INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The discharge patterns of respiratory neurons of the caudal ventral respiratory group (cVRG) are subject to potent GABAergic gain modulation. Local application of the competitive GABA_A receptor antagonist bicuculline (BIC) (MacDonald and Olsen 1994) amplifies the underlying discharge frequency (F_n) patterns that are mediated by endogenous excitatory and inhibitory synaptic inputs (Dogas et al. 1998; McCrimmon et al. 1997). These results imply a tonic GABAergic mechanism that constrains the baseline F_n and reflexly induced activities of these bulbospinal respiratory premotor neurons to ~35-50% of their discharge rate in the absence of this inhibitory input.

The functional location of this form of inhibition on respiratory neurons is not known. On other types of neurons, such as most principal cortical cells and granule cells of the hippocampal dentate gyrus, synaptic inputs arising from distinct groups of inhibitory neurons innervate segregated regions of the target neuron such as the axon initial segment, soma, and various parts of the dendrites (Soltesz et al. 1995). Such spatial segregation of the inputs from GABAergic neurons suggests the possibility that distinct inhibitory inputs may play different functional roles (Nicoll 1994). In other neurons, such as hippocampal granule cells, it has been shown that tonic inhibition is due to GABAergic terminals located near the soma, where it is likely to play an important role in regulating the input-output relations of the neurons (Soltesz et al. 1995). Due to the characteristics of GABAergic gain modulation of respiratory neurons (Dogas et al. 1998; McCrimmon et al. 1997), it is possible that this gain modulation takes place near the soma and/or spike initiation zone. Alternatively, the site of modulation could be on the dendrites. This more distal location is consistent with our preliminary data in which the localized application of glutamate receptor agonists, presumably to the somal region, are not gain modulated while the spontaneous neuronal activity is modulated (Tonkovic-Capin et al. 2001a).

In addition to GABAergic input, an additional mechanism of gain modulation of respiratory neuronal activity is produced by changes in the size of the medium-spike afterhyperpolarizations (AHPs). These AHPs are produced by increases in the conductance of small-conductance, calcium-activated potassium channels (SK channels). Block of the SK channels with the highly selective antagonist apamin abolishes AHPs and increases discharge frequency (F_n) (Viana et al. 1993). Local application of apamin to respiratory neurons produces an increase in their F_n patterns that is proportional to the underlying pattern (i. e., gain modulation). In a previous study, we have shown that GABAergic gain modulation and SK-channel-mediated gain modulation act in a cascade fashion (Tonkovic-Capin et al. 2001b), where the total modulation is the product of the two stages of attenuation. Because AHPs are the direct result of Ca²⁺ entry during the action potential, the AHP mechanism is likely to be located near or at the spike initiation zone. This is in the final output pathway of the neuron and should affect all forms of excitation whether endogenously or exogenously induced.

The purpose of the present study was to determine if, or to what extent, the GABAergic gain mechanism modulates the excitatory responses elicited by activation of excitatory receptors on the soma of respiratory neurons in vivo. Modulation might be expected if the GABAergic input is functionally located near the soma and/or spike initiation zone. On the other hand, it was expected that block of the AHPs with apamin would modulate both exogenously as well as endogenously induced activities. Accordingly, the effects on the net increases in F_n produced by repeated short-duration picoejections (for 2 respiratory cycles) of the glutamate receptor agonist, -amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA), were quantified before and during locally induced antagonism of GABA_A receptors by BIC or SK channels by apamin. The results from these protocols showed that the average AMPA-induced increases in F_n were not significantly altered by BIC but were amplified by apamin in accordance with the gain increase of the respiratory-related activity. These findings suggest that the mechanism for GABAergic gain modulation may be functionally isolated from soma/spike initiation zone, e.g., located on a dendritic trunk.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This research was approved by the Subcommittee on Animals Studies of the Zablocki VA Medical Center in accordance with provisions of the Animal Welfare Act, the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and VA policy. Data were obtained from 15 mongrel dogs of either sex weighing from 8 to 15 kg. Mask induction with a volatile anesthetic (isoflurane or halothane) was used, and anesthesia was maintained during the surgical procedure with isoflurane (1.4-2.0% end-tidal concentration). Airway concentrations of isoflurane, CO₂, and O₂ were continuously monitored with an infrared analyzer (POET II, Criticare Systems, Waukesha, WI). The animals were monitored for signs of inadequate anesthesia (e.g., salivation, lacrimation, and/or increases in blood pressure and heart rate), and if required, the depth of anesthesia was increased immediately.

