INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX I REFERENCES

EXPIRATORY (E) bulbospinal neurons make up a great majority of the E neurons in the caudal portion of the ventral respiratory group (VRG) in the region of the nucleus retroambigualis (4, 21). E bulbospinal neurons are known to have both monosynaptic and polysynaptic connections with contralateral spinal cord thoracic E motoneurons (9, 16, 17). It also appears that lumbar E motoneurons receive mono- and polysynaptic inputs from the contralateral E bulbospinal neurons (22). E bulbospinal neurons are the major source of drive for thoracic and abdominal E motoneurons (3) and, therefore, are responsible for active expiration. In addition, E bulbospinal neurons provide inhibition (presumably via interneurons) to thoracic inspiratory (I) motoneurons during expiration (29).

The excitability of E bulbospinal neurons is highly dependent on arterial CO₂ tension (Pa_CO2) activation of central chemosensory sources with the greatest sensitivity being found over the hypo- to normocapnic range (2). In addition, these neurons receive excitatory inputs from carotid chemoreceptors (18, 30) and both excitatory and inhibitory inputs from pulmonary mechanoreceptors with vagal afferent fibers (4, 9). On the basis of spontaneous discharge patterns and responses to lung inflation, the E bulbospinal neurons in dogs can be divided into two types: type A, augmenting (20-30%), and type D, decrementing (70-80%) (4). Graded inhibition of type A neurons is produced by lung inflation when transpulmonary pressure (P_t) exceeds 3-4 mmHg. For type D neurons, studies using step and slow positive or negative ramp inflations (1-2 mmHg/s) demonstrate that the relationship between neuronal discharge frequency (F_n) and P_t is made up of two major, linear components (4). For 1.5  P_t  4.5 mmHg, the relation is positive (excitatory), whereas for 4.5  P_t  20 mmHg, the relation is negative (inhibitory). The inhibition is strong enough to override the excitation. Because the excitatory and inhibitory responses to step inflations are slowly adapting, it is highly likely that the slowly adapting pulmonary stretch receptors (PSRs) are involved. Two types of canine PSRs have been identified based on the relationship of their discharge frequency to P_t and on their anatomic location in the airways (23, 24). Because the characteristics of these two PSR types coincide with the response characteristics of the type D (decrementing) E bulbospinal neurons, it is possible that these two different types of PSRs mediate the excitatory and inhibitory components.

The activity of E bulbospinal neurons is sensitive to vagal feedback from pulmonary mechanoreceptors and to central and peripheral chemosensory stimulation and thus can influence most of the mechanical properties of ventilation, such as tidal volume, breathing frequency, airflow rate, and end-expiratory lung volume. In response to metabolic demands, they appear to play a significant role in adjusting ventilatory mechanics to provide efficient performance. These studies were undertaken to characterize the interaction between PSR and central chemosensory inputs on the discharge patterns of the type D E neurons of dogs.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX I REFERENCES

Experimental preparation. Experiments were performed on 18 mongrel dogs (10-20 kg) anesthetized with thiopental sodium (induction dose: 15 mg/kg iv; additional doses given as needed during preparation; maintenance dose during data collection, 4-8 mg · kg¹ · h¹ iv continuous infusion). Positive-pressure constant flow ventilation was produced by an alternating two-valve solenoid ventilator through a cuffed endotracheal tube using 100% O₂. Airway CO₂ was measured with an infrared analyzer (Instrumentation Laboratory IL-200), and tracheal pressure was measured from an airway sideport with an air-filled catheter connected to a Gould-Statham P23ID transducer. Arterial pressure was measured from a femoral artery fluid-filled catheter using a Gould-Statham P23ID transducer. Arterial blood samples were obtained hourly for the measurement of pH, Pa_CO2, and Pa_O2 using a Radiometer ABL 1 analyzer. When required, metabolic acidosis (base deficit >5 mM) was corrected with an appropriate amount of sodium bicarbonate in saline. The dogs were monitored for signs of inadequate anesthesia, including movement, salivation, lacrimation, and/or increases in blood pressure and heart rate. The anesthetic depth was increased immediately if such signs were present. Esophageal temperature was maintained at 38 ± 1°C using a servocontrolled heating pad.

