Control of Secretion by Temporal Patterns of Action Potentials in Adrenal Chromaffin Cells

Sensitivity of ion channels to changes of APW shape

Currently, C_m can only be measured under voltage clamp. To apply membrane C_m measurement in AP-induced cell secretion, both current clamp (for APs) and voltage clamp (for C_m) are needed. However, the transfer between voltage clamp and current clamp takes time. Thus, currently no amplifier can monitor real AP-induced secretion by C_m measurement. To avoid using current clamp, the cell is stimulated by a short step pulse or by a mock AP, which consists of several ramps, but a mock AP cannot mimic a real AP precisely. Thus, it is necessary to examine how sensitive the ion channels are to changes in the AP shape. The experiments to test the sensitivities of voltage-gated Na⁺, K⁺, and Ca²⁺ channels are shown in Figure 2. A typical APW was imitated by three ramps. Whereas the first ramp was a slow “foot,” the second and third ramps generated the rapidly rising phase (0.9 msec duration) and the decay phase (4 msec duration), respectively. The peak of fast inactivating inward current was measured as peak of Na⁺ current. The Ca²⁺ influx was measured by the integral of Ca²⁺ current. In Figure 2A, the duration of the second (rising) ramp was changed from 0.4 to 1.4 msec in 0.5 msec steps (Fig. 2A2). The Na⁺ currents (the fast inactivating inward currents in Fig. 2A1) and the Ca²⁺ currents (Fig. 2A3) induced by the mock AP were recorded. Relative to the intermediate ramp duration, whereas Ca²⁺ currents were not sensitive to the slope change in the rising phase (Fig. 2B2), the Na⁺ current was altered dramatically by a factor of 2 (Fig. 2B1). The K⁺ currents were altered as well (part of the tail currents in Fig. 2A1). The duration of the third ramp of negative slope was changed from 2 to 6 msec in 2 msec steps (Fig. 2C2). The Na⁺, K⁺ (Fig. 2C1), and Ca²⁺ currents (Fig. 2C3) induced by the mock AP were recorded. Whereas the Na⁺ current was not sensitive to the slope change in the decay phase, the net ion flux during the tail currents (mixing currents from Na⁺, K⁺, and Ca²⁺ channels) was altered dramatically, by fourfold (Fig. 2D1). These tail currents in Figure 2, B1 and D1, were reproducible, regardless of whether the ramp slopes were changed either from 2 to 6 msec or from 6 to 2 msec. The Ca²⁺ influx, measured by the integral of Ca²⁺ currents shown in Figure 2C3, was significantly altered by 35 ± 4% (decrease) and 28 ± 5% (increase) compared with current at intermediate ramp duration (Fig. 2D2). These experiments suggest that voltage-gated ion channels are very sensitive to minor changes in AP shape. Thus, compared with a real AP, a step pulse or mock AP may cause large differences in currents and secretion. In contrast, an AP template should be a more physiological stimulus, because it was prerecorded from a typical RACC. For a given cell, stimulation by the reconstituted AP code under whole-cell voltage clamp is similar to that of field stimulation applied to the cell under current clamp. Together, it is better to use a real AP (not a mock AP) for the quantitative study of AP-secretion coupling.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Voltage-sensitive channels are sensitive to slight changes in the slope of a mock AP. A, An AP waveform was simulated by three ramps as start potential, end potential, and duration as follows, respectively: -80 mV, -35 mV, 11.7 msec; -35 mV, 48 mV, 0.9 msec; and 48 mV, -70 mV, 4 msec. To test the slope sensitivity, the second ramp duration (0.9 msec) was changed by ±0.5 msec (A2). The corresponding currents induced by these three mock APs were recorded in either physiological solution (high-NaCl bath and high-KCl pipette for Na⁺ currents; A1) or solution for Ca²⁺ currents (200 nm TTX in bath and high-CsCl pipette; A3). All traces in A1 were from one cell. All traces in A3 were from another cell. B, With increased ramp duration, the fast inward Na⁺ current changed by 48 ± 17 and 159 ± 19% (5 cells; B1). The tail currents in A1, induced by the third ramp of the mock AP, were altered as well, but the mock AP-induced Ca²⁺ influx was not sensitive to this slope change (6 cells; B2). C, To test the slope sensitivity of the tail ramp, the third ramp duration (4 msec) was changed by ±2 msec (C2). The corresponding currents by these three mock APs were recorded in either physiological solution (C1) or solution for Ca²⁺ currents (C3). All traces in C1 were from one cell. All traces in C3 were from another cell. D, With increased ramp duration, the tail ion fluxes-tail currents changed from negative (influx) -618 ± 243 to 548 ± 319 and 1889 ± 383 fC, respectively (6 cells; D1). As a component of the tail ion flux, the Ca²⁺ influx also changed by 66 ± 4 and 128 ± 5%, respectively (6 cells; D2).

