Regional differences in hippocampal calcium handling provide a cellular mechanism for limiting plasticity

Abstract

Although much is known about the mechanisms underlying synaptic plasticity, the cellular mechanisms that negatively regulate plasticity in some brain regions are considerably less studied. One region where neurons do not reliably express long-term potentiation (LTP) is the CA2 subfield of the hippocampus. Given the connection between synaptic plasticity and increases in postsynaptic [Ca²⁺], and that CA2 neurons express a large number of calcium-regulating proteins, we tested the hypothesis that the relative lack of LTP in CA2 results from differences in the calcium dynamics of these neurons. By measuring calcium-dependent fluorescence transients in dendritic spines, we show that CA2 neurons have smaller action potential-evoked intracellular Ca²⁺ transients because of a higher endogenous Ca²⁺-buffering capacity and significantly higher rates of Ca²⁺ extrusion when compared with CA1 and CA3 neurons. Perfusion with higher external [Ca²⁺] during induction restores LTP to CA2 neurons, suggesting that they possess the cellular machinery required for plasticity, but that the restriction of postsynaptic [Ca²⁺] limits its expression. Camstatin, an analogue of the calcium-modulating protein Pep-19 strongly expressed in CA2 neurons, blocked LTP and increased Ca²⁺ extrusion in CA1 neurons, suggesting a role for extrusion in the regulation of plasticity in CA2. In agreement with this idea, we found that intracellular introduction of a PMCA pump inhibitor (carboxyeosin) allows for the induction of LTP in CA2 neurons. Our results indicate that regulation of postsynaptic [Ca²⁺] through modulation of extrusion and/or buffering regulates expression of LTP in CA2 and potentially other brain regions.

Long term potentiation (LTP) of glutamatergic synaptic transmission is thought to mediate many aspects of brain plasticity, but it only occurs in some regions and at some stages of development (1). Although many of the molecular mechanisms underlying LTP have been elucidated, little is known about the mechanisms that negatively regulate LTP. An example of a dramatic regional difference in synaptic plasticity has recently been described between the CA1 and CA2 fields of the hippocampus (2). Specifically, pyramidal neurons in CA2 do not reliably express LTP in response to synaptic high-frequency stimulation (HFS) or pairing protocols, unlike their CA1 counterparts.

This relative lack of synaptic plasticity in CA2 raises two interesting questions. First, what is the functional reason for this differential expression of plasticity within a structure where plasticity is especially prevalent, and second, what are the mechanisms that limit plasticity in this and other regions of the brain? In this study, we have addressed the latter of these two issues. Here we demonstrate that differences in calcium extrusion in CA2 neurons can largely explain the relative differences in plasticity in the hippocampus. We suggest that regulation of postsynaptic calcium concentration through differential calcium handling is one way by which plasticity can be modulated across different brain regions.

Results

Synaptic plasticity requires increases in postsynaptic calcium (3, 4), and so we tested whether calcium handling in CA2 pyramidal neurons differs significantly from handling in CA1 or CA3 pyramidal neurons. To compare the calcium dynamics in dendritic spines of pyramidal neurons from each hippocampal CA subfield, we used dual indicator 2-photon laser scanning microscopy imaging of calcium-dependent fluorescence transients. Fig. 1B shows that maximum fluorescence evoked by single action potentials in dendritic spines was consistently higher in CA1 and CA3 neurons when compared with CA2 neurons, indicating smaller increases in intracellular free [Ca²⁺] in CA2 spines at all concentrations of indicator used (Fig. 1C). Similar results were obtained from primary apical dendrites (Fig. S1). To obtain measures of the endogenous buffering capacity (K_E), and Δ[Ca²⁺]_AP in the absence of indicator (Δ[Ca²⁺]_AP^[dye]=0), we plotted Δ[Ca²⁺]_AP⁻¹ as a function of the added buffer capacity (K_b = [Ca²⁺ indicator]) and extrapolated linear fits to the data back to 0 K_b. Because the linear fits to the data at two indicator concentrations were in close agreement with previous studies, no additional concentrations were tested in CA1 and CA3 (Fig. 1 and Fig. S1) (5–7). A third indicator concentration was used in CA2, however, to ensure a valid linear approximation. We observed substantially smaller Δ[Ca²⁺]_AP^[dye]=0 (Fig. 1C; P < 0.001) and a significantly larger endogenous buffering capacity in CA2 spines when compared with CA1 and CA3 spines (P < 0.01). The decay time (τ_decay) was similarly computed by plotting τ_decay as a function of K_b and was not significantly different between spines from any of the three subfields (Table 1). We conclude that CA2 dendritic spines have increases in total [Ca²⁺] similar to those in CA1/CA3 (Table 1) (7), but the higher buffering capacity of CA2 neurons results in smaller free [Ca²⁺] transients. Because CA2 neurons possess a significantly higher buffering capacity and that no measurable differences in decay time were observed, we also conclude that CA2 neurons have stronger extrusion activity [extrusion activity = (1+ K_E)/τ_decay] (7). We propose that the attenuation of calcium because of the higher buffering capacity and/or extrusion activity observed in the apical dendrites and spines of CA2 neurons could play a critical role in restricting LTP induction in CA2.