Data recording. Efferent phrenic activity, spike potentials from the brain stem neurons, airway CO₂ concentration, tracheal pressure, and blood pressure were recorded on an FM tape recorder (Vetter, model D). The above-mentioned parameters and time-averaged phrenic activity (PNG), neural spikes/100 ms, I duration (T_I), and E duration (T_E) were recorded on a Grass model 7 polygraph. Phrenic recordings were obtained with bipolar electrodes from the desheathed central end of the C₅ rootlet, which was immersed in a mineral oil pool formed from a neck pouch. The phrenic nerve signal was amplified with a band pass of 0.1-3 kHz. The online moving time average of the phrenic activity was obtained by full-wave rectification and low-pass filtering (averaging window = 50 ms). The positive PNG slope at the onset of phrenic activity and the negative PNG slope at the onset of the abrupt decline in phrenic activity were used to generate I and E timing pulses, respectively. These timing pulses were used to compute online values of T_I and T_E.

Protocol. During control cycles, the ventilator frequency was adjusted to be near that of the central respiratory rhythm (as indicated by PNG) and entrained the central pattern via PSR feedback. Once a type D expiratory neuron was located in the caudal VRG at normocapnia, ventilatory tidal volume was increased to produce a hypocapnic Pa_CO2 level of ~20 mmHg. After airway end-tidal CO₂, F_n and peak PNG (if central inspiratory rhythm was still present) had reached a steady state (usually after 3-5 min), four slow positive (+) and four negative () test ramp inflations separated by six to ten control ventilator cycles were applied. These test inflations had duration of 6-10 s and produced peak P_t values of 12-20 mmHg. Arterial blood samples were drawn for measurement of Pa_CO2, pH, and Pa_O2. To obtain several steady-state Pa_CO2 levels over the range of 20-80 mmHg, CO₂ was added to the inspired O₂ via a gas blender. Neuronal responses to the ramp inflations were obtained for one to eight steady Pa_CO2 levels, averaging four Pa_CO2 levels per neuron.

Data reduction. Offline data analyses were carried out on a Hewlett-Packard model 360 computer with a data converter interfaced through an IEEE 488 data port. A conversion rate of 100 Hz was used to enter 5- or 10-min epochs of the number of neuronal spikes per 10 milliseconds, phrenic activity, time averaged for 10 ms, tracheal pressure, and a ventilator I phase indicator into the computer memory and subsequently onto a disk file. Software routines assigned a number to each consecutive ventilation cycle and displayed the signals on the monitor. Ventilator cycles were numbered consecutively, and the numbers corresponding to a given test inflation pattern, (+) or () ramps, at a given Pa_CO2 level were identified and used to generate cycle-triggered histograms (CTHs) of unit activity and ensemble averages of both phrenic activity and tracheal pressure patterns. The temporal alignment of the CTHs and ensemble averages was accurate to within 10 ms of the phase onset indicator signal. CTHs and ensemble averages were saved on disk files for further analyses, which included the generation of F_n versus P_t plots for test inflation cycles.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX I REFERENCES

Expiratory neuronal responses to step lung inflations delivered during the expiratory phase. The E neuronal responses to step inflations during the E phase are slowly adapting and related to the magnitude of the step P_t in a biphasic manner. In the example of Fig. 1, the 4-s-long step test inflations (P_t) were delayed 300 ms from the beginning of the E phase and were separated by 8-10 control cycles to minimize changes in Pa_CO2 during subsequent test cycles. During control and test cycles, 1-s duration inflations were delivered during the I phase to provide steady-state ventilation. For P_t values <5 mmHg, step inflations delivered during test respiratory cycles produced reflex increases in T_E and increases in F_n (Fig. 1A). A step inflation of 1.3 mmHg prolonged the E phase and the neuronal discharge (Fig. 1A, S1) of an E neuron with a decrementing control pattern, whereas larger step inflations of 2.6 and 4.8 mmHg increased F_n of the step responses (Fig. 1, A and B, S2 and S3). For P_t >5 mmHg, step inflations reduced F_n below maximum response values and the responses remained relatively independent of time during the step inflation (e.g., Fig. 1B, S4 = 10.6 mmHg). The neuronal responses are mainly dependent on inflation pressure and only to a small degree on time. This allowed us to quantify the biphasic neuronal response to inflation using plots of F_n versus P_t, where P_t is used as an index of slowly adapting PSR activity (Fig. 1B).

