Paradoxical Ca²⁺ rises induced by low external Ca²⁺ in rat hippocampal neurones

RESULTS

Low-Ca²⁺ media activate primed hippocampal neurones

As shown in Fig. 1A, the metabotropic glutamate receptor agonist t-ACPD elicited distinct responses in neurones and astroglial cells when applied at low concentrations (5-10 μm) to hippocampal slices perfused with standard ACSF. While astrocytes displayed a prompt, often oscillatory, elevation in [Ca²⁺]_i, the majority of neurones (> 80 %) failed to respond, and their [Ca²⁺]_i, at least at the level of the soma, did not change significantly (Fig. 1Aa and b). Surprisingly, however, a high number of neurones (> 90 %, see below for statistics) displayed a large [Ca²⁺]_i rise upon switching to a nominally Ca²⁺-free ACSF (Fig. 1Ac and d). This latter response was neurone-specific since the [Ca²⁺]_i of astrocytes returned slowly to resting level, as expected, upon decrease in the [Ca²⁺]_o (Fig. 1B and C, showing single and average traces, respectively). Lowering the [Ca²⁺]_o in the absence of t-ACPD resulted in either no response or a marginal [Ca²⁺]_i rise in neurones (Fig. 2A and B dashed traces). This distinct response of neurones to [Ca²⁺]_o changes was observed with protocols employing both more drastic and milder conditions for lowering the [Ca²⁺]_o (Fig. 2A and B, continuous traces). In particular, in Ca²⁺-free ACSF, containing EGTA (1 mm), the response was much more transient (Fig. 2A) than in ACSF containing a low, but intermediate [Ca²⁺]_o (0.5 mm) (Fig. 2B). The percentage of responding neurones increased almost linearly with decreasing [Ca²⁺]_o, whereas the amplitude of the response displayed a bell-shaped curve: increasing when the [Ca²⁺]_o was decreased from 2 to 0.5-0.25 mm, and decreasing at lower [Ca²⁺]_o values (Fig. 2D). Unlike the experiments presented above, when higher doses of t-ACPD (20-100 μm) were used, the vast majority of neurones underwent a substantial [Ca²⁺]_i increase in standard ACSF and, upon switching to a Ca²⁺-free medium, the [Ca²⁺]_i remained partially elevated (data not shown). Under those conditions, the neuronal response to [Ca²⁺]_o lowering was not easily distinguishable from that to t-ACPD and was thus not further analysed.

In subsequent experiments, we analysed additional parameters that characterise this paradoxical [Ca²⁺]_i response in neurones. First, a reduction in the [Ca²⁺]_o was specifically required. In fact, in the complete absence of extracellular Mg²⁺, but in the presence of a constant [Ca²⁺]_o, only minor [Ca²⁺]_i rises were observed in stimulated neurones. Experiments were carried out with both the standard Ca²⁺/Mg²⁺ concentration ratio of ACSF (2:1) and the opposite ratio (Fig. 2C), in order to obtain the same amount of divalent cation reduction that occurred in the experiments shown in Fig. 1. Notably, large paradoxical [Ca²⁺]_i rises took place in the same neurones upon lowering the [Ca²⁺]_o from 1 mm to Ca²⁺-free (Fig. 2C). In experiments carried out as described in Fig. 1, but in the presence of 3 mm extracellular MgCl₂ (to maintain constant the total concentration of divalent cations in the bathing medium), a [Ca²⁺]_i response in neurones, albeit reduced by 32 ± 5 % (n = 5), was still observed. Raising [Mg²⁺]_o to 5 mm caused ∼70 % inhibition of the [Ca²⁺]_i response, which was fully abrogated by 10 mm[Mg²⁺]_o (data not shown). Second, the neuronal Ca²⁺ response did not specifically require the stimulation of a metabotropic glutamate receptor (mGluR), since it is also evoked by stimulation of a muscarinic receptor. As shown in Fig. 3A, both carbachol (CCh, 10 μm; continuous trace) and pilocarpine (10 μm; dashed trace) mimicked the t-ACPD effect, and this response was blocked by a pre-incubation with pirenzepine (10 μm; grey trace). Pilocarpine is a non-selective muscarinic agonist, while pirenzepine is a selective antagonist of M₁ muscarinic receptors (M₁-AChRs). Notably, CCh at this concentration (in standard ACSF) did not affect the [Ca²⁺]_i of the majority of CA1 pyramidal cells, whereas at a higher concentration (60 μm), it elicited substantial [Ca²⁺]_i rises (data not shown). Henceforth, unless otherwise stated, both t-ACPD and CCh were used at 10 μm. Third, the neuronal response to low [Ca²⁺]_o induced by t-ACPD (Fig. 3B, dashed trace) was mimicked by 5 μm DHPG, a selective agonist of group I mGluRs (Fig. 3B, continuous trace), but not by 50 μm APDC, a selective agonist of group II mGluRs (Fig. 3B, grey trace). Fourth, this phenomenon was observed only in specific brain regions (Table 1). In fact, upon lowering the [Ca²⁺]_o to 0.25 mm - in the presence of t-ACPD or CCh - neuronal [Ca²⁺]_i responses similar to those recorded in CA1 hippocampal pyramidal neurones were barely detectable in the dentate gyrus (DG). The CA3 area was also at variance with the CA1 region: when stimulated with t-ACPD or CCh in standard ACSF, the majority of pyramidal neurones (83 ± 10 %, n = 7) showed [Ca²⁺]_i rises, often oscillatory, which were maintained upon lowering the [Ca²⁺]_o; stimulation with a lower concentration of t-ACPD (1-5 μm) mimicked better the behaviour of CA1 pyramidal neurones. In pyramidal neurones from the neocortex, a [Ca²⁺]_i response was also observed in the presence of t-ACPD or CCh, albeit with a much lower frequency and amplitude when compared to the hippocampus. Cerebellar granule cells, although sensitive to high K⁺ stimulation, failed to respond to t-ACPD or CCh, in either standard or low-Ca²⁺ media.

