J. Palecek: Institute of Physiology, Czech Academy of Sciences, Videnska 1083, 142 20 Praha 4, Czech Republic.
1A quantitative analysis of endogenous calcium homeostasis was performed on 65 motoneurones in slices of the lumbar spinal cord from 2- to 8-day-old mice by simultaneous patch-clamp and microfluorometric calcium measurements.
2Somatic calcium concentrations were monitored with a temporal resolution in the millisecond time domain. Measurements were performed by using a monochromator for excitation and a photomultiplier detection system.
3Somatic calcium signalling was investigated during defined voltage-clamp protocols. Calcium responses were observed for membrane depolarizations positive to −50 mV. A linear relation between depolarization time and free calcium concentrations ([Ca2+]i) indicated that voltage-dependent calcium influx dominated the response.
4Endogenous calcium homeostasis was quantified by using the ‘added buffer’ approach. In the presence of fura-2 and mag-fura-5, calcium transients decayed according to a monoexponential function. Decay-time constants showed a linear dependence on dye concentration and the extrapolated constant in the absence of indicator dye was 371 ± 120 ms (n= 13 cells, 21 °C).
5For moderate elevations (< 1 μm), recovery kinetics of depolarization-induced calcium transients were characterized by a calcium-independent, ‘effective’ extrusion rate γ= 140 ± 47 s−1 (n= 13 cells, 21 °C).
6The endogenous calcium binding ratio for fixed buffers in spinal motoneurones was κB’= 50 ± 17 (n= 13 cells), indicating that less than 2% of cytosolic calcium ions contributed to [Ca2+]i.
7Endogenous binding ratios in spinal motoneurones were small compared to those found in hippocampal or cerebellar Purkinje neurones. From a functional perspective, they provided motoneurones with rapid dynamics of cytosolic [Ca2+]i for a given set of influx, extrusion and uptake mechanisms.
8With respect to pathophysiological conditions, our measurements are in agreement with a model where the selective vulnerability of spinal motoneurones during excitotoxic conditions and motoneurone disease partially results from low endogenous calcium buffering.
Physiological modulation of neuronal activity has been associated with elevations in cytosolic calcium concentrations induced by calcium influx through postsynaptic ligand-activated channels, voltage-dependent calcium channels and calcium release from intracellular stores (e.g. McBurney & Neering, 1987; Blaustein, 1988; Keller et al. 1992; Yuste & Tank, 1996). Voltage-dependent calcium influx occurs through a heterogeneous population of underlying calcium channels, which can be distinguished by their profile of voltage dependence, gating behaviour and subtype-specific pharmacological properties (Hess, 1990; Llinas et al. 1992). For a given influx pathway, the effective elevation in intracellular calcium concentrations depends on several components, including calcium uptake into intracellular stores, extrusion mechanisms and endogenous calcium buffering (McBurney & Neering, 1987; Blaustein, 1988; Baimbridge et al. 1992; Neher, 1995). Endogenous buffering can be quantified by the calcium binding ratio κS, i.e. the ratio between the number of calcium ions bound to endogenous buffers compared to the number of calcium ions that make up the free calcium concentration in the cytosol (Neher & Augustine, 1992; Zhou & Neher, 1993). It can be experimentally determined by the ‘added buffer’ method based on a controlled loading of cells with fluorescent indicator dyes (see Neher, 1995 for review). Such measurements have demonstrated a notable heterogeneity of endogenous buffering in neurones ranging, for example, from κS= 41 for hypoglossal motoneurones (Lips & Keller, 1998) to κS= 2000 for cerebellar Purkinje cells in juvenile rats (Fierro & Llano, 1996).
Voltage-dependent calcium influx into motoneurones is mediated by multiple calcium channel types, which can be distinguished by their specific voltage dependence, modulatory properties and pharmacological profile (Mynlieff & Beam, 1992; Umemiya & Berger, 1995; Westenbroek et al. 1998). Importantly, disruptions of calcium influx and homeostasis have been associated with pathophysiological conditions in motoneurone disease, in particular with a selective vulnerability of motoneurones to glutamate- and calcium-mediated excitotoxic conditions (DePaul et al. 1988; Choi, 1988; Rothstein & Kuncl, 1995; Krieger et al. 1996; Roy et al. 1998). The molecular and cellular factors responsible for the differential vulnerability of motoneurones are not completely understood. It is interesting to note that endogenous calcium buffers such as calbindin-D28k and parvalbumin are expressed at low levels in those motoneurone populations that are particularly affected. This points to the possibility that low expression of calcium-binding proteins is one risk factor magnifying the impact of pathophysiological signal cascades (Baimbridge et al. 1992; Alexianu et al. 1994; Reiner et al. 1995). Accordingly, the loading of vulnerable cells with exogenous buffers and/or the enhanced expression of intrinsic buffers have been suggested as neuroprotective strategies. Indeed, increased concentrations of calcium-binding proteins were able to protect particularly vulnerable motoneurones against neurodegeneration and excitotoxic damage (Alexianu et al. 1994; Ho et al. 1996; Roy et al. 1998).
In the present investigation, we conducted a quantitative analysis of endogenous calcium homeostasis in motoneurones of the mouse spinal cord by performing simultaneous patch-clamp and microfluorometric measurements. The added buffer method (Neher, 1995) permitted us to quantify the endogenous parameters of calcium homeostasis that shape calcium signalling under dye-free conditions. Our results suggest that endogenous calcium buffering capacities in spinal motoneurones are small compared to those found in other neuronal populations in different parts of the central nervous system (Neher, 1995; Fierro & Llano, 1996; Helmchen et al. 1996). In a physiological context, low endogenous buffering provides spinal motoneurones with a rapid, energy-conserving calcium homeostasis at low energy cost. Our results also supported the hypothesis that under pathophysiological conditions, a selective vulnerability of motoneurones results partly from low levels of endogenous calcium buffers. Some of these results have been published in preliminary form (Palecek et al. 1998).
