In pyramidal neurons of the hippocampus and cortex, NMDA-type glutamate receptors (NMDARs) are the predominant sources of synaptically evoked calcium (Ca) signals [1–7] (Figure 9.1A through C). Ca influx through NMDARs regulates diverse processes including kinase and phosphatase activity, protein trafficking, structural and functional synaptic plasticity, cell growth, cell survival, and apoptosis [8–11]. Which of these many Ca-dependent processes are triggered when NMDARs open depend on the context of receptor activation and the magnitude, kinetics, timing, and spatial spread of the resulting Ca transients. This chapter reviews the features of NMDARs that determine Ca influx through the receptors and discusses how the context of NMDAR activation shapes synaptically evoked Ca transients.

FIGURE 9.1

(See color insert following page 212). NMDAR-dependent calcium influx into active spines is modulated by AMPARs and SK channels. (A) Spiny dendrite from a mouse CA1 pyramidal neuron filled through a somatic whole-cell recording electrode with the red (more...)

NMDARs are heteromeric tetramers typically composed of NR1 subunits and NR2 or NR3 subunits [12,13]. Each subunit has multiple isoforms and in some cases multiple splice variants [14,15]. This structural diversity is functionally relevant; the specific subunit composition of a receptor along with the timing and magnitude of local membrane potential fluctuations determines the duration and magnitude of Ca current through NMDARs.

Each of the four NR2 subunits (NR2A through D) has a glutamate binding site [16–18]. However, the affinity for glutamate differs among the isoforms such that NR2A has the lowest affinity, NR2D the highest, and NR2B and NR2C have intermediate affinities. Generally, high binding affinity indicates a low dissociation rate of glutamate from the receptor and more prolonged NMDAR opening following glutamate binding. Thus, differential glutamate affinity may explain some of the variability in NMDAR deactivation kinetics. Receptors containing the NR2A subunit generate currents that decay rapidly (t ~120 msec) in comparison to those containing NR2B and NR2C (t ~400 msec) or NR2D (t ~5 sec) [19–21].

Similarly, receptor affinity for Magnesium (Mg) also varies with the NR2 subunit such that NR2A and NR2B are more susceptible to block by extracellular Mg and show greater voltage dependence than NR2C and NR2D [19,22,23]. Finally, single-channel conductance is subunit-dependent such that NR2A- and NR2B-containing receptors conduct nearly twice as much current as NR2C- and NR2D-containing receptors [20,23–25]. For these reasons, activation of NR2A-containing receptors generates relatively large and fast currents. In comparison, current influx through NR2B-containing receptors is also large but lasts far longer. NR2C- and NR2D-containing receptors generate the smallest and longest lasting currents. For similar reasons, receptors containing different NR2 subunits generate Ca transients with different time courses.

The influence of NMDAR subunit composition on Ca signaling suggests that activation of receptors composed of distinct subunit combinations may trigger different biological pathways. This may partially explain the wide range of physiological outcomes associated with NMDAR signaling. Moreover, regulation of receptor subunit composition may provide a cell or even an individual synapse with an efficient mechanism for determining which Ca-dependent signaling cascades are engaged.

Recent studies employing a wide range of techniques suggest this may be the case. Immunogold electron microscopy suggests that NR2A- and NR2B-containing receptors are often segregated so that most spines express NR2A or NR2B but not both [26]. Stimulation of single postsynaptic terminals using two-photon uncaging of glutamate has shown that the contributions of NR2A- and NR2B-containing receptors to NMDAR-dependent currents and evoked Ca transients vary widely from spine to spine [27]. Furthermore, antagonism of NR2B-containing receptors with ifenprodil reduced the intraspine variability and the amplitude of NMDAR-mediated calcium transients, indicating that NR2A- and NR2B-containing receptors are present in spines and that NR2B-containing receptors flux more calcium.

