Capillary fragments were isolated from guinea pig hearts, and their electrical properties were studied using the perforated-patch and cell-attached mode of the patch-clamp technique. A voltage-dependent K+ current was discovered that was activated at potentials positive to −20 mV and showed a sigmoid rising phase. For depolarizing voltage steps from −128 to +52 mV, the time to peak was 71 ± 5 ms (mean ± SE) and the amplitude of the current was 3.7 ± 0.5 pA/pF in the presence of 5 mM external K+. The time course of inactivation was exponential with a time constant of 7.2 ± 0.5 s at +52 mV. The current was blocked by tetraethylammonium (inhibitory constant ∼3 mM) but was not affected by charybdotoxin (1 μM) or apamin (1 μM). In the cell-attached mode, depolarization-activated single-channel currents were found that inactivated completely within 30 s; the single-channel conductance was 12.3 ± 2.4 pS. The depolarization-activated K+current described here may play a role in membrane potential oscillations of the endothelium.
- coronary circulation
- membrane potential oscillations
- electrophysiology of endothelium
endothelial cells show complex changes in membrane potential on the application of vasoactive agonists (20-22) and are able to produce membrane potential oscillations (18,19). Many of the functions of the endothelium, for example, the release of vasoactive compounds and the regulation of the hydraulic conductivity of the vascular wall, are influenced by the free intracellular Ca2+ concentration, which increases with hyperpolarization and often shows pronounced oscillations (2, 11, 13, 15, 29). Thus the electrical activity of the endothelial cells will almost certainly have profound effects on endothelial function. As yet, the role of different ion channels in generating the electrical responses of the endothelium is still far from clear.
Most of our present knowledge of the electrophysiology of vascular endothelial cells has been derived from studies on1) cultured macrovascular endothelial cells, 2) cell lines derived from endothelium, and 3) intact endothelium of large vessels (6, 23). There is, however, relatively little information available on the electrical activity of capillaries. ATP-sensitive K+channels (12), Ca2+-activated K+ channels (30), and voltage-gated Ca2+ channels (1) have been found in cultured microvascular endothelial cells. The significance of these findings is uncertain because it is well known that endothelial cells may undergo profound changes in metabolism and electrical properties during culture (28). Gögelein and co-workers managed to obtain single-channel recordings from freshly isolated cerebral capillaries and found nonselective cation channels (25) and inwardly rectifying K+ channels (10).
Here we report the first whole cell recordings from fragments of freshly isolated coronary capillaries. We give a quantitative description of the most prominent current in these cells, a voltage-dependent K+ current that activates rapidly on depolarization and inactivates very slowly.
Isolation of capillaries.
Coronary capillaries were obtained from guinea pig hearts by enzymatic dispersion. Guinea pigs weighing 250–450 g were decapitated and their hearts rapidly excised. The isolated heart was perfused at a constant flow rate of 10 ml/min using a peristaltic pump. The heart was submerged in a small organ bath warmed to 37°C. The perfusing solution contained (in mM) 130 NaCl, 15 KCl, 2 CaCl2, 0.8 MgCl2, 1 NaH2PO4, 2 Na-pyruvate, 10 glucose, and 10 HEPES. The pH was 7.4 (adjusted with NaOH); the temperature was 37°C. The elevated K+ concentration of the perfusate caused cardiac arrest. Within 15–20 min, coronary perfusion pressure increased to a steady level between 60 and 100 mmHg, indicating recovery of energy metabolism. Subsequently, as a test of adequate perfusion of the coronary blood vessels, 1 μM adenosine was applied for 1–2 min, which elicited a reversible reduction of perfusion pressure by ∼50%.
To initiate dissociation of the cells, the heart was perfused for 5 min with nominally Ca2+-free solution, which otherwise had the same composition as described above. The heart was then perfused for 10 min with Ca2+-free solution to which 30 μM Ca2+ and 1 or 1.5 mg/ml collagenase blend (type H, Sigma) was added. Subsequently, the heart was removed from the organ bath and washed briefly in a solution containing (in mM) 65 K-glutamate, 45 KCl, 30 KH2PO4, 3 MgSO4, 0.5 EGTA, 20 taurine, and 10 glucose (pH adjusted to 7.4 with KOH). The heart was then disintegrated by being gently shaken with a forceps. Drops of the suspension were transferred immediately to 35-mm petri dishes (Nunc, Roskilde, Denmark) containing the same solution. After 30 min, nonadhering cells were washed away with normal physiological salt solution containing 5 mM K+ (for composition, see Electrophysiology, solutions, and reagents). Intact myocytes, spherical cells of 10–15 μm in diameter, and capillary fragments remained attached to the bottom of the petri dishes. In seven experiments, capillary fragments prepared by the method of Langheinrich and Daut (14) were used. At least 30 min before the start of electrophysiological recording, 350–400 units of deoxyribonuclease (DNase I, type IV, Sigma) were added to each petri dish to clean the surface of the cells.