Preparation of spinal motoneurones
In vitro spinal cord preparations were obtained from 2- to 8-day-old mice as described previously (Palecek et al. 1999). Experiments were carried out in accordance with the guidelines of the Ethics Committee of the Medical Faculty at the University of Göttingen, Germany. Animals were decapitated with a small guillotine, and spinal cords were removed and subsequently cooled to 4°C. Transverse slices with a thickness of 200 μm from lumbar segments L4-L6 of the mouse spinal cord were prepared according to previously described methods (Keller et al. 1991). An oxygen supply was achieved by continuous carbogen (95 % O2, 5 % CO2) bubbling of saline, with Tygon tubing connecting saline reservoirs with the recording chamber and saline perfusion rates up to 10 ml min−1. Slices were maintained at 33°C for 30 min and then at room temperature in bicarbonate-buffered saline (mm:118 NaCl, 3 KCl, 1 MgCl2, 25 NaHCO3, 1 NaH2PO4, 1.5 CaCl2, 20 glucose) at pH 7.4. Prior to recordings, slices were incubated for at least 1 h to allow recovery. If not indicated otherwise, 0.5 μm TTX and 40 mm TEA were added to perfusion solutions, and 10 μm CNQX, 50 μm D-APV, 10 μm bicuculline and 10 μm strychnine were added to bath solutions to block spontaneous synaptic activity. For whole-cell recordings (Edwards et al. 1989), slices were placed in the recording chamber under a Zeiss upright microscope and continuously superfused with the carbogen-bubbled solution. Experiments were carried out at room temperature (21 ± 1°C), unless stated otherwise.
In patch-clamp experiments, suitable motoneurones were selected by their intact overall shape, their ability to fire action potentials in drug-free solution and the occurrence of spontaneous synaptic activity. Slices displaying mechanically or metabolically disturbed motoneurones, which could be detected visually by a gradual degeneration of their overall shape and electrically by decreasing input resistances, were not further investigated. The intracellular pipette solution contained (mm): 140 CsCl (or alternatively, 140 KCl), 20 TEA, 10 Hepes, 2 MgCl2, 4 MgATP, 0.4 NaGTP (adjusted to pH 7.3 with KOH or CsOH). Fura-2 and mag-fura-5 were purchased from Molecular Probes (Eugene, OR, USA) and applied in concentrations ranging from 40 μm to 1 mm in the pipette solution. Patch pipettes as well as stimulation pipettes were pulled from borosilicate glass tubing (Hilgenberg, Malsfeld, Germany). When filled with intracellular solution, they had resistances of 1.5-2.5 MΩ. Voltage-clamp recordings were performed with a patch-clamp amplifier (EPC-9, HEKA Elektronik, Lambrecht, Germany), employing optimal series resistance compensation as previously described (Titz & Keller, 1997; Weigand & Keller, 1998). The series resistance of spinal motoneurones before compensation was typically 8-15 MΩ. Cells with series resistances higher than 15 MΩ were not included in the analysis. Input resistances of spinal motoneurones were 184 ± 36 MΩ (mean ±s.e.m.; n= 40 cells) under our experimental conditions. Rs and Cm were monitored throughout the experiment. Series resistance compensation was set to 50-60 %. No compensation was made for liquid junction potentials. Unless stated otherwise, whole-cell currents were recorded with sampling frequencies of 100 Hz-5 kHz and filtered (3-pole Bessel filter, 10 kHz; 4-pole Bessel filter, 2.9 kHz) before analysis.
Microfluorometric calcium measurements
Electrophysiological and microfluorometric signals were recorded simultaneously with EPC-9 hardware and software (PulseFit, HEKA, Germany). Intracellular calcium concentrations were measured using a computer-operated monochromator (TILL Photonics, Munich, Germany), which was controlled by EPC-9 fura extension software. Fluorescence signals were collected and spatially integrated over the soma of the cell (Lips & Keller, 1998). They were detected by a photomultiplier mounted on a viewfinder (TILL Photonics) that defined an adjustable, rectangular region of interest over which the fluorescence intensity was integrated. Rapid recordings of membrane current, F360 and F390 were obtained with the Pulse software (EPC-9) at a sampling rate of 10 kHz. For each recording interval lasting up to 70 s, fluorescence signals F390 and F360 were recorded at short intervals of 25 ms each, then F390 was collected for the rest of the interval to monitor calcium changes during voltage stimulation protocols (see, for example, Fig. 3A). Calculations of intracellular calcium concentrations and further analysis were performed off-line by using PulseFit (HEKA) and IGOR (Wavemetrics, OR, USA). Calibration constants for fura-2 were determined according to Grynkiewicz et al. (1985) by patch-clamping cells with the following intracellular solutions (mm): Rmin: 140 KCl, 10 Hepes, 2 MgCl2, 4 MgATP, 0.4 NaGTP, 10 BAPTA (adjusted to pH 7.3 with KOH); Rmedium: 140 KCl, 10 Hepes, 2 MgCl2, 4 MgATP, 0.4 NaGTP, 9.9 BAPTA, 6.6 CaCl2 yielding a final concentration of 450 nm[Ca2+]i; and Rmax: 140 KCl, 10 Hepes, 2 MgCl2, 4 MgATP, 0.4 NaGTP, 10 CaCl2. In this case, Rmin, Rmedium and Rmax denote the calcium-dependent F360/F390 fluorescence ratio R in nominally free, 450 nm and 10 mm intracellular free calcium concentrations, respectively. Under our experimental conditions the dissociation constant for fura-2 (Kd) was determined by using the equation:
Calibration constants Kd, Rmax and Rmin were adjusted after several days of experiments to account for small fluorescence changes in the microfluorometric system. Typical values for Kd, Rmin and Rmax were 240 nm, 0.26 and 1.54, respectively. Fura-2 concentrations were checked by spectroscopic analysis, which demonstrated that nominal fura-2 concentrations reflected real concentrations with a deviation around 10 %. Fluorescence intensity is given in bead units (BU) (catalogue no. 18340, Polysciences Inc., PA, USA).
