INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recent experiments have shown that repetitive synaptic activation evokes regenerative calcium waves in the apical dendrites of hippocampal CA1 pyramidal neurons (Nakamura et al. 1999). The [Ca²⁺]_i increases associated with these waves often reach a level of several micromolar, much larger than the measured, spatially averaged [Ca²⁺]_i increases due to Ca²⁺ entry through N-methyl-D-aspartate (NMDA) receptor channels or from Ca²⁺ entry through Ca²⁺ channels opened by backpropagating action potentials. Pharmacological experiments (Nakamura et al. 1999, 2000) established that these increases are due to Ca²⁺ released from intracellular stores. The primary mechanism for opening these stores is activation of inositol 1,4,5-trisphosphate (IP₃) receptors by mGluR (metabotropic glutamate receptor)-mobilized IP₃. In some cases, this release is enhanced by coactivation of the IP₃ receptors by Ca²⁺ entering through voltage-gated Ca²⁺ channels (Nakamura et al. 1999) or NMDA receptors (T. Nakamura, N. Lasser-Ross, K. Nakamura, and W. N. Ross, unpublished details). The large magnitude of these [Ca²⁺]_i increases and their location near the soma suggest that these waves might trigger critical signaling events in pyramidal neurons as has been suggested for Ca²⁺ waves in other cell types (e.g., Berridge 1998).

The requirement for repetitive stimulation to evoke this regenerative event is intriguing. It appears to resemble one protocol for evoking long-term potentiation (LTP) in these cells (Bliss and Lomo 1973). To establish a firmer foundation for this comparison, we decided to do a more systematic analysis of the conditions evoking Ca²⁺ release in these cells. We reasoned that the analysis could be made quantitative because release is generally an all-or-none event (although there are variations in amplitude and time course) and therefore can be scored unambiguously.

A critical analysis of the conditions necessary for evoking release also could contribute answers to two other issues. The first is a more detailed understanding of how synaptic activation causes Ca²⁺ release in these cells. The apparent requirement for repetitive stimulation at a certain rate suggests that a buildup and decay of a second messenger (probably IP₃) might be involved. The second relates to the question of why these waves were not seen until recently even though [Ca²⁺]_i changes have been measured in many experiments in hippocampal neurons for more than 10 yr, beginning with the first imaging paper by Regehr, Tank, and Connor (1989). An exploration of different stimulus protocols could establish whether there are special conditions required to observe synaptically activated Ca²⁺ release.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Preparation of hippocampal slices, patching of pyramidal neurons, synaptic stimulation, and fluorescence imaging of [Ca²⁺]_i changes were done similarly to previous experiments measuring Ca²⁺ release in our laboratory (Nakamura et al. 1999). Briefly, 14- to 18-day-old Sprague-Dawley rats were anesthetized with methoxyflurane before decapitation. Submerged and superfused transverse slices (300 µm thick) were mounted on a stage rigidly attached to an air table and were viewed with a ×40 water-immersion lens on an Olympus BX50WI microscope mounted on an X-Y translation stage. Somatic whole cell recordings were made using patch pipettes pulled from 1.5-mm-OD thick-walled glass tubing (1511-M, Friderick and Dimmock, Millville, NJ). Slices were maintained in artificial cerebral spinal fluid (ACSF) consisting of (in mM) 124 NaCl, 2.5 KCl, 2 CaCl₂, 2 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 or 20 glucose, bubbled with a mixture of 95% O₂-5% CO₂, making the final pH 7.4. The same solution superfused the slices during the experiments. Pyramidal neurons were patched under visual control using the "blow-and-seal" technique (Sakmann and Stuart 1995). The pipette solution consisted of (in mM) 140 K-gluconate, 4 NaCl, 4 Mg-ATP, 0.3 Na-GTP, and 10 HEPES, pH adjusted to 7.2-7.4 with KOH, supplemented with 200-300 µM bis-fura-2 (Molecular Probes, Eugene, OR). Time-dependent [Ca²⁺]_i measurements from different regions of the pyramidal neuron were made as previously described (Lasser-Ross et al. 1991; Nakamura et al. 1999). Electrical traces were recorded simultaneously and matched to the optical recordings. Data were taken and analyzed with Windows-based software written in our laboratory. Images were taken at 33-ms intervals. Electrical records were sampled at 200-µs intervals.