Surgical procedure

Dogs were intubated with a cuffed endotracheal tube and mechanically ventilated with an air-O₂-isoflurane mixture. The surgical procedures, monitoring, and maintenance of body homeostasis have been previously described in detail elsewhere (Dogas et al. 1998). Briefly, after cannulating the femoral artery (for blood pressure recording and blood-gas sampling) and vein (for continuous infusion of maintenance fluids and drugs), a bilateral pneumothorax was performed to reduce motion artifacts. A bilateral vagotomy was performed to remove the ventilator-induced effects of pulmonary mechanoreceptors on the breathing pattern. The animal was then decerebrated (Tonkovic-Capin et al. 1998). This procedure leads to an anatomically well-defined, midcollicular decerebration. After completion of the decerebration, isoflurane was discontinued, and the dogs were ventilated with an air-O₂ mixture and maintained in hyperoxic normocapnia (PO₂>400 mmHg, PCO₂ 35-45 mmHg). The dorsal surface of the medulla oblongata was exposed by an occipital craniotomy.

Phrenic nerve activity was recorded from the central end of the desheathed right C₅ rootlet. The phrenic neurogram (PNG) was obtained from the moving-time average (100 ms) of the amplified phrenic nerve activity and was used to produce timing pulses corresponding to the beginning and end of the inspiratory phase. The neuromuscular blocker pancuronium (0.1 mg/kg, followed by 0.1 mg · kg¹ · h¹) was then given to reduce motion artifacts during neuronal recordings. Four-barrel micropipettes, (10-30 µm composite tip diameter), consisting of one recording barrel containing a carbon filament and three drug barrels, were used for extracellular neuronal recordings and pressure ejection of nanoliter/picomol amounts of drug solutions. BIC (200 µM; Sigma), apamin (0.125-0.150 µM; Alomone Labs), and -amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA, 7 and 20 µM; Sigma) were dissolved in an artificial cerebrospinal fluid (ACSF) vehicle that also served as a control solution. Due to the potent effect of 20 µM AMPA observed in the first two experiments (8 neurons) we reduced the AMPA concentration to 7 µM (last 20 neurons). Further details of the picoejection technique and its limitations have been previously described (Dogas et al. 1998; Krolo et al. 1999; Stuth et al. 1999). Unit activities from cVRG inspiratory (I) and expiratory (E) neurons were recorded from a region extending 2-4 mm caudal and 2-4.5 mm lateral from the obex and 2-4.5 mm below the dorsal medullary surface. >88% of the respiratory-related neurons within this region have been shown to be bulbospinal premotor neurons (Bajic et al. 1992; Stuth et al. 1994).

Protocols

After decerebration and discontinuation of the anesthetic, a period of 1 h was allowed for washout of isoflurane (airway concentration <0.05%) and for stabilization of the neural breathing pattern. On establishing a stable recording of single-unit activity, repeated automatic short-duration picoejections (for 2 respiratory cycles) of the glutamate receptor agonist AMPA were made before and after picoejection of BIC (protocol 1) and apamin (protocol 2). Short-duration picoejections of AMPA were set to occur every 7-14 respiratory cycles and triggered at the onset of I phase (for E neurons) or E phase (for I neurons) of the respiratory cycle. Volume-ejection rate was measured as a change in meniscus height/time with a ×100 microscope equipped with a reticule. The dose rate (i.e., volume-rate) of AMPA was constant throughout each neuron study. Dose rates during the two respiratory cycle period were determined before and after picoejection of BIC or apamin. The total volume ejected divided by the sum of the two-cycle durations was used to determine the AMPA ejection rates. Picoejections of antagonists (i.e., BIC and apamin) were done in dose rates that were known to produce a near maximal increase in F_n (Tonkovic-Capin et al. 2001b). Picoejections of ACSF were routinely used to verify that the ACSF constituents and/or ejected volumes were without effect.

In Protocol 1, the average net increases in F_n of I and E cVRG neurons produced by the short-duration picoejection of AMPA were quantified before and during locally induced antagonism of GABA_A receptors by BIC. Protocol 2 was similar to protocol 1, but instead of inducing antagonism of GABA_A receptors, we used apamin to block SK channels.