AP code and cell secretion monitored by C_m measurement

Because voltage-gated currents are highly sensitive to the AP shape (Fig. 2), the mock AP, which cannot match the AP shape precisely, could cause unpredictable changes in stimulus-secretion coupling. Thus, we developed a method for AP-induced secretion by using real APW as the basic stimulus unit under voltage-clamp conditions.

A single APW template was prerecorded in an RACC under current-clamp mode (Fig. 3A). An AP train with defined time intervals between individual APs is called an AP code, defined with the general “code function” F(n, m, f, d), where the four coding parameters (n, m, f, d) are defined as follows (Fig. 3C). A series of consecutive APs is defined as a “burst”: n is the number of APs in a burst, m is the number of bursts in the whole AP pattern, f is the frequency of APs in a burst, and d is the interval between two adjacent bursts (when m = 1, d has no meaning). Clearly, this equation defines only a simple kind of AP code. However, a more complex AP code can be built with several simple AP codes from this equation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Defining protocols for measuring secretion in conjunction with stimulation by complex patterns of naturally occurring AP waveforms. A, Example of an AP elicited in an RACC during constant current injection (20 pA with a holding current of -10 pA). B, Sustained current injection results in repetitive firing in many RACCs. Bottom, The instantaneous frequency for each pair of APs in such bursts is plotted for a set of 20 cells. C, A prerecorded AP (as in A) was used to construct burst patterns as defined in Results. Sequences of AP waveforms were linked to the 1000 Hz sine wave illustrated on the right, and the precise timing between sine wave application and AP waveforms was defined as described in Materials and Methods. Bottom, An example of four bursts of APs, each containing 10 APs occurring at 10 Hz, which were used to stimulate a RACC resulting in the C_m trace shown below. With perforated-patch recording, a decrement in C_m indicative of endocytosis was typically observed after stimulation.

A prerecorded AP from a current-clamp recording in an RACC (Fig. 3B) was used as the elementary APW template and programmed with Igor software. This APW template (Fig. 3A) was used to produce an AP code F(n, m, f, d) with a laboratory-made macro program in Igor. A macro procedure from HEKA Elektronik was used to transfer the protocol to a file template of Pulse software. To get defined time intervals between APs and AP bursts, we made the whole F(n, m, f, d) as a single AP code template file for Pulse. As required by C_m measurement, a sine wave of 40 mV (peak-to-peak) and 1 kHz was inserted throughout the protocol, except during AP bursts. Using this AP-C_m protocol, the AP-induced C_m signals are recorded when AP code F(10, 4, 10 Hz, 5 sec) was applied to an RACC (Fig. 3C).

Frequency dependence of secretion (the f effect)