Because the main focus of these experiments concerned variations in the parameters of synaptic stimulation, particular attention was paid to the construction and placement of the stimulating electrode. We used a bipolar tungsten electrode that had one tip (WPI, model TM33B01KT) about 1 mm in front of the other. The tip of this electrode was sharpened to a point and was pressed on the surface of the slice near the patched pyramidal neuron. The tip could be placed repetitively in the same position relative to the soma in different experiments to an accuracy of about ±3 µm. Stimulating pulses had a duration of 100 µs. The number, interval, and intensity of these pulses were varied during the experiments. Trials were separated by 1-2 min.

For some experiments, QX-314 (0.25 mM; lidocaine n-ethyl bromide) was added to the pipette solution to prevent action potential generation by the synaptic potentials. QX-314, K-gluconate, Mg-ATP, and Na-GTP were purchased from Sigma Chemical. All other reagents were from Fisher Scientific.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Stimulation with different frequency and pulse protocols

Figure 1 shows a typical experiment where repetitive synaptic stimulation evoked an all-or-none Ca²⁺ wave in the apical dendrites of a pyramidal neuron. The time-dependent fluorescence traces (right) show increases in [Ca²⁺]_i in different regions of interest that began with a delay following the beginning of synaptic stimulation and that occurred at different times in nearby regions (Nakamura et al. 1999). In this trial, we gave a train of 15 stimuli at 100 Hz. The stimulation current was 100 µA for 100 µs. The tip of the stimulating electrode was placed 20 µm to the side of the pyramidal neuron and 40 µm from the center of the soma. The lateral distance was measured relative to an axis perpendicular to the cell body layer and was established before patching the soma. Because the exact configuration of the dendrites varied from cell to cell and was not known in advance, the exact distance of the electrode from individual oblique processes or the dendritic shaft differed among experiments.

View larger version (32K): [in this window] [in a new window]   Fig. 1. Repetitive synaptic activation evokes all-or-none regenerative Ca2+ waves in the apical dendrites of pyramidal neurons. The cell image shows the patch pipette on the bis-fura-2-filled neuron and the point of the stimulating about 20 µm lateral from the main dendrite and 40 µm distal from the soma. Stimulation with 15 pulses at 100 Hz () using 95-µA current evoked only a small [Ca2+]i increase in the 3 regions of interest (ROIs). Increasing the current to 100 µA evoked large [Ca2+]i increases that began and peaked at different times in the 3 ROIs.

To determine the threshold, we lowered the stimulation current until the same train failed to evoke a wave (Fig. 1, middle). In most cells, we could repeat the process and get more than one value for the threshold that were then averaged together to estimate the threshold. In some cells, the ability to evoke a wave gradually deteriorated after many trials and even substantially higher currents failed to generate a regenerative event later in the experiment. For these cells, we used the earliest determined threshold current.

Although there was clearly some arbitrariness in determining the position of the electrode and the estimation of the threshold current, we tried to overcome this problem by repeating the experiments many times on different cells using the same standards for estimating these parameters. For most cells, we tested the threshold current for only one set of stimulation parameters. In some cells, we tested two protocols, returning to the first configuration after determining the threshold for the second. As shown previously (Nakamura et al. 1999), Ca²⁺ waves often could be evoked many times in the same cell. In the example shown in Fig. 1, a stimulation current of 100 µA evoked Ca²⁺ release but 95 µA did not. This sequence was tested twice. From these data, we estimated a threshold current of 100 µA for this cell.