A developmental switch from NR2B- to NR2A-containing NMDARs occurs in many brain areas [28,29]. However, a recent study indicates that the subunit composition is also rapidly regulated in response to plasticity inducing stimuli [30]. Thus, induction of long-term potentiation at CA3 to CA1 synapses in hippocampus of young rats is accompanied by a switch from NR2B- to NR2A-containing synaptic NMDARs. This switch accelerates the decay kinetics of NMDAR-mediated synaptic currents and, although not measured directly in the cited study, should also alter the time course of NMDAR-dependent Ca influx and Ca transients in spines.

Differential coupling to downstream signaling systems [31,32] may allow opening of NR2A- versus NR2B-containing receptors to have different functional implications for plasticity induction [33–35] (but see references [36–38]). Therefore, developmental and activity-dependent changes in NMDAR subunit composition, through regulation of synaptically evoked Ca influx, may constitute a form of metaplasticity that regulates the induction of activity-dependent synaptic plasticity.

Regulation by phosphorylation provides a rapid means to alter the Ca permeability of NMDARs. PKA activity enhances the Ca permeability of both NR2A- and NR2B-containing receptors [39]. Furthermore, NMDAR Ca signaling is controlled by a negative feedback loop by which repetitive activation of NR2B-containing NMDARs activates a serine–threonine phosphatase that decreases Ca permeability of NMDARs [31]. These data suggest that the Ca permeability of NMDARs may be regulated by an AKAP protein complex [40] associated with the NR2B subunit, although this has not been explicitly demonstrated.

As discussed above, the amount of Ca entering via open NMDARs is governed by many intrinsic features of the receptors including Ca permeability, glutamate affinity, and Mg affinity. When a synapse is stimulated, NMDAR-dependent Ca transients will also be shaped by extrinsic factors such as the context of receptor activation. For example, Ca current through a receptor is greatly regulated by membrane potential. Since the membrane potential is controlled by a wide array of ion channels, the activities of many channels have the capability to shape Ca influx through NMDARs. Furthermore, the concentration of Ca reached in postsynaptic terminals is determined by the Ca buffering capacity, Ca extrusion rate, and diffusional isolation of the terminals. This section and the following one consider the impacts of these extrinsic receptor factors on NMDAR dependent Ca transients.

The most powerful and rapidly adjustable factor that regulates NMDAR-dependent Ca flux is membrane potential. Changes in membrane potential alter the driving force for Ca entry and the degree of Mg block of the receptor [41]. The vastly asymmetric distribution of Ca across membranes results in a high Ca reversal potential (~125 mV assuming Ca concentration is 2 mM outside and 100 nM inside the cell). Because of this large driving force, a 20 mV depolarization from rest will reduce the driving force for Ca entry by roughly 10 to 15%. This depressive effect is modest compared to the large enhancement of Ca entry caused by partial relief of Mg block. As demonstrated in classic studies, the affinity of Mg for the NMDAR is decreased nearly 10-fold by a 20 mV depolarization in the subthreshold range [42]. Thus, depolarization from −70 to −50 mV, despite decreasing the driving force for Ca entry, increases current flow through NMDARs and the magnitude of evoked Ca transients.

The effects of voltage-dependent Mg block on synaptically evoked Ca transients can be seen in several contexts. In vivo, many neurons show large fluctuations in the resting membrane potential that, when mimicked by in vitro whole cell recordings, exert large effects on NMDAR-dependent Ca influx. For example, in striatal medium spiny neurons (MSNs) the amplitude of NMDAR-mediated Ca transients in active dendritic spines increases nearly four fold with depolarization from −80 to −50 mV [43]. This effect may contribute to the dependence of the induction of long-term synaptic plasticity on resting membrane potential in these cells [44–46]. As expected for NMDAR-mediated signals, similar effects of resting or holding potential on synaptically evoked Ca transients have been described in other cell types [1,3,4,47].

Transient changes in spine membrane potential occur during back-propagating action potentials (bAPs) that, in many cell types, can travel from the soma into the proximal dendrite and dendritic spines. Back-propagation of an action potential into the spine transiently relieves the Mg block of NMDARs. The rapid kinetics of Mg block [48] provides a brief enhancement of Ca influx through NMDARs that lasts approximately as long as the bAP and can be seen as a nonlinear increase in Ca entry into active spines [5,6,43,49–51].