View larger version (25K): [in this window] [in a new window]   Fig. 1.   Example of the slowly adapting responses of an expiratory (E) neuron to step inflation patterns. PNG, phrenic neurogram; Fn, cycle-triggered histograms (CTHs) of discharge frequency; Pt, ensemble average of transpulmonary pressure, TE0, control E duration. Four test cycles/CTH. A: responses to low-Pt step inflations. con, Control pattern without inflation (thick line). S1: 1.4 mmHg; S2: 2.6 mmHg. B: Fn, neuronal responses to higher Pt step patterns. S3: 4.8 mmHg, S4: 10.5 mmHg. B, top: plot of Fn vs. Pt for mean ± SD of CTH data during the step input (between 1.5 and 4.5 s). Note the biphasic nature of the E neuronal response.

E neuronal responses to slow ramp lung inflations delivered at different times during the E phase. The relationship between F_n and P_t can also be obtained when slow ramp inflations are used to scan the entire P_t range from 0 to 15-20 mmHg in a single test cycle. This is possible because time-dependent effects on these responses are minimal. This is illustrated for an E neuron where slow ramp test inflations were delivered in the E phase with different delays with respect to the onset of the PNG (Fig. 2A). The biphasic nature of the response can be seen for the ramp inflation with the larger delay (Fig. 2A). This delay was the largest one that could be used, because the control T_E was <2 s (T_E0, Fig. 2A) and the next I phase would start before the inflation reflexly prolonged the E phase. Near the beginning of the E phase, the early portion of the control decrementing pattern can be seen before the ramp inflation starts (Fig. 2A, F_n). As the ramp P_t increased, F_n increased and reached a maximum at P_t  5 mmHg, then decreased with increasing P_t. Note that the amount of inhibition is more than able to suppress the P_t-induced excitation. When the ramp inflation was terminated, F_n rapidly increased and then decreased as the subsequent I phase began. For those cycles in which the slow ramp inflations began 2 s earlier, the early portion of the response was not seen because I-phase inhibition produced neuronal silence (Fig. 2A). The lack of time dependence in these neuronal inflation responses, separated by 2 s, is demonstrated by the high degree of overlap in the F_n traces when the two P_t traces are superimposed (Fig. 2B). The plot of F_n versus P_t for the more delayed ramp inflation quantifies this biphasic relationship (Fig. 2B). This relationship is similar to the one for step inflations (Fig. 1B).

View larger version (25K): [in this window] [in a new window]   Fig. 2.   Typical E neuronal responses to delayed slow ramp inflations. A: inflations have delays separated by 2 s. Earlier ramp started late in the inspiratory (I) phase (Pt, thin line) and later ramp started just before the end of the E phase indicated by the dashed line at upstroke of control PNG. B, middle and bottom: same data displayed with Pt traces superimposed with time shift. Note the high degree of overlap in the Fn CTHs indicative of response dependence on Pt rather than time. Earlier portion of the excitatory response to the earlier ramp is missing due to I phase inhibition of E activity; 6 test cycles/CTH. B, top: plot of Fn vs. Pt shows a typical biphasic response similar to that of Fig. 1B, top. Fn vs. Pt data were obtained from the 50-ms bin data of the CTH during the more-delayed ramp pattern.

The effects of Pa_CO2 on E neuronal responses to positive and negative slow ramp inflations. We used both positive and negative test ramp inflations for the same neuron to better isolate P_t-dependent effects from time-dependent effects. By using "contrasting" pressure-time profiles, the amount of time dependence affecting the neuronal response to inflation will be reflected in the difference between F_n-P_t plots for each type of inflation pattern. In the absence of time dependence, F_n-P_t plots obtained from positive and negative ramps should coincide.