In many of these experiments, especially those that required higher currents, the stimulation protocols evoked action potentials. Because backpropagating action potentials can synergistically combine with tetanic stimulation to enhance the probability of release (Nakamura et al. 1999), we were concerned that thresholds measured when spikes occurred in the synaptic responses would be different from thresholds determined when spikes did not occur. To control for this possibility, we tested the thresholds in some cells with 0.25 mM QX-314 included in the pipette to prevent action potentials. Using a protocol of 25 pulses at 50 Hz, we found a threshold current of 137 ± 16 µA (SD; n = 5). When QX-314 was not included in the pipette, the threshold current was 113 ± 27 µA (n = 5). These thresholds are not significantly different (P > 0.2, t-test). The most likely reason that spikes did not affect release probabilities in these experiments is that the spikes almost always occurred at the beginning of the train causing voltage-gated Ca²⁺ entry to occur too early to combine with the mGluR-mediated release of IP₃ to enhance Ca²⁺ release (Nakamura et al. 1999). Consequently, we ignored the occurrence of spikes in the synaptic response in determining thresholds if the spikes occurred early in the train.

Threshold estimates from many cells with the electrode positioned 20 µm to the side of the main apical dendrite are summarized in the histogram in Fig. 2. Two trends are clear. First, using the same frequency of stimulation, higher currents were required to evoke release if there were fewer pulses. Second, using the same number of pulses, higher currents were required for lower stimulation frequencies. Release could be evoked reliably with all the combinations of parameters shown in this histogram if the stimulation current was sufficiently high. Experiments using stimulation parameters outside of this range were less successful. Only 3/7 cells tested with 3 pulses at 100 Hz evoked release with any current up to 1,000 µA and only 1/5 cells tested at 5 Hz using 25 pulses showed regenerative release.

View larger version (40K): [in this window] [in a new window]   Fig. 2. Required threshold current increases with decreasing stimulation frequency or decreasing number of pulses in the train. The histogram shows the mean threshold current in each condition. The error bars represent SE. The numbers of cells tested in each configuration were: 100 Hz (50 pulses, 5; 25 pulses, 6; 15 pulses, 9; 10 pulses, 10; 5 pulses, 5); 50 Hz (4; 5; 7; 6; 4); 20 Hz (5; 6; 8; 5; 4).

In addition to these standard protocols for tetanic stimulation, we tested theta-burst stimulation, another common protocol used to evoke LTP in pyramidal neurons (Larson et al. 1986). In all cells tested with 10 bursts of four pulses at 100 Hz, with bursts separated by 200 ms, we evoked regenerative release. Figure 3 shows a typical example of this kind of experiment. In this figure, the Ca²⁺ wave is shown as a "line scan" along the main apical dendrite of the pyramidal neuron (Nakamura et al. 2000). The mean threshold current was 152 ± 12 (SE) µA (n = 9). This current is near the center of the range of thresholds found with tetanic stimulation (Fig. 2).

View larger version (89K): [in this window] [in a new window]   Fig. 3. Theta burst stimulation evokes Ca2+ waves in pyramidal cell dendrites. The cell image shows a patch pipette on the bis-fura-2-filled soma and the tip of the stimulating electrode. The string of small black boxes ("superpixels") indicates the positions of the "line scan" representation of the Ca2+ wave (right). In this representation, the abscissa of the image corresponds to the time axis of the electrical trace below. The ordinate corresponds to the positions of the superpixels in the cell image. The amplitude of the [Ca2+]i increases at each time and position is represented on a gray scale. The scale (black to white) corresponds to 0-30% F/F. The image shows that the Ca2+ wave began at a position above the electrode and spread in both directions along the apical shaft. Stimulation was 10 repetitions of 4 stimuli at 100 Hz; each repetition was separated by 200 ms. Stimulation current was 140 µA. In this trial, the wave did not begin until more than 1 s after the beginning of the train.