Several lines of evidence indicate that, even in the absence of bAPs, the membrane of an active spine or a stretch of dendrite with multiple active synapses can experience large depolarizations that shape NMDAR-dependent Ca influx. These dynamic effects modulate Ca influx during the synaptic potential. In both hippocampal CA1 and lateral amygdala pyramidal neurons, blockade of SK-type Ca-activated K channels (SKs) modulates NMDAR-mediated synaptic currents in a Mg-dependent manner [52,53]. These studies suggested that SKs either repolarize or hyperpolarize the membrane near the active synapse and rapidly alter the Mg block of synaptically activated NMDARs.

This signaling cascade has been examined in more detail in spines of hippocampal CA1 pyramidal neurons [2] (Figure 9.2). In these cells, the blockade of SK channels with apamin nearly doubled the amplitude of NMDAR-mediated Ca transients in active spines. Opening of SKs in the spine is triggered by the entry of Ca through SNX-482 sensitive voltage-sensitive Ca channels (VSCCs, presumably Ca_V2.3). Since these are high, voltage activated VSCCs, this suggests that the spine must be depolarized many tens of millivolts to reach the threshold for activation of the channels. Furthermore, since the opening of other VSCCs that are known to be present in the spine does not activate SKs, SK channels must lie in the Ca microdomain of Ca_V2.3 VSCCs, possibly due to a physical association of the two ion channels.

FIGURE 9.2

Model of regulation of spine head Ca transients by glutamate receptors and ion channels. Glutamate release activates AMPARs and NMDARs in the spine head. Opening of AMPARs depolarizes the spine head, enhancing current flow through NMDARs by relief of (more...)

In addition to shaping single synapse responses, interactions of VSCCs and NMDARs determine synaptic responses and nonlinearities during near-synchronous stimulation of groups of synapses. Rapid activation of ~20 synapses on an individual segment of a radial oblique dendrite of a CA1 pyramidal neuron generated a Ca spike in the dendrite that was detectable in the soma as a rapid, all-or-none rising phase to the compound EPSP [54]. Interactions of NMDARs, VSCCs, and voltage-sensitive Na channels in these dendrites boosted the somatic potential and the dendritic Ca influx, presumably through increased relief of Mg block of synaptically activated NMDARs. Such locally confined dendritic spikes may play a role in the induction of associative plasticity at distal synapses [55].

In cortical layer 5 pyramidal neurons, activation of clustered synapses on basal dendrites produced an NMDAR-mediated spike detectable as an all-or-nothing ~5 to 10 mV depolarization at the soma and Ca transient in the dendrite [56]. The proposed mechanism for this spike is that synaptic depolarization relieves Mg block of the NMDARs, and this increases inward current flux through the receptors, further depolarizing the dendrite and further relieving Mg block. This positive feedback loop is enhanced by VSCCs and voltage-sensitive Na channel opening but neither of these channels is strictly necessary for spike initiation.

Finally, in striatal medium spiny neurons (MSNs), clustered activation of synapses on a short stretch of dendrite (5 synapses in ~10 microns) also boosted synaptic potentials and Ca transients in an NMDAR- and VSCC-dependent manner [57]. However, no dendrite spike was elicited and graded increases in the amplitude and duration of the EPSP were seen. Ca influx into the active spine was enhanced, presumably due to increased relief of Mg block during the enhanced potential.

The mechanisms described above nonlinearly boost synaptic signals by modulating the Mg block of NMDARs. They either require or are enhanced by the activation of VSCCs. The depolarization of 20 to 40 mV required for opening high-voltage-activated VSCCs such as Ca_V2.3 suggests that the submillivolt unitary EPSPs measured at the soma represent highly filtered versions of much larger depolarizations in the spine. Since synaptic depolarization is principally mediated by current flow through AMPARs, these results predict that blocking AMPARs should produce a significant impact on synaptically evoked Ca transients in spines. The reported effects of AMPAR blockade on synaptic Ca transients range broadly from small reductions to complete blockades [1,51,58].