Data analysis

Cycle-triggered histograms (CTHs; bin width: 50 ms), triggered from either the onset of the E or I phase and based on 5-19 respiratory cycles were used to quantify the discharge frequency patterns before and during picoejection of AMPA. The values of F_n for each bin were calculated as the number of spikes per bin/bin duration in seconds. For each time increment (bin) within the triggered cycle, these values were averaged over the number of cycles used to generate the CTH. In addition, the SD and SE of each bin were calculated. Because plots of CTHs ± SDs indicated that the size of the bin SDs appeared to be uniform throughout the active phase of the respiratory cycle, the bin SDs were averaged over the active phase of each CTH. These values served as an index of F_n variability and for calculations of the coefficient of variation for peak F_n. The highest mean bin value was used as an estimate of peak F_n.

Drug-induced changes in the gain and offset of the discharge frequency pattern were analyzed via plots of the F_n(drug) versus F_n(control) values obtained from the CTHs. This method has the advantage of being insensitive to the geometric shape of the pattern. It can detect the amount by which a pattern is shifted up or down, i.e., the amount of parallel shift or offset (y intercept) and the amount by which a pattern is amplified, i.e., the change in gain (slope), regardless of pattern trajectory. The net increase in F_n produced by AMPA was time-averaged over the neuron's active phase, and these values were computed for the period before and the period during locally induced antagonism of GABA_A receptors by BIC or SK channels by apamin. The time-average values during the antagonist effect were then normalized relative to their control values, which were assigned a value of 100%. These values were separately collected for each neuron and pooled according to neuron type (I or E neurons) and according to protocol (BIC and apamin). In addition, for each protocol, gain factors were determined for either BIC or apamin from respiratory cycles without AMPA-induced excitation, i.e., the gains of the endogenously induced activity were determined. These gain factors were then multiplied by the corresponding AMPA-induced net responses during the control period to estimate the expected increases in the net response, if in fact such responses were subject to gain modulation. Normalized values of the average AMPA-induced net increase during the control period, during the period of enhanced activity in response to BIC or apamin (termed BIC or apamin effect), and corresponding predicted gain-modulated values were compared with each other with one-way, repeated-measures ANOVA procedures. Differences were considered significant for P < 0.05, using the false discovery rate procedure for multiple comparisons (Curran-Everett 2000). Values are expressed as means ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Protocol 1: effect of bicuculline on AMPA-induced responses

Complete protocols with BIC were obtained for seven I and seven E neurons. Figure 1 shows a typical example of the responses of a cVRG I neuron to repeated short-duration picoejections of AMPA before and during the BIC-induced effect. The effect of AMPA was usually visible almost immediately from the onset of picoejection, and full recovery usually required 10-20 s. As can be seen, AMPA produced about the same net increase in F_n before and during the BIC effect. CTHs were used to quantify the magnitude of the AMPA-induced responses. Data for the control CTHs were taken from the respiratory cycles immediately preceding each of the seven AMPA picoejections in the pre-BIC period (e.g., Fig. 1, bottom, control). Data for the CTHs of the AMPA-induced effect were taken from the second cycle during AMPA picoejections (Fig. 1, bottom, AMPA). Data during the BIC-induced effects were obtained using a similar procedure. For this example, the shaded areas between the CTHs of Fig. 2 quantify the net AMPA-induced responses. Plots of the difference between CTHs (F_n, Fig. 2, bottom) indicate that the AMPA-induced responses are relatively constant throughout the I phase and that BIC had little affect on the responses. The time-averaged net increases in F_n (shown as horizontal lines through difference CTHs) were 49 and 55 Hz for pre- and post-BIC periods, respectively. AMPA also increased activity during the normally silent phase both before and after BIC, but this activity was not constant because the subthreshold drive varies with time. This activity was not analyzed because the subthreshold drive level cannot be measured with extracellular recordings. The constant nature of the AMPA-induced response during the neuron's active phase is also confirmed by the plot of F_n(AMPA) versus F_n(control) where the slope is nearly one (i.e., 1.02, Fig. 2, top right). The average BIC-induced gain increase for the endogenously mediated activity, obtained from the F_n(BIC) versus F_n(control) plot (Fig. 2, middle right) was 1.96. If the AMPA-induced response had been gain modulated, the expected response magnitude would have been 1.96 × 4996 Hz rather than the observed 55 Hz.