Electrophysiology, solutions, and reagents.
Patch-clamp recordings were carried out on the stage of an inverted microscope (Zeiss, IM 35) at room temperature (23°C) within 10 h after the suspension was seeded on the petri dishes. After a capillary fragment adhering to the bottom of a petri dish was selected, a Perspex frame was mounted over the capillary, as described previously (5). The frame formed a perfusion chamber of 0.75 mm in height, 1.5 mm in width, and 18 mm in length. The perfusion rate was 5–10 ml/h. The normal physiological salt solution contained (in mM) 140 NaCl, 5 KCl, 1 MgCl2, 1 NaH2PO4, 2 CaCl2, 10 glucose, and 10 HEPES (pH 7.4, adjusted with NaOH). In solutions containing higher K+ concentrations or tetraethylammonium (TEA; up to 20 mM), Na+ was reduced to keep osmolarity constant. Charybdotoxin (CTX) and apamin were purchased from Alomone (Jerusalem, Israel). The solutions containing CTX or apamin were prepared immediately before the experiments from stock solutions (prepared with physiological salt solution) stored at −20°C for no longer than 4 wk. All other reagents were obtained either from Merck (Darmstadt, Germany) or from Sigma (St. Louis, MO).
Voltage-clamp experiments were carried out in both the conventional whole cell mode of the patch-clamp technique (n = 28) and the perforated-patch mode (n = 52). For perforated-patch measurements, pipettes of 1- to 2-μm tip diameter were made of thin-walled glass 1.5 mm in diameter without filament (Clark, Reading, UK). They were coated with Sylgard to reduce capacitance and heat-polished directly before use. The resistance of the pipettes was 5–8 MΩ. The pipette solution contained (in mM) 45 KCl, 100 K-aspartate, 1 MgCl2, 0.5 EGTA, and 10 HEPES (pH 7.2, adjusted with NaOH). With this solution, Donnan potentials at the perforated patch were assumed to be negligible. The tip of the patch electrode was first filled with amphotericin-free pipette solution by aspiration, and then the pipette was backfilled with the same solution to which 300 μg/ml amphotericin B had been added from a stock solution. Sonication was applied to improve solvation of amphotericin B. The stock solution contained 20 mg/ml amphotericin B in DMSO and was prepared freshly every day. Perforation started shortly after seal formation and reached a steady-state level within 5–10 min. The pipette solution used for conventional whole cell recordings contained (in mM) 45 KCl, 100 K-aspartate, 10 EGTA, 1 CaCl2, 3 MgCl2, 2 Na2ATP, 0.1 Na3GTP, and 10 HEPES (pH 7.2, adjusted with NaOH). After a gigaseal was formed, the patch membrane was ruptured by application of suction.
The seal resistance was determined by slow voltage ramps from −120 to +40 mV and back to −120 mV before the patch was broken by suction or before amphotericin started to perforate the patch. The membrane capacitance was calculated from the current offsets observed during the ascending and descending voltage ramps in the whole cell mode.
Recordings were carried out with an EPC-7 patch-clamp amplifier (List, Germany) and a modified digital audio tape recorder (DTC-55ES, Sony; sampling rate 44 kHz). In all whole cell recordings, the capacitance compensation circuitry of the EPC-7 amplifier was used. For analysis, the data were filtered with an eight-pole Bessel filter before sampling. The cutoff frequency was one-half the sampling rate, which ranged from 100 to 5,000 Hz. Potentials were corrected for liquid junction potentials (−5 to −8 mV). The liquid junction potentials were determined separately for the recording electrode and for the reference electrode because changes in extracellular K+ also affected the potential of the reference electrode.
