Differential distribution of NCX1 contributes to spine–dendrite compartmentalization in CA1 pyramidal cells

Abstract

Compartmentalization of Ca²⁺ between dendritic spines and shafts is governed by diffusion barriers and a range of Ca²⁺ extrusion mechanisms. The distinct contribution of different Ca²⁺ clearance systems to Ca²⁺ compartmentalization in dendritic spines versus shafts remains elusive. We applied a combination of ultrastructural and functional imaging methods to assess the subcellular distribution and role of NCX1 in rat CA1 pyramidal cells. Quantitative electron microscopic analysis of preembedding immunogold reactions revealed uniform densities of NCX1 along the shafts of apical and basal dendrites, but densities in dendritic shafts were approximately seven times higher than in dendritic spines. In line with these results, two-photon imaging of synaptically activated Ca²⁺ transients during NCX blockade showed preferential action localized to the dendritic shafts for NCXs in regulating spine–dendrite coupling.

Hippocampal pyramidal cells form glutamatergic synapses on dendritic spine heads. Imaging experiments using Ca²⁺ sensitive dyes revealed that synaptic stimulation raised intracellular Ca²⁺ to higher levels in spines than in dendrites, suggesting that spines are individual compartments able to accumulate Ca²⁺ (1–3). This compartmentalization offers the basis for selective regulation of single synapses and synaptic plasticity.

Several factors are involved in the compartmentalization of Ca²⁺ into spines. Ca²⁺ spatiotemporal dynamics is controlled by biophysical factors, influx, efflux, buffers, pumps, and stores (4, 5). Synaptic activation restricts Ca²⁺ influx to the spine head by activating NMDA receptors, Ca²⁺-permeable AMPA receptors, and voltage-gated Ca²⁺ channels (6). Dendritic spine necks serve as diffusion barriers by separating the spine head from its parent dendrite (7, 8), as demonstrated by the slow diffusion of endogenous Ca²⁺ buffers and rapid Ca²⁺ clearance through Ca²⁺-extrusion mechanisms (4, 5, 9). However, the distinct contribution of different extrusion mechanisms to Ca²⁺ compartmentalization in dendritic spines versus shafts remains elusive.

Na⁺/Ca²⁺ exchanger (NCX) and plasma membrane Ca²⁺-ATPase (PMCA) molecules play an essential role in Ca²⁺ clearance across the plasma membrane in excitable cells (10). Although PMCAs have higher affinity to Ca²⁺ than do NCXs, the much lower turnover rate of PMCAs may cause quick saturation at elevated levels of Ca²⁺ (10). The limited capacity of PMCAs suggests a more prominent role for NCXs in the process of rapid Ca²⁺ extrusion; indeed, NCXs drive Ca²⁺ extrusion after synaptic stimulation (11, 12). In addition to the turnover rate, the capacity of a transporter molecule is determined by its density on the plasma membrane. Distinct subcellular distributions could assign different roles to NCXs in Ca²⁺ homeostasis along with their exposure to specifically located regulatory systems.

There are three known NCX proteins (NCX1–3) encoded by different genes (13–15). Hippocampal pyramidal cells and astrocytes were shown to express NCX1 and NCX2 mRNAs (16–18), but they seem to have differences at the protein level. By using monoclonal antibodies, the NCX1 protein was localized to pyramidal cells, whereas NCX2 was predominantly found on glial cell plasma membranes in cell culture (19). Owing to the lack of selective NCX subtype inhibitors, pharmacology is insufficient to study the contribution of different NCX subtypes to Ca²⁺ homeostasis, especially in cells with complex dendritic structures, such as pyramidal cells. Thus, we applied a combination of ultrastructural and functional imaging methods to assess the subcellular distribution and role of NCX1 in rat CA1 pyramidal cells. Quantitative electron microscopic analysis of preembedding immunogold reactions revealed uniform densities of NCX1 along the shafts of apical and basal dendrites, but densities in dendritic shafts were Approximately seven times higher than in dendritic spines. In line with these results, two-photon imaging of synaptically activated Ca²⁺ transients during NCX blockade showed preferential action localized to the dendritic shafts for NCXs in regulating spine–dendrite coupling.

Results

Light Microscopic Distribution of NCX1 in Hippocampus and Neocortex.