Chromaffin cells fire APs during stimulation from cholinergic neurons and release catecholamine (Kidokoro and Ritchie, 1980). To test the range of AP frequency in native firing, a current step (1 sec, to 20 pA) was injected to an RACC to evoke APs. By measuring the interval between two adjacent APs, frequency was calculated. The distribution of AP frequencies is shown in Figure 3B. Because 80% of firing occurs at frequencies lower than 20 Hz, only frequencies below that value were used in this work. Using amperometric measurements, Zhou and Misler (1995) reported that, with a given AP number, an increase of AP frequency from 1 to 4 Hz increases the amount of secretion by twofold in RACCs. As shown in Figure 4A, using membrane capacitance measurements, we confirmed that secretion increased when AP frequency f was increased from 1 to 7 Hz. However, we found no further increase at higher frequencies. To investigate the f saturation, the intracellular calcium concentration ([Ca²⁺]_i) was monitored by combined patch-clamp and fura-2 fluorescence. In the calcium imaging experiments (Fig. 4B), a given number (n = 20) of APs was applied to a cell in one burst but at different frequencies. The peak [Ca²⁺]_i change induced by the AP trains increased from f = 2 Hz to f = 7 Hz and nearly saturated after f ≥ 7 Hz (Fig. 4C). This result was consistent with the secretion result (Fig. 4A). The saturation of the [Ca²⁺]_i change was not attributable to saturation of the fura-2 dye, because [Ca²⁺]_i could be further increased with a stronger pulse stimulus. During stimulation with APs, calcium ions passed through calcium channels and accumulated in the cytoplasm. At lower frequencies such as 1 or 2 Hz, the accumulation was weak because of the activity of Ca²⁺ clearance mechanisms. Only the vesicles near the Ca²⁺ channel could be released. At higher frequencies, the residual Ca²⁺ could accumulate. This allowed the release of vesicles far from the calcium channels. When frequency was higher than 7 Hz, there was no further accumulation, probably because the clearance effect is negligible during the short interval between two adjacent APs. The similar f effects for both secretion and [Ca²⁺]_i, particularly the similar saturation of the f effect at f > 7 Hz, is consistent with the hypothesis that AP-induced secretion is exclusively Ca²⁺ dependent in chromaffin cells (Augustine and Neher, 1992; Zhang and Zhou, 2002).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Influence of f, action potential frequency, and n, number of action potentials in a burst, on secretion. A, Effect of frequency on secretion. Twenty APs were applied to a cell at different frequencies. Secretion was normalized to that evoked by a 10 Hz stimulus (at 10 Hz, ΔC_m = 28 ± 3 fF). Normalized ΔC_m was plotted for a set of seven cells and shows that secretion increased as f increased from 1 to 7 Hz. Additional increases in frequency did not evoke more secretion. For f ≥7 Hz, the interval between adjacent APs was too short to measure C_m, so it was measured after the 20 APs. For f = 1, 2, and 4 Hz, to eliminate possible effects of endocytosis during the intervals between adjacent APs, total secretion (C_m) was taken as the sum of the 20 individual AP-induced C_m. B, Effect of frequency on [Ca²⁺]_i. Twenty APs were applied to cells at different frequencies. [Ca²⁺]_i increased when f increased from 2 to 7 Hz. However, the increase in [Ca²⁺]_i saturated at frequencies higher than 7 Hz. The [Ca²⁺]_i saturation at f ≥7 Hz was not attributable to dye saturation, because stronger stimulation (40 APs applied at 10 Hz in a single burst; SS) could increase [Ca²⁺]_i further. C, Statistics of B. [Ca²⁺]_i is nearly saturated for f ≥7 Hz (6 cells). D, Effect on secretion of bursts of different numbers of APs, all applied at 10 Hz. For n = 30, ΔC_m = 28 ± 3 fF. The relationship between total ΔC_m and the number of APs is plotted for eight cells.

The relationship between secretion and number of APs (n effect)