Quantitative model of calcium homeostasis in spinal motoneurones
Calcium homeostasis was quantified by using the ‘added buffer’ method (Neher & Augustine, 1992). In this case, calcium is assumed to bind to a fast, endogenous buffer S, and an incremental binding ratio (κs) is defined by:
and the calcium binding ratio κB′ of an indicator dye is given by:
(Neher, 1995) and [Ca2+]T denotes the average total calcium concentration. In the same model, the inverse amplitudes of calcium transients are determined by the equation:
where qCa denotes the associated, calcium-mediated charge influx per cytosolic volume and F represents the Faraday constant (Neher, 1995). To experimentally determine the decay-time constants, a single exponential was fitted from peak to baseline using either PulseFit or IGOR software. For determinations of 1/A, calcium recordings were low-pass filtered with 50 Hz to reduce noise (Lips & Keller, 1998).
To illustrate the importance of calcium homeostasis for energy consumption in rhythmically active motoneurones, it is useful to consider a cell where rhythmic action potential discharges (bursts) are paralleled by voltage-dependent calcium elevations (Lev-Tov & O'Donovan, 1995; Ladewig & Keller, 1998). By using the above model, the calcium-mediated charge influx per burst is given by:
To restore resting conditions after a burst, a corresponding amount of calcium has to be transported against electrochemical gradients of cellular membranes. For a burst frequency f, the necessary charge transport per time interval is:
By assuming that electrochemical gradients are constant in the time interval of interest, the corresponding energy consumption per time interval PCa is given by (F= constant):
It is important to note that according to this concept, calcium-related energy consumption per time interval is directly proportional to the endogenous buffering capacity of the cell provided that all other parameters (electrochemical gradients, amplitude, frequency) are constant. Correspondingly, low endogenous buffering capacities provide rhythmically active motoneurones with rapid recovery of individual calcium transients at relatively low energy cost (see also Lips & Keller, 1998).
Calcium diffusion and microdomains in patch-clamped motoneurones
To evaluate the impact of pipette perfusion of the cytosol on the spatiotemporal profile of calcium signals in motoneurones, it is useful to consider an elementary model where the distance r that a calcium ion diffuses in a time interval t is given by (Neher, 1986):
Deff denotes the ‘effective’ calcium diffusion constant. In a patch-clamped neurone, Deff is approximated by (Zhou & Neher, 1993):
where DCa= 220 μm2 s−1 represents the diffusion constant of calcium in the absence of buffers and Dfura-2= 170 μm2 s−1 is the diffusion constant of fura-2. κB‘, κmob and κS denote the calcium buffering capacities of fura-2, mobile and fixed cellular buffers, respectively (Zhou & Neher, 1993; Klingauf & Neher, 1997). During whole-cell recordings, the contribution of mobile buffers to the overall buffering capacity of motoneurones is gradually reduced, as mobile buffers are washed out by pipette perfusion of the cytosol. In earlier reports (Zhou & Neher, 1993; Klingauf & Neher, 1997) buffering capacities of mobile buffers κmob, most likely represented by diffusable organic anions, have been estimated as 25 % of the κS characteristic for fixed or hardly mobile buffers. According to this approximation, values of κS= 50 characteristic for spinal motoneurones provided an estimate κB′ >> κS > κmob for fura-2 concentrations above 300 μm (e.g. κB′= 440 for 400 μm fura-2). Accordingly, Deff approximated Dfura-2 in our whole-cell recordings and calcium diffusion was essentially determined by the mobility of fura-2.
In quantitative terms, the above equation provides an estimated Deff= 153 μm2 s−1 for calcium diffusion in spinal motoneurones in the presence of 400 μm fura-2. During our whole-cell recordings, this predicted an equilibration time of < 3.3 ms for calcium transients in dendrites (radius < 1 μm), 0.33 s for somatic transients (somatic radius = 10 μm) and 2.9 s for the diffusion time between the soma and a proximal dendritic compartment at a distance of 30 μm. Interestingly, this suggested that the soma of motoneurones could be considered as one compartment with respect to calcium homeostasis, at least for rapid events that occur in the subsecond time domain. For example, during 500 ms stimulation intervals utilized to characterize calcium homeostasis (Fig. 7), somatic calcium transients could diffuse a distance of 12 μm along dendrites. This indicates that somatic calcium amplitudes during stimulation pulses mainly reflect calcium influx through somatic channels. Even for longer time intervals, the previous observation of voltage-dependent calcium channels both in somatic and dendritic membranes of motoneurones (see for example Ladewig & Keller, 1998) suggested that depolarization-induced calcium elevations were comparable in both compartments. In this case, calcium diffusion from the soma into dendrites was largely compensated by inverse fluxes resulting from simultaneous rises in dendritic [Ca2+]i. For our present purposes, these arguments provide reasonable support for the approximation of somatic calcium homeostasis in spinal motoneurones by a one-compartment model (Neher, 1995). For calcium signals lasting several seconds, however, the ‘somatic’ model could include signal components from proximal dendritic compartments.
To evaluate the impact of cytosolic perfusion for localized calcium signalling in the submicrometer domain, it is useful to consider calcium profiles around open calcium channels (Neher, 1986; Klingauf & Neher, 1997). For moderate influx rates (non-saturated local buffers; Roberts, 1994), the average time a calcium ion can diffuse before it is ‘captured’ by a buffer molecule is given by:
where DCa has the above-mentioned meaning. Low-affinity calcium-binding proteins like calbindin (Kd around 10 μm) with kon rates around 5 × 108 M−1 s−1 are thought to represent the fixed or hardly mobile buffer population in intact cells. In this case, an endogenous calcium binding ratio of 50 found for spinal motoneurones predicted an endogenous buffer concentration ([Ca2+]i << Kd):
According to this concept, the presence of fura-2 notably increases relevant buffer concentrations during whole-cell recordings (kon= 5 × 108 M−1 s−1 for fura-2; Klingauf & Neher, 1997). For example, 400 μm fura-2 reduces effective diameters of calcium domains from control values of L < 59 nm (500 μm total buffer concentration) to L < 44 nm (900 μm). It is important to point out that this approximation was based on homogeneously distributed buffers which were not saturated by local calcium gradients, two assumptions that were not necessarily valid in submembrane compartments of intact cells (Roberts, 1994; Klingauf & Neher, 1997). For our present purposes, however, these approximations provide a quantitative tool to evaluate disruptions of localized calcium domains by indicator dyes in patch-clamped cells. If not indicated otherwise, statistics are given throughout the text as means ±s.d.