Components of the neuronal [Ca²⁺]_i response

Given the requirement for the stimulation of metabotropic receptors and the fact that the driving force for Ca²⁺ entry is significantly reduced upon lowering of [Ca²⁺]_o, release of Ca²⁺ from intracellular Ca²⁺ stores might be expected to play a major, if not a dominant, role in the total response. To investigate this issue, the intracellular Ca²⁺ stores were initially emptied before challenging hippocampal slices with the protocol described in Fig. 1. Slices were pre-incubated in Ca²⁺-free ACSF (containing 0.2 mm EGTA) for 30 min at 37 °C, in the presence of cyclopiazonic acid (CPA; 50 μm), an inhibitor of sarco(endo)plasmic reticulum Ca²⁺-ATPases (SERCAs). Subsequent application of a mixture of InsP₃ -generating agonists (CCh 60 μm, t-ACPD 50 μm and ATP 100 μm) failed to induce any [Ca²⁺]_i rise, in either astrocytes or neurones, demonstrating that intracellular Ca²⁺ stores were effectively depleted by this protocol (Fig. 4A). Furthermore, switching to a Ca²⁺-containing ACSF, in the continuous presence of CPA, caused a marked increase in the [Ca²⁺]_i of astrocytes (Fig. 4A, dashed trace) and a much lower, but detectable, increase in neurones (Fig. 4A, continuous trace), indicating that capacitative Ca²⁺ entry was fully activated (Pizzo et al. 2001). Lowering the [Ca²⁺]_o at the end of this protocol, resulted in a marked increase of [Ca²⁺]_i in neurones and a prompt [Ca²⁺]_i decrease in neighbouring astrocytes (Fig. 4A). From the quantitative analysis shown in Fig. 4B, the CPA treatment reduced the neuronal [Ca²⁺]_i rise by 30 ± 10 % (n = 5), which indicates that intracellular Ca²⁺ stores play only a minor role in the whole response. Pre-treatment with ryanodine (20 μm) was largely ineffective (Fig. 4B), suggesting that if ryanodine-sensitive stores were involved they contributed only modestly to neuronal [Ca²⁺]_i rises. It is worth noting that ryanodine was delivered for 15 min before challenging the slice with the standard protocol of Fig. 1. Within that period, the slice was also stimulated briefly with KCl (40 mm), which serves a double purpose: it allows the differential identification of neurones and astrocytes while, at the same time, permiting ryanodine binding to intracellular Ca²⁺ channels (opened by [Ca²⁺]_i rises). The effectiveness of ryanodine was checked in parallel experiments where the transient [Ca²⁺]_i peak caused by DHPG (20 μm) plus pilocarpine (20 μm) was blunted both in neurones and in astrocytes (Fig. 4C).

