular composition

NMDA receptors assemble as tetramers of two obligatory GluN1 subunits and two GluN2 and/or GluN3 subunits. Differential RNA splicing produces eight molecularly distinct GluN1 subunits (14), and transcriptional regulation of separate genes results in four GluN2(A-D) and two GluN3(A-B) subunits (15). Because both these regulatory mechanisms, splicing and translation, are dynamic, it is presumed that the NMDA receptor proteome has substantial diversity (16). The bulk of the literature on the biophysical and pharmacologic properties of NMDA receptors consists of reports on principal neurons from either brain slices or dissociated cultures of brain or spinal neurons, and on recombinant receptors expressed in Xenopus oocytes or HEK cells. Overall, studies on recombinant receptors explain behaviors observed for neuronal receptors, and, therefore, it is assumed that NMDA receptors do not require auxiliary proteins for proper signaling (17). However, they interact dynamically with a large complement of proteins (18,19,20). Some of these interactions are likely necessary for the proper transport, targeting, and turnover of NMDA receptors; others will likely be found to play roles in signal transduction and receptor regulation. Relevant to the topic reviewed here are the interactions of NMDA receptors with the intracellular proteins CaM and PSD95, which occur through the intracellular domains of the GluN1 and GluN2 subunits, respectively.

NMDA receptors have modular structures

Structural models of functional NMDA receptors based on crystallographic and cryo-EM data illustrate tetramers where each monomer has two stacked extracellular domains that form binding sites for the coagonists glutamate and glycine and a variety of allosteric modulators, and a transmembrane domain that forms the agonist-controlled gate and the ion-permeable pore (21). More than a third of the receptor mass lies intracellularly and appears to be largely unstructured. Each domain has distinct evolutionary origins and imparts specific functional properties (22). Calcium ions influence the NMDA receptor current by targeting several sites that are structurally, functionally, and mechanistically distinct. An extracellular site located at the pore entrance on the GluN1 subunit is responsible for direct calcium-dependent effects on gating, permeation, and conductance (23,24,25). In contrast, intracellular sites are responsible for indirect effects of calcium on gating and permeation; these are mediated by a series of calcium-dependent effectors, including CaM, α-actinin, calcium-dependent protein phosphatases, and protein kinases (26,27,28). Therefore, both extracellular and intracellular calcium levels influence NMDA receptor signals.

NMDA receptor currents are sensitive to extracellular calcium

Fluctuations in extracellular calcium around the average values reported for the cerebrospinal fluid (1 mM) (29) affect NMDA receptor gating kinetics, channel conductance, and calcium permeability. In this concentration range, the relative calcium permeability increases with increasing levels of external calcium (30,31) and can result in nontrivial changes in calcium flux (32,33,34). In contrast, similar increases in extracellular calcium reduce receptor gating and channel conductance (23,35). The mechanisms and physiological relevance of these mechanisms remain poorly understood.

NMDA receptor currents are sensitive to intracellular calcium

Elevations in in intracellular calcium, whether produced by NMDA receptor activation or by other calcium sources (calcium-permeable AMPA receptors, BK channels, TRP channels, etc.) control NMDA receptor output through indirect mechanisms (36,37,38). A CaM-mediated process, referred to in the early literature as “calcium-dependent inactivation,” has been studied extensively. It requires CaM binding to residues in the intracellular portion of the GluN1 subunits and reduces channel gating kinetics (39). Furthermore, calcium-bound CaM can activate other local effectors such as calcineurin and CaM-dependent protein kinase II (40), which in turn modulate NMDA receptor activity and trafficking. PKA-mediated phosphorylation of the GluN2A C-terminal tail, which can be regulated by the metabolic status of the cell through phosphorylation, controls the NMDA receptor calcium permeability (27,41,42).

Given the multiplicity of calcium-dependent regulatory processes, which often work in parallel, and have unknown relative contributions to the measured NMDA receptor output, rigorous mechanistic evaluation of calcium-dependent channel regulation has been technically challenging. Here, we focus specifically on CaM-dependent inhibition (CDI) to briefly outline its present understanding and to highlight recent advances that set the precedent for future work on these elusive processes.

Historical perspective

The first observation that Ca²⁺ flux produces reduction in channel activity came from studies of Paramecium (43). Such Ca²⁺-dependent inactivation of voltage-gated Ca²⁺ channels was found to represent a reduction in channel activity due to accumulation of free intracellular Ca²⁺ (44). Importantly, CaM was found as a required mediator between the intracellular Ca²⁺ transient and the subsequent inactivation of the Ca²⁺ channel (45). In similar fashion, in both native (46) and recombinant NMDA receptors (47), extracellular Ca²⁺ reduced channel activity and induced currents to desensitize faster and deeper. Moreover, the magnitude of the Ca²⁺ influx correlated directly with the extent of current reduction and could be prevented by intracellular Ca²⁺ buffers (48). Importantly, as for voltage-gated Ca²⁺ channels, endogenous CaM is required for the Ca²⁺-induced reduction in NMDA receptor current (39). Since this observation, a reduction in NMDA receptor activity that is dependent on intracellular Ca²⁺, CaM, and the C-terminus of GluN1 subunit has been referred to in the literature as “Ca²⁺-dependent inactivation” (39,47,49,50,51,52,53,54,55,56,57,58,59,60,61). Importantly, within the ionotropic glutamate receptor family, this mode of inhibition is unique to the NMDA receptor subgroup (62).

Although at this time, the mechanism by which CaM binding to the C-terminus of GluN1 increases the occupancy of desensitized states to reduce the current passed by NMDA receptors is incompletely understood, it is clearly distinct from the mechanism by which CaM reduces the activity of voltage-gated Ca²⁺ channels. Specifically, the CaM-dependent NMDA receptor effect is saturable, incomplete, and manifests as faster and deeper macroscopic desensitization due to slower recovery from intrinsic desensitized states (50). For this reason, the term “inactivation” is inaccurate. Others have used the term Ca²⁺-dependent desensitization in recognition of this (63,64,65), which is a more appropriate description of the observed effect on whole-cell current kinetics. However, a requirement for CaM and or for the GluN1 C-terminus have not been established for these observations. In this article, we use the more generic term CDI, which simply indicates an allosteric reduction of the macroscopic current, as a result of reduced gating kinetics.