The results obtained with the conventional whole cell mode and with the perforated-patch mode were very similar. Therefore, the results obtained with both approaches were combined. The data are given as means ± SE, and n denotes the number of capillaries from which the data were obtained.
Electrical properties of coronary capillary fragments.
The capillary fragments could be easily identified by their morphology. They appeared as thin translucent tubes of ≤300 μm in length and contained a regularly spaced string of nuclei. As measured between the nuclei, the diameter of the vessel was 3–4 μm. Occasionally the capillaries had side branches of similar diameter. The capillary fragments most suitable for whole cell recording were nonbranched, contained only three to five nuclei, and had an overall length of 30–80 μm (Fig. 1). The endothelial nature of the cells in the capillary fragments was ascertained in control experiments by multicell RT-PCR (26) with the use of a hydraulic “cell picker” and specific primers for endothelin-1.
The tip of the patch pipette was sealed in most cases to the perinuclear region near the middle of the capillary fragment; the mean seal resistance was 76 ± 6 GΩ (n= 20). The capacitance of the capillary fragments was 42 ± 5 pF (n = 31). The mean membrane potential in physiological salt solution containing 5 mM K+ was −34 ± 2 mV (n = 32). In the high-K+ solution (145 mM K+), the membrane potential was −3.5 ± 0.8 mV (n = 25). The mean steady-state input resistance of the capillary fragments (at voltages near 0 mV) was 1.8 GΩ (range 0.7–40 GΩ). However, in capillary fragments with a calculated input resistance >10 GΩ, the true input resistance could not be determined precisely, because it was on the same order of magnitude as the seal resistance.
Single freshly isolated endothelial cells (in the same preparation) had an average membrane capacitance of 9.2 ± 0.8 pF (n = 30) (N. von Beckerath, M. Dittrich, and J. Daut, unpublished observations), i.e., ∼22% of the capacitance of our small capillary fragments consisting of three to five cells. This is consistent with the assumption of electrical cell coupling in the capillary fragments. In view of the high input resistance of single endothelial cells (6, 32), we assume that a spatially uniform voltage clamp of the entire capillary fragment could be achieved.
Depolarization-activated K+ current.
The most prominent current found in the capillary fragments was a voltage-activated outward current. A typical recording is shown in Fig.2 A. Depolarizing voltage steps were applied from a holding potential of −98 mV to test potentials between −38 and +52 mV. The initial phase of the current is shown on a faster time scale in Fig.2 B. At +52 mV, the time to peak was 71 ± 5 ms (n = 23) with 5 mM external K+ and 49 ± 4 ms (n = 16) with 145 mM external K+. The activation was followed by a very slow inactivation. The time course of inactivation could be fitted with a single exponential. At +52 mV, the mean time constant of inactivation (determined by voltage steps >30 s) was 7.2 ± 0.5 s in the presence of 5 mM external K+(n = 26). With 145 mM external K+, the time constant of inactivation (at +52 mV) was 6.4 ± 0.7 s (n = 9), which is not significantly different from the values obtained with 5 mM external K+.
As can be seen by comparing Fig. 2, Aand B, the amplitude of the initial current (after decay of the capacitive transient) was similar to the amplitude of the steady-state current, which suggests that the current inactivated completely. The residual time-independent current (Fig. 2) showed a variable degree of outward rectification, which was most likely due to a Cl−conductance (23). No attempt was made to eliminate this current, because it did not interfere with the time-dependent current investigated here (see discussion). The amplitude of the inactivating outward current was determined by subtracting the steady-state current observed after inactivation (measured 30 s after the depolarizing voltage step) from the peak current. At +52 mV, the amplitude of the depolarization-activated current was 3.7 ± 0.5 pA/pF with 5 mM external K+(n = 31) and 3.3 ± 1.0 pA/pF with 145 mM external K+(n = 9). If we assume a reversal potential of 83 mV in 5 mM external K+ (0 mV in 145 mM K+), we obtain a conductance of 27 pS/pF in 5 mM external K+ and 63 pS/pF in 145 mM external K+.
To investigate the ionic nature of the current, the external K+ concentration was varied between 5 and 145 mM. The outward current was activated by short depolarizing voltage steps to +52 mV, and the reversal potential of the deactivation tails was determined. In the experiment shown in Fig.3 A, the tail currents reversed at −49 mV in the presence of 20 mM external K+. Figure3 B shows that the dependence of the reversal potential on external K+was in close agreement with the prediction of the Nernst equation for a K+-selective current.