Fig. 1.

Ca²⁺ buffering in dendritic spines of hippocampal neurons. (A) CA1, CA2 and CA3 spines; linescans (dashed line) were performed on spines located <150 μM from the soma on secondary or tertiary dendrites. (B) Ca²⁺-dependent fluorescence traces evoked by single and high frequency (40 AP at 83 Hz; 30 AP at 63 Hz) trains of action potentials, bars under high frequency traces indicate time of AP firing (∼500 ms). Traces represent averages of 15 and 5 trials for single and high frequency conditions respectively. Single AP traces are fit with an exponential decay function (dashed line) used in calculation of τ_decay. (C) Plots of Δ[Ca²⁺]_AP⁻¹ as a function of indicator concentration (K_b). Linear fits extrapolated back to 0 K_b (dotted lines) provided estimates of Δ[Ca²⁺]_AP^[dye]=0 and the endogenous buffering capacity (K_E = -K_b0). Endogenous τ_decay was computed using the same method. Error bars, SEM.

Table 1.

Calcium handling in hippocampal dendritic spines

	[Ca2+]0, nM	Δ[Ca2+]AP, nM	KE	τdecay, ms	γ, s−1	Δ[Ca2+]total, μM
CA1 (n = 19)	116 ± 16	1,055 ± 152	11 ± 2	35 ± 12	343	12.6
CA2 (n = 28)	89 ± 18	227 ± 70*	41 ± 3*	34 ± 10	1,235*	9.5
CA3 (n = 15)	107 ± 14	1,253 ± 135	9 ± 2	24 ± 9	417	12.5
CA2 10 mM [Ca2+]ext (n = 13)	105 ± 10	1,309 ± 114**	27 ± 4	106 ± 14**	264**	36.7*

Calcium handling in hippocampal dendritic spines. Resting calcium ([Ca²⁺]₀), changes in free calcium in response to single action potentials (Δ[Ca²⁺]_AP), endogenous buffering capacity (K_E), and decay time constants (τ_decay) were calculated at zero added buffer. Extrusion activity (γ ) and total increases in calcium (Δ[Ca²⁺]_total) were computed using previously used methods (5, 7, 34).

* indicates statistical significance (P < 0.001) compared against CA1 controls, and

** indicates statistical significance (P < 0.001) compared against CA2 controls.

If the achievable or sustainable levels of calcium in CA2 neurons are the restricting factor in this region's expression of LTP, then it follows that elevation of the external calcium concentration could overcome these deficits and rescue LTP in CA2. To test this hypothesis, we reexamined the calcium handling characteristics and response to LTP inducing protocols of CA2 neurons while perfusing slices with artificial cerebrospinal fluid containing higher [Ca²⁺]. During perfusion with 10 mM external [Ca²⁺], apical dendrites and spines of CA2 neurons showed a substantial increase in Δ[Ca²⁺]_AP^[dye]=0 that matched or exceeded increases in calcium seen in CA1 and CA3 neurons under control conditions (Fig. 2A and B, Table 1, and Fig. S1). Perfusion with 10 mM [Ca²⁺] also restored LTP in CA2 neurons (Fig. 2C). Calcium buffering remained relatively unchanged, whereas τ_decay was only substantially longer in CA2 spines, suggesting saturation of Ca²⁺ pumps/exchangers or Ca²⁺-dependent inhibition of extrusion (8) (Table 1). Sufficient elevation of extracellular [Ca²⁺] is proved to be critical in the expression of stable LTP in CA2 given that perfusion with 4 mM extracellular [Ca²⁺] (double the concentration of normal ACSF) resulted in essentially normal LTP in CA1, but did not allow induction of lasting LTP in CA2 neurons. The NMDA receptor antagonist APV effectively blocked the induction of LTP in high calcium experiments indicating a dependence on calcium influx through NMDA receptors (Fig. 2C). Based on these results, we conclude that CA2 neurons possess the cellular machinery needed to express LTP, but fail to do so under normal physiological conditions because of the restriction of postsynaptic calcium via higher intracellular buffering or extrusion activity.