Variable experimental conditions and metrics of CDI

Mathematical definitions

The gap in our understanding of both mechanism and purpose of CDI may be due to the variety of and inconsistent experimental approaches by which CDI has been defined and studied. Classically, CDI is elicited by prolonged agonist application in whole-cell recordings. This nonsynaptic-like condition raised questions of the physiological relevance. Subsequent studies revealed that this process occurs even upon brief synaptic stimulation to significantly shape the slow component of the postsynaptic current (56,57,66). Nevertheless, the magnitude of measured CDI varies widely between studies making it challenging to infer broad mechanistic insights. Contributing to this problem are inconsistencies in both the metrics used to quantify CDI and the conditions used to study it.

Presently the literature documents several metrics to quantify the extent of NMDA receptor CDI but the reported values are difficult to reconcile across experimental conditions, most likely due to uncontrolled confounding factors. Aside from effects on response kinetics, external Ca²⁺ controls NMDA receptor conductance (23); and, aside from Ca²⁺, the kinetics and extent of NMDA receptor macroscopic desensitization vary with external glycine concentrations (67,68), mutations (69), and posttranslational modifications (41,70,71).

A lack of consistency in the mathematical definition of CDI across existing studies has likely contributed to the high variability of values reported in the literature and hampered rigorous investigation of this process. Some studies quantified CDI as the “percentage of macroscopic desensitization” (52,53,58,59):

$$CDI=\left(\frac{I_{\text{ss},\mathrm{Ca}}}{I_{\mathrm{pk},\mathrm{Ca}}}\right)\times 100$$

where I_ss,Ca is the steady-state current amplitude and I_pk,Ca is the peak current amplitude both in the presence of Ca²⁺. Unless comparison is made to currents elicited in Ca²⁺-free conditions, for this metric to quantify true CDI specifically, it necessarily assumes that no desensitization occurs in the absence of Ca²⁺. Therefore, this approach most likely overestimates CDI. To address this limitation, subsequent studies aimed to eliminate the contribution of glycine-dependent desensitization by focusing on currents elicited with varied agonist concentrations (46). However, agonist concentration also influences channel P_o and thus the influx of Ca²⁺ that triggers CDI (51). As a result, this approach may likely underestimate CDI.

Most commonly, CDI has been defined as the “percent of inhibition” in several studies (46,52,53,59) as:

$$CDI=\left(\frac{I_{\mathrm{ss},\mathrm{Ca}}-I_{\mathrm{pk},\mathrm{Ca}}}{I_{\mathrm{pk},\mathrm{Ca}}}\right)\times 100$$

where I_ss,Ca is the steady-state current amplitude of a whole-cell macroscopic signal at the end of a prolonged application of agonist (glutamate or NMDA). While intuitive, this metric quantifies CDI as the degree of change in current amplitude between I_ss and I_pk. This equation assumes that, in the absence of Ca²⁺, the change in I_ss from I_pk at the end of agonist application should approximate 0 given that no CDI occurs in Ca²⁺-free solutions. However, because these measures are made exclusively from a current signal recorded in extracellular Ca²⁺ with no reference to a Ca²⁺-free baseline, the final measured value of CDI reflects not only CaM-mediated inhibition, but intrinsic receptor desensitization that occurs even in the absence of Ca²⁺. This would presumably overestimate the extent of “pure” CDI. In other work (54), this same metric was used but applied to macroscopic currents elicited with much shorter agonist applications (∼2 s) where traces are often not allowed to proceed to steady state. Therefore, the extent of CDI may be underestimated.

In an attempt to remedy these confounds, previous work (46) has applied the same metric to currents elicited with low concentrations of NMDA (1–10 μM) and high glycine (10 μM) to attenuate the contribution of glycine-independent desensitization to the measured CDI. While more stringent, the complete elimination of glycine-dependent desensitization is unlikely given the persistent residual, low-level decay in macroscopic current still apparent using these conditions in Ca²⁺-free solutions. Thus, measured CDI may still be somewhat overestimated.

CDI has also been quantified using very different configurations (72). Brief (50–150 ms) test pulses of agonist were applied to recombinant cells bathed in physiological Ca²⁺ solutions. A 5–15 s conditioning pulse was applied to induce CDI followed by additional test pulses. CDI was then quantified as the time course of recovery of the test pulse amplitudes following the conditioning pulse by:

$$CDI\left(t\right)=\frac{I_o-I\left(t\right)}{I_o}$$

where I_o is the amplitude of the test pulse before the conditioning pulse and I is the amplitude of the test pulse at time t after the conditioning pulse.

To mathematically correct for intrinsic Ca²⁺-independent desensitization without the need to introduce conditions that impact gating, our work adapted the expression used for quantifying pure CDI in voltage-gated Ca²⁺ channels corrected for intrinsic voltage-dependent inactivation (73,74,75,76,77). The terms have been modified to reflect our recording conditions:

$$CDI\left(t\right)=1-\frac{I_{\mathrm{Ca}}\left(t\right)/I_{\mathrm{pk},\mathrm{Ca}}}{I_{\mathrm{Na}}\left(t\right)/I_{\mathrm{pk},\mathrm{Na}}}$$

where I_pk is peak current amplitude of a macroscopic signal in response to a prolonged application of saturating agonists in Ca²⁺-free (Na⁺ only) or physiological Ca²⁺ solutions, and I(t) is the macroscopic current amplitude within the macroscopic signal at time t after I_pk in either Ca²⁺-free (Na⁺ only) or physiological Ca²⁺ solutions. In our conditions, this measure of CDI at any time t during an agonist pulse allows a ratiometric correction for baseline desensitization (Fig. 1 a). If currents are to be normalized by their peak amplitudes before analysis, this equation simplifies to the ratio of the normalized I_ss current amplitudes. This may be warranted if Ca²⁺-induced changes in the peak amplitude are expected to be due to processes distinct from CDI (e.g., Ca²⁺-dependent block).

Figure 1.