Simultaneous patch-clamp and rapid microfluorometric measurements
Motoneurones in the spinal cord of 2- to 8-day-old mice were investigated using simultaneous patch-clamp and microfluorometric recordings in spinal cord slices. They were located in the lateral ventral horn of the spinal cord corresponding to Rexed lamina IX (Fig. 1A). Using infrared differential interference contrast optics (Dodt & Zieglgänsberger, 1994), cells could be visually identified by their large somata with diameters ranging from 20-30 μm (Fig. 1B), and their characteristic dendritic arborization previously illustrated by retrograde labelling (Takahashi, 1990; Palecek et al. 1998). After the whole-cell recording configuration had been established, a rectangular window was adjusted to cover the surface of the soma to permit spatially defined fluorescence recordings of calcium signals (Fig. 1B, see also Methods). During whole-cell recordings, spontaneous synaptic activity was observed at a rate of 0.1-0.3 s−1 (Takahashi, 1990; Jonas et al. 1998), confirming good viability of the slice preparation.
In the presence of 140 mm CsCl or KCl in the pipette solutions, the functional diversity of inwardly directed postsynaptic currents reflected synaptic activation of glutamate, GABA and glycine receptor types (Fig. 2A). Even under these conditions of enhanced excitability, no significant change of somatic calcium levels could be detected during bursts of synaptic activity (upper traces in Fig. 2A). Spontaneous synaptic currents were completely suppressed by adding bicuculline, strychnine and CNQX to extracellular solutions (not shown). The response of motoneurones to changes in membrane potential was investigated by voltage ramp protocols illustrated in Fig. 2B. We found an input resistance of 184 ± 36 MΩ (mean ±s.e.m.; n= 40 cells) at resting potential conditions around -65 mV. Input resistances were lower for recordings in intracellular KCl compared to CsCl solutions with values of 107 ± 21 MΩ (s.e.m.; n= 10 cells) and 210 ± 48 MΩ (s.e.m.; n= 30 cells), respectively. Activation of inward cation currents, presumably through voltage-activated calcium channels, was clearly observable in the current-voltage relationship for the positive voltage range (Fig. 2B). Simultaneous measurements of somatic calcium concentrations (upper trace in Fig. 2B) supports the view that voltage-operated calcium channels are strongly activated during this protocol.
Depolarization-induced calcium responses were further studied by measuring changes in calcium-dependent fluorescence of fura-2 as illustrated in Fig. 3. For this, somatic patch-clamp recordings were performed in voltage-clamp mode and depolarization-induced calcium responses were monitored as a function of cytosolic buffer concentration. After the whole-cell recording configuration had been established, exchanges between pipette and cytosolic solutions occurred with a time constant of 3.4 ± 1.7 min (n= 10). This was monitored by calcium-independent fluorescence of fura-2 at an excitation wavelength of 360 nm exemplified in Fig. 3. To avoid disruptions of voltage-induced calcium signals by synaptic responses in distal dendritic compartments, synaptic activity was blocked by including 10 μm CNQX, 10 μm bicuculline and 10 μm strychnine in bath solutions during depolarization protocols. Repetitive stimulation episodes were composed of somatic depolarizations from -70 to 0 mV with depolarization intervals lasting 500 ms. Figure 3A illustrates corresponding calcium responses by fluorescence recordings at 390 nm for different time intervals after establishing the whole-cell recording configuration. The magnification of calcium-dependent deflections ΔF390 over time reflected the larger amount of calcium ions captured by increasing fura-2 concentrations. Saturation of ΔF390 during ongoing dye loading indicated that the exogenous buffer fura-2 had completely overcome the endogenous buffering capacity of the cell.
After a stable cytosolic dye concentration had been reached, calcium signals were investigated under steady-state conditions during a series of step depolarizations starting at -70 mV (Fig. 4). Resting calcium concentrations were found to be [Ca2+]rest= 112 ± 33 nm (n= 8 cells). For depolarizations positive to -50 mV, notable increases in intracellular calcium concentrations were observed with a mean elevation of 77 ± 52 nm at -40 mV (n= 8 cells). Depolarization-induced calcium currents measured in voltage-clamp mode displayed amplitudes around 2 nA for voltage steps to +10 mV and were characterized by significant current inactivation. During dye loading, inactivation time constants increased from 0.16 ± 0.03 s (n= 6 cells; depolarization to 0 mV; monoexponential fit) at the beginning of whole-cell recordings in the presence of < 40 μm fura-2 to 0.8 ± 0.24 s (n= 5 cells) after complete filling of cells (1 mm fura-2). A likely interpretation is that high-voltage-activated (HVA) channel inactivation was gradually disrupted by increasing buffer concentrations, as recently suggested by Nägerl & Mody (1998). The quantitative approximation valid for low calcium influx rates (described in the Methods section), indicates that disruption of HVA channel inactivation resulted from a reduction in localized calcium domains from control diameters of L= 59 nm (‘dye-free’ conditions) to L= 34 nm in 1 mm fura-2.
Figure 4 shows a superimposed set of calcium responses for membrane depolarizations from -70 to +50 mV. Voltage steps up to -35 mV were necessary to reach half-maximum calcium signals (mean = -33 ± 10 mV, n= 9 cells) that saturated at +10 mV. Calcium concentrations displayed a monotonic increase upon membrane depolarization (Fig. 4, top traces); this was found even for depolarizing pulses lasting several seconds, indicating that voltage-dependent calcium influx was an essential determinant. It was of interest that delays between onset of depolarizations and calcium responses were shorter than 50 ms (Fig. 4). As illustrated in the Methods section, this provided an estimate of 3 μm for the average distance between calcium influx channels and the fluorescence detection window. Combined with the linear increase of calcium responses, this strongly suggested that elevations in [Ca2+]i resulted from influx through somatic calcium channels located close to the fluorescence recording site (Fig. 1). Another important parameter for the analysis of calcium signals was the minimum time interval needed to achieve complete recovery of calcium transients between consecutive depolarizations. Within the resolution of our measurements, a complete recovery of cytosolic calcium levels was achieved when subsequent stimulations were separated by time intervals of 60 s. After recovery, robust and reproducible calcium signals were obtained during standard experiments lasting up to 1 h. A component of the recovery time for stimulated calcium transients was most likely associated with calcium uptake into mitochondria and calcium stores of the endoplasmic reticulum, where recovery times of calcium loads could be significantly slower compared to the cytosol.