View larger version (49K): [in this window] [in a new window]   Fig. 1. AMPA-Induced responses of a caudal ventral respiratory group (cVRG) inspiratory neuron pre- and postbicuculline picoejection. Repeated (every 12 respiratory cycles) short-duration (2 respiratory cycles) picoejections of AMPA produced similar net increases (50 Hz) in Fn before and during the bicuculline (BIC) effect. Thick arrow: time-expanded view of the phrenic neurogram (PNG), picoejection marker (PEM), neuronal activity (NA), and neuronal discharge frequency (Fn) during the short-duration picoejection of AMPA. Net effects were determined from the difference of the 2nd cycle (AMPA) after the start of picoejection and the cycle labeled control. Dashed horizontal lines: reference line through the peak Fn of the control cycles as an aid to visualizing the net AMPA-induced response.

View larger version (25K): [in this window] [in a new window]   Fig. 2. Cycle-triggered histograms (CTHs) of unit activity of the inspiratory (I) neuron of Fig. 1. Top left: pre-BIC period: AMPA increased Fn by 49 Hz throughout the I phase of the control period as seen by the parallel shift in the CTH (shaded area). This is confirmed by the lack of change in the slope of the plot of Fn(AMPA) vs. Fn(control) (top right; slope: 1.02). - - -, line of identity. Middle, left: BIC postejection period. Picoejection of BIC produced a gain modulated increase in the Fn (thick line CTH). The plot of Fn(BIC) vs. Fn(control) estimated the gain of the endogenously mediated activity to be 1.96 with an offset of -27.5 (Middle right, lower plot). Picoejection of AMPA again produced a parallel shift in the Fn pattern during the BIC effect (shaded). CTHs of the differences (Fn, bottom) show that over the time period indicated (vertical dashed lines) the average net increases in Fn are similar for the pre- and post-BIC periods, i.e., the AMPA-induced response is not gain modulated (49 vs. 55 Hz; thin and thick lines through difference CTHs, respectively). 7 cycles/CTH.

Pooled data for bicuculline-induced gain modulation

The average AMPA-induced increase in F_n obtained from seven I neurons during the BIC response was 96.0 ± 5.2% of control, indicating that BIC had no effect on the net neuronal response to AMPA (bar "a," Fig. 3, top, bicuculline). The average predicted AMPA-induced response, based on gain factors obtained from the effects of BIC on endogenous activity, was significantly larger than the actually observed responses during the BIC effect (172.6 ± 18.5%, bar "p," Fig. 3 top, bicuculline).

View larger version (28K): [in this window] [in a new window]   Fig. 3. Summary data of the effects of bicuculline or apamin on the net AMPA-induced responses of cVRG I and E neurons. Data are presented as net responses normalized to their respective control levels before application of either bicuculline or apamin (Fn). Data are from 7 I and 7 E neurons. Bars labeled "a": actual net responses; Bars labeled "p": predicted net responses. Predicted responses were obtained from the product of the control period AMPA response and the gain factor obtained from plots of Fn(BIC or apamin) vs. Fn(control). Top: I neurons. The net responses during the effects of bicuculline (a bar, bicuculline) were not significantly different from control (100%, dashed horizontal line) but significantly less than their respective predicted responses (p bar, bicuculline). The net response during apamin was significantly greater than control (a bar, apamin) and no different from their respective predicted responses (bar p, apamin). Net responses during apamin were significantly greater than those during bicuculline (a bars). Similar results were found for the net AMPA-induced responses of E neurons (bottom). P values for comparisons with control responses: ##, P < 0.01 and ###, P < 0.001. P values for comparisons indicated by brackets: *, P < 0.05, **, P < 0.01, ***, P < 0.001. See text for further explanation.

Similar results were obtained for the pooled data from 7 E neurons (Fig. 3, bottom, bicuculline). The average AMPA-induced response of 112.6 ± 5.9% during the BIC-induced effect was not different from control and was significantly less than the predicted gain modulated response of 144.2 ± 4.3%.

Protocol 2: effect of apamin on AMPA-induced responses.