Ventilation patterns synchronized with central rhythm. Examples of the responses to positive and negative ramps for the same neuron at different levels of Pa_CO2 are shown in Figs. 3 and 4. Both inflation patterns were alternately presented at each Pa_CO2 level. Increases in Pa_CO2 increased the peak F_n of control cycles and enhanced the inflation-induced responses. For the positive ramps (Fig. 3), the biphasic nature of the response is evident and similar to those of Fig. 2. One-second duration control inflations were delivered during the I phase. However, to produce a slower central rhythm, a 1-s delay from the onset of the PNG was used to shift the inflation later into the I phase and subsequent deflation later into the E phase. This prevented too much reflex shortening of T_I and produced longer T_E values. On test cycles, no delay was used for the I-phase inflation to allow sufficient time for deflation and P_t to return to baseline before the slow test ramp was initiated (1.8-s delay from onset of E phase, Fig. 3) that had to occur early enough to reflexly prolong T_E. The negative ramps were produced by an increase in E flow resistance during the test cycle (P_t, Fig. 4). Because P_t is highest at the onset of the E phase, F_n is initially reduced and gradually increases as the P_t-induced inhibition decreases, unmasking the P_t-induced excitation. The neuronal response peaks as P_t passes through the inhibitory threshold and then declines as the P_t-induced excitation decreases. The peak F_n is again highly dependent on Pa_CO2 (Fig. 4).

View larger version (25K): [in this window] [in a new window]   Fig. 3.   Effect of arterial PCO2 (PaCO2) on E neuronal responses to slow positive ramp inflations. Fn: ratemeter output (0.1-s intervals). Vertical dashed lines: onset of the E phase of the test cycle. Slow ramps delayed 1.8 s from onset of E phase. Fn of both control and biphasic inflation response patterns increased with increased PaCO2.

View larger version (27K): [in this window] [in a new window]   Fig. 4.   Effect of PaCO2 on E neuronal responses to slow negative ramps for the same neuron shown in Fig. 3. Vertical dashed lines: onset of the E phase of the test cycle. E airflow resistance added during test cycle to produce slow negative ramps with no delay. As Pt decreased, the decline in inhibition resulted in increased Fn. With further decreases in Pt, excitation was also removed. Peak Fn of both control and test cycles increased with increased PaCO2.

Central rhythm entrained by ventilation pattern. To investigate the effects of Pa_CO2 both above and below the apneic threshold, the phenomenon of entrainment was used to time lock the central respiratory pattern to the ventilation pattern. For this purpose, the ventilator rate for control cycles is set close to the central respiratory rhythm rate, which can be determined from a few PNG cycles without ventilation. In the example of Fig. 5, the control inflation (P_t) can be seen to consistently occur during the early part of the E phase. For the test cycles, slow ramps replaced the control inflation pattern. Slow negative ramps were produced by adding an E flow resistance during test inflations. In addition, the I flow rate was increased to produce greater peak P_t levels during the negative ramp test cycles (15-20 mmHg).

View larger version (41K): [in this window] [in a new window]   Fig. 5.   Effect of PaCO2 on inflation responses above and below the apneic threshold. The central respiratory pattern was entrained by the ventilation pattern. Both E neuronal activity (ENA) and PNG decreased as PaCO2 decreased. At a PaCO2 of 35, PNG was abolished but central rhythm remained entrained. In this case, small-amplitude E unit activity was also present. The biphasic characteristic of the inflation response was preserved at all PaCO2 levels.

View larger version (33K): [in this window] [in a new window]   Fig. 6.   Typical examples of Fn vs. Pt plots as a function of PaCO2 for 2 E neurons (A and B). Top: data for positive (+) ramp inflations. Bottom: data for negative () ramp inflations. Data obtained from CTHs of Fn and ensemble averages of Pt. All CTHs were triggered from the onset of the test Pt patterns, because no PNG is present at low PaCO2 levels.