Multiple types of voltage-dependent calcium channels have been identified in spinal motoneurones, including calcium channels of the P/Q, N, L and R types (Westenbroek et al. 1998). To minimize the complexity of depolarization-induced calcium responses, we started our voltage-clamp protocols from holding potentials of -70 mV, thus restricting the population of open calcium channels essentially to HVA types. Figure 5 illustrates HVA-mediated responses with calcium elevation and recovery for the standard stimulation protocol. Positive to -50 mV, somatic depolarizations evoked massive calcium responses with amplitudes of 300 ± 149 nm at 0 mV (n= 8 cells), which recovered after several seconds. Prepulses to -90 mV performed to optimally activate low voltage-activated (LVA) calcium channels induced only minor responses after consecutive depolarizations to -50 mV (not shown), indicating that LVA-mediated calcium influx was relatively small. This was not surprising, as LVA-mediated calcium currents display rapid inactivation, thus restricting the absolute amount of charge transport and associated calcium influx.
Calcium homeostasis has been investigated in great detail in other neurones, such as cerebellar Purkinje cells (Fierro & Llano, 1996), providing an important reference point for comparison. To demonstrate that variations in experimental conditions did not account for differential results, we investigated calcium responses in cerebellar Purkinje cells in mice of the same age group (2-8 days). As exemplified in Fig. 6, standard stimulation episodes evoked notably different calcium responses in spinal motoneurones and cerebellar Purkinje cells under identical experimental conditions. A clear difference was observed for the recovery time constant of calcium transients, which was 2.5 s in spinal motoneurones compared to 7.5 s in cerebellar Purkinje cells (Fig. 6). Other differences included the retardation of calcium responses by increasing fura-2 concentrations, which was less pronounced in cerebellar Purkinje cells. Together with previous findings (Fierro & Llano, 1996), these results confirmed that there are marked differences in somatic calcium buffering and transport mechanisms between cerebellar Purkinje cells and spinal neurones.
Quantitative analysis of calcium homeostasis in spinal motoneurones
Calcium homeostasis in spinal neurones was further quantified according to the added buffer approach (Neher & Augustine, 1992). Experimentally, this was done by recording calcium responses during repetitive somatic depolarizations in voltage-clamp mode. During fura-2 loading of cells, competition between endogenous and exogenous calcium buffers reveals essential parameters of neuronal calcium homeostasis (Neher, 1995). For example, the quantitative model illustrated in the Methods section predicts a linear relation between decay-time constants τ of calcium transients and exogenous calcium buffering capacities κB′ of fura-2, according to:
(Neher & Augustine, 1992; Helmchen et al. 1997; Lips & Keller, 1998), where κS and γ denote the endogenous buffering capacity of the cell and the effective calcium extrusion rate constant, respectively. For the analysis of κS, calcium responses were evoked by depolarizations lasting 500 ms. In this case, recovery phases of calcium transients were well approximated by a single exponential, as exemplified in Fig. 6A.
By monitoring decay times of calcium transients as a function of fura-2 loading, endogenous buffering capacities κS were quantified as illustrated in Fig. 7B. It shows the result from 13 cells based on the calcium indicator dyes fura-2 and mag-fura-5, respectively. For fura-2, we found decay-time constants between 0.5 and 15 s for exogenous buffering capacities ranging from 16 to 1300. As predicted by the linear model, decay times were well correlated with the dye concentration in the cytosol (Pearson's r= 0.84). Endogenous calcium binding capacities were calculated from the x-axis intercept and were found to be 50 ± 17 (s.e.m.; n= 13 cells). This indicated that only 1 out of 50 calcium ions contributed to the free cytosolic calcium concentration and 98 % of calcium ions were taken up by endogenous buffers. Extrapolation of Fig. 7B to κB′= 0 yielded a time constant of 0.37 ± 0.12 s (s.e.m.; n= 13 cells, 21°C), thus providing an estimate for recovery of calcium transients under dye-free conditions. An alternative strategy to determine this parameter was based on calcium indicator dyes with small buffering capacities. For example, the low-affinity calcium indicator dye mag-fura-5 displayed a buffering capacity of κB′= 16 (100 nm[Ca2+]i) at a cytosolic concentration of 500 μm. Under these conditions, calcium recovery was approximated by an exponential with a decay-time constant of 0.57 ± 0.26 s (n= 5 cells, average identified by the open circle in Fig. 7B). As this result was similar to decay times of 0.37 s extrapolated from fura-2 measurements, two independent lines of evidence support the validity of our quantitative analysis.
Calcium homeostasis of spinal motoneurones was further quantified by an approach based on calcium-dependent fluorescence at 390 nm. An advantage of this strategy is that it provides an estimate of calcium homeostasis that is independent from ratiometric calibration parameters required for quantitative calcium measurements. As illustrated in Fig. 3A, deflections in calcium-dependent fluorescence (ΔF390) notably increased with exogenous buffer concentrations, thus providing a ratio-independent monitor of calcium dynamics and buffering. In spinal motoneurones, deflections ΔF390 saturated for exogenous calcium binding capacities around 100 (Fig. 8A), demonstrating that this amount of buffer was sufficient to override the endogenous buffers of the cell. As described in the Methods section, the fura-2 dependence of ΔF390 permitted an estimate of endogenous buffering capacities from the relation:
where ΔFmax denotes the saturating value of ΔF390. Experimentally, this analysis was performed by plotting ΔF390 as a function of κB′ as illustrated in Fig. 8A. A least-squares fit identified ΔFmax= 0.62 ± 0.04 BU (n= 24) and an endogenous buffering capacity κS= 39 ± 7 (n= 24), which was in reasonable agreement with the value of κS= 50 previously determined by decay-time analysis (Fig. 7).