Complete protocols with apamin were obtained for seven I and seven E neurons. Figure 4 shows a typical example of the responses of a cVRG I neuron to repeated short-duration picoejections of AMPA before and during the apamin-induced effect. The net AMPA-induced responses during the apamin effect are significantly larger than the responses during the control period. AMPA produced a parallel upward shift in the F_n patterns [shaded area between CTHs, Fig. 5, left, and slope of 1.00 for F_n(AMPA) vs. F_n(control) plot, Fig. 5, top right]. The CTH analysis of this data shows that the time-averaged AMPA-induced response increased from 39 Hz to 72 Hz post apamin picoejection, an 84.6% increase (F_n, horizontal lines, Fig. 5, bottom left). This increase is commensurate with the 91% increase in the gain of the endogenous activity produced by apamin [1.91, slope of F_n(apamin) vs. F_n(control) plot, Fig. 5, middle right].

View larger version (45K): [in this window] [in a new window]   Fig. 4. AMPA-induced responses of a cVRG inspiratory neuron, pre- and postapamin picoejection. Repeated (every 11 respiratory cycles), short-duration (2 respiratory cycles) picoejections of AMPA produced larger net increases in Fn during the apamin block of Ca2+-activated K+ channels. Thick arrow: time-expanded view of PNG, PEM, NA, and Fn before and during picoejection of AMPA. Net effects were determined from the difference of the 2nd cycle (AMPA) after the start of picoejection and the cycle labeled control. Dashed horizontal line: reference line through the peak Fn of the control cycles as an aid to visualizing the net AMPA-induced response.

View larger version (24K): [in this window] [in a new window]   Fig. 5. Cycle-triggered histograms (CTHs) of unit activity of the I neuron of Fig. 4. Top left: preapamin period; AMPA increased Fn by 39 Hz throughout the I phase of the control period as seen by the parallel shift in the CTH (shaded). This parallel shift is verified by the slope of 1.00 in the plot of Fn(AMPA) vs. Fn(control) (top right). Dashed line: line of identity. Middle, left: picoejection of apamin produced a gain modulated increase in Fn that is most noticeable near the end of the I phase (thick line CTH). The plot of Fn(apamin) vs. Fn(control) estimated the gain to be 1.91 with an offset of -65.9 (middle right). Picoejection of AMPA now produced a large parallel shift in Fn (shaded CTH). CTHs of the differences (Fn, bottom left) show that the average net increase in Fn over the time period indicated (vertical dashed lines) is much greater during the apamin effect than that of the preapamin period, i.e., the response is amplified by 1.8 (i.e., 72 vs. 39 Hz; thick and thin lines through difference CTHs, respectively). 7-10 cycles/CTH.

Pooled data for apamin-induced gain modulation

The average AMPA-induced response of 161.7 ± 8.4% of control, obtained from seven I neurons during the apamin-induced effect, was not significantly different from the predicted response of 146.6 ± 7.8%, based on gain factors obtained from the effects of apamin on endogenous activity (Fig. 3, top right, apamin).

Similar results were obtained for the pooled data from seven E neurons. The average AMPA-induced response of 151.9 ± 13.6% during the apamin-induced effect was significantly different from control, and not significantly different from the predicted response of 152.3 ± 9.6% (Fig. 3, bottom, apamin).

Comparison of bicuculline and apamin effects

Even though the gain increases in endogenous neuronal activity produced by BIC and apamin were similar in magnitude (58.1 ± 10.0 and 47.1 ± 6.4%, respectively, n = 14 each), the effects on the AMPA-induced responses were markedly different for BIC and apamin. This differential effect is illustrated by the net AMPA-induced responses for two cVRG E neurons in Fig. 6. The net time-averaged responses pre- and post-BIC application were essentially the same, 22.6 and 20.5 Hz, respectively (shaded regions, Fig. 6A). In contrast, the net time-averaged response postapamin of 29.3 Hz was markedly larger (45%) than the time-averaged control response of 20.4 Hz (shaded regions, Fig. 6B). The pooled AMPA-induced response data for BIC and apamin are contrasted in Fig. 3 ("a" bars, bicuculline vs. apamin).