Pressure-dependent and time-dependent response components. Although it is clear that the canine E neuronal inflation response is highly dependent on P_t via the PSRs, the response also appears to depend to a minor degree on time (t), i.e., F_n = F(P_t,t). To separate these two effects, we assumed, as a preliminary hypothesis, that the time component was linear or F(t) = t, where  is the slope, which can be positive or negative and has the units of Hertz per second. This assumption appears reasonable based on the neuronal responses to relatively constant inputs such as those of Fig. 1 and those in which step frequency, electrically induced PSR inputs were used (32). From a practical viewpoint, it would be very difficult to separate the pressure and time-dependent components, if a more complex form of time function is used. Thus F_n = F(P_t) + t, and the pressure-dependent component is F(P_t) = F_n  t. Data for the responses to both the positive and negative ramps were use to calculate an average value of  over the time span of the test inflations (see APPENDIX I for details). Figure 7 shows an example of F_n-P_t plots before (Fig. 7, top) and after (Fig. 7, bottom) the removal of the time-dependent component, t. The latter provides a better estimate of the P_t-dependent relationship. Thus, if the neuronal response is dependent only on P_t, then the same F_n-P_t plot should be obtained whether response data from (+) or () test ramps are used. To estimate the ability of this method to reduce the differences between the F_n-P_t plots for the (+) and () ramp inflations, an average error index was calculated before and after the time compensation procedure. The error index was the average of the absolute difference between the two plots for the range of overlap. The average error between plots was reduced by 25.3% from 7.94 ± 0.42 to 5.93 ± 0.34 Hz (P < 0.0001) for the 41 neurons with multiple Pa_CO2 levels.

View larger version (16K): [in this window] [in a new window]   Fig. 7.   Example of Fn vs. Pt plots for both positive (P) and negative (N) ramp inflation patterns. Top: before time correction, average error between plots: 11.1 Hz. Bottom: after time correction procedure, average error between plots: 4.0 Hz. PaCO2: 79 mmHg; : 5.5 Hz/s.

Time-dependent component. The  values at each Pa_CO2 level for each neuron were also analyzed for their dependence on Pa_CO2. The slope and intercept values for plots of  vs. Pa_CO2 were obtained by linear regression for each of the 41 neurons. The mean values of the slopes and intercepts, weighted according to the number of Pa_CO2 levels per neuron (average no. of levels = 2.8) yielded the following average relationship:  = 0.62 + 0.027 · (Pa_CO2  40) Hz/s, where the slope (i.e., 0.027 ± 0.008) and the intercept at Pa_CO2 = 40 mmHg (0.62 ± 0.19) were significantly different from zero. This relationship indicates that the time-dependent effect is relatively small. For example, at a Pa_CO2 of 40 mmHg,  = 0.62 Hz/s and for a 10-s duration inflation, this component would contribute 6.2 Hz at the end of the response. At a Pa_CO2 of 60, the contribution would be 11.6 Hz.

Pa_CO2 effect on F_n-P_t relationship. To analyze the Pa_CO2 effect on the F_n-P_t relationship, the salient features of the plots were quantified using a piecewise linear approximation of the relationship after the time-dependent component was removed. To facilitate the analysis, the F_n-P_t data were sorted according to P_t value and placed in a histogram format and cursors were used to define the analysis ranges. The F_n-P_t relationship was then divided into three linear segments: one with a positive slope (Slp₀) and two with negative slopes (Slp₁ and Slp₂; Fig. 8). Standard linear regression techniques were used to obtain the best-fit lines. The intersection of the positive and negative slope lines was used to determine the P_t value or threshold (P_thr1) where the PSR-mediated inhibition began to reduce F_n. The actual F_n value, which corresponds to P_thr1 was defined as F_max. A second P_t value (P_thr2) was defined by the intersection of the two negative slope line segments. In addition, a third P_t value (P_thr0) was defined as the P_t at which lung inflation began to produce excitation of the E type-D neurons. This value was easiest to obtain at the lower Pa_CO2 levels when central inspiratory rhythm was slower or absent. However, it was also obtained at higher Pa_CO2 levels in cases where I-phase inhibition did not coincide in time with the low-P_t portion of the test ramp inflations. Typical plots of these parameters versus Pa_CO2 for an E neuron are shown in Fig. 9. Linear regression was used to quantify the relationship between each parameter and Pa_CO2 (lines through data points). The plots of F_max versus Pa_CO2 were best fit by a sigmoidal function of the form: F_max = F₀[x/(1 + x)], where x = (Pa_CO2/Pa_CO2*)^k, k is a noninteger exponent related to steepness, Pa_CO2* is the value of Pa_CO2 at which F_max = F₀/2, and F₀ is the asymptotic maximum of the sigmoid function.