Defining and quantifying NMDA receptor CDI. (a) Experimental protocol used to measure CDI requires paired recordings of agonist-elicited currents in the absence and presence of external Ca²⁺. To calculate Ca²⁺-dependent reduction in current, four parameters are measured: peak Na⁺ current (I_pk,Na), steady-state Na⁺ current (I_ss,Na), peak current in Ca²⁺ (I_pk,Ca), and steady-state current in Ca²⁺ (I_ss,Ca). (b) Currents recorded in the absence (black) and presence of Ca²⁺ (red) are normalized to peak and overlaid in cells expressing CaM_WT (left) or CaM_MT (right) to reveal the magnitude of CDI quantified with this protocol (shaded red). Overexpressing CaM_MT prevents Ca²⁺-dependent reduction in current and thus identifies this modulatory mechanism as CDI. (c) Top: systematic simulations with the CDI equation show the dependency of the calculated CDI strength as a function of the baseline receptor desensitization magnitude in the absence of Ca²⁺ (x axis) for a given condition (e.g., channel isoform, channel variants, presence of additional modulators, etc.). Each curve represents the CDI strength for a given observed reduction in current magnitude in the presence of Ca²⁺. Bottom: simulated currents with different baseline desensitization kinetics and similar CDI strength highlight the ratiometric nature of this method. To see this figure in color, go online.

Desensitization can be observed and measured in Ca²⁺-free solutions from the steady-state fractional residual current (1 − I_ss/I_pk), and can theoretically take values from zero (nondesensitizing) to unity (fully desensitizing). In external Ca²⁺, currents are smaller and desensitize deeper and faster. We define CDI specifically as the increased desensitization of the macroscopic response observed in Ca²⁺ relative to Ca²⁺-free conditions; it can also take values from zero (noninhibiting) to unity (fully inhibiting). By using a metric that is relative to the channel’s baseline activity in the absence of Ca²⁺, we are able to quantify with greater precision the degree of inhibition that is specifically Ca²⁺ and CaM dependent (Fig. 1 b). This becomes increasingly important when investigating the protein structural mechanism of CDI. Invariably, this requires perturbation of the structure in some way and evaluating how CDI changes. Given that structural perturbations may themselves alter receptor function at baseline, without knowing this, the degree of observed inhibition attributable to Ca²⁺ is unknown, precluding any rigorous inferences about underlying mechanisms of CDI.

This approach is not without limitations and assumptions. In particular, as the baseline (Ca²⁺-independent) extent of desensitization (I_ss/I_pk) increases, small fluctuations in I_ss/I_pk in the presence of Ca²⁺ can produce large changes in the calculated CDI magnitude. This becomes a pragmatic challenge if the normal variation or noise of the recording system is large (Fig. 1 c). Another challenge with this approach is the assumption that the I_ss/I_pk measured in the presence of Ca²⁺ will always be smaller compared with that measured in the absence of Ca²⁺. If Ca²⁺ were to specifically reduce the peak amplitude (e.g., a mutation that enhances Ca²⁺ influx), any calculation may be erroneously calculated as a potentiation by Ca²⁺. Another potential confounder is the conditions under which baseline desensitization is measured. Our work utilized Na⁺-only conditions and accounted for Ca²⁺ allosteric inhibition and block by first pre-equilibrating resting channels with external Ca²⁺ and normalizing resulting currents by peak amplitudes (50,51,78). Others have substituted Ca²⁺ for Ba²⁺, which does not activate CaM (49,79). This approach is viable, however, the allosteric and blocking effects of Ca²⁺ and Ba²⁺ may not be equivalent (80). Until the effects of Ba²⁺ are more precisely characterized, caution should be used interpreting results using this approach.

Experimental conditions

In addition to the variety of numerical methods employed in the literature to quantify CDI, there has been an equally wide array of experimental conditions used to study it (Table 1). As with all Ca²⁺-permeable channels, NMDA receptors generate high amplitude Ca²⁺ fluctuations near the pore (81). The Ca²⁺ concentration decays as a function of distance and time from the channel pore (Fig. 2 a). Within this microenvironment, cell-type-specific factors such as compartmentalization and membrane electrostatic effects (82), molecular crowding/volume exclusion (83), channel clustering (84), and altered diffusion (85) can significantly impact the local Ca²⁺ concentration (Fig. 2 b). In particular, differences in the amount and type of intracellular Ca²⁺ buffers, whether or not additional ATP/GTP is used as a “regenerating” solution to mitigate channel rundown (86), and the cell type used (which will have its own unique endogenous buffering properties) make a profound difference on the intracellular Ca²⁺ concentration experienced by the CaM/NMDA receptor complex (Fig. 2 c). While it is important to study the effects of CDI in physiological conditions such as neurons, it becomes a challenge to systematically evaluate results from multiple studies to gain insight into the underlying mechanisms. Using a recombinant expression system to have better control of the molecular composition of both receptor type and cellular CaM levels, we find that, in HEK293 cells, overexpression of CaM boosts the observed strength of CDI, suggesting that, at least in this system where endogenous CaM levels are low relative to overexpressed receptors, the receptor pool is not saturated by CaM (Fig. 2 d). Channel proximity also impacts the local free Ca²⁺ concentration and degree of buffer saturation. Stronger CDI is observed in cells with higher charge density owing to nearby cross talk within channel clusters (78). Finally, endogenous Ca²⁺ extrusion mechanisms in the cell types used to study CDI can have profound influence on CDI magnitude (63).

Table 1.