Altogether, these experiments indicate that the major part of the [Ca²⁺]_i response activated by [Ca²⁺]_o reduction depends on Ca²⁺ influx from the extracellular medium. As shown in Fig. 2, this paradoxical [Ca²⁺]_i increase, albeit much more transient, was observed even when the medium contained no Ca²⁺ and was supplemented with 1 mm EGTA, i.e. when [Ca²⁺]_o was lower than [Ca²⁺]_i. The simplest explanation is that, in the brain slice, the diffusion of the medium is relatively slow and a sudden reduction of [Ca²⁺]_o in the perfusion fluid results in a relatively slower, but progressive, reduction of the [Ca²⁺]_o close to the neurones. Note that, in the same time window, [Ca²⁺]_i rises induced in astrocytes slowly decayed upon external Ca²⁺ removal (see Fig. 1B). In other words, when a Ca²⁺-free EGTA medium is perfused, [Ca²⁺]_o is reduced, but in the proximity of the neurones it remains sufficiently elevated for a few minutes to allow the influx of Ca²⁺ through plasma membrane channels.

The nature of the Ca²⁺ channels involved in this neuronal [Ca²⁺]_i increase was next investigated. Perfusion with NBQX (30 μm) and d-AP5 (50 μm), to block activation of Ca²⁺-permeable ionotropic glutamate receptors (AMPA/ kainate and NMDA receptors) failed to inhibit the [Ca²⁺]_i rise induced by switching to a Ca²⁺-free ACSF in the presence of t-ACPD (not shown). Conversely, 30 min pre-incubation with verapamil (20 μm, not shown) or nimodipine (10 μm), two classical inhibitors of L-type VOCCs, almost completely abolished the [Ca²⁺]_i rise induced by [Ca²⁺]_o lowering (0.5 mm) in the presence of t-ACPD (Fig. 5A, continuous trace). A partial block was observed upon pre-treatment with ω-conotoxin-GIVA (0.3 μm, a selective P-type Ca²⁺ channel inhibitor (not shown), and ω-conotoxin-MVIIC (3 μm, a broad spectrum inhibitor of N-, P- and Q-type Ca²⁺ channels (Fig. 5A, grey trace). The role of VOCCs was further confirmed by the observation that Ni²⁺ (0.5 mm), when added acutely, inhibited the neuronal [Ca²⁺]_i rise induced by lowering the [Ca²⁺]_o to 0.25 mm (Fig. 5B). When different cations were tested at 0.5 mm with the protocol described in Fig. 5B, the time course of the decay of the neuronal [Ca²⁺]_i rise was taken as a rough indication of their order of potency (Ca²⁺≈ Cd²⁺ > Ni²⁺ > Zn²⁺≈ Co²⁺ >> Mg²⁺). At that concentration, Mg²⁺ was practically ineffective, and full inhibition was reached only at 10 mm[Mg²⁺]_o (not shown). Surprisingly, Ca²⁺ at 0.5 mm was as effective as Cd²⁺, suggesting that recovery of the [Ca²⁺]_o is sufficient to switch off the response (Fig. 5C). This latter finding may suggest the involvement in the paradoxical [Ca²⁺]_i rise of Ca²⁺ receptors, which are known to be expressed in the CNS (Ruat et al. 1995). The classical Ca²⁺ receptor (CaR) is activated by increases in the [Ca²⁺]_o and its response is mimicked by micromolar concentrations (50-300 μm) of Gd³⁺, La³⁺ and polyvalent cations, such as spermine (Brown & MacLeod, 2001). However, none of the trivalent cations, when tested at the same concentration (0.5 mm) employed in the protocol of Fig. 5B, were able to acutely block the [Ca²⁺]_i increase (not shown), nor was spermine (up to 0.5 mm), which, however, at higher concentrations (1 mm) partially reduced the neuronal [Ca²⁺]_i rise (Fig. 5D).

Mechanism of coupling [Ca²⁺]_o changes to neuronal [Ca²⁺]_i rises

It has been demonstrated that lowering the [Ca²⁺]_o depolarises cultured hippocampal neurones by activating voltage-operated Na⁺ channels and non-selective cation channels (NSCs) (Xiong et al. 1997). Pre-treatment with TTX for 3 min (0.5 μm), a selective blocker of voltage-gated Na⁺ channels, fully abrogated the neuronal response to reduced [Ca²⁺]_o. The inhibition was observed in neurones stimulated with either t-ACPD (Fig. 5E, continuous trace) or CCh (data not shown). In contrast, classical inhibitors of NSCs such as econazole (10 μm) and SKF96365 (30 μm) (Fasolato et al. 1994) did not affect the neuronal response (data not shown).