Position of stimulating electrode

In all of these experiments, the sharpened point of the stimulating electrode was placed laterally about 20 µm from the main apical dendritic shaft and 40 µm distally from the cell body. The lateral distance is closer than in most hippocampal slice experiments. In addition, the use of an electrode with a sharpened point placed in front of the second wire is different from the typical symmetric pair of blunt wires. With our electrodes, most of the current density is along the axis of the electrode. This configuration probably activates a narrower strip of presynaptic fibers than the symmetric electrode. To test the significance of these differences, we measured the threshold for regenerative Ca²⁺ release when the electrode was positioned at various distances from the apical shaft. We used the protocol of 50 pulses at 100 Hz that was very effective at 20 µm. Figure 4 shows that the required current increased significantly when the electrode was positioned 70 and 120 µm from the dendrite. In additional experiments, we used a standard bipolar electrode 500 µm from the dendrite. The required current in this case was even higher, almost 10 times the current needed with a sharp electrode 20 µm from the dendritic shaft. The success rate declined with distance. We could always find a threshold current when a sharp electrode was positioned 20, 70, or 120 µm from the shaft. But only 3/10 cells demonstrated regenerative release with the bipolar electrode 500 µm from the cell. In another series of experiments, some cells were tested with 100 pulses at 100 Hz at different distances from the dendrites. Consistent with the data in Figs. 2 and 4, the thresholds measured in these experiments were slightly lower than the thresholds measured with 50 pulses but increased with distance (data not shown).

View larger version (10K): [in this window] [in a new window]   Fig. 4. Required threshold current increases with distance of the stimulating electrode from the cell dendrite. The histogram shows the mean threshold current at each distance. In all trials the stimulation was 50 pulses at 100 Hz. For 20, 70, and 120 µm, the stimulating electrode was a sharpened asymmetrical bipolar electrode. For 500 µm, the electrode was a symmetrical blunt bipolar electrode. The error bars represent SE. The numbers of cells tested in each configuration were: 20 µm, 5; 70 µm, 5; 120 µm, 5; 500 µm, 10.

Some of the required increase in threshold current when the electrode was further from the dendrite was probably due to the decreasing probability that stimulated Schaffer collateral axons contacted the tested pyramidal neuron. We tried to control for this parameter by measuring the peak synaptic potential at each distance. However, the relative contribution of inhibitory responses varied from experiment to experiment. Because we could not block inhibition without altering the conditions of the experiment, we did not check this hypothesis.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of these experiments are straightforward. Synaptic activation of regenerative Ca²⁺ release in pyramidal neurons required a minimum of three to five stimuli at 100 Hz. Increasing the number of pulses at this frequency reduced the required stimulation current. Release could also be generated reliably if the stimulation frequency was reduced. However, either increasing the stimulation current or increasing the number of pulses in the train had to compensate for this frequency reduction. The effective frequency could not be reduced to 5 Hz when the other parameters in our protocol were kept within reasonable limits.

These results are valid using a sharp pointed tungsten stimulating electrode positioned about 20 µm lateral to the proximal apical dendrites. Moving the stimulating electrode more laterally (170 µm from the dendrites) made it more difficult to evoke release. Stimulation at 500 µm with a standard bipolar electrode was rarely effective.

In previous experiments (Nakamura et al. 1999), we established that neurotransmitter was released by stimulating action potentials in presynaptic fibers. In addition, we found that the threshold current for evoking Ca²⁺ release was always higher than the threshold current required for generating an excitatory postsynaptic potential (EPSP). These results show that many presynaptic fibers must be activated together to evoke release (cooperativity). Otmahkov et al. (1993) estimated that 19-26 fibers must be activated together to reach action potential threshold. This number, therefore is an upper limit for the minimum number of fibers required to evoke regenerative Ca²⁺ release. Together with the results concerning frequency and number of pulses this cooperativity requirement restricts the kinds of models that can explain regenerative Ca²⁺ release.

While we were able to evoke Ca²⁺ release with just 3-5 stimuli in some experiments, reliable release only was observed with 10 or more stimuli. This result suggests that some quantity is accumulating to a threshold level during the train to initiate regenerative Ca²⁺ release. One possibility is an increase in membrane potential. However, voltage cannot be the essential parameter because release was observed even when the membrane hyperpolarized following stimulation in the presence of APV and 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX) (Nakamura et al. 1999). Although an increase in membrane potential (Nakamura et al. 1999, 2000) can enhance the generation of calcium waves, the importance of EPSPs or EPSP-evoked spikes is likely to be minimal in these experiments because these depolarizations occurred at the beginning of the synaptic train. Blocking the early spikes with QX-314 did not affect the threshold current. Ca²⁺ entry through the NMDA receptor (or some factor produced by NMDA receptor activation) cannot be critical because release occurs when these receptors are blocked. Activation of these receptors can influence the generation of calcium waves (Nakamura et al., unpublished data), but the magnitude of this effect has not been determined. A third possibility is that repetitive stimulation is necessary to produce enough glutamate to activate mGluRs that are positioned perisynaptically on the dendritic spine (Lujan et al. 1996).