Voltage dependence of activation.
To determine the voltage dependence of activation, the potential was clamped to −98 mV for 2.7 s and then the outward current was activated by depolarizing voltage steps to various test potentials, as shown in Fig. 4. The currents activated by depolarizing voltage steps showed a sigmoid rising phase. The time required to reach the peak current decreased with increasing depolarization. The degree of sigmoidicity of the rising phase depends on the number of sequential conformational changes required to allow opening of the channel (35, 36). The best fits of the time course were obtained with an exponential function raised to the fourth power; exponents of 1, 2, and 3, respectively, gave inferior fits (see legend to Fig. 4). As shown in Fig. 4 B, the time constants (τ) of the fitted curves decreased from 40 ms at 18 mV to 11 ms at +12 mV (n = 4).
The extent of activation of the K+current at different test potentials was determined by analyzing the deactivation tails, with the assumption that the tail current amplitude was proportional to the conductance at the (preceding) test potential. At the peak of the current, the potential was stepped back to −34 mV to induce deactivation. This potential, close to the Cl− equilibrium potential (E Cl), was chosen to minimize any contamination by a time-dependent Cl− current (seediscussion).
In Fig. 5 the normalized conductance derived from the tail currents at −34 mV is plotted against the potential at which the current was activated. The voltage dependence of the conductance change was sigmoid and could be described by a Boltzmann function raised to the fourth power [which is consistent with an activation mechanism involving four independent and identical conformational changes (35, 36)]. The potential for half-maximum activation (V ½) was 7 mV (n = 7). A simple Boltzmann function gave a slightly inferior fit (not shown). We have also analyzed the voltage dependence of activation of the K+ current in the presence of 145 mM external K+. Under these conditions, V ½was 5 mV (n = 6) and the shape of the activation curve was very similar to that shown in Fig. 5 for 5 mM external K+.
Voltage dependence of inactivation.
The voltage dependence of steady-state inactivation was determined by applying depolarizing voltage steps from holding potentials between −128 and +37 mV (Fig. 6). Holding potentials were maintained for at least 6 s before the application of depolarizing steps (of ≥30 s duration) to +52 mV, which maximally activated the outward current. The voltage dependence of steady-state inactivation could be described by a simple Boltzmann function (Fig.5). With 5 mM external K+, the half-inactivation potential (V ½) was −50 mV and the slope factor (k), indicating thee-fold change in conductance per millivolt (see legend to Fig. 5), was 13 mV (n = 21). With 145 mM external K+,V ½ was −53 mV and k was 13 mV (n = 8). Thus the inactivation curve, like the activation curve, was almost unaffected by changes in external K+.
Recovery from inactivation was studied in eight experiments using the protocol shown in Fig.7 A. First, the membrane potential was clamped to +52 mV for at least 28 s to induce complete inactivation. The potential was then stepped back to −88 mV for various periods of time. With increasing duration of the hyperpolarization, the peak outward current recorded at +52 mV became successively larger. The time course of recovery from inactivation could be described by a single exponential (Fig.7 B); the time constant was 551 ± 94 ms (n = 8).
K+ channel blockers.
As a first step toward a pharmacological characterization of the outward current, we tested the K+channel blockers TEA, CTX, and apamin. TEA reversibly inhibited the voltage-activated current. In the presence of 10 mM TEA, the current amplitude was reduced to 23 ± 6% of control during depolarizing voltage steps from −98 to +52 mV (n = 4). Figure8 shows a dose-response curve in which the (interpolated) concentration required for half-maximal inhibition (IC50) was 3.5 mM. In a second complete dose-response curve IC50was 2.5 mM. These results suggest that the current described here was TEA-sensitive with an IC50 of ∼3 mM.
Some members of the Kv family of K+ channels are sensitive to CTX (3, 24). When we tested the effects of CTX on the voltage-activated K+ current, we found that 1 μM CTX had no effect. Endothelial Ca2+ activated K+ channels are sensitive to either apamin or CTX (23). Application of 1 μM apamin did not have any effect on the outward current transient or the steady-state current after complete inactivation. These findings suggest that Ca2+-activated K+ currents and charybdotoxin-sensitive Kv channels (3) did not play a role under our experimental conditions.