View larger version (30K): [in this window] [in a new window]   Fig. 6. Examples of the effects of bicuculline and apamin on the AMPA-induced responses of 2 cVRG E neurons. A: CTHs of activity from an E neuron with an augmenting discharge frequency pattern. 5-12 cycles/CTH. Picoejection of AMPA produced an upward parallel shift in the pattern (top, shaded area: net response). Bicuculline increased the augmenting slope of the CTH (BIC, thick line, bottom). However, the net AMPA-induced response (shaded region) is very similar to the pre-BIC net response. B: CTHs of activity from an E neuron with a decrementing discharge frequency pattern. 6-9 cycles/CTH. The shaded area (preapamin) shows the control net AMPA-induced response. Apamin increased the decrementing slope of the CTH (apamin, thick line, bottom). Note that the net AMPA-induced response (shaded area, postapamin) is significantly greater that the control AMPA response.

Consistency of the transient AMPA-induced responses

The average AMPA-induced increase in activity was 43.2 ± 2.1 Hz (n = 28). The average AMPA picoejection rates prior to and after the BIC or apamin application were not significantly different, 0.144 ± 0.015 and 0.146 ± 0.016 pmol/min, respectively. To evaluate the consistency of the responses to the transient picoejections of AMPA both during the control period and during the effects of BIC and apamin, the time-averaged bin SD of each CTH was calculated. For the control cycles immediately preceding the AMPA picoejection, the SD serves as an index of the cycle-to-cycle variation of the spontaneous activity. The SD for the cycles with increased activity due to AMPA reflects the variation of the spontaneous activity plus the variation in the net AMPA-induced responses. From the pooled data of 28 neurons, the average SD value for the AMPA excited cycles (14.7 ± 0.7 Hz) was greater than that of control cycles (11.6 ± 0.5 Hz, P < 0.0001) prior to the local application of BIC or apamin. The average difference of 3.1 Hz represents a 27.0 ± 3% increase in SD for the AMPA excited cycles. Similar data for the period after BIC or apamin showed that the increase in SD for the AMPA excited cycles of 30.0 ± 4% was not significantly different from that prior to BIC or apamin.

To appreciate the magnitude of the cycle-to-cycle variations, coefficients of variation (COV: 100*SD/peak F_n) were also compared. Prior to any picoejections, the average baseline peak F_n value for the I neurons of 93.2 ± 13.1 Hz was not significantly different from the average baseline peak F_n value of 74.1 ± 6.2 Hz for the E neurons. In addition, the average COV values for I and E neurons were not significantly different, and pooled COV values are given. Prior to BIC or apamin application, the COV of cycles with AMPA-induced excitation (11.8 ± 0.6%) was less than the COV for the control cycles (15.6 ± 1.2%). Thus even though the SD increased with AMPA, the peak F_n increased relatively more.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The major finding of this study is that the AMPA-induced increases in I and E neuronal activities were not gain modulated by BIC but were gain modulated by apamin. This differential effect occurred even though BIC and apamin produced similar increases in the gain of the baseline endogenous activities (58.1 ± 10.0 and 47.1 ± 6.4%, respectively).

AMPA-induced response variability

To determine the presence or absence of gain modulating effects on the AMPA-induced responses, it is necessary that these transient submaximal AMPA responses can be consistently reproduced. One important contributing factor is that of consistent dose rates during each response both before and after the application of BIC or apamin. Comparison of the measured picoejected volumes of AMPA showed that there were no significant differences (<2%) in dose rates before and during the effects of BIC or apamin. Other factors contributing to the variability of the responses may include time-dependent alterations in the diffusion path and natural variation in the neuronal response to AMPA. By repeating the AMPA picoejections several times (5-19) before and during the responses to BIC or apamin, the variability of the responses of each neuron could be estimated. The finding, that the average SD of the AMPA stimulated cycles (14.7 Hz) was greater than the averaged SD of the control cycles (11.6 Hz), indicates that the additional variation was due to the variation of the net AMPA response. That is, the variability of the AMPA stimulated cycles is made up of two components: one due to the variability of the control cycles and the other due to the variability of the net AMPA-induced response. Using the principle that the variance of the sum of two random variables is equal to the sum of each variance, the SD of the net response was found to be 9.0 Hz. This is 22% less than the SD of the control cycles and indicates that the AMPA-induced net responses were highly reproducible using the picoejection protocol of this study. Every neuron tested responded consistently to the AMPA picoejections. The effect of desensitization was not observed (e.g., Figs. 1 and 4). This may be due to the rapidity of AMPA receptor desensitization (1 ms < time constant < 5 ms), and what was observed may have been the postdesensitization steady-state response (Dingledine et al. 1999). Alternatively, our AMPA stimuli may have been brief enough and/or less intense so as not to produce long-lasting changes in neuronal excitability as reported for NTS neurons in vitro (Zhou et al. 1997).