We used a subtype-specific mouse monoclonal antibody (R3F1) raised against NCX1 to evaluate the cellular and subcellular distribution of NCX1. The antibody recognizes two intracellular epitopes located within the amino acid sequences 560–629 and 649–705, covering all NCX1 splice variants. The R3F1 antibody proved to be specific for NCX1 on Western blot (19) experiments. To further test the specificity of the antibody in brain slices, another mouse monoclonal antibody (MA3–926) raised against a different intracellular epitope (371–525) was used.

An intense immunolabeling was found in hippocampal CA1–CA3 areas with the R3F1 antibody (Fig. 1A and B). Although the MA3–926 antibody produced a much weaker signal (Fig. 1 C and D), the labeling patterns of the two antibodies were strikingly similar, suggesting that they specifically recognize NCX1. Astrocytes were also immunoreactive with the MA3–926 but not with the R3F1 antibody. Areas containing the dendritic compartments of pyramidal cells, such as stratum oriens (SO), stratum radiatum (SR), and stratum lacunosum moleculare (SLM) displayed the strongest immunoreactivity in CA1 and CA2, as well as SO and SR in CA3, followed by a less intense labeling in SLM of the CA3 area and the inner third of the molecular layer in the dentate gyrus (DG). Only a moderate level of immunoreactivity was observed in the CA3 stratum lucidum (SL) in the outer two thirds of the molecular layer of the DG and in the hilus. No immunolabeling was found in the somatic cytoplasm of pyramidal cells and granule cells.

Fig. 1.

Distribution of NCX1 immunoreactivity in the hippocampal formation. (A and B) Although the R3F1 antibody produced more intense signals than the MA3–926 antibody (C and D), labeling patterns of the two antibodies were strikingly similar, suggesting that they specifically recognize NCX1. (A and C) Immunolabeling is strongest in SO, SR, and SLM in the CA1 and CA3 areas. Moderate levels of immunoreactivity can be seen in the inner third of the molecular layer (m) in the DG. Lighter labeling is present in the outer part of the molecular layer of the DG and in the hilus (h). The weakest labeling was found in the granule cell layer (g) of DG, SL of the CA3 area, and stratum pyramidale of CA1 and CA3 areas. (B and D) Higher magnification images of the CA1 area. (Scale bars: A and C, 400 μm; B and D, 200 μm.)

In the somatosensory cortex, the most intense immunoreactivity was also associated with pyramidal cells [see supporting information (SI) Fig. 7]. The labeling outlined the dendritic plasma membranes (SI Fig. 7 A–C) of long apical dendrites of layer V pyramidal cells, whereas only few immunoreactive spines were found along the dendrites (SI Fig. 7C Inset). In addition to the moderate cytoplasmic labeling in dendrites and its lack in cell bodies, we observed immunoreactivity in the neuropil.

To characterize the subcellular distribution of NCX1 in pyramidal cells, we used preembedding immunogold reactions.

Subcellular Distribution of NCX1 in Pyramidal Cells.

Electron-microscopic evaluation of immunogold reactions confirmed our light-microscopic observations in pyramidal cells. Indeed, in neocortical layer II/III, most of the gold particles representing NCX1 labeling associated with plasma membranes of large apical dendrites and smaller oblique dendrites of pyramidal cells (SI Fig. 8). In rare cases, immunoreactive dendritic spines were also observed with gold particles attached to the plasma membrane of spine heads and spine necks.

To characterize the subcellular distribution of NCX1 in CA1 pyramidal cells, we took electron micrographs in different layers of the CA1 area. Immunoparticles were found on the somatic plasma membranes of pyramidal cells in the pyramidal layer (Fig. 2D). Some of the immunoparticles were located in plasma membrane regions closely apposed to underlying junctional endoplasmic reticules as suggested in cultured cortical neurons (20). Gold particles were also associated with rough endoplasmic reticulum (Fig. 2D). Dendritic domains of pyramidal cells were the most intensely labeled (Fig. 2 A–C) profiles. The majority of the gold particles was attached to the plasma membranes of thick apical dendrites in SR (Fig. 2B) as well as small-caliber pyramidal dendrites in SLM and SR (Fig. 2 A and C). Few gold particles were found in the cytoplasm, many of them associated with microtubules. Immunoreactive dendritic spines were also observed with gold particles taking place on the plasma membrane of spine heads as well as spine necks (Fig. 2 A Inset and C). Although previous studies have emphasized the role of Na⁺/Ca²⁺ exchangers in neurotransmitter release (20–22), we could hardly find immunoreactive inhibitory (SI Fig. 9A) or excitatory (SI Fig. 9B) axon terminals. It is possible that NCX1 is expressed in axon terminals at low levels barely detectable with our detection method. In the case of the rare examples of immunolabeled axon terminals, gold particles were homogeneously distributed along the plasma membrane, either in the vicinity or distal from the active zone. Immunoparticles were never associated with synaptic vesicles, in agreement with findings in synaptosomes (22). Axon initial segments and small axons were also immunonegative.