Finally, we tested the possible involvement of ATP-sensitive cation channels. In fact, ATP could be released during the stimulation period from both neurones and astrocytes (Fields & Stevens, 2000; Verderio & Matteoli, 2001). Moreover, at a fixed concentration of total ATP, lowering the extracellular concentration of divalent cations decreases the amount of ATP complexed with Mg²⁺ and Ca²⁺ while increasing that of ATP^4-, the agonist of ionotropic P2X receptors (Fasolato et al. 1990). Therefore, a rise in ATP^4- is expected immediately upon switching to low-Ca²⁺ media. By activation of ionotropic receptors, the latter compound may depolarise the cell membrane sufficiently to activate VOCCs. Notably, the neuronal response was not reproduced by simply lowering the [Mg²⁺]_o, and was still observable in the presence of a compensating concentration of MgCl₂ (to avoid the increase in ATP^4-). Furthermore, challenging the slice with benzoylbenzoyl-ATP (100 μm), a specific agonist of P2X ionotropic receptors (El-Sherif et al. 2001; Khakh, 2001), failed to mimic the neuronal [Ca²⁺]_i rise (data not shown).

Quite surprisingly, lowering the [Ca²⁺]_o was not sufficient, by itself, to elicit a neuronal response, but required the presence of t-ACPD or CCh. Furthermore, the experiments with pirenzepine (Fig. 3A), DHPG and APDC (Fig. 3B) indicate that metabotropic receptors, coupled to PLC but not to adenylate cyclase, are involved. However, we were unable to test a direct involvement of PLC since perfusion with U73122 (1-20 μm), a broad spectrum PLC blocker (Chuang et al. 2001), while being ineffective on the neuronal response induced by [Ca²⁺]_o lowering, also failed in blocking the [Ca²⁺]_i rise induced by t-ACPD in astrocytes. Pre-incubation with U73122 (1 μm) for 5–20 min resulted in toxic effects (i.e. massive increases in [Ca²⁺]_i of neuronal cells). Recent evidence suggests that, in hippocampal neurones, different metabotropic receptors couple to non-receptor tyrosine kinase (non-RTK) via a G-protein-independent pathway (Heuss et al. 1999; Heuss & Gerber, 2000). We have therefore tested the effect of PP2, a selective inhibitor of the Src-family of RTK (Hanke et al. 1996). The neuronal response to low [Ca²⁺]_o was drastically reduced by 20 min pre-incubation with PP2 (20 μm) (Fig. 5F, continuous trace). Notably, in the same slice, PP2 did not affect the astrocyte response to t-ACPD (Fig. 5F, grey trace). The signalling cascade leading to Src activation may be triggered also by means of PKC via a G-protein-dependent pathway (Lu et al. 1999; Benquet et al. 2002). However, 20 min pre-incubation with the PKC inhibitor Ro 31–8220 (10 μm) failed to block the neuronal [Ca²⁺]_i rise, whereas a higher concentration could not be tested because of its ability to also block voltage-operated Na⁺ channels (Lingameneni et al. 2000).

To investigate the mechanism by which metabotropic receptor agonists increase neuronal sensitivity to low-Ca²⁺ media, we analysed the time window within which the two stimuli must be delivered to be effectively coupled. Experiments were thus carried out with a protocol different from that described in Fig. 1. Slices were perfused with standard ACSF and then challenged with ACSF containing a low [Ca²⁺]_o (0.25 mm). t-ACPD was delivered for 3 min, at the same time at which the bath was switched to the low [Ca²⁺]_o (zero time, not shown) or, at later times (Fig. 6A, 9 min). The amplitude of the neuronal response, when plotted as a function of the application time of t-ACPD, increased within the first 5 min and slowly declined by delaying the stimulation with t-ACPD (Fig. 6B). After 40 min in low-Ca²⁺ medium, t-ACPD was much less effective in inducing a neuronal [Ca²⁺]_i rise (Fig. 6B). We next investigated the effect of the stimulation period with t-ACPD in Ca²⁺-containing ACSF. As shown in Fig. 6C, and at variance with the protocol described in Fig. 1, t-ACPD was delivered for only 3 min in standard ACSF, before challenging the slice with low [Ca²⁺]_o (0.25 mm) in the absence of t-ACPD. This stimulation period was not sufficient to elicit a significant response when [Ca²⁺]_o was subsequently lowered, and t-ACPD application was further required. Note, however, that 3 min exposure to t-ACPD in low-Ca²⁺ medium was sufficient to elicit a neuronal [Ca²⁺]_i rise (compare Fig 6C and A). Conversely, when the stimulation period with t-ACPD in standard ACSF was prolonged from 3 to 12 min, as shown in Fig. 6D, the neuronal sensitivity to a reduction in [Ca²⁺]_o was enhanced. In fact, a [Ca²⁺]_i response was evident a few seconds after switching to the low [Ca²⁺]_o even in the absence of t-ACPD and only marginally increased upon t-ACPD application. Accumulation of an intracellular messenger, or long-lasting modification of the effector/target in the absence of continuous receptor(s) occupancy might occur under those conditions.