Fig. 2.

High bath Ca²⁺ restores LTP and significantly increases free calcium in CA2 neurons. (A) Ca²⁺ handling in CA2 dendrites and spines with 10 mM bath calcium. Ca²⁺ dependent fluorescence traces evoked by single and high frequency (40 AP at 83 Hz; 30 AP at 63 Hz) trains of action potentials. (B) Plots of Δ[Ca²⁺]_AP⁻¹ as a function of indicator concentration (K_b). Linear fits extrapolated back to 0 K_b (dotted lines) provided estimates of Δ[Ca²⁺]_AP^[dye]=0 and the endogenous buffering capacity (K_E = -K_b0). Error bars indicate SEM. (C) Response of CA2 neurons to high frequency stimulation (HFS) using 4 mM (gray circles, n = 10), 10 mM (black circles, n = 12), and 10 mM Ca²⁺ ACSF with 50 μM APV (open circles, n = 6) during LTP induction. Similar experiments in CA1 neurons using 4 mM Ca²⁺ ACSF are shown for comparison (gray triangles, n = 8). Time of perfusion (0–3 min) with high Ca²⁺ ACSF is indicated by solid gray bar. Arrow indicates onset of HFS. Error bars, SEM.

Although it is known that intracellular injection of added calcium buffers (i.e., EGTA, BAPTA) can block the expression of LTP (3), we find it unlikely that the higher buffering capacity described in CA2 neurons alone can explain the lack of LTP in this region. For example, the increased buffering capacity at the spine in CA2 neurons would only equate to the addition of approximately 20 μM BAPTA in CA1 neurons [ΔK_E = ([BAPTA] · K_d(BAPTA))/(Ca²⁺₀ · K_d(BAPTA))(Ca²⁺₀ · Ca²⁺_AP· K_d(BAPTA))], a concentration that proved insufficient to block LTP on its own (Fig. S2) (4, 9). This suggests that increased Ca²⁺ extrusion may in fact play a larger role in limiting the expression of LTP in CA2.

What factors could account for the increased extrusion in CA2? Immunohistochemistry studies looking at the expression of different extrusion pumps (10) or exchangers (11) have been performed but do not reveal any differences that could explain the robust calcium extrusion in CA2 relative to CA1. Another possibility is that additional proteins or modulators with higher expression in CA2 could increase the efficiency of already present systems for calcium extrusion. Pep-19 is 1 such candidate protein (12, 13, 24) (Fig. S3) that has been shown to influence [Ca²⁺] dynamics through its interaction with calmodulin (CaM) (14, 15). To determine whether Pep-19 could account for any of the differences observed in calcium handling or the expression of plasticity in CA2 neurons, we added to the internal electrode solution a functional fragment of Pep-19 (camstatin) containing a similar IQ motif for binding to CaM (16). We then tested its effect on Ca²⁺ handling and the expression of LTP in CA1 neurons, where Pep-19 is not endogenously expressed at high levels.

Although the addition of camstatin did not significantly impact Δ[Ca²⁺]_AP^[dye]=0 or buffering in CA1 dendrites, the decay time constantwas significantly shorter (50 ms - camstatin; 140 ms - controls), resulting in a much higher calculated extrusion activity (780/s - camstatin vs. 329/s - controls; Table S1). The addition of 10 μM camstatin completely blocked LTP induction after high frequency stimulation (HFS) and had no apparent effect on baseline synaptic responses or on calcium influx through voltage gated calcium channels (Fig. 3C and Fig. S4). Heat inactivated camstatin was tested as a control and it failed to block LTP in CA1 neurons (n = 5).