Fig. 2.

Subcellular distribution of NCX1 immunreactivity in hippocampal CA1 pyramidal cells. (A–C) The majority of gold particles is associated with plasma membranes of dendritic shafts (d; arrowheads), and some are present in dendritic cytoplasm (double arrows) in SO (A), SR (B), and SLM (C). Immunoparticles are attached to the plasma membranes of dendritic spines (s; arrows in A Inset and C), including spine heads and necks, but are not associated with asymmetric synapses (∗). (D) NCX1 immunoreactivity in the soma (so) of a pyramidal cell. Cytoplasmic (arrows) as well as plasma membrane attached (double arrowheads) immunogold labeling was observed. A gold particle is attached to the endoplasmic reticulum (double arrow). Nonspecific labeling was determined from the density of gold particles (open arrowheads) over nuclei (nuc). (Scale bars: A–C, 0.5 μm; D, 1 μm.)

Quantitative Evaluation of NCX1 Immunoreactivity in CA1 Pyramidal Cells.

To reveal differences in the subcellular distribution of NCX1 immunolabeling in the somatodendritic domain of CA1 pyramidal cells, we quantitatively analyzed the density of gold particles in the plasma membrane and cytoplasm of dendritic spines, dendritic shafts, and cell bodies. Samples representing distal dendrites, thick apical dendrites, and basal dendrites were taken in SLM, SR, and SO, respectively. Distal parts of apical dendrites in SLM receive inputs from the entorhinal cortex via the perforant pathway, whereas proximal parts of the apical dendrites are innervated by axons of CA3 pyramidal cells, called Shaffer-collaterals. This arrangement of synaptic inputs allowed us to investigate the distance and input dependence of the NCX1 distribution in the apical dendrites at the same time.

First, we tested whether immunoparticle densities obtained from different subcellular compartments are significantly different from nonspecific labeling over nuclei. The results are summarized in Fig. 3A and in SI Table 1. Twelve of 14 compartments contained significantly higher densities of gold particles than nuclei. Cytoplasmic pools of NCX1 immunolabeling were smaller in each measured compartment of dendritic shafts than NCX1 density in accompanying plasma membranes. Next, we compared the density of gold particles in plasma membranes between compartments (Fig. 3B) that contained significantly higher amounts of gold particles than the background. The density of immunoparticles labeling NCX1 in small distal dendritic shafts in SLM didn't differ significantly from densities in proximal apical dendritic shafts in SR or basal dendrites in SO (P > 0.05). Similarly, dendritic spines in SLM didn't contain significantly different amounts of gold particles than dendritic spines in SR or SO (P > 0.05). Furthermore, somatic plasma membranes of pyramidal cells contained similar densities of gold particles when compared with dendritic spines in SLM, SR, and SO. However, dendritic shafts contained 7.51 ± 1.15-fold (SLM: 6.49; SR: 8.75; SP: 7.29) a higher density of immunoparticles, than dendritic spines (P < 0.01).

Fig. 3.

Quantitative evaluation of immunogold reactions in CA1 pyramidal cells. (A) Dendritic cytoplasm (DeCy) as well as dendritic and spine plasma membranes (DeM and SpM, respectively) in the different layers contained significantly more gold particles than the nonspecific level measured over nuclei (Nuc). (B) Dendritic shafts of different layers contained significantly higher density of NCX1 than dendritic spines or somata (P < 0.01 for both). SpCy, spine cytoplasm; SoM, somatic plasma membrane; SoCy, somatic cytoplasm. (C) Distribution of gold particle densities in peri- and extraspine regions of dendritic shaft plasma membranes (see Materials and Methods). The horizontal lines indicate 5th, 25th, 50th, 75th, and 95th distribution percentiles, filled circles indicate minimum and maximum values, and open rectangles indicate means. Mean NCX1 density was ≈1.6 times larger in perispine regions than in extraspine regions. (C Inset) Cumulative probability plot of gold density distribution in peri- (filled square), and extraspine (open square) regions shows a shift to larger values in perispine regions. (D) Lateral distribution of gold particles in the plasma membrane of dendritic spines as a function of distance measured from the closer edge of the postsynaptic density (PSD). (D Inset) Pooled data from all layers is shown. The majority of gold particles avoided the perisynaptic region (within 200 nm) and were localized to the extrasynaptic membrane at a mean distance of 0.369 ± 0.02 nm.