Because the effect of frequency saturated when f was higher than 7 Hz, an AP burst of 10 Hz was used in the following experiment to investigate the effect of another coding parameter, “n,” on RACC. As shown in Figure 4D, the amount of secretion increased when the number of APs increased, but the relationship between secretion and n was not linear. It can be fitted by a single-exponential function. The rundown of AP efficiency could be for two possible reasons. One is that the Ca²⁺ influx evoked by each AP decreases during stimulation. In the latter part of this paper, the rundown of Ca²⁺ influx is described (Fig. 5C). This suggests that rundown of Ca²⁺ influx might participate in this process. Another possible reason is a limit defined by the size of the readily releasable pool (RRP) of vesicles. In this case, the calcium ions entering with later APs have less chance to trigger release of vesicles (see Discussion). Note that, at n = 40 APs, C_m was significantly larger than the predicted plot value (p < 0.01; eight cells). This is probably because a facilitation of secretion starts after 30 APs at f = 10 Hz (Zhou and Misler, 1995).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Modulation of secretion and [Ca²⁺]_i by number of AP bursts, m. A, Secretion induced by one- or four-burst stimuli. Secretion was measured as ΔC_m (change of C_m after AP code stimulation). The apparent ΔC_m was twofold larger when the same total number of APs (n = 40) was presented in four bursts separated by 4 sec. To compare the AP code-induced secretion signals, segments of the C_m before stimulation were placed on the same baseline. The ΔC_m evoked by each burst was added to obtain total secretion evoked by the protocol. B, Effect of the two protocols in A on changes in cytosolic [Ca²⁺]_i. In this case, the single burst of 40 APs resulted in a larger apparent increase in peak [Ca²⁺]_i. Similar results were obtained in 17 cells. C, The net Ca²⁺ influx per AP plotted for the two protocols illustrated in A. In one case, a continuous burst of 40 APs was applied at 10 Hz, or AP code F(40, 1, 10 Hz, d). In the other, four bursts each of 10 APs at 10 Hz were applied at an interburst interval of 4 sec, or AP code F(10, 4, 10 Hz, 4 sec). For continuous stimulation, a gradual decrease in net Ca²⁺ influx was observed. In the protocol with m = 4, net Ca²⁺ influx recovered between bursts. Thus, 40 APs applied as four separate bursts elicited more Ca²⁺ influx than 40 APs applied at m = 1. D, The statistical results of m = 1 versus m = 4 for 12 cells. For m = 1, total ΔC_m = 34 ± 9 fF. The four-burst stimulus evoked 2.2 ± 0.1 times more secretion (p < 0.01).

Parameters of inter-AP bursts modulated secretion dramatically (m and d effects)

Parameters m and d define the inter-AP bursts of an AP code. To test the effect of number of bursts or m, the other parameters were set as n = 40/m, f = 10 Hz, and d = 4 sec. The total number of APs in the AP code was fixed at 40 = n^*m. The result of m = 1 versus m = 4 is shown in Figure 5A. The left and right panels show control and recovery of C_m induced by F(40, 1, 10 Hz, 4 sec). The middle panel shows the C_m trace induced by F(10, 4, 10 Hz, 4 sec). These results are summarized in Figure 5D. Although the total number of APs and AP frequency were fixed (n^*m = 40; f = 10 Hz), the four-burst AP code evoked secretion 2.2 ± 0.1 times greater than that caused by a single burst (p < 0.01). The [Ca²⁺]_i signals responsible for these differences in secretion are illustrated in Figure 5B. The single-burst stimulus evoked a much higher but narrower Ca²⁺ spike than the four-burst stimulus. A stronger stimulus (80 APs applied at 20 Hz in a single burst) was applied at the end of the experiment to make sure that there was no saturation of the dye. In the [Ca²⁺]_i spike evoked by the four-burst stimulus, the peaks induced by each burst could be distinguished. Although these peaks were smaller than the maximum [Ca²⁺]_i, the AP code of m = 4 evoked more secretion than the AP code of m = 1. The m effect was further confirmed by m = 2 versus m = 4. To make error induced by endocytosis minimal, the total stimulus duration was fixed. The protocols F(10, 4, 10 Hz, 1.36 sec) and F(20, 2, 10 Hz, 4 sec) were applied. The secretion evoked by F(10, 4, 10 Hz, 1.36 sec) was 1.3 ± 0.1 times larger than that evoked by F(20, 2, 10 Hz, 4 sec) (data not shown; 12 cells).

As shown in Figure 5C, the calcium current caused by each AP was counted and plotted against its ordinal number. A clear rundown of calcium current was observed (Fig. 6A). During the interburst interval, the Ca²⁺ channels recovered, so that the four-burst stimulus could induce more Ca²⁺ influx. On the other hand, it is likely that the RRP would also recover during the interval between bursts. Thus, both of these factors might contribute to the larger response elicited by the four-burst protocol. This m effect could have important physiological consequences (see Discussion).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Effects of interval time d on [Ca²⁺]_i and secretion. A, To record Ca²⁺ currents, Na⁺ and K⁺ currents were blocked by TTX in the bath and high Cs⁺ in the patch pipette. The AP waveform shown in the top elicited the Ca²⁺ currents (1st and l0th AP in a burst) shown in the middle. The Ca²⁺ current was integrated to determine the total amount of Ca²⁺ influx (bottom). B, Effects of stimuli in which the interval time between bursts, d, was either 1 or 10 sec. ΔC_m in both cases was similar as summarized below. C, Effect of protocols with different d (1 or 10 sec) on changes in cytosolic [Ca²⁺]_i. The protocol with the shorter interval caused a higher peak of cytosolic [Ca²⁺]_i, even in the absence of more secretion (Fig. 5). Similar results were obtained in six cells. D, The statistical result of d = 1 sec versus d = 10 sec for 15 cells. For d = 1 sec, total ΔC_m = 65 ± 9 fF. The paired t test showed no difference.