The quantitative model of calcium homeostasis was refined by investigating signal amplitudes as a function of fura-2 concentrations in the cytosol. As described in the Methods section, the inverse amplitude 1/A of individual calcium transients is derived from the equation:
(Neher & Augustine, 1992; Helmchen et al. 1997), where qCa denotes the calcium-mediated charge influx per cytosolic volume element and F represents the Faraday constant. Following this concept, inverse amplitudes of calcium responses were plotted as a function of fura-2 buffering capacity as illustrated in Fig. 8B. The x-axis intercept yields an estimate of κS= 54 ± 9 (n= 20) for the endogenous buffering capacity, which is in reasonable agreement with approximations obtained from decay times and ΔF390. The intercept with the y-axis corresponds to a value Ao= 1.8 ± 0.3 μm (n= 20), thus providing an estimate of calcium amplitudes under dye-free conditions. Furthermore, Fig. 8B demonstrates that a somatic, calcium-mediated charge influx of qCa= 21 ± 5 pC pl−1 (n= 20) is associated with a 500 ms depolarization interval. Similar results were found in four additional cells.
Further experiments were performed to verify the linear, one-compartment model for somatic calcium homeostasis in spinal motoneurones. One assumption of this model is that the absolute amount of calcium flowing into the cell per stimulation pulse remains constant over the time course of an experiment (typically 1 h). This condition was checked by calculating the product of calcium amplitudes and decay times from microfluorometric measurements (‘calcium integrals’, Fig. 9). Calcium influx was stable during dye-loading of cells, which was demonstrated by constant integrals for the first minutes after establishing the whole-cell configuration (Fig. 9A). Figure 9B shows a similar result by displaying calcium integrals as a function of exogenous buffering capacity, covering a time interval of 1 h in whole-cell mode. The stability of calcium responses was surprising in the light of earlier observations showing that inactivation times of voltage-activated calcium currents were prolonged during dye loading (see above). The most likely explanation is that increased calcium influx resulting from retarded inactivation was largely compensated by a moderate ‘rundown’ of calcium current amplitudes. This view was indeed supported by patch-clamp recordings in voltage-clamp mode, which showed a gradual decrease of current amplitudes over the time course of an experiment.
Another assumption of the one-compartment model is that the recovery phase of calcium transients is described by a single extrusion rate:
Indeed, the accurate approximation of calcium recovery by a single exponential component provides strong support for this hypothesis (Figs 6A and 7). As illustrated in the Methods section, γ characterizes the superposition of several mechanisms, including active calcium transport across the plasma membrane and uptake into intracellular organelles such as the endoplasmic reticulum or mitochondria (Neher, 1995). A numerical estimate was obtained from Fig. 7, identifying a value of γ= 140 ± 47 s−1 (s.e.m.; n= 13 cells, 21°C) in spinal motoneurones. To rule out a potential calcium dependence, we performed a series of voltage-clamp experiments with varying depolarization time intervals as illustrated in Fig. 10A. For depolarization time intervals lasting 400-1200 ms, we found somatic calcium amplitudes of 200-1000 nm. Decay-time constants were similar for different calcium elevations (Fig. 10B), indicating that the assumption of a calcium-independent extrusion rate was justified. Also, calcium integrals increased linearly with depolarization time as illustrated in Fig. 10C, providing further support for the validity of the linear model. Similar results were obtained for n= 5 cells.
In a methodological context, an important question is related to the potential disruption of calcium transport by the experimental conditions of our patch-clamp recordings, for example by pipette perfusion of the cytosol. To investigate this, calcium extrusion rates were further analysed as a function of recording time in the whole-cell configuration. As illustrated in Fig. 10D, extrusion rates were determined from the decay phases of calcium transients, which were well approximated by a single-exponential component even in the first minute after establishing the whole-cell mode. For a perfusion time constant of 3.4 min, characteristic for our patch-clamp recordings (see above), this indicated that no more than 25 % of the cytosolic components with diffusion constants comparable to that of fura-2 had been ‘washed out’ into the pipette. Extrusion rates were only mildly retarded upon further pipette perfusion in whole-cell mode as illustrated by data from four cells in Fig. 10D. More specifically, extrapolation of the function γ(t) to t= 0 s identified a starting value of γ(0) = 145 ± 40 s−1 (n= 5), which was close to the value previously determined from fura-2 measurements shown in Fig. 7. Taken together, these results support the view that (i) the linear, one-compartment model provides a reasonable description of somatic calcium homeostasis for moderate calcium elevations below 1 μm, (ii) calcium influx is relatively constant for recording time intervals of interest (ca. 1 h), (iii) the superposition of calcium extrusion and uptake mechanisms is adequately represented by a single, calcium-independent extrusion rate, and (iv) the effective extrusion rate is not dramatically altered by pipette perfusion of the cytosol during whole-cell recordings.
Simultaneous electrophysiological and microfluorometric recordings from spinal motoneurones
Our results demonstrate that simultaneous patch-clamp and microfluorometric calcium measurements can be performed on visually identified motoneurones in the mouse spinal cord. We utilized a microfluorometric system (Lips & Keller, 1998; Ladewig & Keller, 1998) that permits ratiometric calcium measurements with a temporal resolution in the millisecond time domain to perform a quantitative analysis of calcium homeostasis in spinal motoneurones. Calcium dynamics were probed by voltage activation of somatic calcium channels in the presence of varying concentrations of ‘exogenous’ calcium buffers in the cytosol (Neher & Augustine, 1992; Lips & Keller, 1998). By using this ‘added buffer’ approach, we found that endogenous calcium buffering capacities were small in spinal motoneurones compared to primary neurones in the cerebellum or hippocampus (Fierro & Llano, 1996; Helmchen et al. 1996).
To investigate calcium homeostasis under stable and well-controlled conditions, a suitable calcium signal had to be functionally isolated. Somatic membrane depolarizations from -70 to 0 mV provided reproducible calcium responses both with respect to the overall amount of calcium flowing into the cell and the recovery to basal resting conditions. A necessary requirement was that consecutive pulses were separated by recovery intervals of 45-60 s. Protocols starting from holding potentials of -70 mV reduced activation of calcium channels to high-voltage-activated (HVA) types while low-voltage-activated (LVA) channels were mostly inactivated (Plant et al. 1998; Westenbroek et al. 1998). Responses from kinetically slower HVA channels were preferable to those from LVA channels as they reduced the potential variability of calcium signals resulting from imperfect space-clamp conditions expected in motoneurones with long dendrites. Even under these reduced conditions, depolarization-induced calcium signals were shaped by many factors, including different HVA channel types (Viana et al. 1993; Plant et al. 1998), their subcellular distribution, electrotonic distances of dendritic compartments, distribution of buffers, extrusion mechanisms and the local surface-to-volume ratio of the cell.