The difference in the NCX1 density between dendritic shafts and spines suggests that NCX1 has a predominant role in Ca²⁺ homeostasis in dendritic shafts. We further analyzed whether NCX1 in dendritic shafts accumulates in the vicinity of spine origins. Densities of gold particles were determined over plasma membrane compartments within 300 nm of the base of spines (perispine regions) in dendritic shafts (n = 88.3 ± 17.7 spine bases per layer) in each layer. NCX1 densities around the spine origin didn't significantly differ from each other between different layers (Mann–Whitney U test; P > 0.05: SLM: 1.01 ± 0.62 g/μm; SR: 0.57 ± 0.2 μm; SO: 0.65 ± 0.17 μm), therefore, we pooled the data from all layers and compared (Fig. 3C; Mann–Whitney U test, 0.01 < P < 0.05) to gold densities calculated over the rest of dendritic shaft plasma membranes (extraspine regions). NCX1 densities were shifted to higher values in perispine regions (0.74 ± 0.41 g/μm) compared with the extraspine regions (0.48 ± 0.10 g/μm) of the dendritic shafts, suggesting a slight preference of NCX1 distribution to the base of spines.

In dendritic spines we investigated the distribution of NCX1 as a function of distance from the postsynaptic density (PSD) in NCX1 immunoreactive dendritic spines (Fig. 3D). In each layer, 10.1 ± 6.7% of the total number of gold particles was located in spine necks, whereas 89.9 ± 6.7% was located in spine heads. In all three layers, the majority of gold particles was localized to the extrasynaptic membrane at a mean distance of 0.369 ± 0.02 μm (SLM: 0.385 ± 0.258 μm, SR: 0.374 ± 0.257 μm; SO: 0.348 ± 0.192 μm) from the PSD.

Na⁺/Ca²⁺ Exchange Does Not Shape Ca²⁺ Signals Evoked by Single Synaptic Events.

Our anatomical results indicate a preferential role of NCX1 in dendritic shafts over spines. Therefore, in the next set of experiments, we further investigated the involvement of NCXs in the extrusion of synaptically evoked Ca²⁺ signals recorded simultaneously in these compartments (Fig. 4A) during extracellular stimulation of Schaffer collateral synapses in CA1 pyramidal neurons. We placed line scans on spines and parent dendrites and free line scans along the dendrites to find the active inputs and to exclude the simultaneous activation of additional neighboring synaptic contacts during the experiments (see Materials and Methods). Similar to earlier experiments (4), subthreshold synaptic stimulation produced intracellular Ca²⁺ concentration ([Ca²⁺]_i) increases restricted to individual spines (Fig. 4 A–C) without spreading to the dendritic shaft. This distinct, compartmentalized spatiotemporal Ca²⁺ dynamics remained preserved in spines and parent shafts after blocking Na⁺/Ca²⁺ exchange with benzamil (23). Benzamil did not elevate the amplitude of the shaft Ca²⁺ transients (control, 8.4 ± 2.6%, ΔF/F; benzamil, 8.8 ± 2.3%, ΔF/F; n = 8; P = 0.74; Fig. 4 B and C). Similarly, the decay (τ_control, 268 ± 27 ms, τ_benzamil to 247 ± 46 ms) and the amplitude (control, 60 ± 8%, ΔF/F, benzamil 56 ± 10%, ΔF/F) of Ca²⁺ transients in spines did not change significantly (n = 9; Fig. 4 B and C), suggesting that NCXs do not contribute to Ca²⁺ compartmentalization at single synaptic inputs. The measured Ca²⁺ signals were within the dynamic range of the indicator (see SI Methods). We could also reproduce these results in low-affinity indicator (OGB-5N), excluding the possibility of perturbation of spine–dendrite coupling with the high affinity OGB-1 (Fig. 4E). Although the resolution of two-photon microscopy is not high enough to selectively record from the neck of the spine, we checked the transient regions between the spine head and parent shaft and could not detect changes in the Ca²⁺ signal in benzamil (Fig. 4D).