To monitor pure calcium current and Ca influx during a single AP stimulus, TTX was added to the bath solution to block the sodium channels (Fig. 6A). With a given number of APs (i.e., n^*m) and a given stimulation time (time between first and last APs), d is a parameter depending on m. Thus, the m effect shown in Figure 5A should be considered as the combined effects of m and d. However, if the stimulation time is not fixed, d is independent of the other three coding parameters. When d is shortened to zero, several bursts (m > 1) will form one continuous burst (m = 1), which is treated in Figure 5A as the m effect. In the following, we only consider the situation in which d was longer than 1 sec. To test the d effect, the other parameters were set as n = 20, m = 3, and f = 10 Hz. As shown in Figure 6, B and C, two AP codes of d = 1 sec and d = 10 sec produced a similar ΔC_m but different [Ca²⁺]_i (Fig. 6D, 15 cells). The AP code of d = 1 sec caused a higher but briefer [Ca²⁺]_i spike. An intrinsic limitation of C_m measurement is that C_m is the net capacitance change resulting from total exocytosis and endocytosis. A decay phase is seen after the AP-induced C_m increase. This decay is most likely attributable to membrane retrieval or endocytosis. Thus, it is possible that endocytosis may occur simultaneously with exocytosis during the AP burst, and the apparent AP-induced C_m increases are probably underestimated.

Effect of different types of native AP bursts

Two typical subtypes of AP firing were encountered in cultured RACCs. Pattern 1, which accounted for ∼40% of RACCs, was firing with consecutive APs (Fig. 7A, inset). Pattern 2, which accounted for an additional 40% of RACCs, was firing in which one AP was followed by a prolonged plateau (Fig. 7B, inset). The remaining 20% had AP patterns intermediate between the two patterns. Each cell displayed a specific AP pattern in response to current injection (5-10 pA for a good seal), which was slightly above the threshold to generate APs. The AP pattern was relatively insensitive to injection currents of up to 10 times threshold. Using these two native bursts as templates, we built two types of AP codes with given m and d (m = 3; d = 1 sec). As shown in Figure 7, A and B, the secretion induced by the AP code of pattern 2 was much greater than that with the AP code of pattern 1. On average, the C_m induced by pattern 2 was at least 3.6 ± 0.5 times greater than that by pattern 1. This drastic difference in secretion was apparently attributable to Ca²⁺ influx during the long plateau of pattern 2 (Fig. 7B, inset, bottom trace), which was much larger than that of pattern 1 (Fig. 7A, inset, bottom trace). Statistically, the Ca²⁺ influx induced by pattern 2 was 4.5 ± 0.4 times larger than that of pattern 1. This experiment demonstrated that, like reconstituted AP codes, native AP codes (not just the number of APs and duration of AP bursts) strongly modulate cell secretion in RACCs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Modulation of secretion by more complex natural action potential waveforms. A, B, The bottom traces of the insets show two types of complex natural AP burst waveforms (patterns 1 and 2) observed in RACCs during 1 sec of 10 pA current injection. The top traces of each inset show the Ca²⁺ currents evoked by the application of such AP burst stimuli. In A, the AP burst of pattern 1 was used to build the AP code (m = 3, d = 1 sec), which was applied to the cell and induced a C_m increase of 45 fF. In B, the AP burst of pattern 2 was used to build the same AP code (m = 3, d = 1 sec), which was also applied to the same cell of A and induced a C_m increase of 180 fF, or four times larger than the AP code using pattern 1. C, The statistical results from six cells. Pattern 2 evoked 3.59 ± 0.50 times more secretion than pattern 1 (p < 0.01). For pattern 1, total ΔC_m = 79 ± 22 fF. D, Calculated calcium influxes during the stimuli. Note that pattern 2 evoked 4.52 ± 0.40 times more Ca²⁺ influx than pattern 1 (p < 0.01).