Electrical activity of hippocampal pyramidal neurones in situ

It was previously reported that lowering [Ca²⁺]_o depolarises hippocampal neurones both in situ and in culture (Xiong et al. 1997; Su et al. 2001). We confirmed this observation in CA1 pyramidal neurones from hippocampal slices using the patch-clamp technique in the current-clamp configuration (see Methods). Slice perfusion with a low-Ca²⁺-containing ACSF (0.25 mm) triggered a small depolarisation (4.0 ± 0.4 mV) and the discharge of action potentials (1.2 ± 0.7 spikes s⁻¹; n = 3). Subsequent exposure to the standard [Ca²⁺]_o (2 mm) allowed recovery of the resting membrane potential (60.7 ± 7.5 mV; n = 3) and switched off spike generation (Fig. 7A). In the presence of DHPG (10 μm), the rate of action potential discharge, triggered by a second challenge with the low-Ca²⁺-containing ACSF, was drastically increased (up to 4.2 ± 1 spikes s⁻¹; n = 3), on top of a more depolarised plateau potential (average decrease of 9.8 ± 0.5 mV with respect to 4.0 ± 0.4 mV, respectively in the presence and in the absence of DHPG; n = 3; Fig. 7B).

Paradoxical [Ca²⁺]_i rises in primary cultures of cortical neurones

As shown in Fig. 6B, perfusing the slices with an ACSF at low [Ca²⁺]_o, induced paradoxical [Ca²⁺]_i rises in hippocampal neurones when challenged with t-ACPD even at longer times (10-20 min). Under those conditions, an almost complete equilibration of the extracellular space with the bathing medium should have been reached, as also indicated by the effectiveness of different drugs employed under perfusion. Experiments, similar to those reported above, were carried out in primary cultures of neurones obtained from neonatal rat cortices. In the presence of t-ACPD, immediately upon switching to a low [Ca²⁺]_o (0.2 mm), neuronal cells showed paradoxical [Ca²⁺]_i rises of amplitudes significantly larger than those recorded in its absence (Fig. 8A). In cell cultures, at variance with slices, perfusion with a Ca²⁺-free medium failed to evoke any response, whereas [Ca²⁺]_i rises induced by low [Ca²⁺]_o media were evoked more rapidly than in slices. Notably, this neuronal response was similarly evoked by DHPG, CCh and pilocarpine, and was sensitive to nimodipine (5 μm), as well as to TTX (1 μm) pre-treatment (data not shown). At variance with slices, in neuronal cultures, 20 min pre-incubation with the PLC inhibitor U71322 (3 μm) revealed no major toxicity, but failed to inhibit the neuronal response while significantly reducing the astrocyte response to DHPG (10 μm) (Fig. 8B). As shown in slices, the paradoxical [Ca²⁺]_i rise of neurones was inhibited by pre-incubation with PP2 (10 μm; Fig. 8C, continuous trace), being, however, insensitive to its inactive analogue PP3 (10 μm) (Fig. 8C, dashed trace). On the other hand, 20 min pre-incubation with genistein (20 μm) was without effect on the neuronal response (data not shown). At this dose, genistein has been reported to be rather selective for RTK (compared to non-RTK), being totally ineffective on at least two types of Src-kinase (Meggio et al. 1995).