Fig. 4.

NCX1 activity does not shape Ca²⁺ transients compartmentalized to dendritic spines during single synaptic stimulation. (A Upper) Z-stack of a CA1 pyramidal neuron. The red box and Inset indicates the region of interest selected for measurement. The multiple line-scan method was used to detect selectively activated synapses. The scanning path was repeated every 3 ms and is dotted on the picture of the dendrite; red segments were scanned with a moderate constant speed (4.2 μm/ms), whereas yellow segments were jumped over quickly (100 μs). Line scans collected in control conditions (Lower Left) and in the presence of the NCX blocker benzamil (30 μM; Lower Right) show that only one of the three spines was activated (sp₁). The red triangle indicates the time point of synaptic stimulation. (B) Ca²⁺ transients in spine sp₁ (Top) and the parent dendrite (Middle) in control condition (black traces) and in the presence of benzamil (red traces; failures are not shown). The Ca²⁺ transient was restricted to the dendritic spine without spreading into the dendritic shaft. (B Bottom) Average responses in the dendrite (d) and spine (sp) in the presence (red) and absence (black) of benzamil. (C) Amplitudes of Ca²⁺ transients in dendrites and spines did not change significantly in the presence of benzamil. (D) Benzamil did not affect the amplitude of Ca²⁺ transients in the transient region between the head of the spine and the dendritic shaft (neck; red traces). (E) The same experiment as in B but in the presence of low-affinity Ca²⁺ indicator (100 μM Oregon green BAPTA5 N).

We observed spontaneous synaptic activity in three cells. Benzamil did not change the properties of the spontaneous Ca²⁺ transients (control, 8.4 ± 2.6%; benzamil, 8.5 ± 4.1% ΔF/F).

Na⁺/Ca²⁺ Exchange Contributes to Biochemical Compartmentalization of Ca²⁺ in Individual Spines at Repeated Single Synaptic Events.

We have shown that Na⁺/Ca²⁺ exchange is not crucial for Ca²⁺ extrusion in dendritic spines at single synaptic events. However, because of their high turnover rate, the contribution of NCXs may increase at elevated levels of Ca²⁺ (10). Thus, we applied synaptic stimulation at individual spines with short trains (five to seven stimuli; length, 0.1 ms; amplitude, 10–20 V; frequency, 100–200 Hz). We scanned long (>100 μm) dendritic branches to find a single, individually active spine in response to the repeated stimulation. Adjacent to the active spine (sp1 in Fig. 5A), a segment of the parent dendrite and neighboring spines were simultaneously monitored for data collection (Fig. 5A). Dendritic Ca²⁺ transients were compartmentalized to a short segment of the parent dendrite peaking at the base of the active spine (d1 in Fig. 5C). Ca²⁺ transients were normalized to the peak of the spine transients to exclude the effect of the time-dependent decrease of Ca²⁺ signals. Benzamil broadened the spatiotemporal distribution of dendritic Ca²⁺ (Fig. 5C). Dendritic Ca²⁺ transients had a larger amplitude and a slower decay in the presence of benzamil (control: 24 ± 7% and 733 ± 155 ms; in benzamil: 32 ± 5% and 949 ± 165 ms; P = 0.036 and P < 0.01 respectively; n = 6; Fig. 5B). Benzamil also elongated Ca²⁺ transients in the individually active spine (τ_control: 280 ± 72 ms; τ_benzamil: 369 ± 71 ms; P < 0.04; n = 6; Fig. 5B). The change in the area of the dendritic Ca²⁺ transients correlated well with that of the spine transients (linear regression, R = 0.93; area was measured in the first 1,500-ms interval after stimulus onset), suggesting that the elevated dendritic [Ca²⁺]_i in the presence of benzamil may also contribute to the elongation of the spine transient by lowering the [Ca²⁺]_i gradient through the neck.

Fig. 5.