For spinal motoneurones, the surprisingly linear dependence of calcium responses on depolarization time argues that voltage-dependent calcium influx plays a dominant role under our experimental conditions (Fig. 4). This view is supported by the immediate recovery of calcium transients at the end of a depolarizing pulse (Fig. 5). From such data, three conclusions can be stated with respect to calcium responses in spinal motoneurones: (i) simultaneous patch-clamp and microfluorometric recordings provide stable recording conditions for at least 1 h, (ii) depolarization-induced calcium signals were reproducible upon repetitive depolarizations for time intervals of interest, and (iii) steady-state resting calcium conditions in the cytosol were maintained within the resolution of our calcium measurements if repetitive depolarizations were separated by time intervals of 1 min.
Calcium dynamics and buffering in somatic compartments
The added buffer method (Neher & Augustine, 1992) permitted us to perform a quantitative analysis of endogenous calcium homeostasis in spinal motoneurones (see also Neher, 1995; Fierro & Llano, 1996; Helmchen et al. 1997). During controlled dye-filling of cells, endogenous calcium binding ratios were found to be κS= 50, indicating that only 2 % of calcium ions in the cytosol contributed to the free calcium concentration. This value is comparable to previously determined binding ratios of κS= 41 for hypoglossal motoneurones in mice of similar age (Lips & Keller, 1998), 40 for the calyx of Held (Helmchen et al. 1997) and 40 for adrenal chromaffin cells (Zhou & Neher, 1993). It is interesting to note that spinal motoneurones displayed significantly smaller buffering capacities compared to other primary neurones with similar sizes and comparably complex dendritic arborizations, such as cerebellar Purkinje cells (κS= 900 for 6-day-old rats; Fierro & Llano, 1996; for review see Neher, 1995) and CA1 pyramidal neurones (κS= 168-207; Helmchen et al. 1996).
The expression of calcium-binding proteins has been shown in other studies to depend significantly on ontogenetic development. For example, a more than two-fold increase in endogenous calcium buffering capacity has been described for cerebellar Purkinje cells at age P6 compared to P14 (Fierro & Llano, 1996). Accordingly, it is important to evaluate the impact of postnatal stage for buffering capacities in motoneurones determined at age P2-8. One difficulty in investigating this question experimentally is the pronounced vulnerability of motoneurones in older animals, resulting in impaired neurones in spinal cord slices taken from animals older than P12. However, immunocytochemical studies have demonstrated that the most prominent calcium-binding proteins, calbindin and parvalbumin, are expressed in spinal motoneurones in a development-dependent way, where expression levels gradually increase during embryonic stages to peak levels around birth (Zhang et al. 1990). For our purposes, these results suggest that expression of calcium-binding proteins had reached maximum levels in the animals utilized in this study, and that values around 50 most likely represented an upper limit for endogenous calcium buffering capacities above P8. A decrease in calbindin and parvalbumin levels with postnatal age also provides a plausible cellular basis to account for the pronounced vulnerability of motoneurones in slice preparations of animals older than P12. Other explanations for this vulnerability may be attributed to increases in dendritic arborization with postnatal age, presumably leading to higher fractions of impaired dendrites during preparations of spinal cord slices.
As illustrated in the Methods section, the presence of calcium indicator dyes in cytosolic compartments influences the kinetic properties of calcium signals measured in patch-clamped cells. According to the added buffer approach, the kinetics of calcium transients under dye-free conditions can be approximated by extrapolating them to κB′= 0 (Fig. 8; Neher, 1995). We found decay-time constants in the soma of 0.37 ± 0.12 s (n= 9 cells; 21°C), which is slow compared to those found in presynaptic terminals (91 ms at 21°C, Helmchen et al. 1997), relatively fast compared to those in hypoglossal motoneurones (0.7 s; Lips & Keller, 1998), but significantly accelerated compared to decays in adrenal chromaffin cells (7.2 s; Neher & Augustine, 1992). For a given decay time, the extrusion rate γ was calculated by assuming a one-compartment model for calcium homeostasis in the soma as described in the Methods section (Neher, 1995). For spinal motoneurones, we determined an effective extrusion rate γ= 140 ± 47 s−1 (n= 13 cells, 21°C), which is small compared to values found for the calyx of Held (400 s−1, 21°C; Helmchen et al. 1997), faster than values of γ found in hypoglossal motoneurones (60 s−1, 21°C; Lips & Keller, 1998), but several times faster compared to those found in adrenal chromaffin cells (Neher & Augustine, 1992).
Potential implications of calcium homeostasis for somatic calcium signalling
A particularly interesting aspect of calcium homeostasis relates to the temporal profile of calcium transients under physiological conditions. For example, locomotion is associated with bursts of action potential discharges in motoneurones with burst frequencies up to 10 Hz (James et al. 1995). As notable calcium elevations have been shown to occur during rhythmic motoneurone activity (Lev-Tov et al. 1995; Ladewig & Keller, 1998), recovery of calcium transients in vivo has to occur on the time scale of tens of milliseconds to avoid a potentially dangerous accumulation of basal calcium levels during consecutive bursts. Recovery phases of calcium transients in patch-clamped motoneurones are significantly slower, and differences in temperature (21°C for our measurements) most likely account for this discrepancy. In addition, we cannot exclude that whole-cell recordings were paralleled by a simultaneous retardation of calcium recovery dynamics, for example as a result of ‘washout’ of cytosolic components responsible for fast calcium transport in intact cells. A first step to address this question was taken by measuring calcium extrusion rates as a function of whole-cell recording time, where our results suggested that extrusion rates were only mildly retarded (Fig. 10D). This was noteworthy since the analysis of extrusion rates started within the first minute of recording time, when only 25 % of the cytosolic components with diffusional properties comparable to those of fura-2 had been washed out (see Methods).