Summary of experimental conditions used to study CDI

Study	Cell typec	[Ca2+]ob (mM)	pH	Agonistb (μM)	Buffersa,b (mM)	Comments
(46)	rat hippocampal neurons (1 day old)	0–50	7.3	10 glycine 0.1–300 NMDA	10 EGTA or 15 BAPTA, 4 ATP	Vh = −70 mV
(142)	rat hippocampal neurons (1 day old)	0.2 or 2	7.3	10 glycine 10 or 100 NMDA	5 EGTA or 5 BAPTA or 5 EDTA	Vh = −70 mV
(47)	HEK293	2	7.4	10 glycine 100 glutamate	1.1 EGTA, 4 ATP	Vh = −50 mV, 0.25 mM Ca2+ intracellular
(57)	rat hippocampal neurons (neonate)	0.2–2.7	7.2	10 glycine 10 NMDA	10 EGTA, 2 ATP, 0.3 GTP
(54)	rat dorsal horn neurons (1–2 days old)	2	7.3	5 glycine 30 NMDA	None	0.1 mM Ca2+ intracellular
(72)	rat hippocampal neurons (2 days old)	2	7.4	10 glycine 50–300 NMDA	1.1 BAPTA, 4 ATP, 0.6 GTP	Vh = −60 mV, 0.1 mM Ca2+ intracellular
	rat cerebellar granule neurons (7–8 days old)	2	7.4	10 glycine 50–300 NMDA	1.1 BAPTA, 4 ATP, 0.6 GTP	Vh = −50 mV, 0.1 mM Ca2+ intracellular
(52)	HEK293	0.2–2	7.25	10 glycine 10–100 NMDA 100–1000 glutamate	0.1–10 EGTA or 2.4–10 BAPTA, 4 ATP	FK-506 used in some experiments
(61)	HEK293	2	7.4	20 glycine 100 glutamate	1.1 EGTA, 4 ATP	0.25 mM Ca2+ intracellular
(71)	HEK293	10	7.25	50 glycine 10 NMDA	0.1 EGTA, 6 ATP
(39)	HEK293	1.8	7.3	10 glycine 10 glutamate	10 EGTA, 4 ATP
	rat hippocampal neurons (1–2 days old)	1.8	7.3	10 glycine 20 NMDA	10 EGTA, 4 ATP
(53)	HEK293	10	7.25	100 glycine 100 NMDA	0.1 EGTA, 6 ATP	Vh = −50 mV
(56)	rat hippocampal neurons (2 days old)	2	7.4	20 glycine	1.1 BAPTA, 4 ATP, 0.6 GTP	Vh = −60 mV, 0.6 mM Ca2+ intracellular, monosynaptic cultures
(55)	rat hippocampal neurons (12–20 days old)	1.3	7.4	3–10 glycine 300 NMDA	11 EGTA or 20 BAPTA, 4 ATP	1 mM Ca2+ intracellular
(90)	rat hippocampal neurons (12 days old)	1	7.4	10 glycine 10 NMDA	10 EGTA	Vh = −60 mV, 12 nM free Ca2+ intracellular
(59)	HEK293	1	7.25	100 glycine 10–100 NMDA or 1000 glutamate	0.1 EGTA, 4 ATP	Vh = −50 mV
(91)	rat hippocampal neurons (12 days old)	1	7.4	10 glycine 0.1–10 NMDA	10 EGTA	Vh = −30 to −80 mV 800 nM free Ca2+ intracellular
(92)	rat hippocampal neurons (12 days old)	2.5	7.4	10 glycine 0.01–10 NMDA	10 EGTA	Vh = −30 to −80 mV
(143)	HEK293	2	7.3	100 glycine 50 NMDA	10 EGTA, 4 ATP 0.5 GTP	Vh = −50 mV
(58)	HEK293	1.8	7.3	100 glycine 100 glutamate	10 EGTA, 2 ATP	Vh = −70 mV
(66)	mouse hippocampal neurons (14–21 days old)	2	7.2	none	4 ATP 10 EGTA or 1 BAPTA	synaptically evoked currents in slice preparations Mg2+ free or 1.25 mM extracellular
(50)	HEK293	0.1–5	8.0	100 glycine 1–1000 glutamate	0.5–5 EGTA or 5 BAPTA, 4 ATP, 0.3 GTP	Vh = −80 mV YFP-CaM overexpression
(63)	rat cortical neurons (E16)	1	7.2–7.4	10 glycine 100 NMDA	10 EGTA or 1 BAPTA and 10 EGTA	corrected for LJP, 0.1 mM Ca2+ intracellular
(78)	HEK293	2	8.0	100 glycine 1000 glutamate	5 EGTA, 4 ATP 0.3 GTP	Vh = −80 mV YFP-CaM overexpression
	rat hippocampal neurons (E18)	2	8.0	100 glycine 1000 glutamate	5 EGTA, 4 ATP 0.3 GTP	Vh = −80 mV
(51)	HEK293	2	8.0	100 glycine 1000 glutamate	5 EGTA or 10 BAPTA, 4 ATP, 0.3 GTP	Vh = −80 mV YFP-CaM overexpression
(49)	HEK293	2	7.4	1000 glycine 1000 glutamate	01 EGTA, 4 ATP 0.3 GTP	Vh = −60 mV

LJP, liquid junction potential; Vh, holding potential; [Ca²⁺]_o, extracellular Ca²⁺ concentration.

^aIntracellular Ca²⁺/divalent buffers.

^bDose responses are reported as a range of values.

^cReported ages are the age of cell harvest (not days in vitro).

Figure 2.

Experimental conditions that confound the measurement of CDI. (a) Schematic of the Ca²⁺ transient microdomain (red shaded) illustrates how the concentration of free Ca²⁺ varies as a function of distance from the Ca²⁺ source (i.e., channel pore). At more proximal distances, only Ca²⁺ chelators with fast binding rates and high affinities (e.g., BAPTA) control Ca²⁺ concentration effectively. (b) The predicted free Ca²⁺ level generated per unit of Ca²⁺ current also varies with the local microdomain conditions in the cell type being studied. Volume exclusion (yellow) was set to 90%; slow Ca²⁺ diffusion was set to 0.1× diffusion coefficient; membrane electro diffusion was assumed to be 0.3 q/nm². Measured NMDA receptor gain (dashed line) allows estimation of the distance between the Ca²⁺ sensor (i.e., CaM) and the Ca²⁺ source in the 10–20 nm range (red shaded). (c) Top: predicted time-averaged free Ca²⁺ levels in the 10–20 nm range, as a function of total intracellular buffer concentration ([B_T]) and buffer selection (assuming P_o = 0.4 and Pf = 12%). Bottom: predicted time-averaged free Ca²⁺ level as a function of channel (P_o) and fractional Ca²⁺ current (Pf), assuming 8 pA unitary amplitude. (d) Whole-cell currents recorded from HEK293 cells expressing GluN1/GluN2A receptors and endogenous CaM (CaM_endo) or overexpressing CaM_WT or CaM_MT. To see this figure in color, go online.