NCX1 activity shapes Ca²⁺ transients in dendritic shafts and spines during repeated stimulation of single synapses. Synaptic stimulation was elicited by seven consecutive stimuli delivered at 200 Hz. (A Upper) Z-stack of a dendritic segment imaged with a scanning path similar to Fig. 4. (A Lower) Average of four single traces. Only one spine (sp₁) and the corresponding parent dendritic segment were responding. (B) Normalized Ca²⁺ transients in a spine (Upper) and parent dendrite (Middle) in control conditions (black traces) and in the presence of 30 μM benzamil (red traces). (B Inset) Line-scan images of the responding spine show elongated Ca²⁺ transients in the presence of benzamil. (B Bottom) Simultaneously measured excitatory postsynaptic potentials (somatic recording). Traces are averages of four responses. (C Left) Normalized dendritic Ca²⁺ transients as a function of space and time revealed only one activated compartment in the dendritic shaft peaking at the base of the single responding spine (sp₁). Line scans were spatially normalized to the basal fluorescence measured along the dendrite [F₀(x)]. (C Right) The spatiotemporal distribution was wider in the presence of benzamil.

Na⁺/Ca²⁺ Exchange Shapes Functional Dendritic Compartments in Dendritic Shafts.

Earlier results show that increased stimulus number and intensity in focal extracellular synaptic stimulation may recruit several inputs, induces Ca²⁺ transients peaking in spines near the stimulating electrodes, and gradually decays to baseline along the dendritic segment forming a “computational subunit” (24). We therefore studied whether NCX channels, which are expressed at higher density in shafts, are involved in shaping the Ca²⁺ dynamics of these dendritic subunits. Using free line scans, we imaged long dendritic segments (Fig. 6A). Subthreshold synaptic stimulation elicited by two stimuli (0.1 ms, 10–20 V) at 100 Hz activated several spines with inhomogeneous distributions, forming active compartments in shafts (n = 11 from n = 18). Line scans were spatially normalized to the basal fluorescence measured along the dendrite [F₀(x)] and were integrated. The integral of the first 150 ms of the spatially normalized transients increased by 49 ± 14% (from 172 ± 32 ms·μm·% to 229 ± 39 ms·μm·%, P < 0.002), revealing a wider spatiotemporal spread of dendritic Ca²⁺ signal in the presence of benzamil. Benzamil also increased the amplitude (Fig. 6B, from 19.0 ± 2.0% to 28.8 ± 2.8%, ΔF/F, P < 0.0001) and the time constant of decay (from 837 ± 126 ms to 1138 ± 214 ms, P < 0.02) of dendritic Ca²⁺ transients (from 20 ± 2.5% to 24 ± 3.4%, ΔF/F, P < 0.05, Fig. 6D, n = 17) but did not change the amplitude of the excitatory postsynaptic potentials recorded at the soma (control, 10.1 ± 1.21 mV; benzamil, 9.97 ± 1.18 mV; P > 0.05, data not shown). We successfully reproduced these results using another potent NCX blocker, SEA0400 (SI Fig. 10) (25).

Fig. 6.

NCX1 contributes to Ca²⁺ elimination in “computational subunits” activated by increased subthreshold synaptic stimulation. (A) Z-stack of a dendritic segment (Top) scanned in free line scan mode (Middle). (A Bottom) Spatial and temporal broadening of Ca²⁺ transients in the presence of benzamil shown by spatially normalized fluorescent traces (Left, control; Right, benzamil). (B) Benzamil increased the amplitude of Ca²⁺ transients induced by synaptic stimulation.

The preferential dendritic effect of NCXs in Ca²⁺ extrusion was also observed on Ca²⁺ transients evoked by single or five repetitive backpropagating action potentials (bAPs; SI Fig. 11) in the presence of benzamil.

Discussion

Subcellular Distribution of NCX1.

In this study, we mapped the precise subcellular distribution of NCX1 immunoreactivity for the first time in the intact brain and revealed its preferential dendritic distribution in CA1 pyramidal cells. Available immunocytochemical data with the same antibody (R3F1) in hippocampal cell cultures are a bit confusing. Some studies suggest dominant somatic or somatodendritic localization (11, 19), whereas others emphasize preferential NCX1 presence in presynaptic terminals (21, 26) in pyramidal cells. Similarly, astrocytes have been shown either to have (11) or to lack (19) NCX1 labeling. We found labeled excitatory or inhibitory terminals only occasionally, which may suggest a low level of axonal expression in intact tissue. We did not observe labeling in glial plasma membranes with the R3F1 antibody. Astrocytes express exon B exclusively, whereas neurons have only exon A-containing NCX1 isoforms, thus, it is possible that the R3F1 antibody has much lower affinity to B variants, especially because the antigen used for raising R3F1 contained only A variants.