A capacitance method to study AP-induced secretion

The physiological stimuli in neurons are APs, which can be recorded by current clamp. Quantitative secretion can be recorded in real time by using the voltage-clamp method to measure C_m. In the AP-C_m method, we prerecorded a real AP or AP burst under current clamp. Then we used it as a template waveform to reconstitute a complex firing pattern (AP code). This computer generated AP code template was then applied to a cell in voltage-clamp mode. With a built-in sine wave stimulus, the secretion of the cell was recorded by the lock-in C_m measurement. This method allowed us to analyze the “AP information” contained in a given AP code, which can be “read-out” by recording the corresponding C_m of a cell.

The AP-C_m method has two drawbacks. First, the capacitance response is a mixture of exocytosis and endocytosis. Thus, the capacitance response induced by an AP code may be contaminated by endocytosis during AP stimulation. The endocytosis would cause underestimation of total secretion. Because endocytosis is coupled with exocytosis, this estimation error is larger for stronger stimulation (Neher and Zucker, 1993; Thomas et al., 1994). Second, the AP-C_m method can only be applied to cells that can be voltage clamped. This limitation sets requirements for the shape and size of a cell (Gillis, 1995). As shown in this report, RACCs meet these conditions.

Mechanisms for modulation of secretion by AP codes

The amount of AP code-induced secretion depends on the size of the RRP and its refilling rate after depletion. In the bovine chromaffin cell, RRP size is estimated to be between 20 and 45 fF (Gillis et al., 1996). In the rat chromaffin cell, an estimate of 34 fF has been made (Horrigan and Bookman, 1994). The maximum recovery rate of the RRP is 0.007 sec^-1 (Smith et al., 1998). In our experiments, a 10 AP burst (at 10 Hz) released half of the RRP. After secretion induced by an AP code, a proportion of the released vesicles will be recycled from the cell membrane and reloaded for the next release.

Depletion of the RRP during the AP burst provides a possible mechanism for the m effect: the AP code F(10, 4, 10 Hz, 4 sec) (m = 4; total APs = 40; f = 10 Hz; d = 4 sec) evokes twice the secretion induced by the AP code F(40, 1, 10 Hz, 4 sec) (m = 1; total APs = 40; f = 10 Hz; d = 4 sec). For m = 1, the long AP burst might deplete the RRP during the first 20 APs, so fewer vesicles are available for later APs. For m = 4, the RRP could be refilled during the intervals between two shorter AP bursts, so more RRP vesicles are available for the next short AP burst of the AP code.

An alternative mechanism for the m effect, as well as the f effect, could be the difference in Ca²⁺ influx generated by two AP codes. During a burst of APs, a decreased calcium current was induced by later APs attributable to inactivation of Ca²⁺ channels (Fig. 5C). During the interval between two bursts, the Ca²⁺ currents, including T-, N-, and L-types (Hollins and Ikeda, 1996; Prakriya and Lingle, 2000), exhibited recovery. Thus, the four-burst stimulus induced more Ca²⁺ influx.

Finally, the large dense-core vesicles are docked at a substantial distance (300 nm) from the site of calcium influx (Zhou and Misler, 1995; Klingauf and Neher, 1997; Elhamdani et al., 1998; Mansvelder and Kits, 2000). In the presence of mobile and immobile Ca²⁺ buffers, the local Ca²⁺ concentration decays sharply along the 300 nm diffusion pathway (Nowycky and Pinter, 1993; Zhou and Neher, 1993). The Ca²⁺ concentration needs to reach a threshold level before triggering secretion (Kao and Schneider, 1986; Neher, 1998). The Ca²⁺ gradient determines when and where a vesicle fusion occurs. When the frequency is high, the calcium accumulation is fast enough to compete with the Ca²⁺ clearance system. This “gradient hypothesis” explains the f effect but not the m effect.