An important aspect of calcium homeostasis of spinal motoneurones is related to the molecular entities that account for endogenous buffering capacities. As previously pointed out by Zhou & Neher (1993), endogenous buffers that persist for more than 1 h of whole-cell recordings are probably hardly mobile calcium-binding proteins with a low affinity for calcium under physiological conditions (KD around 10 μm). Parvalbumin and calbindin represent suitable candidates, since both proteins are expressed in spinal motoneurones in significant amounts during different stages of development (Zhou et al. 1990). With respect to our quantitative analysis, they probably represented the dominant population of buffers underlying κS, the buffering capacity determined during long-term whole-cell recordings. Other components which are suitable candidates for hardly mobile buffers include calcium binding sites in mitochondrial and endoplasmic reticulum membranes, which presumably remain in the cytosol during pipette perfusion. Interestingly, as described in the Methods section, a value of κS= 50 provided an estimate of 500 μm for the total concentration of hardly-mobile calcium buffers with low affinity, including components from the cytosol, mitochondria and endoplasmic reticulum (Zhou & Neher, 1993).
The impact of endogenous calcium buffers with higher mobility could not be identified directly in our present analysis and future experiments involving perforated-patch recordings are necessary to clearly identify their role. Earlier work established that in other systems, such as chromaffin cells, mobile buffers represent approximately 25 % of the total buffering capacity (Zhou & Neher, 1993; Klingauf & Neher, 1997). If the conditions were comparable, this provides an estimate of κmob= 13 for spinal motoneurones and κS+κmob= 63 for the total buffering capacity. Molecular components that could account for mobile buffers include small organic anions or free concentrations of ATP, which presumably diffuse through the cytosol with an estimated diffusion constant around 15 μm2 s−1 (Zhou & Neher, 1993). As this is one order of magnitude smaller than the diffusion constant of fura-2, it indicates that there was only a mild washout of mobile calcium buffers during the first minutes of whole-cell recording (perfusion time constant of fura-2 = 3.4 min; see also Methods). A potential washout of mobile buffers would have directly affected the numerical value of effective extrusion rates determined during whole-cell recordings (Fig. 10D). However, the moderate retardation of extrusion rates observed in the first minutes of whole-cell recording time further supports the view that diffusion of mobile calcium buffers into the pipette solution was relatively slow.
Several parameters of calcium homeostasis in spinal motoneurones appear to be optimal for rapid calcium kinetics. First, as demonstrated by the one-compartment model, calcium decay times were linearly related to endogenous buffering capacities κS. The low endogenous buffering capacities found in spinal motoneurones therefore account directly for high-speed calcium recovery kinetics if all other parameters are held constant. Theoretically, fast decay times could also be achieved for higher buffering capacities through elevated calcium extrusion rates. A direct consequence of this strategy would be that calcium transients with comparable amplitudes would be associated with greater amounts of calcium influx and subsequent extrusion across the cell membrane. This would require higher ATP-dependent energy consumption during rhythmic motoneurone activity. Hence, low endogenous buffering capacities provide spinal motoneurones with rapid calcium dynamics at relatively low energy cost. A second feature provided by low buffering capacities versus higher ones is that they account for larger domains of elevated calcium concentrations around open channels when all other parameters are comparable (Roberts, 1994; Neher, 1995; Klingauf & Neher, 1997). To a first approximation, low buffer concentrations therefore facilitate localized calcium signalling and presumably improve the control of calcium-dependent second messenger systems in defined subcellular compartments.
Potential role of exogenous calcium buffers as neuroprotective agents during excitotoxic conditions and motoneurone disease
The quantitative analysis of calcium homeostasis presented in this report provides a useful tool to evaluate how exogenous buffers influence motoneurone protection. First, calcium-dependent inactivation of HVA channels was clearly disrupted for fura-2 concentrations around 400 μm (κB′= 440), but remained essentially intact for concentrations below 40 μm (κB′= 44). By using approximations valid for small calcium influx rates (i.e. non-saturated endogenous buffers; Roberts, 1994), this suggests that reduction of localized calcium domains to ‘effective’ diameters of 44 nm (400 μm fura-2) impairs HVA channel inactivation in spinal motoneurones. In contrast, low doses of fura-2 (40 μm) with domain diameters around 60 nm appear to be tolerated without causing a potentially dangerous disruption of HVA channel inactivation (Chad, 1989; Abdel-Hamid & Baimbridge, 1997; Nägerl & Mody, 1998). Second, buffers could have a damaging effect by retarding calcium recovery (see for example Fig. 7). For burst frequencies up to 10 Hz measured in spinal motoneurones during locomotion in mice (James et al. 1995), even a modest retardation of calcium recovery can result in a substantial accumulation of cytosolic calcium during consecutive bursts. Our quantitative analysis suggests that recovery times are notably prolonged in the presence of exogenous buffer concentrations equivalent to tens of micromolar fura-2. During high-speed locomotion, this could presumably enhance calcium accumulation during consecutive bursts and thus magnify the risk for excitotoxic damage.
Other arguments provided by our study further illustrated the potential role of exogenous buffers as neuroprotective compounds. In addition to prolonging calcium decay times, exogenous buffers reduced the amplitude of calcium transients when total calcium influx was held constant. Thus, exogenous buffers presumably acted as neuroprotective compounds by avoiding large-amplitude calcium transients and associated activation of high-threshold, calcium-dependent second messenger systems commonly linked to neuronal degeneration and cell death (Choi, 1988). Taken together, our results therefore indicate that calcium buffers as neuroprotective agents need to be evaluated quantitatively by different parameters of calcium signalling. Such parameters include the endogenous buffering capacity of the cell, the ratio between endogenous and exogenous buffer concentrations, extrusion rates, the calcium dependence of ‘excitotoxic’ processes, the temporal profile of electrical activity and the physical chemistry of the buffer. Quantitative analysis of calcium homeostasis such as was carried out in this investigation constitutes a valuable tool for future evaluations of neuroprotective strategies to preserve spinal neurones during motoneurone disease and associated excitotoxic conditions.
We thank D. Crzan for excellent technical assistance. Also we thank Drs P. Lalley and D. W. Richter for valuable discussions. This research was supported by DFG grants Ke 403/5-2, Ke 403/6-1, Sonderforschungsbereich 406 and Czech Republic grant GACR 305/96/0680.