Fig. 3.

Camstatin blocks LTP and dramatically increases extrusion in CA1 neurons. (A) Ca²⁺ dependent fluorescence traces evoked by single and high frequency (40 AP at 83 Hz; 30 AP at 63 Hz) trains of action potentials. Exponential decay fits to traces of single AP evoked fluorescence demonstrate a significant shortening in τ_decay with camstatin (50 ms) vs. controls (140 ms). (B) Plots of Δ[Ca²⁺]_AP⁻¹ as a function of indicator concentration (K_b). Linear fits extrapolated back to 0 K_b (dotted lines) provided estimates of Δ[Ca²⁺]_AP^[dye]=0 and the endogenous buffering capacity (K_E = -K_b0). Error bars, SEM. (C) The response of CA1 neurons to HFS was measured using normal internal solution (gray triangles, n = 9), and internal solution containing 10 μM camstatin (black circles, n = 11). Heat inactivated camstatin failed to block LTP (open circles, n = 5). The response to continuous 0.05-Hz stimulation was also measured with 10 μM camstatin (gray circles, n = 6).

Because the rates of Ca²⁺ extrusion in CA1 neurons with camstatin were, similar to CA2 neurons, significantly elevated from CA1 control cells, we hypothesized that this difference is primarily responsible for the limitation of LTP observed in CA2. Another observation in support of this view is that in our high external calcium experiments restoring LTP, extrusion of calcium in CA2 spines was effectively overwhelmed (Table 1). Because the impact of increased calcium extrusion on the expression of LTP has not been studied, we investigated the effect of a plasma membrane calcium ATPase (PMCA) antagonist (carboxyeosin) on LTP in CA2. We have focused our attention on this pump because of its interaction with CaM, a target of Pep-19 action (17). Fig. 4A shows that the introduction of 10 μM carboxyeosin (CE) into CA2 neurons significantly reduced rates of Ca²⁺ extrusion (lengthened τ_decay) in spines and dendrites. Using the ratio of τ_decay measured with and without CE, we can estimate the impact of CE on the endogenous τ_decay in spines (τ_decay = τ_decay* ratio). This calculation gives us a τ_decay of 260 ms and computed extrusion activity of 161/s, a rate closer to the rates of extrusion measured in CA1 spines (Table 1). CE had no significant effect on calcium influx or resting calcium levels. The addition of CE ‘rescued’ the expression of LTP in a majority (n = 10/13) of CA2 neurons but did not alter the long term stability of baseline synaptic responses (Fig. 4B). Control experiments using 0.01% DMSO vehicle failed to show a restoration of LTP in CA2 neurons (n = 6). Together with the results from experiments using high extracellular [Ca²⁺], these data indicate that extrusion of calcium is an effective regulator of synaptic plasticity in neurons.

Fig. 4.

Ten micromolar intracellular carboxyeosin (CE) reduces rates of Ca²⁺ extrusion and rescues LTP in CA2 neurons. (A) Single AP responses measured with the calcium indicator X-Rhod 5F (300 μM) in CA2 neurons with and without 10 μM CE. Average τ_decay (n = 8) are indicated for the single dye concentration shown and are significantly longer with CE. (B) Synaptic responses from CA2 neurons loaded with the PMCA pump inhibitor CE (black circles, n = 10) after HFS. Arrow indicates timing of HFS (100 Hz). The impact of CE on baseline synaptic responses (gray circles, n = 8) and the response to 0.01% DMSO vehicle controls (n = 6) over the same time period is also shown. Error bars, SEM.

Discussion

In this study, we have demonstrated that CA2 pyramidal neurons display smaller action potential-evoked free calcium transients, have a higher Ca²⁺-buffering capacity, and have significantly faster rates of Ca²⁺ extrusion than pyramidal neurons in either CA1 or CA3. Further, we show that a functionally similar fragment of one protein, highly expressed in CA2, substantially increases Ca²⁺ extrusion and is sufficient to block LTP induction when introduced into CA1 neurons. Higher concentrations of extracellular Ca²⁺ or inhibition of PMCA pumps is sufficient to overcome these differences in Ca²⁺ handling and restores LTP to CA2 neurons. Determining how and whether the Ca²⁺ dynamics might be modulated to allow plasticity in CA2 neurons under some physiological conditions will require further investigation.