In the time evolution of CDI, there are likely two populations of channels contributing to the recorded output: CaM-bound/associated (Ca²⁺ sensitive) and CaM-unbound (Ca²⁺ resistant). Because this proportion is unknown in a given system, the degree of measured CDI may vary by an unknown degree between different cell systems. Furthermore, both in vitro and in vivo estimates of apoCaM affinity for the channel suggest affinity is low, which supports the hypothesis that, under physiological conditions, the degree of receptor saturation by CaM is variable (39,50,53,60,87). This also has implications for the choice of intracellular buffer used in a given experiment. Given that at least a fraction of the receptor population is not bound with a resident CaM, their inhibition will depend on CaM diffusion from the bulk cytoplasm. Because the amplitude of the Ca²⁺ transient is a function of distance from the channel, lower affinity/slower binding buffers will have greater influence in mitigating activation of this bulk cytoplasmic pool of CaM compared with local CaM positioned proximally to the high amplitude Ca²⁺ spike near the channel (Fig. 2, a and b). These differences may offer some insight into the discrepancies between studies as to the strength of exogenous intracellular buffering needed to eliminate CDI.

Another inherent challenge interpreting studies from native cell preparations is the lack of control of receptor subtype composition. It is well known that different subtypes function with unique calcium permeability (Pf, fractional calcium current) and gating kinetics (P_o, open probability). Together, these terms govern the amount of Ca²⁺ that fluxes through an individual channel over time (<Ca²⁺>). We predict that the rise in Ca²⁺ at a given distance from the pore is linearly related to these permeation and kinetic properties (Fig. 2 c). However, we have previously found a nonlinear relationship between P_o and CDI (51) consistent with the nonlinear dose-response relationship observed between intracellular Ca²⁺ and CDI (50).

Finally, another often uncontrolled factor that influences CDI strength is the density of channels present on the cell at the time of recording. Previous work has shown that NMDA receptor responses are transiently inhibited by Ca²⁺ influx from neighboring Ca²⁺ channels and Ca²⁺-permeable AMPA receptors (38,88,89). Physiologically, channels are often found in clusters rather than homogeneously distributed across the cell membrane. We identified that NMDA receptors can also be inhibited by neighboring NMDA receptors in a manner that is mitigated by intracellular buffers and CaM levels (78).

Kinetic mechanism of inhibition

Upon discovery of this new mode of NMDA receptor regulation, defining the exact mechanism by which CaM reduces channel activity has been an elusive pursuit. This has in part been due to the technical challenges of performing extended single-channel recordings under conditions that isolate CDI from other Ca²⁺-dependent processes in intact cells.

Early insights from single-channel recordings in excised membranes argued against a “blocking” mechanism by CaM as the reduced channel open probability was due specifically to shortened open durations while burst closed durations remained unchanged (90,91). This pattern of activity was more consistent with an allosteric mechanism of inhibition. CDI has classically been studied using macroscopic recordings with prolonged agonist applications, which demonstrate a “slow” evolution of CDI similar to channel desensitization. Thus, it has been assumed that CDI represents a strengthening and/or acceleration of channel desensitization. However, observations that CaM could also accelerate deactivation in response to brief glutamate pulses both in excised patches (48,92) and synaptic preparations (56,57,66) suggest a more complex kinetic mechanism.

Resolving this question is best addressed by prolonged single-molecule electrophysiology to allow statistical identification of energetically distinct functional states explored by the channel during gating and subsequent derivation of the kinetics (93). Some have asserted that CDI may represent a distinct functional “state” the channel enters (65). However, single-channel recordings suggest that inhibited channels continue to explore all functional states that uninhibited channels do without the presence of extra states. Thus, CDI represents a tiered model by which a full kinetic model composted of five closed and two to five open state transitions to a second model with distinct gating kinetics that determine each respective model’s P_o (Fig. 3 a). Furthermore, results from biochemical and functional studies suggest that cytoskeletal and other components may play a role in inhibition (91,94), thus recording from intact cells is preferable to maintain the intracellular milieu. To achieve this while circumventing the direct allosteric inhibitory effects of Ca²⁺ without introducing disruptive mutations, our group recently utilized extended, on-cell, single-molecule recordings of channel activity in Ca²⁺-free conditions with ionophore application to raise intracellular Ca²⁺. Using this approach, it was found that channels recorded in conditions of high intracellular Ca²⁺ exhibited a higher probability of dwelling in desensitized states (Fig. 3 c). This was achieved by increasing the energy barriers to activation rather than altering the desensitization rates directly (51). Importantly, this mechanism was conserved for both GluN2A- and GluN2B-containing receptors.

Figure 3.

Mechanism and physiological role of CDI. (a) Schematic of the CDI process illustrates stochastic fluctuation between an active gating mode (CDI = 0) and an inactivated mode (CDI = max). Modes differ in the magnitude of channel P_o and the level of bound CaM control the transition between kinetic modes. (b) Top: time evolution of CDI (red curve, with standard error shaded) often shows biphasic time course before reaching steady state. Inset illustrates whole-cell current recordings used to measure CDI. Bottom: in any given experiment, three populations of NMDA receptors may produce the observed current: primed receptors are bound with apoCaM, which inactivate rapidly (left); naive receptors, which are CaM-free, may transition more slowly into the inactivated mode as the concentration of CaM increases (middle); or they can remain unaltered by Ca²⁺ influx (right). (c) Left: on-cell single-channel Na⁺ currents (recorded without added Ca²⁺ in the recording pipette) before and after bath application of a Ca²⁺ ionophore, reveal drastic change in gating pattern characteristic to active and inactivated modes, respectively. Scale: 5 pA and 200 ms. Right: corresponding kinetic models inferred from currents recoded before (gray) and after (red) ionophore application. Sphere diameters illustrate changes in relative state occupancies. Inset: relative changes in energy barriers of activation moving from C₃ to open (O). (d) Differences in CDI between GluN2A and GluN2B receptors predict developmental changes in the sensitivity of NMDA receptor currents to CDI. To see this figure in color, go online.

Structural determinants of inhibition

Among the ion channels modulated by direct interaction with CaM, no consistent structural mechanism or CaM binding motif has been identified, suggesting that the process of CDI is the result of convergent evolution among different ion channel families. Thus, it is likely that the structural mechanism of CDI for NMDA receptors is also unique. The structural determinants of how CaM-induced changes in the GluN1 C-terminal domain structure transmit to facilitate closure of the pore, thus far, remained elusive owing to the experimental challenge of isolating CDI from other Ca²⁺-dependent mechanisms of regulations. However, several recent studies offer new information on this process.