Our results suggest a subcellular compartment-dependent distribution of NCX1 in the plasma membrane of CA1 pyramidal cells. Dendritic shafts contained approximately seven-times-higher density of NCX1 than dendritic spines and somata. The compartment-dependent distribution of NCX1 is paralleled by the segregation of glutamatergic and GABAergic synapses on the dendritic spines and shafts of hippocampal pyramidal cells, respectively (27). Moreover, we found comparable NCX1 densities in dendritic shafts and spines sampled in different layers, rendering NCX1 distribution distance-independent. Dendritic compartments in different layers are innervated by different glutamatergic and GABAergic inputs (27). Entorhinal inputs composing perforant pathways innervate dendritic spines in the SLM, Schaffer collateral inputs from CA3 pyramidal cells terminate in SR and SO, and commissural afferents arrive in SO; therefore, our results also indicate an excitatory input-independent distribution of NCX1.

Gold particles showed a relatively even distribution without any particular preference to peri-, or extrasynaptic plasma membrane in dendritic spines. However, we found a slight accumulation of gold particles around the base of spines on dendritic shafts, which together with the high NCX1 density on dendritic shafts, implies a role for NCX1 in the restriction of spatial spread of synaptic Ca²⁺ signals, as was shown in neocortical interneurons (23).

The Function of NCXs in Ca²⁺ Compartmentalization in Spines.

In line with the relatively low density of NCX1 in spines, we could not detect significant changes in the decay kinetics of Ca²⁺ transients in dendritic spines, neither when NCX activity was inhibited during small intensity synaptic stimulation, resulting in Ca²⁺ signals restricted to spines, nor when bAPs were applied. Accordingly, the contribution of Na⁺/Ca²⁺ exchangers to Ca²⁺ signaling in spine heads does not appear to be central in extrusion. This finding can be explained by the slow activation kinetics of NCXs or the higher Ca²⁺ concentrations required for their activation. The involvement of NCXs in Ca²⁺ extrusion was indeed significant only at longer repetitive activation of synaptic NMDA receptors, producing a larger and longer Ca²⁺ influx to dendritic spines.

BAPs in pyramidal cells evoke fast Ca²⁺ transients in dendrites requiring fast activation of extrusion systems. Smooth endoplasmic reticulum has been proposed to sequester the majority of Ca²⁺ in this process, with little contribution of NCXs in dendritic shafts (28) as well as dendritic spines (7), but see ref 5. We have also found a small but significant contribution of NCXs to the dendritic, but not the spine, Ca²⁺ clearance in bAP-evoked Ca²⁺ transients.

The major strength of our newly developed imaging method lies in the fact that, by simultaneous scanning of neighboring spine–dendrite pairs, we are easily able to isolate the cases where only single spines are activated. This results in smaller Ca²⁺ signals restricted to spines compared with some reported larger spine signals (2, 29) invading also to dendritic shafts. Even with carefully chosen focal extracellular synaptic stimulation it is hard to selectively activate single synaptic inputs within the same experiment. It cannot be excluded therefore, that these larger reported calcium signals are contaminated by the activation of neighboring inputs, which may lead to an increased Ca²⁺ influx as a result of the removal of the magnesium block of NMDA channels or to the opening of voltage-gated Ca²⁺ channels. In more recent experiments, similar to our results, single synaptic inputs elicited well compartmentalized Ca²⁺ signals, which were restricted to spines without spreading to neighboring shafts (4).

The Function of NCXs in Ca²⁺ Compartmentalization in Dendritic Shafts.

Repetitive activation of single, individual synapses resulted in Ca²⁺ diffusion from the activated spines to the parent dendrite, where the Ca²⁺ signal is likely to be controlled by dendritic NCXs, as shown in our experiments with benzamil.

Our results propose that Ca²⁺ extrusion via dendritic NCXs can also shape the Ca²⁺ compartments formed in dendritic shafts after activation of neighboring inputs. These dendritic compartments are believed to function as individual integrative subunits (30).