View larger version (20K): [in this window] [in a new window]   Fig. 8.   Example illustrating method used to quantify the inflation response. A piecewise linear approximation of the relationship was used after time correction. See PaCO2 effect on Fn-Pt relationship for further details. Fmax, peak Fn.

View larger version (26K): [in this window] [in a new window]   Fig. 9.   Typical plots of the Fn-Pt relationship parameters vs. PaCO2. Data shown after time correction has been applied. Linear relationships were used to summarize parameter dependencies on PaCO2, except for Fmax where a sigmoidal relationship was used. Right: data for positive ramps. Left: data for negative ramps. Note the general agreement in parameter responses to PaCO2 for the 2 inflation patterns.

Average Pa_CO2-dependent parameters of the F_n-P_t relationship. After correction for time dependence, the corresponding parameters obtained from the best fits of the F_n-P_t plots (e.g., Fig. 8) for the (+) and () ramp responses were averaged at each Pa_CO2 level for each neuron. Table 1 summarizes the pooled data of 41 neurons. The intercepts have been translated to a Pa_CO2 value of 40 mmHg. Thirty-nine of forty-one neurons exhibited an Slp₂ segment. Data for the sigmoidal type relationship indicates that the average maximum F_n at high Pa_CO2 was 91.9 Hz and that 50% of this value was achieved at a Pa_CO2 = 37.1 mmHg. Significant linear relationships were found for Slp₀, Slp₂, P_thr0, P_thr1, and P_thr2, but not for Slp₁ (Table 1).

Average F_n-P_t relationship as a function of Pa_CO2. With the use of the mean values of Table 1, it is possible to produce a family of F_n-P_t relationships in which Pa_CO2 is the family parameter. Similar to the analysis, the empirical model for F_n-P_t relationship is comprised of line segments that are functions of P_t (F_a, F_b, F_c, and F_d, Fig. 10, top). These explicit P_t-dependent functions, which were used to approximate the F_n-P_t relationship, are given in Fig. 10, middle. The point P_thr1,F_max is the key starting point from which the F_n-P_t relationship is constructed. Because the line segments are also functions of Pa_CO2, the slope, intercept, and intersection point parameters of these line segments are functions of Pa_CO2 as defined in Fig. 10, bottom. These analytic relations were used to generate a three-dimensional surface plot of F_n as a function of both P_t and Pa_CO2 (Fig. 11). The strong dependence of F_n on both of these variables can be appreciated. The inflation-mediated excitation is much more sensitive to Pa_CO2 than the inflation-mediated inhibition (compare Slp₀ with Slp₁ and Slp₂, Fig. 10, bottom, and Fig. 11). The F_n-P_t relationship is maintained in the absence of central respiratory rhythm at Pa_CO2 levels typically <30 mmHg. The increases in threshold levels for P_thr0, P_thr1, and P_thr2 with increases of Pa_CO2 can also be seen.

View larger version (30K): [in this window] [in a new window]   Fig. 10.   Description of the empirical model for the E neuronal response to lung inflation and PaCO2. The mathematical description of each segment of the Fn-Pt relationship is given below the line plot (box, center). The parameters of the linear functions are themselves functions of PaCO2 as defined by the average relationships in the bottom box, which are based on the data from 41 E neurons. With this model, the Fn of an average E neuron can be calculated for any desired values of Pt and PaCO2.

View larger version (80K): [in this window] [in a new window]   Fig. 11.   Graphical summary of Fn as a function of both Pt and PaCO2 for the average E neuron.

View larger version (18K): [in this window] [in a new window]   Fig. 12.   Estimated relationship between Fn and PaCO2 at various Pt levels, as indicated (parentheses at right of traces). Data computed from the model of Fig. 10. Top, low Pt range; bottom, high Pt range.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX I REFERENCES

This study characterizes the integration of mechanosensory and chemosensory inputs by canine E neurons of the caudal VRG. Most of these neurons are presumed to be bulbospinal neurons on the basis of previous studies of E neurons in the same location, where >93% of those neurons were antidromically activated from the cervical spinal cord (5). An empirical model has been developed that can predict the instantaneous discharge frequency, F_n, of an average E bulbospinal neuron for given Pa_CO2 and transpulmonary pressure, P_t, values. In addition, the profile of the discharge pattern is also predictable for a given trajectory of P_t.