When the electrical activity was measured in cultured neurones, reduction of the [Ca²⁺]_o from 2 to 0.25 mm depolarised the cells by 17 ± 5 mV from an average resting membrane potential of 70 ± 3 mV (n = 3) and, on top, increased spiking activity (Fig. 8D). The depolarisation, however, was rapidly reversible in spite of the presence of low external Ca²⁺. Challenging the same neurone with t-ACPD in the Ca²⁺-containing medium (2 mm) slightly increased basal activity (Fig. 8D). Switching to a low-Ca²⁺ medium in the continuous presence of t-ACPD induced a long-lasting depolarised plateau of 25 ± 7 mV (n = 3), with intense spiking activity (Fig. 8D). In cultured cortical neurones, agonist-dependent paradoxical [Ca²⁺]_i rises and long-lasting depolarised plateaus were observed at rather low frequency, on average, 20 and 25 %, respectively, similar to that found in cortical slices (see Table 1). In fact, of the other neurones analysed, three of them showed, when switched to a low Ca²⁺ medium, a depolarisation that was higher in the presence of t-ACPD, but without spike generation. In contrast, in six additional neurones, t-ACPD did not increase, or even reduced, the electrical activity in both the standard and the low-Ca²⁺-containing medium.

Discussion

Our results show that in the presence of metabotropic receptor agonists coupled to PLC activation, hippocampal neurones from CA1-CA3 regions possess the unique property of responding to low [Ca²⁺]_o by raising their [Ca²⁺]_i. The novelty of this phenomenon is twofold. First, in spite of the observed increase in neuronal firing, lowering the [Ca²⁺]_o does not lead to widespread neuronal [Ca²⁺]_i rises. This finding indicates that the reduced driving force, together with the cell capability to tightly control the [Ca²⁺]_i, limits Ca²⁺ entry. Secondly, and totally new and surprising, a decrease in the external Ca²⁺ elicits prompt Ca²⁺ rises only in selective brain regions upon concomitant activation of metabotropic receptors. Notably, the phenomenon observed in slices can be reproduced in cultured cortical neurones indicating that it is not due to a poor control of the composition of the extracellular environment but relies on intrinsic neuronal properties.

As far as the process involved in this phenomenon is concerned, we demonstrate that intracellular Ca²⁺ stores very modestly contribute to the response, whereas both Na⁺ and Ca²⁺ voltage-operated channels (VOCs) are primarily involved. Indeed, selective blockers of these channels completely, or partially, abolished the rise in [Ca²⁺]_i. Treatment with a SERCA pump inhibitor (CPA) reduced the [Ca²⁺]_i rise by 30 %, whereas ryanodine did not completely mimic the CPA effect, suggesting that both ryanodine- and InsP₃-sensitive stores are involved. Notably, selective antagonists of VOCCs fully abrogated the [Ca²⁺]_i response, indicating that Ca²⁺ release from intracellular Ca²⁺ stores is triggered by Ca²⁺ influx.

It is well known that lowering the [Ca²⁺]_o reduces the screening effect on surface charge and shifts the activation curve of VOCs to the left (Frankenhauser & Hodgkin, 1957; Hille 2001). A contribution of charge shielding to the observed phenomenon was also apparent from the capability of divalent cations to reverse, acutely, the neuronal [Ca²⁺]_i rise. The reduction of the neuronal response, occurring when the [Ca²⁺]_o was lowered in the presence of a high concentration of extracellular Mg²⁺ (10 mm), is compatible with surface charge effects (Piccolino et al. 1996). Yet, loss of shielding effects is probably necessary, but not sufficient, to explain the [Ca²⁺]_i rise that occurs in neurones upon lowering [Ca²⁺]_o. In fact, the phenomenon was observed only in neurones stimulated with agonists of group I mGluRs or M₁-AChRs, but not in non-stimulated cells.

Large neuronal [Ca²⁺]_i rises could not be evoked by lowering the extracellular Mg²⁺ concentration, but were clearly observed at different levels of reduced [Ca²⁺]_o. The bell-shaped dose-response curve for different [Ca²⁺]_o indicates that external Ca²⁺ may bring about opposite effects on Ca²⁺ influx: decreasing the [Ca²⁺]_o is required to trigger the response but, at the same time, reduces the driving force for Ca²⁺ entry across VOCCs sensitised, directly or indirectly, by activation of G-protein-coupled receptors. Our data confirm that the response is specifically set for sensing changes in the [Ca²⁺]_o. Regarding the mechanism by which activated neurones sense low [Ca²⁺]_o, it is interesting to note that Ca²⁺ itself was among the most potent inhibitors of the [Ca²⁺]_i rise. While this might simply be explained on the basis of shielding effects, we cannot exclude that other factors are likely to be at work. Involvement of a classical CaR (Brown & MacLeod, 2001) is unlikely, however, since known agonists - such as La³⁺ and Gd³⁺ (but not spermine) - failed to mimic the Ca²⁺ inhibitory effect. The presence of a Ca²⁺-binding receptor different from the classical CaR in CA1 pyramidal neurones was suggested by Su et al. (2001). These authors demonstrated the capability of CA1 pyramidal neurones to increase intrinsic bursting upon reduction of the [Ca²⁺]_o, and suggested that the channel responsible for the persistent, TTX-sensitive Na⁺ current is the Ca²⁺ sensor.