A single GluN1/GluN2A receptor may contain as many as six CaM binding sites, which can be controlled by GluN1 splicing and the presence of GluN2A (95). However, the sites on GluN1 C1 cassette and GluN2A do not appear to be essential for CDI (39,50). The overlap of these non-CDI CaM sites with other posttranslational modification sites may hint at other roles for CaM such as controlling receptor trafficking or tuning the Ca²⁺ by altering PKA binding (42,87). Early evaluation of GluN1 splice variants and engineered constructs found that the inhibitory effect of CaM was due to binding of calcified CaM to the C0 region of the GluN1 subunit C-terminal domain (60). This region is common to all GluN1 splice variants and, therefore, present on all functional NMDA receptor subtypes. CaM can also bind with high affinity to the C1 region immediately adjacent to C0. While it is clear that C1 is not required for CDI, some evidence suggests a modulatory role on CDI, although the exact mechanism is lacking and this effect of C1 could be phosphorylation dependent (55,87). While each receptor subtype has the potential to undergo CDI, comparison of GluN2 subtypes suggested that only GluN2A and GluN2D receptors were capable of CDI (52). However, recent work utilizing Ca²⁺ dialysis revealed that GluN2B receptors can also achieve robust CDI. In both subtypes, the magnitude of CDI increases in a P_o-dependent manner offering a mechanistic explanation why external Ca²⁺ induces minimal CDI in GluN2B receptors (51,52).

The structural changes in the cytoplasmic domains of GluN1 induced by CaM have been challenging to study owing to the intrinsically disordered nature of the domain. Based on our current understanding, three models may be proposed to explain how CaM initiates structural changes in the GluN1 C-terminal domain that ultimately result in CDI.

Electrostatic model

The C0 element contains a number of positive charges that facilitate electrostatic interactions with the membrane (96). CaM binding may initiate structural changes in the C-terminal domain by disrupting this resting conformation. Swapping the basic residues in C0 to negative glutamates is sufficient to abolish CDI, whereas mutations to alanine did not have this effect (39). This same mutation when expressed in vivo results in altered neural development of mouse GABAergic synapses (97).

Dimerization model

Further attempts to define the molecular mechanism have suggested that CaM is capable in vitro of binding two isolated C0 peptides (58). Functional evidence in support of this model is limited to use of highly truncated receptor constructs with altered activity at baseline. Of note from this study, however, truncation of the M4 helix resulted in currents with no apparent desensitization supporting the likely role of the peripheral M4 helix as a transduction element for CaM.

Displacement model

Others have proposed that CaM primarily functions to displace actinin from C0 which results in reduced activity (53,91,94,98). In this model, resting NMDA receptors are bound with actinin which promote a high P_o mode of gating and apoCaM may exist either preassociated or diffusible (91). The influx of Ca²⁺ triggers the resident CaM molecule to fully bind C0 and displace actinin from C0, thus rendering the channel to gate in a low P_o mode. Either lobe of CaM is sufficient to displace actinin (94). This holds true in functional recordings where either lobe of CaM is sufficient to initiate CDI (78).

In the above models, it is yet unclear to what extent each model may contribute to CDI initiation or whether they represent distinct processes. Furthermore, no model attempts to provide explanations for the structural endpoints of the channel in CDI after initiation by CaM.

CaM decodes complex Ca²⁺ signals

Within the dendritic spine, there are multiple Ca²⁺ sources. It has largely been assumed that plasticity depends on the magnitude of the intracellular Ca²⁺ elevation. However, the spatial arrangement of channels at a single spine couples specific Ca²⁺ sources with distinct effector signaling pathways that differentially regulate plasticity (99,100). Thus, NMDA receptor CDI is part of a complex interplay of CaM responding to the range of Ca²⁺ signals that ultimately determine the final output.

Although it is well established that CaM, in response to intracellular elevations in Ca²⁺, binds and inhibits the receptor, how CaM responds to the complex intracellular Ca²⁺ landscape to mediate this process has only recently been partly revealed (50,78). CaM itself is composed of an N-terminal lobe and C-terminal lobe, each with two Ca²⁺ binding sites with distinct kinetics. Specifically, the N-terminal lobe has low-affinity, fast binding kinetics, whereas the C-terminal lobe has slower, high-affinity kinetics making this molecule well suited to decoding kinetics and amplitude changes in Ca²⁺ (101). In NMDA receptors, either lobe is sufficient to initiate CDI (78,94). Early studies in excised outside-out patches revealed that desensitization could be reduced by intracellular BAPTA. Substitution with a low-affinity dinitro-BAPTA did not mitigate desensitization (48), suggesting that the intracellular sensor is associated in close proximity to the channel and responds to local Ca²⁺ signals. In addition to local Ca²⁺, NMDA receptors are sensitive to proximal sources of Ca²⁺ from neighboring channels and NMDA receptors (102). BAPTA titration reveals that this coupling can be exceedingly close in proximity (78).

With our current understanding of CDI, it is possible for us to construct a preliminary model of CaM dynamics in vivo and predict how this will impact synaptic function (Fig. 3 b). Biochemical and functional studies suggest that Ca²⁺-free CaM associates with the GluN1 C0 cassette with low affinity (μM) compared with the Ca²⁺-bound CaM which binds at high affinity (nM). Therefore, within a defined population of receptors at rest, a fraction will already be preassociated with CaM. This CaM molecule is privileged to the Ca²⁺ fluxed immediately through the channel pore resulting in fast autoinhibition of the channel. The remaining portion of channels are not preassociated with CaM and, thus, are able to conduct Ca²⁺ initially without a reduction in activity. Depending on the availability of free CaM pool in the local cytoplasm environment (103), a second population of CaM-naive channels are quickly bound by Ca²⁺-bound CaM as local intracellular Ca²⁺ levels rise from continued channel activity. Finally, some portion of channels may never be bound and inhibited by CaM during the time course of the synaptic event. Together, these populations determine the ensemble NMDA receptor output.