The most reasonable candidate for an accumulating, fundamental quantity is IP₃ because generation of this molecule is required for Ca²⁺ release in these cells (Nakamura et al. 1999). IP₃ has a measured lifetime in some cells of less than 10 s (Wang et al. 1995), which certainly permits its accumulation during a train. The lifetime in pyramidal cell dendrites actually may be <1 s because action potentials must occur within 1 s of synaptic stimulation to synergistically enhance release (Nakamura et al. 1999). This short lifetime is consistent with our experiments since increasing the duration of stimulation beyond 0.5 s had little effect on the threshold current (Fig. 2).

In Xenopus oocytes, where regenerative, IP₃-mediated Ca²⁺ release has been studied extensively (e.g., Lechleiter et al. 1991; Marchant et al. 1999), regenerative waves result from the coalescing of nearby, localized release events ("puffs"), each initiated by IP₃ interacting with a cluster of IP₃ receptors. Our results are consistent with this model of Ca²⁺ release. In this interpretation, multiple stimuli at high frequency are needed to produce enough IP₃ to trigger release from IP₃ receptors. Stimulation close to the dendrite might be necessary to ensure that the activated IP₃ receptors are close enough to each other to generate positive feedback from the release of Ca²⁺ from one cluster to another. Stimulation of multiple presynaptic fibers might be necessary to activate enough IP₃ clusters within the activated region if each synapse can produce only a limited amount of postsynaptic IP₃. Further experiments with more spatial precision are needed to test this model in more detail.

Comparison with stimulus parameters used to evoke LTP

Superficially, the stimulation protocols that effectively evoke Ca²⁺ release resemble those used to evoke LTP in pyramidal neurons (Madison et al. 1991). Tetanic and theta-burst stimulation, as used in our experiments, are classic techniques for evoking LTP. In the original experiments of Bliss and Lomo (1973) stimulation was continued for several seconds using frequencies of 10-100 Hz, but later experimenters usually evoked LTP at these frequencies with trains lasting less than 1 s. Stimulation at frequencies lower than 5 Hz often leads to long-term depression (LTD) instead of LTP (e.g., Dunwiddie and Lynch 1978). This frequency also was the lower limit for generating release in our tests. We had a few successes in evoking Ca²⁺ release with only three pulses at 100 Hz. But even single stimuli at high strength have been shown to be effective in evoking LTP in some experiments (Abraham et al. 1986). Therefore the range of effective frequencies and train durations are comparable for the two processes.

It is hard to compare the intensities used in experiments causing Ca²⁺ release and in experiments evoking LTP because the kind of electrode, its position, and the condition of the slice all can affect the required stimulus intensity for either kind of experiment. However, our observation that Ca²⁺ release usually can be evoked with intensities subthreshold for spike generation suggests that the intensity range for the two classes of experiments are not significantly different.

Generation of LTP (Lee 1983) and generation of Ca²⁺ waves both require stimulation intensities significantly higher than needed to generate a threshold EPSP. This "cooperativity" requirement for LTP is usually interpreted as necessary to make the EPSP large enough to remove the voltage-dependent Mg²⁺ block of NMDA receptors (e.g., Collingridge et al. 1988). This cannot be the explanation for the cooperativity requirement for Ca²⁺ release because release occurs even when the pyramidal neurons hyperpolarize during synaptic activation (Nakamura et al. 1999). As suggested in the preceding text, activation of many synapses may be needed to mobilize IP₃ in a region larger than a single spine in order achieve regenerative activation of IP₃ receptors.