The most striking feature of this paradoxical [Ca²⁺]_i rise induced by low-Ca²⁺ media is the absolute requirement for contemporary stimulation with metabotropic receptor agonists. In hippocampal slices, lowering the [Ca²⁺]_o while delaying the stimulation with t-ACPD reduced the neuronal [Ca²⁺]_i rise, suggesting that stronger responses can be obtained only through the concomitant reduction of [Ca²⁺]_o and mGluR stimulation (see Fig. 6). These findings are consistent with the electrophysiological data showing that only the combination of the two stimuli results in a prolonged and intense bursting activity.

In brain slices, at the level of the glial-neuronal network, the tight cell packaging is probably responsible for a poor control of the composition of the extracellular medium and therefore of the paradox of nimodipine-sensitive [Ca²⁺]_i rises occurring at theoretical nanomolar [Ca²⁺]_o. In addition, Ca²⁺ pumping across the plasma membrane of activated astrocytes might represent a relevant source of Ca²⁺ for the narrow extracellular space surrounding the neurones and, accordingly, delay the equilibration of the [Ca²⁺]_o in the extracellular space with that in the medium. To understand whether this phenomenon is a property of the tissue as a whole or of its single components, we also tested primary cultures of cortical neurones. With a low frequency, similar to that found in cortical slices, cultured neurones show fast, nimodipine- and TTX-sensitive [Ca²⁺]_i rises when switched to a low-Ca²⁺ medium containing t-ACPD or CCh. Notably, the neuronal [Ca²⁺]_i rise was blocked by PP2 at a lower concentration. Under current-clamp conditions, lowering the [Ca²⁺]_o induced spike generation and cell depolarisation, which were increased and prolonged by t-ACPD. Altogether, these findings are compatible with surface charge effects and VOC activation induced by low external Ca²⁺, reinforced by agonists of group I mGluRs or M₁-AChRs.

The phenomenon we described was strictly dependent on the activation of receptors coupled to PLC but, at the same time, was insensitive to the PLC inhibitor U71322. This suggests that less classical pathways are probably involved. In fact, the neuronal response was blocked by PP2 (a selective inhibitor of the Src-family of tyrosine kinases), but was insensitive to PP3 (the inactive analogue). The Src-kinase cascade is activated by metabotropic receptors by means of either PKC- (Lu et al. 1999; Benquet et al. 2002) or G-protein-independent pathways (Heuss et al. 1999; Heuss & Gerber, 2000). Our data favour the latter, given the insensitivity to PLC and PKC inhibitors (U71322 and Ro 31–8220, respectively). It is worth noting that U71322, while being ineffective on the neuronal response, significantly reduced the DHPG-induced [Ca²⁺]_i rise in astrocytes. This finding not only confirms the selectivity of the drug, but also suggests that the neuronal response is probably independent of the activation of surrounding astrocytes.

A hypothetical model of the signalling pathway described above, is depicted in Fig. 9. Metabotropic receptors might couple to the Src-kinase both directly (not shown) or indirectly by means of an adaptor molecule such as a homer- or β-arrestin-like protein. Possible targets of the Src-kinase, in addition to VOCs themselves, are different channels whose opening (or closure) will help depolarisation; external Ca²⁺ reduction will finally allow the activation of VOCs.