Implications for physiology and pathology

Since its discovery, the exact physiological “purpose” of NMDA receptor CDI has been difficult to ascertain. Because it imparts activity-dependent inhibition, it is asserted that CDI prevents excitotoxicity from excess Ca²⁺ influx (104). Indeed, mice treated with W-7, a CaM inhibitor, before NMDA show hyperexcitability and tremors (105). However, no direct evidence thus far has linked excitotoxicity specifically to impaired NMDA receptor CDI. Therefore, this argument, while logical, is largely teleological and does not address other hypothetical roles such as tuning synaptic responses for optimal plasticity (106), how CaM-induced changes in the cytoplasmic domain may alter nonionotropic signaling (107), or receptor recycling by cytoskeletal uncoupling (94). Of note, NMDA receptor homologs may be found in Drosophila that mediate transmission at the neuromuscular junction and appear to also display greater desensitization in Ca²⁺ compared with Ba²⁺ (108). Ca²⁺-permeable plant glutamate receptor-like channels, which mediate the plant stress response, also exhibit robust desensitization and are hypothesized to be inhibited by CaM (109) like other plant ligand-gated Ca²⁺ channels (110). Thus, CDI of NMDA receptors may serve a multitude of roles given the diverse physiological functions glutamate receptors serve within their evolutionary lineage.

Regardless of these putative roles, given that CDI occurs even upon brief synaptic stimulation to reversibly decrease channel P_o and shape the slow component of the postsynaptic current (56,57,66), it is reasonable to hypothesize that the ability for intracellular Ca²⁺ to reversibly decrease channel open probability may be a critical, physiological negative feedback mechanism to tune excitatory currents and curb unchecked Ca²⁺ entry (104). This regulation is governed by both local autoinhibition by a resident CaM molecule and by the bulk Ca²⁺ levels set by the balance between proximal Ca²⁺ sources, Ca²⁺ extrusion and buffering mechanisms, and availability of CaM.

Within neurons, the differential spatial organization of NMDA receptors may play a role in drug efficacy (111). Previous work has identified close coordinated inhibition by neighboring receptors (78). Compared with synaptic receptors which reside in close proximity, extrasynaptic receptors along dendrites are more spaced out and thus expected to be less susceptible to CDI. This could result in fluxing more Ca²⁺ per channel resulting in enhanced nNOS activation and toxicity. Extrasynaptic NMDA receptors, which experience tonic glutamate exposure under these conditions, are likely to be modulated by CDI. Given that extrasynaptic NMDA receptors are coupled with apoptotic pathways, CDI provides a necessary protective mechanism against toxicity (112). The reduced CDI in this population also means that open channel blockers, such as aminoadamantanes, will inhibit this population more specifically to a higher degree (113). Similarly, recent work has suggested that memantine may possess differential affinities to receptors that have been inhibited by intracellular Ca²⁺ (65). Thus, inhibited channels, as might be expected in pathological conditions, may be more amenable to the therapeutic effects of memantine.

During brain development and synaptic maturation, the balance of NMDA receptor subtypes shifts from GluN2B to GluN2A. While GluN2B receptors are capable of robust CDI, in physiological conditions they do not exhibit significant CDI, in contrast to GluN2A (Fig. 3 d). During synaptic development, robust Ca²⁺ elevations are needed to initiate appropriate downstream effectors for spine growth and maturation. Ca²⁺ imaging has shown that both juvenile animals as well as immature spines generate robust Ca²⁺ elevations that trigger further release of Ca²⁺ from intracellular stores necessary for spine growth. Older animals and mature spines exhibit Ca²⁺ elevations largely restricted to the spine head (114). Therefore, the reduced sensitivity of GluN2B to CDI may be crucial in preventing excessive channel inhibition and allowing sufficient Ca²⁺ entry during synaptic activity to facilitate spine growth and maturation. Subsequently as a spine matures during plasticity, for example, more CDI is favored as the priority transitions from synaptic growth to synaptic maintenance, requiring different effectors.

Remaining unknowns and necessary future research

With nearly 30 years of research into NMDA receptor CDI, we continue to learn how this evolutionary ancient process contributes to physiology. Nevertheless, there is still much to be discovered. Previous extensive functional characterization of CDI at a single-molecule and population level set a strong groundwork for defining the precise structural changes that occur in both the receptor and CaM that underlie CDI. Further, understanding how pathogenic mutations in NMDA receptors, CaM, and related postsynaptic proteins alter CDI in the context of disease will inform our knowledge of disease mechanism. Together, these insights will help design novel therapeutic modalities for conditions that largely have no treatment.

Molecular endpoints of CDI

While it is established that CaM initiates CDI upon binding the GluN1 intracellular domain in response to local and proximal fluctuations of intracellular Ca²⁺ levels, the molecular endpoints of CDI in the receptor are still to be determined. Specifically, what structural changes in the receptor induced by CaM result in CDI. Insights from single-molecule recordings show that CDI represents an increased occupancy of desensitized states resulting from impaired activation kinetics (51). FRET studies show that the glycine-bound GluN1 LBD adopts a more closed configuration and dissociates glycine at a slower rate when the channel is bound to CaM (49). While the mechanism of desensitization in NMDA receptors remains elusive, tryptophan scanning of the M4 helices profoundly alters channel desensitization (115). With the GluN1 C0 element contiguous with the M4 helix, it is likely that these helices represent important transduction elements for the actions of CaM to be communicated to the pore. The S2-M4 linker region may also have a role in desensitization (116). Decreasing the mechanical tension in the S2-M4 linker by inserting glycine residues increases the efficacy of channel activation (117). This is consistent with the observation that CaM induces GluN1 S1-S2 cleft closure, which is expected to increase tension on the S2-M4 linker and inhibit channel activation. With greater barrier to opening, channels will gradually dwell in desensitized states. This process is consistent with single-channel recordings of CDI, whereby desensitization rates are largely unaffected while activation rates are reduced leading to population of desensitized states (51). Thus, CDI may represent a complex bidirectional interplay of structural motions between the C-terminal domain and the ligand binding domain. Given the close proximity of the GluN1 M4 helix with the pore-forming GluN2 M3 helix, CaM facilitation of pore closure may involve direct allosteric interactions between these helices and associated linkers.