Differential effects

The finding that BIC did not modulate the AMPA-induced responses, whereas apamin did, is in agreement with our previous study, which demonstrated that bicuculline methochloride does not block SK channels via a nonspecific mode of action in vivo (Tonkovic-Capin et al. 2001b). The fact that apamin gain modulated both the baseline endogenously mediated and AMPA-induced neuronal activities by the same degree is consistent with the assumption that the location of the mechanism for the AHP is in the final common neuronal output pathway (i.e., spike initiation zone), where all inputs are expected to be similarly affected. The fact that apamin modulated the AMPA-induced response, but BIC did not, suggests that the picoejected AMPA may be acting at receptors located near the soma and/or spike initiation zone. Also, due to the rate limitation of the diffusion process, the rapidity of the onset and recovery of AMPA-induced responses suggests receptor locations near the recording electrode, which presumably is located near the soma where the extracellular action potentials are largest. At greater distances from the electrode tip, responses would be expected to be more delayed, blunted, and prolonged. Furthermore, because of the steep diffusion-dependent concentration gradient, the brisk responses to low pipette concentrations of AMPA (e.g., 7 µM) suggest receptor locations near the electrode tip. It is also likely that AMPA receptors are located distal to the soma, but they are unlikely to have been stimulated by the low pipette concentrations used (Krolo et al. 1999).

The finding that the AMPA-induced responses were not modulated by antagonism of the BIC-sensitive GABA_A receptors suggests that the GABA receptors are located distal to both the spike initiation zone/soma and the activated AMPA receptors. Our previous studies suggest that this location is relatively close to the micropipette tip because the BIC-induced responses also have rapid onset times. A possible location would be on a major dendritic trunk, where the GABAergic mechanism may gain modulate the various more distal respiratory-related dendritic inputs to the same extent.

The data from this study and our previous studies (McCrimmon et al. 1997) do not rule out the possibility of GABAergic presynaptic inhibition, which could also produce F_n patterns that are proportional to the underlying baseline patterns. However, our evaluation of the various characteristics of GABAergic gain modulation suggests a postsynaptic site. This suggestion is based on the observation that the spontaneous phasic F_n patterns and their modification by reflexly induced excitatory and inhibitory inputs appear to be all modulated by the same gain factor. Although this phenomenon could be explained by presynaptic inhibition, it would require that each of the excitatory and inhibitory synaptic inputs that produce the F_n pattern would have a corresponding presynaptic gain modulating input of equivalent strength. As previously mentioned, the onset effects of BIC are relatively rapid suggesting antagonism of GABA_A receptors relatively close to the soma of these neurons that have large dendritic systems (Bianchi et al. 1995).

Other evidence for differential receptor distribution

While our data suggest functionally different locations for the GABAergic and AHP-mediated gain modulation mechanisms, there is additional pharmacological evidence consistent with this interpretation. Champagnat et al. (1982) suggested that fast inhibitory postsynaptic potentials (IPSPs) mediated by GABA_A and glycine receptors are spatially segregated in brain stem respiratory neurons. In inspiratory neurons, glycine-sensitive IPSPs are preferentially located on the soma and are responsible for rapid inhibition at the beginning of expiration. In contrast, GABA_A-mediated IPSPs are primarily located on distal dendrites and serve to maintain synaptic inhibition throughout the expiratory phase. Additional evidence from anatomically based studies corroborates a physical separation of the receptors responsible for these mechanisms. Using combined immunohistochemistry and retrograde labeling in adult rats, Robinson and Ellenberger (1997) found that bulbospinal VRG neurons and phrenic motoneurons showed positive immunolabeling for N-methyl-D-aspartate (NMDA), AMPA, and kainate receptor subunits. Furthermore, there was a unique distribution for each receptor subtype along the neuronal membrane. Immunoreactivity for AMPA receptor subunits was distributed throughout the somata and proximal dendrites; NMDA receptor subunit immunolabeling was localized to the soma, while kainate subunit immunolabeling was confined mainly to dendrites. If a similar distribution of glutamate receptors exists for canine respiratory neurons, it would suggest that the AMPA responses of the current study were due to receptors located on or near the soma. Similar immunolabeling studies for GABAergic synaptic inputs and GABA_A receptor subunits on respiratory neurons have yet to be performed.