As far as the physiological relevance of this phenomenon is concerned, it is conceivable that a reduction of the [Ca²⁺]_o occurs in the microenvironment surrounding the synaptic terminals during high neuronal activity. It has been demonstrated, by different experimental protocols, that intense electrical or chemical stimulation induces profound alterations in the extracellular milieu of specific brain regions. [Ca²⁺]_o as low as 0.7-0.1 mm was measured by ion-selective microelectrodes in cortical layers that show tight cell packaging (and, therefore, small extracellular volumes), such as those from the hippocampus, the motor and the cerebellar cortex (Nicholson et al. 1978; King & Somjen, 1981; Pumain & Heinemann, 1985; Kovacs et al. 2001). Low-Ca²⁺ media increase intrinsic burst-firing of hippocampal pyramidal cells, thus enhancing neuronal excitability and synchronisation (Haas & Jefferys, 1984). Localised [Ca²⁺]_o depletion was estimated from Ca²⁺ currents analysis at the level of calyx-type synapses upon prolonged depolarisation (Borst & Sakmann, 1999; Stanley, 2000). Furthermore, peri-dendritic Ca²⁺ fluctuations may occur during normal activity, because of back-propagating action potentials (Egelman & Montague 1999; during the revision of the present paper, a first measurement of extracellular Ca²⁺ depletion at the CA1 level was elegantly performed by Rusakov & Fine, 2003). All these phenomena might have profound effects on the efficacy of the synapses directly involved. In this context, our results appear relevant under at least two different aspects. First, the assumption that low-Ca²⁺ media work by non-synaptic mechanisms must carefully be considered. Augmentation of Ca²⁺ entry via VOCCs upon switching to low-Ca²⁺ media might contribute to explaining the lower threshold of excitability of hippocampal neurones (Lian et al. 2001). Second, the fact that this phenomenon exclusively occurs in cells activated by agonists of metabotropic receptors suggests that the capability of pyramidal cells to initiate a Ca²⁺ rise in response to low [Ca²⁺]_o is strictly dependent on the type and timing of stimulation. From the functional point of view, the phenomenon reported in this work might be relevant to integrating single spikes into synchronised bursts, which are considered the optimal input/output for information coding and induction of long-term plasticity (Lisman, 1997). The fact that it is not widespread but occurs in specific brain regions and is strictly dependent on the coincidence of two events, i.e. external Ca²⁺ depletion and the presence of low doses of agonist, might be suggestive of a new coding modality. We can hypothesise that glutamate/acetylcholine spillover from active synapses, together with Ca²⁺ depletion, influences the same or neighbouring synapses; a similar role for Ca²⁺ depletion alone has already been proved (von Gersdorff & Borst 2002; Rusakov & Fine, 2003). Depending on receptor localisation and type of stimulation, we can envisage either facilitation by increased release probability at the presynaptic level, or depression by increased external Ca²⁺ depletion at the postsynaptic side.

In addition, [Ca²⁺]_i rises induced by low [Ca²⁺]_o might favour, under pathological conditions, the development of seizure-like activity specifically in hippocampal regions (low-Ca²⁺ model of epilepsy; Haas & Jefferys, 1984; Konnerth et al. 1986; Bikson et al. 1999). It has recently been demonstrated that spontaneous, pharmacologically evoked seizures, increased Src-kinase activity (Sanna et al. 2000). The frequency of epileptiform discharges was reduced by the TK inhibitor PP2 (Sanna et al. 2000). Interestingly, we found that paradoxical [Ca²⁺]_i elevations were inhibited by the same inhibitor, thus suggesting a possible linkage between increased electrical activity and neuronal [Ca²⁺]_i rises induced by low-Ca²⁺ media.

Pathological conditions that elicit strong alterations in the extracellular milieu of the CNS may also be associated with a decrease in the [Ca²⁺]_o. Acute phenomena include focal cerebral ischaemia and prolonged anoxia (Puka-Sundvall et al. 1994; Ohta et al. 2001). In particular, peri-infarct depolarisation, when associated with decreases in the [Ca²⁺]_o, contributes to expanding infarction in focal cerebral ischaemia (Ohta et al. 2001). Moreover, spreading depression, which was associated with seizure, migraine aura and head injury, may also occur as a consequence of extracellular Ca²⁺ fluctuations (Martins-Ferreira et al. 2000). Finally, it has been shown that hypocalcaemia, caused by parathyroidectomy, increases cortical hyperexcitability and the amplitude of somato-sensory evoked potentials (Kanda et al. 1989).

In summary, reductions of the [Ca²⁺]_o that occur under both physiological and pathological conditions at the level of the CNS, or even systemically, influence the activity of selective brain regions in very different ways. The present data may contribute to elucidating the molecular mechanisms involved in processes ranging from synaptic plasticity to brain injury and, potentially, to designing new pharmacological interventions.