The control of Ca²⁺-dependent secretion by AP codes, particularly parameters m and d, is a sum of the kinetics of [Ca²⁺]_i and the RRP. Both [Ca²⁺]_i (influx and clearance) and the RRP (exocytosis and refilling) also depend on many positive and negative regulators. Thus, depending on the native ion channels of a neuron, the two variables of the AP codes illustrated in Figure 5 could provide four different effects: (1) m positive and d positive; (2) m positive and d negative; (3) m negative and d positive; and (4) m negative and d negative. It so happens that the behavior of the RACCs was m positive and d negative under the conditions of Figures 5 and 6, respectively. For example, Figure 5 shows that the RACC has effect type 2: the total secretion was dependent on m with the combination (n, f, d) of Figure 5A, although it was independent of d with another combination (n, m, f) of Figure 6B. We should emphasize that, for given total time (T) of AP codes, the d and m are actually related by the following formula: T = m^*n/f + d^* (m - 1). Thus, the d effect can also be considered another type of m effect. This means that the m effect on secretion could differ for different combinations of parameters.

Physiological impact of AP code

Using the AP-C_m method, we studied the secretory response of a cell to different AP codes. With a given number of APs, cell secretion is modulated not only by AP frequency (f) but also by burst pattern (m and d). The m effect is important, because it increases the ability of a neuron to process AP information. By changing parameters (n, m, f, d) of the AP code, the output-secretion of a neuron can be modulated. Interestingly, a variety of AP codes exist in different types of neurons, in same type of neurons, or even in the same neuron. Our experiments provide evidence that AP codes contain distinct types of information. This cellular coded information is an addition to structure-based synaptic plasticity [i.e., long-term potentiation (LTP)] (Otani and Ben-Ari, 1993).

Short-term plasticity includes facilitation and depression. The f effect, or secretion facilitation by higher AP frequencies, has been found not only in RACC but also in many other neurons (for review, see Zucker, 1989). However, the m effect on secretion was less known. In RACC, the m effect is likely a combination of facilitation (Fig. 4) (Charlton and Bittner, 1978; Zhou and Misler, 1995, their Fig. 1) and depression (during the first 30 APs in Fig. 4D) (Zucker, 1989). We believe, like the f effect, the m effect probably exists in most other neurons. It is possible that the m effect differs among different cell types, because the final secretion depends on many factors, including the shape of a single AP, the AP code, ion channels, fusion proteins, colocalization of channels and vesicles, vesicle pools, and Ca²⁺ buffers. All of these factors vary among different types of cells. It is thus possible that, for an A-type cell, pattern F1 evokes more secretion than pattern F2. However, for a B-type cell, pattern F1 evokes less secretion than pattern F2. In fact, our preliminary data in DRG neurons support this prediction: the same two AP codes as those shown in Figure 5 produced a completely opposite m effect (our unpublished observation). Using postsynaptic potential (PSP) as the bioassay, it was found that dividing a given number of stimuli into several groups to produce a “temporal pattern stimulation” can induce LTP more efficiently than a constant frequency stimulus train (Tsukada et al., 1994; Dobrunz and Stevens, 1999). One important advantage of the present work is that our C_m signal is exclusively attributable to presynaptic events, although the effect on LTP measured by PSPs could not exclude postsynaptic contamination (i.e., receptor desensitization). The fact that the m effect of RACC using C_m assay is consistent with effect of temporal patterns of APs using PSP assay (Tsukada et al., 1994; Tsodyks and Markram, 1997; Dobrunz and Stevens, 1999) suggests that those PSP facilitation-LTPs might be dominated by a presynaptic mechanism.

Because of technical challenges, capacitance measurement cannot be applied to many cell types. However, our AP code protocol can be extended to cells beyond patch-clamping conditions. Combined with AP code stimulation, another secretion assay (i.e., EPSCs, FM 1-43 dye, or amperometry) might be used. Furthermore, the AP code stimulation protocol can even be used as extracellular field stimulation. In this way, AP codes can be applied to virtually all types of neurons in brain slices or even in in vivo recordings. It would be of great interest to study stimulus-secretion coupling using AP code protocols in CNS neurons.