According to our results, increased stimulus number and intensity of single inputs as well as multiple synaptic activation lead to increased Ca²⁺ concentration in dendritic shafts (31), possibly as a result of saturation or depression of the calcium-clearance mechanisms and enhanced Ca²⁺ diffusion from spines. The higher Ca²⁺ concentration in dendritic shafts may be the substrate of NCX1 located primarily on dendrites and suggests an increasing contribution of NCX1 as a function of synaptically elicited dendritic Ca²⁺ concentration.

Materials and Methods

Immunocytochemistry.

Tissue preparation and immunocytochemical reactions were conducted as described (32). Briefly, sections were incubated with one of two types of mouse monoclonal anti-NCX1 antibodies, either R3F1 (1:100; kindly provided by H. Porzig), or MA3–926 (1:100; Affinity Bioreagents, Golden, CO). Sections for immunogold reactions were incubated in goat anti-mouse IgG coupled to 0.8-nm gold particles (Aurion Immunoresearch, Wageningen, The Netherlands) and silver enhanced by using R-Gent SE-LM Silver kit (Aurion Immunoresearch). No specific immunoreactivity could be detected when either the primary or the secondary antibodies were omitted and the sections were treated with the silver kit.

NCX1 densities were determined in different subcellular compartments of CA1 pyramidal cells. Electron micrographs were taken within the same section of randomly chosen areas from each hippocampal layer in 1-μm depth from the surface. Nonspecific labeling was determined over nuclei. Quantitative evaluation of the ImmunoGold reactions was similar as described in our previous study (32), and it is detailed in the SI Methods section.

Electrophysiology and Imaging.

Slices were prepared and maintained according to the guidelines of the Institute of Experimental Medicine Protection of Research Subjects as described (33). Pyramidal cells in the CA1 subfield of the hippocampus were visualized under 900-nm infrared differential interference contrast (DIC; 900 nm; Olympus, Budapest, Hungary). Electrodes (6–9 MΩ) were filled with 125 mM potassium gluconate, 20 mM KCl, 10 mM Hepes, 10 mM Di-Tris-salt phosphocreatine, 0.3 mM Na-GTP, 4 mM Mg-ATP, 10 mM NaCl, and 60 μM Oregon green BAPTA-1. Recordings were made at 32°C by using a MultiClamp 700B Amplifier (Axon Instruments, Foster City, CA). Cells with a resting membrane potential more negative than −65 mV were accepted. Data acquisition was performed by using pClamp8 (Axon Instruments) and a MATLAB program written in our laboratory. We used 6- to 9-MΩ patch electrodes filled with ACSF for synaptic stimulation.

Imaging was performed by using a two-photon laser scanning system (see SI Methods) built in our laboratory (33), following the design of our modified confocal microscope (34). Regions of interest were scanned with constant speed, whereas intermediate sections were jumped over within 100 μs, by using a spline interpolated path (multiple line scan mode of scanning). This scanning mode increased the signal-to-noise ratio and also improved the probability of finding individual synaptic inputs. Data recording started 20–30 min after break-in. At the end of each experiment, a series of images across the depth of the volume encompassing the imaged neuron was taken. Measurement control, real time data acquisition, and analysis were performed with a MATLAB program (using C++ and assembly routines) developed in our laboratory. Fluorescence traces are expressed as relative fluorescence changes [ΔF/F = (F−F₀)/F₀], where F₀ is the background-corrected prestimulus fluorescence. If not indicated otherwise, data are presented as means ± SEM. To get the amplitude and decay of the transients, three sweeps were averaged before and after the perfusion of benzamil at high stimulus intensities.

During the experiments, the amplitude of Ca²⁺ transients decreased with time because of the continuous loading, in agreement with ref. 35. To exclude this effect, benzamil (30 μM) was applied by using continuous-flow capillary perfusion. The advantage of this method compared with the bath application is rapid drug delivery, so the effects of the drugs on the Ca²⁺ transient can be readily distinguished from the cell loading artifact. The measured parameters (amplitude and decay time constant of the Ca²⁺ transients) appeared as sharp changes as a function of time.

Abbreviations

DG

dentate gyrus

NCX

Na+/Ca2+ exchanger

SL

stratum lucidum

SLM

stratum lacunosum moleculare

SO

stratum oriens

SR

stratum radiatum.