Based on the observations that most forms of synaptic plasticity require the influx of Ca²⁺ through NMDA receptors (18), several models have attempted to explain bidirectional synaptic plasticity. One recent model suggested that when postsynaptic [Ca²⁺] is restricted to lower concentrations (0.4–0.5 μM), LTD is produced, whereas only higher concentrations (0.6–1.0 μM) result in LTP (19). Unlike pyramidal neurons in CA1, however, neurons in CA2 do not reliably express LTP in response to high frequency synaptic stimulation but more frequently display either no change in the synaptic response or LTD (2). In this study, we have shown that the increase in free calcium in response to an action potential in CA2 dendritic spines is significantly smaller than what is achieved in spines of CA1 or CA3 neurons (∼0.2 μM vs. >1.0 μM, respectively). Furthermore, the enhanced calcium extrusion in CA2 may prevent the sustained increase in intracellular [Ca²⁺] required for LTP (20). These limitations can be overcome by raising the external [Ca²⁺], which increases postsynaptic [Ca²⁺] in spines by apparently overwhelming the Ca²⁺ extrusion to restore LTP in these neurons.

Ca²⁺-Regulating Proteins in CA2.

The differences in calcium handling in CA2 neurons can likely be explained in part by the higher expression of a number of calcium-binding proteins (CaBPs) in this region, including but perhaps not limited to calbindin, efhcbp2, and parvalbumin (13, 21, 22). Our data confirm that calcium buffering is in fact higher in CA2 neurons. Previous work has shown that CA2 neurons display a surprising resistance to trauma resulting from epileptic activity or hypoxia, and the higher calcium buffering of CA2 neurons could play a major role in this resistance (22). Calbindin has been hypothesized to be an important contributor in this regard (22).

Although Pep-19 does not directly bind Ca²⁺, it has been shown to impact Ca²⁺ dynamics through its interaction with CaM (16). In the present work we have demonstrated that camstatin, a functional fragment of Pep-19, is capable of blocking LTP and dramatically increasing calcium extrusion in CA1 pyramidal neurons. Pep-19 is a cytosolic polypeptide that belongs to a family of proteins including neurogranin and GAP-43, which share an IQ motif for binding to CaM (16). When bound, Pep-19 exerts a number of effects on CaM that make it particularly relevant to the discussion of Ca²⁺ dynamics and synaptic plasticity. First, binding of Pep-19 to CaM has been shown to increase the rates of association and dissociation of calcium to CaM by as much as 40- to 50-fold (25, 27). Models have predicted that this would reduce the increase in free calcium and rapidly saturate and release Ca²⁺ from the C-domain of CaM during high frequency stimulation (14, 25). The PMCA pump is strongly regulated by the C-domain of CaM, and its rate of calcium efflux is enhanced by >30-fold when bound to calcium loaded CaM (23). Pep-19 may therefore indirectly influence Ca²⁺ extrusion through its interaction with CaM, and so it could contribute to the profound differences in calcium dynamics and synaptic plasticity in CA2. While it is unlikely that camstatin produces all of the same biochemical effects in the target cell as Pep-19 (26, 27), our data indicate that the fragment is sufficient to produce a robust increase in calcium extrusion. More work will be required to determine the mechanism by which camstatin confers this increase in extrusion. Curiously, dialysis of camstatin into the postsynaptic neuron also blocked post tetanic potentiation (PTP), which is widely considered to be a presynaptic phenomenon (28). Postsynaptic calcium chelators have been shown to block PTP (3), and although camstatin does not bind Ca²⁺ directly, it does increase the binding kinetics of Ca²⁺/CaM and thus could potentially impact PTP (25). Another possibility that we cannot rule out is that camstatin retrogradely diffuses across the synaptic terminal and reduces the presynaptic calcium accumulation thought to drive PTP (28), although that a peptide would be capable of this seems less likely.

One potentially confounding property of Pep-19/camstatin is that because it acts through CaM (16), it might be expected to inhibit CaM dependent protein kinases necessary for the induction of LTP (29). Our experiments with 10 mM external calcium during LTP induction, as in our experiments rescuing LTP in CA2 neurons, successfully restored LTP in CA1 neurons with camstatin (Fig. S5). Therefore, we conclude that the restriction of postsynaptic Ca²⁺ was the limiting factor for LTP induction rather than the antagonism of CaM-dependent kinases.