In 4 of 80 experiments, steady-state inward rectification was observed at potentials negative to the resting potential. Thus it seemed appropriate to exclude a possible contribution of inward rectifier channels to the kinetics of the depolarization-activated K+ current (see legend to Fig. 3).
In 3 of 80 measurements in capillary fragments, voltage-dependent single-channel currents could be resolved in the cell-attached mode of the patch-clamp technique. An example obtained with 5 mM K+ in the external solution is shown in Fig.9 A. The top trace shows the transmembrane potential of the patch (inside − outside) calculated from the applied extracellular patch potential and the membrane potential of the capillary fragment (recorded later in the whole cell mode). When the patch membrane was depolarized from 138 to +62 mV, single-channel outward currents were observed. In the first 3 s, two channels opened simultaneously. The channel activity subsided within 30 s after the depolarizing voltage step, i.e., the open-state probability of the channels had a time course similar to that of the whole cell currents. The average single-channel conductance calculated from recordings at different potentials in the three capillary fragments was 12.3 ± 2.4 pS (symmetrical K+). The extrapolated reversal potentials of the single-channel currents measured in the cell-attached mode were equal to the membrane potentials of the capillaries, which is consistent with a K+-selective channel.
To get further information on the function of the current, we carried out current-clamp experiments. Depolarizing current pulses were injected into capillary fragments in the conventional whole cell current-clamp mode. Figure 10 shows a typical experiment with a capillary that had a resting potential of ∼20 mV. At this potential the voltage-dependent K+ current was completely inactivated. The current pulses caused voltage displacements that showed no obvious time dependence (the passive membrane time constant was too fast to be resolved on the time scale of the record). In contrast, when the membrane potential was changed to about −40 mV by the application of hyperpolarizing current, a qualitatively different pattern of voltage changes was observed. Figure10 B shows that, in the presence of a holding current of −23 pA, superimposition of the same depolarizing current pulses gave rise to a slow, time-dependent depolarization if the threshold of activation of the K+ current (−20 mV) was crossed. When the applied current was switched off, the membrane hyperpolarized with no obvious delay.
The most likely interpretation of the slow voltage changes shown in Fig. 10 B is that the activation of the voltage-dependent K+ current initially impeded the depolarization and that, subsequently, the cells depolarized slowly as the K+current inactivated. In four current-clamp experiments similar to that shown in Fig. 10, the time constant of the change in membrane potential between 0 and +20 mV was 8.1 ± 2.3 s. This is similar to the mean time constant of inactivation of the voltage-clamp current at +6 mV, which was 6.6 ± 1.2 s (n = 8). It should be noted that the time course of the membrane potential change on current injection and the time course of inactivation during a voltage step to a fixed potential are not necessarily identical. Nevertheless, the good agreement of the time constants measured in voltage-clamp and current-clamp experiments (in the same potential range) is consistent with the hypothesis that both effects may be attributable to the activation and subsequent inactivation of a voltage-activated K+ current. Thus, when the membrane potential of capillaries is more negative than −40 mV, depolarizing currents may be antagonized by activation of an outward current as soon as the threshold of −20 mV is crossed.
Patch-clamp recording from freshly isolated capillary fragments.
Because microvascular endothelial cells in situ are not directly accessible with patch electrodes, the electrophysiology of microvascular endothelium has so far been investigated mostly in cultured cells. However, gene expression of channels, receptors, and metabolic pathways may change rapidly under cell culture conditions (9,28). On the other hand, in freshly isolated endothelial cells, cell identification can pose a problem, because single endothelial cells round up and may not be readily distinguished from rounded up fibroblasts or smooth muscle cells. In the present study we have used freshly isolated capillary fragments that can be easily identified by their morphology. In many vascular beds, endothelial cells were found to be electrically coupled by gap junctions (7, 8, 17), although in capillaries it has been difficult to identify gap junctions by electron microscopy (34). Our finding that the membrane capacitance of single endothelial cells was about five times smaller than the average capacitance of the capillary fragments is consistent with the idea that the endothelial cells in the capillary fragments were electrically coupled and that a spatially homogeneous voltage clamp could be achieved. The electrical length constant of capillaries is probably ∼1 mm (6), whereas the length of the capillary fragments was maximally 300 μm.