Measuring the amplitude of Ca transients may miss large effects of AMPARs on NMDAR-dependent Ca influx. It is likely that the potential in the spine during synaptic activation rises quickly and, due to the activation of repolarizing currents such as SK channels, falls quickly. For these reasons, the depolarization in the spine is likely brief and blocking AMPARs may only affect NMDAR-dependent Ca influx in a short period immediately after synaptic stimulation. Since Ca indicators slow Ca clearance and report the integrated Ca influx through NMDARs over hundreds of milliseconds [4,27,59,60], the effects of AMPARs on Ca influx in active spines may be largely obscured.

To fully reveal the effects of spine depolarization of synaptically evoked Ca transients, quantitative models of the action of Ca buffers [61–65] on Ca handling may be used to calculate the Ca current into the spine. This can be accomplished by deconvolving the Ca transient with the impulse response of the spine to a brief, small increase in Ca [4,59]. In regimes of linear Ca handling, this approach reveals the time course of the Ca current into the spine. In preliminary studies using this approach, the Ca current that underlies synaptically evoked Ca transients in spines of mouse CA1 pyramidal neurons is comprised of fast and slow components (Figure 9.1D through F; authors’ unpublished data).

The amplitude and time course of the fast component are regulated by the actions of ion channels in the spine such as SKs and AMPARs (Figure 9.1D through F), whereas the slow component depends only on the number of open NMDARs. The fast phase represents Ca influx through VSCCs and NMDARs and is modulated by the amplitude and duration of synaptic depolarization in the spine. The slow component outlasts the EPSP and reflects Ca influx through NMDARs after the spine returns to its resting potential. Regulation of these two distinguishable phases of Ca influx may have important functional consequences for activation of downstream Ca-dependent processes such as synaptic plasticity, although this has yet to be demonstrated.

The morphology of the spine may impact Ca transient in the spine head in two ways. First, since the surface-to-volume ratio of a sphere is inversely proportional to the radius, changes in the size of the spine head may impact both the amplitude and kinetics of Ca transients. If Ca channels are present on the spine head at constant density or number, the amplitudes of Ca transients should be smaller in larger spines than in smaller spines. However, this simple relationship is not found experimentally and the sizes of synaptically evoked Ca transients and spine volumes are only poorly correlated [27]. Similarly, at a constant density of Ca transporters and pumps, the clearance of Ca should be slower in larger spines. If Ca diffusion across the spine neck plays an important role in clearing Ca from the head, spines with longer and thinner necks or larger heads should clear Ca more slowly than spines with shorter and thicker necks or smaller heads, leading to longer lasting and larger Ca transients [27,66–68].

The role of Ca diffusion across the spine neck in clearing Ca from the spine head is contentious. Several studies of Ca handling in the spine at room temperature led to the conclusion that Ca diffusion across the neck is a significant mechanism of Ca clearance and that differences in spine morphology directly account for interspine variability in the amplitude and kinetics of Ca transients [60,66,67,69]. In contrast, studies performed at near physiological temperatures revealed that variability in spine neck morphology does not significantly impact synaptically evoked Ca transients in dendritic spines and that most Ca clearance occurs via pumping or transport from the cytoplasm [4,27,59].

NMDARs contribute most synaptically evoked Ca influx into dendritic spines. Ca influx through NMDARs depends on many factors intrinsic to receptors such as subunit composition and phosphorylation state. Ca influx through receptors and the properties of synaptically evoked Ca accumulations in the spine are also regulated by many extrinsic receptor factors. These include relatively stable parameters such as the number of ion channels and Ca pumps and the buffer capacity and morphology of the spine. In addition, changes in the resting potentials of neurons and rapid changes in membrane potential at synapses during synaptic potential strongly influence the amplitude and kinetics of Ca influx through NMDARs. These factors act together to determine the amplitude and kinetics of synaptically evoked Ca transients in dendritic spines and, in ways that are not yet clear, determine the coupling of NMDAR opening to the activation of downstream Ca-dependent processes.