One aspect of our experiments that suggests that there are some differences between the mechanisms generating LTP and the mechanisms causing Ca²⁺ release is the dependence on the location of the stimulus electrode. We found that it became significantly harder to evoke regenerative Ca²⁺ release as the electrode was moved away from the pyramidal cell dendrites. In contrast, LTP can be generated easily in most hippocampal slice experiments by positioning the bipolar stimulating electrode almost anywhere in the stratum radiatum. In one series of experiments using glass stimulating pipettes, Kauer (1999) did find a strong dependence on electrode position. However, somewhat counter intuitively, she found that stimulation close to the recorded cell produced little or no LTP but robust LTP when the electrode was further away.

A second important difference between the requirements for generating Ca²⁺ release and the requirements for evoking LTP in the CA1 region is that NMDA receptor activation is required for LTP (Bliss and Collingridge 1993) while NMDA receptor activation is not required for Ca²⁺ release (Nakamura et al. 1999). Interestingly, activation of NMDA receptors makes it easier to evoke Ca²⁺ release (Nakamura et al., unpublished observations). Whether this synergistic role of NMDA receptor activation is related to LTP induction is not known.

In summary, while the similarities between the induction of LTP and the generation of Ca²⁺ waves are intriguing, it is clear that they are not identical. It is possible that a closer correlation would be observed if the conditions for LTP induction were compared with the conditions evoking the elementary events underlying the waves. These elementary events have not been observed yet in neurons in slices. It also is possible that Ca²⁺ waves contribute to the known requirement for a postsynaptic [Ca²⁺]_i increase in the induction process (Lynch et al. 1983), but they cannot be the entire triggering mechanism.

Why synaptically activated waves have been difficult to observe

Until our recent experiments (Nakamura et al. 1999), synaptically activated Ca²⁺ waves were not observed even though postsynaptic [Ca²⁺]_i changes in CA1 pyramidal neurons have been measured for more than 10 yr (Regehr et al. 1989) and waves have been described often in other preparations for almost as long (e.g., Lechleiter et al. 1991). Jaffe and Brown (1994) observed waves in pyramidal cells following focal application of t-ACPD, but their observations were never followed up. Pozzo-Miller et al. (1996) detected Ca²⁺ release from intracellular stores in CA3 pyramidal neurons but found that the stores were "either insufficiently filled to provide appreciable Ca²⁺ increases, or are inaccessible to triggering events at the synapse."

In retrospect it seems that there are five possible reasons why regenerative events were not seen easily in previous experiments. First, early experiments (e.g., Miyakawa et al. 1992; Regehr et al. 1989) usually used high concentrations of fura-2 in the pipette with the goal of detecting [Ca²⁺]_i changes in all parts of the dendritic arbor including the distal branches. While it was appreciated that this high indicator concentration would buffer and blunt the transients from voltage-dependent Ca²⁺ entry, it was not known that fura-2 concentrations above 0.5 mM would completely block Ca²⁺ waves (Nakamura et al. 1999). Second, many early experiments (e.g., Regehr et al. 1989) detected dendritic Ca²⁺ transients with low time resolution (typically 1-s frame intervals), high spatial resolution recordings that might have missed these events or failed to distinguish them from transients due to bursts of backpropagating action potentials. Third, most experiments stimulated the cells with bipolar electrodes positioned relatively far from the examined cell. Our experiments (Fig. 4) show that positioning the stimulating electrode far from the cell makes it difficult to evoke waves. Fourth, experiments using confocal microscopy (both single and 2 photon) centered attention on the region around spines on oblique dendrites. However, the Ca²⁺ waves do not propagate out to the oblique branches (Nakamura et al. 1999, unpublished results). Finally, except for a few early recordings (e.g., Miyakawa et al. 1992; Regehr et al. 1989) most recent experiments (e.g., Koester and Sakmann 1998; Kovalchuk et al. 2000; Yuste and Denk 1995) have described [Ca²⁺]_i changes resulting from single or only a few synaptic stimuli when many stimuli at relatively high frequency are usually required to evoke Ca²⁺ release (Fig. 2). Although we have been able to observe these release events with almost 100% reliability, it required a select combination of stimulation parameters. In contrast, we observed [Ca²⁺]_i changes resulting from electrical events with almost any combination of stimulation parameters.