The steady-state current-voltage relations between −60 and +52 mV showed a variable degree of outward rectification, similar to the results summarized by Nilius et al. (23). This finding and the mean resting potential of −34 mV, which is close toE Cl, suggest that Cl− channels were also present. Recently it has been shown that the volume-activated Cl− current may also show some inactivation at potentials greater than +40 mV with time constants of maximally 200 ms (31). However, our current exhibits much longer time constants of inactivation, and most of our experiments were carried out at potentials of up to +52 mV, where the extent of inactivation of the volume-activated Cl− current would be negligibly small. Thus we consider it unlikely that Cl− currents interfered with our analysis of the voltage-activated K+ current.
Voltage-dependent K+ current in coronary capillaries.
We have found a voltage-activated K+ current in coronary capillaries that has not been previously described in endothelial cells. This current activates and inactivates much more slowly than the A-type current found by Takeda et al. (27) in cultured aortic endothelial cells (the only other endothelial voltage-dependent K+ current reported so far). Because most previous patch-clamp studies were carried out on cultured macrovascular endothelial cells, it appears possible that the current reported here is restricted to capillary endothelium. Alternatively, its expression may be lost during cell culture.
The capillary endothelial K+current was activated at potentials positive to −20 mV (Fig. 5) and was characterized by a sigmoid onset. The time to peak was on the order of 50–80 ms. The rising phase of the current could be described by an exponential function raised to the fourth power, and steady-state activation could be fitted by a Boltzmann function raised to the fourth power. This gives a minimum for the number of sequential conformational changes required to open the channel (4, 35, 36). The time course of inactivation could be fitted by a single exponential with time constants between 5 and 10 s. The time constant of recovery from inactivation was ∼0.5 s. The current showed a relatively high sensitivity to TEA but was not affected by CTX. The kinetics of the current showed only minor changes when external K+ was increased from 5 to 145 mM, whereas the conductance increased more than twofold. The single-channel conductance was ∼12 pS. The properties of the endothelial K+ current described here differ substantially from eag or HERG currents and from the slowly activating delayed rectifier in cardiac muscle, but show some similarities withShaker-type voltage-activated K+ currents (3, 24).
Possible function of voltage-activated K+ current.
The activation and inactivation curves (Fig. 5) showed some overlap in the voltage range from −0 to −10 mV. Multiplication of relative activation and relative inactivation suggests that ∼2% of the maximal current may flow in the window around −20 mV. Theoretically, this might contribute to the setting of the resting potential in depolarized cells. However, we have not seen a corresponding “hump” in the steady-state current-voltage relation. More importantly, however, the voltage-activated K+ current may have a pronounced influence on the dynamic behavior of the membrane during membrane potential oscillations. Depolarization of the cell membrane from a resting potential of −40 to −45 mV, where 30–40% of the current is available for activation, to 0 mV may activate ∼20% of the maximum current. The rapid activation of the current would counteract any further depolarization. Thus the effect of depolarizing current pulses may be delayed for several seconds until the outward current is inactivated (see Fig. 7). This may play a role during activation of the cells with vasoactive agonists, which often elicit pronounced membrane potential oscillations (11, 13, 15, 33). The mechanisms underlying the membrane potential oscillations in endothelial cells have been widely studied (11, 16, 30, 34). The available evidence suggests that oscillatory changes in intracellular Ca2+ are accompanied by synchronized changes in membrane potential, which are partially, but not entirely, attributable to the opening of Ca2+-activated K+ channels. On the other hand, changes in membrane potential may also influence transmembrane Ca2+ movements (2). An outward current that activates rapidly and inactivates slowly would be expected to modulate the shape and frequency of membrane potential oscillations and thus, via the effect of membrane potential on Ca2+ influx, influence oscillations in intracellular Ca2+.
We thank B. Burk, R. Luzius, R. Graf, A. Mazzola, and E. Hoffmann for technical and secretarial help and Dr. C. Walther for useful comments on the manuscript.
Address for reprint requests and other correspondence: J. Daut, Institut für Normale und Pathologische Physiologie der Universität Marburg, Deutschhausstrasse 2, D-35037 Marburg, Germany (E-mail:).
This work was supported by the Deutsche Forschungsgemeinschaft (Da 177/4-4 and Da 177/7-1).
Present address of M. Dittrich: Institut für Neurophysiologie, Universität zu Köln, Robert-Koch-Strasse 39, D-50931 Köln, Germany (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 1999 the American Physiological Society