However, there is evidence for a differential distribution of GABA inputs and receptors from studies of nonrespiratory neurons. For example, granule cells of the hippocampal dentate gyrus receive inhibitory inputs from a least five types of GABAergic neurons (Halasy and Somogyi 1993; Han et al. 1993). These interneurons mainly terminate on mutually exclusive domains along the longitudinal axis of granule cells. Chandelier cells form axo-axonic contacts exclusively with the initial segment of the axon, whereas basket cells make contact with the somata and proximal dendrites. Various types of hilar cells innervate the different levels of the granule cell's dendritic system. In addition, the dendritic trees of these GABAergic interneuron types can occupy nonoverlapping domains (Soltesz et al. 1995). This strict spatial segregation of the inputs and outputs of these GABAergic neurons suggests that they may play different functional roles (Nicoll 1994). In view of these observations, it is possible that the GABA_A receptors responsible for gain modulation are strategically located on dendritic shafts and/or trunks that are sufficiently close to the soma to affect the respiratory related synaptic inputs, but distal enough so as not to affect the responses to exogenously applied AMPA.

Hypothetical model for gain control of respiratory neurons

Based on our current and past studies, we propose the following hypothetical model to aid in summarizing and explaining the findings of this study (Fig. 7). The working hypothesis is that both excitatory glutamatergic and inhibitory GABA_A and glycinergic respiratory-related tonic and phasic synaptic inputs are located on dendrites. Together these inputs generate the dendritic current I_DEN. As I_DEN flows toward the soma, it is subjected to an attenuation that is controlled by a BIC-sensitive tonic GABAergic input. This attenuation may be mediated by shunting inhibition (Koch et al. 1983; Vu and Krasne 1992). This mechanism is assumed to be strategically located on a dendritic trunk because BIC produces a discharge pattern that is an amplified replica of the underlying spontaneous as well as reflexly induced discharge patterns (McCrimmon et al. 1997). It is also electrically isolated from the soma/spike initiation zone to prevent interactions that could alter the magnitude of the other inputs and/or AHPs. This latter assumption is supported by the fact that the AMPA-induced responses were unaffected by BIC.

View larger version (17K): [in this window] [in a new window]   Fig. 7. Hypothetical model for gain control of respiratory neurons. Scheme to aid the explanation of the current and past observations regarding GABAergic gain modulation and the role of the afterhyperpolarizations. See text for details.

A proportional, but attenuated I^*_DEN is supplied to the soma, where it is combined with other inputs (e.g., behavioral). The latter may or may not be gain modulated. We assume that the excitatory current responsible for the AMPA-induced response is also an input to the soma. Together these input currents make up the somatic current I_SOMA that provides the input to the spike generating process. The discharge frequency of the neuron (F_n) is assumed to be proportional to I_SOMA. However, this proportionality constant is highly dependent on the magnitude of the AHPs that are mediated by small conductance Ca²⁺-activated K⁺ (SK) channels. Thus neuromodulatory inputs or drugs such as apamin, which can affect or block SK channels, alter the overall excitability of the neuron and result in gain modulation of F_n. Because this mechanism is located in the final common output pathway of the neuron, all inputs, including those exogenously induced, are subject to its effect. The SK channels may also play a fundamental role in limiting the discharge frequency of the neuron and protect the neuron from the deleterious effects of continuous high rates of activity (Vergara et al. 1998).

Summary

The results of this study suggest that gain modulation of the discharge frequency of respiratory neurons can be mediated by two distinctly different mechanisms that operate in a cascade manner, at least on endogenously mediated respiratory-related activity. The functional isolation of the GABAergic mechanism from the SK channel mechanism suggests that the processing of inputs to respiratory neurons may be more complex than frequently thought of in that processing may take place in multiple neuronal compartments rather than in a single somatic compartment. Distributive processing of neuronal signals could provide a possible mechanism for the reconfiguration of neuronal activities during, for example, coughing (Shannon et al. 1998) or for gating different central pattern generators to premotor neurons as might occur during the transitions from breathing to vomiting (Fukuda and Koga 1997). GABAergic gain modulation may also provide a means for adaptive control, optimizing the respiratory neuronal discharge frequency patterns for changes in conditions or states (e.g., sleep-wake cycles).