Calcium Extrusion in CA2.

Our data implicate Ca²⁺ extrusion as an important mechanism for the regulation of LTP expression. Recent work has shown that PMCA pump isoforms tend to cluster around the postsynaptic density (PSD-95) (30), and that their migration into dendritic spines near the second week postnatal possibly signifies a maturation of hippocampal neurons (31). This could contribute to the high relative impact of CE observed in dendritic spines in our study. Our data show that manipulations increasing or prolonging the rise in postsynaptic Ca²⁺ allow for the expression of LTP in CA2 neurons and suggest that the rapid extrusion of postsynaptic Ca²⁺ in CA2 is sufficient to limit plasticity. Interestingly, manipulations that were successful in restoring LTP to CA2 neurons (i.e., addition of CE and 10 mM external [Ca²⁺]) actually prevented the induction of LTP in CA1 neurons, suggesting that there is a critical window of calcium required for the induction of plasticity, above which plasticity is not induced (Fig. S6). One study has similarly shown that increasing the external calcium concentration may actually reduce the magnitude of LTP expressed in CA1 (32). Importantly, our results also demonstrate that CA2 neurons possess the biochemical machinery for expressing LTP, but fail to do so in response to physiological ranges of stimuli. Although we find it less likely, the possibility exists that other differences may be responsible for the lack of plasticity in CA2. One possibility could be that downstream signaling molecules such as CaMKII are localized in CA2 neurons further from the site of calcium entry and thus might not be sufficiently activated before the clearance of calcium from the spine. Likewise, some other unknown modulators highly expressed or absent in CA2 may necessitate higher concentrations of calcium for induction of LTP. Future work will be required to determine whether the increased Ca²⁺ buffering and extrusion of CA2 neurons can be overcome under some behavioral circumstances, or whether the CA2, like other areas of the brain, does not express certain forms of plasticity.

Materials and Methods

Slice Preparation and Electrophysiology.

Whole-cell voltage clamp recordings were made from pyramidal neurons in rat hippocampal slices using standard protocols (SI Materials and Methods). Experiments using CA2 neurons were only performed when the CA2 could be visually distinguished from CA1. Because of the larger size of CA2 pyramidal neurons (compared with those from CA1), CA3 pyramidal neurons were also studied to control for differences in cell volume. In some experiments, camstatin was added to the internal solution for a final concentration of 10 μM. For PMCA inhibition, carboxyeosin (CE) was dialyzed into CA2 neurons through the patch pipette in standard internal electrode solution at a final concentration of 10 μM, a level consistent with intracellular block of PMCA (33). CE was stored as a stock solution in DMSO with the DMSO concentrations never exceeding 0.01%. LTP experiments shown represent averages of several neurons ± SEM, with the number indicated in the figure legends. Examples of experiments from individual neurons and synaptic responses are shown in Fig. S7.

Calcium Imaging.

Dual indicator calcium imaging was performed as described (5, 34) using the calcium-sensitive indicator Oregon green BAPTA-1 (OGB1; green, 30–100 μM) and the calcium-insensitive indicator Alexa 594 (red, 30 μM) under a custom-built 2-photon microscope (35). Neurons in each hippocampal subfield were patched in whole-cell mode and the fluorescence from both indicators recorded in their dendrites and/or spines. Once loaded, line scans (500 Hz) were conducted across the neuronal compartment of interest and the fluorescence from each indicator was separated by a dichroic mirror at 565 nm, filtered by bandpass filters (center wavelength/bandwidth = 510/70 nm for green and 620/90 nm for red), and detected with separated photomultipliers. Resting calcium and calcium changes evoked by single action potentials were calculated from fluorescence measurements using previously published methods (3, 4). Because CE is a fluorophore with a peak emission spectra (542 nM) similar to Oregon Green BAPTA-1 (523 nM), we used the calcium-sensitive indicator X-Rhod 5F (Molecular Probes/Invitrogen) for calcium imaging experiments involving CE. In these experiments, the fluorescence emitted from CE was used as the calcium-insensitive signal.