Channelopathies, calmodulinopathies, and auxiliary protein mutations effect on CDI

Calmodulinopathies are a class of diseases which in the past few years are only beginning to be understood. They are caused by de novo mutations in one of the three redundant CALM1-3 genes encoding identical CaM protein (118). These mutations impair CaM affinity for Ca²⁺ or target effector proteins such as ion channels (119). Calmodulinopathies manifest with fatal cardiac arrhythmias attributed to impaired voltage-gated Ca channel regulation. However, patients also experience syncope and seizures (120,121). Animal models also exhibit cognitive deficits reminiscent of autistic behaviors (122) as well as epileptic symptoms (123,124). While it is still disputed whether these neural abnormalities represent a primary neurological dysfunction or are secondary to impaired perfusion from cardiac arrhythmias, the undisputed role of NMDA receptors in neuronal excitability suggests that CDI may be an unexplored mechanism in these pathologies. Notably, reduced CaM binding to GluN1 has been observed in other epileptic disorders (125). This removes negative feedback on NMDA receptors allowing unchecked Ca²⁺ influx and maintaining cell hyperexcitability.

Mutations in NMDA receptor-encoding genes (GRIN) genes have been causally implicated in disease pathogenesis (126,127). However, determining the functional and physiological impact of such mutations continues to be a challenging task given the discrepancy between recombinant and native preparations (128). Several mutations localize to the GluN1 M4 helix that alters channel function and permeation (129). The direct impact of such mutations on regulatory mechanisms, such as CDI, is not known. We previously found a nonlinear relationship between channel P_o and CDI strength (51). Thus, pathogenic mutations that impact the channel P_o and/or Ca²⁺ permeability may differentially engage CDI. For example, channels with low P_o may insufficiently engage CDI and, thus, flux an aberrantly large amount of Ca²⁺ over time leading to an apparent “hyperexcitability” phenotype. Therefore, it will be of high importance to study how both CaM and channel mutations impact CDI (Fig. 4 a). Available treatments for many of these disorders are limited and rates of treatment resistance to available therapy are high. This will likely remain so until the pathophysiological mechanism is more thoroughly elucidated.

Figure 4.

Remaining unknowns in understanding CDI mechanism. (a) Both NMDA receptors and CaM can harbor numerous pathogenic mutations (red). How these mutations impact receptor CDI and how this change impairs synaptic regulation remains unexplored. (b) The stoichiometry of CaM binding to a channel is unknown. Given the presence of two CaM binding domains (CBD) on a functional receptor, three hypothetical models of CaM channel dynamics are shown (0 = unoccupied, 1 = occupied). Top: two CaM necessary for CDI. Middle: one CaM triggers intermediate CDI strength and two CaM necessary for full CDI. Bottom: one CaM sufficient for full CDI. To see this figure in color, go online.

In addition to mutations that directly impact functions of NMDA receptors and CaM, several other proteins in which mutations have been linked to various neurological illnesses may exert their pathophysiological effects, in part, through alteration of NMDA receptor function through CDI. PSD-95 is found in the postsynaptic spines and is involved in channel anchoring and clustering (130). Our group previously found evidence that the strength of CDI is modulated by PSD-95 likely through the adjustment of interchannel proximity. PSD-95 mutations have been linked to increased incidence of schizophrenia and autism (131). Although disrupting the PSD-95/NMDA receptor interaction has no apparent synaptic functional phenotype (132), alteration of the spatial distribution of channels can have important ramifications in physiology, such as coordinated signaling and plasticity (78,100), which alter tissue function. Neurogranin also plays an important role in setting the postsynaptic CaM concentration and modulating the strength of plasticity (103,133). Mutations in neurogranin are reported in schizophrenia. Mutations that impact CaM binding influence the baseline CaM concentration and the proportion of channels capable of being modulated by CDI (79).

Dominant negative effects of distinct CaM binding modes

In vertebrates, CaM is encoded by three separate genes on different chromosomes. This redundancy of the CALM gene family is a testament to its functional importance. However, even mutations in a single gene results in severe disease. One mechanism that explains this is the ability of CaM to bind effector proteins in the absence of calcium. By preassociating with a target protein, a mutant CaM can still compete for effector sites on targets preventing wild-type CaM from acting on them thus acting as a dominant negative phenotype. In contrast, if only Ca²⁺-bound CaM can associate with targets, then the mutant CaM unresponsive to Ca²⁺ will not bind to targets and exert no dominant negative effect. This phenomenon was demonstrated in the case of voltage-gated Ca²⁺ channels where CaM preassociates with channels as a resident Ca²⁺ sensor. Adjusting the ratio of mutant and wild-type CaM in cardiac myocytes, the strength of CDI in voltage-gated Ca²⁺ channels conformed to a saturating relationship consistent with a small fraction of mutant CaM significantly impairing channel function (134). Several biochemical studies have suggested that CaM preassociates with the NMDA receptor at C0 in the absence of Ca²⁺ (87,135). Although others have not recapitulated this in vitro (58). Electrophysiological data comparing various binding models showed that a preassociation model best recapitulates the strength of CDI across various activity levels (50). Consistent with this, the estimated distance of CaM from the channel pore is ∼10 nm, which positions CaM well within the high amplitude Ca²⁺ nanodomain. An important avenue for future work will be utilizing in vivo methods such as FRET to determine more definitively whether CaM can preassociate with NMDA receptors at rest and under what physiological or pathological conditions this interaction is disrupted. In addition, given that NMDA receptors contain two GluN1 subunits each with a C0 element, a useful avenue will be to explore the in vivo stoichiometry (Fig. 4 b). Previous work has suggested a 2:1 stoichiometry in vitro but electrophysiological work to test this hypothesis is limited (58). By using the novel method employed to study triheteromeric receptors, one can gain insight as to the functional stoichiometry of CaM with the receptor (136). This will inform our ability to predict symptom severity in conditions where this interaction is impaired.

Historically, pharmaceutical targeting of NMDA receptors in clinical trials has been unsuccessful owing, in part, to the indispensable role NMDA receptors play in multiple neural processes (137,138,139). However it is this quality that also makes them desirable targets given their increasingly recognized role in neurological and psychiatric illness. As more nuanced pharmaceuticals continue to be developed that target specific receptor functions (140,141), a viable alternative is to target the peripheral regulatory mechanisms such as CDI. With continuing new developments and advances in our understanding of NMDA receptor structure and function, CDI represents a viable avenue for exploring disease mechanism and novel therapeutic strategies.

Editor: Meyer Jackson.