Previous studies have investigated the response of these neurons to each input in isolation (2, 4, 9, 31), but not to the combination of both inputs. Open-loop conditions were used to minimize confounding inputs from other sources or secondary effects. In this regard, neuromuscular blockade and bilateral pneumothorax were used to eliminate phasic afferent inputs from areas other than the lung, and test inflation patterns, separated by several control respiratory cycles, were used during hyperoxia (Pa_O2 >300 mmHg) to minimize the effects of transient changes in Pa_CO2 during test cycles. The delayed responses of neuronal and phrenic activities to step changes in inspired CO₂ concentration, which required 3-5 min to reach steady state, suggest that the hyperoxic conditions minimized inputs from the peripheral chemoreceptors that have a fast response time. Thus, in this study, the major source of chemosensory input to the E bulbospinal neurons is of central origin. The PSR-mediated reflex is also controlled by a GABAergic gain-modulating mechanism (20) that may be affected by barbiturates. However, a constant infusion of the thiopental anesthetic was used to maintain a stable blood concentration and hence a relatively constant level of anesthesia. Thus modulation of the GABAergic input by the anesthetic, if present, would be at a constant level and the nature of the interaction between Pa_CO2 and the PSR input should be unaltered.

On the basis of the sustained E neuronal responses to step inflation patterns, our previous (4) and current study (e.g., Fig. 1) suggest that both the excitatory and inhibitory components of the inflation response are mediated by the slowly adapting PSRs. P_t was used as an index of PSR activity, and vagotomy eliminates the inflation response (4). It is possible that the excitatory and inhibitory components are mediated by the two types of PSRs with appropriate characteristics that have been described in dogs (23-25) or by a single set of PSRs in conjunction with a central mechanism (4).

Pressure and time dependence of the E neuronal response. Although our analysis shows that a time-dependent factor contributes to the discharge pattern of these E neurons, its effect is relatively small and appears to be overridden by inputs from the PSRs, such that the time course of the discharge pattern is highly dependent on the time course of transpulmonary pressure (4). The small contribution of the time related changes in F_n is best illustrated by the E neuronal responses to step inflations, delayed ramp inflations, and by overlap of the F_n versus P_t plots for data from positive and negative ramp inflations, as demonstrated in Figs. 1, 2, and 7, respectively. When the F_n-P_t plots were corrected for time-dependent effects, the average error between plots for the positive and negative ramp inflations for the 41 neurons at the various Pa_CO2 levels was reduced from 7.9 to 5.9 Hz.

1

1

Central chemodrive dependence of the E neuronal response. The activities of both the control and test inflation cycles increased with increases in Pa_CO2 (e.g., Figs. 3 and 4). Plots of the peak F_n (F_max, Fig. 8), as well as F_n at various P_t levels (Fig. 12), versus Pa_CO2 were sigmoidal in shape (Fig. 9, middle). The steepest part of the F_max curve, which occurred at 50% of maximum, was located at an average ± SE Pa_CO2 of 37.1 ± 1.5 mmHg, and the Pa_CO2 range for the 20-80% response was 26-62 mmHg. In chloralose-urethane-anesthetized cats, Bainton and Kirkwood (2) also noted that plots of F_n versus alveolar PCO₂ (PA_CO2) for E bulbospinal neurons were steep and sigmoidal; however, the greatest sensitivity was found to occur at PA_CO2 values from 20 to 30 mmHg. These results suggest that the greatest sensitivity of the caudal VRG E neurons to Pa_CO2 occurs over a Pa_CO2 range somewhat lower than that for inspiratory (phrenic) activity (13).

Nature of interaction between chemosensory and mechanosensory inputs. On the basis of the plot of F_n vs. P_t and Pa_CO2 of the average E neuron (Fig. 11), the general impression is that these two inputs are mainly additive. That is, an increase in Pa_CO2 shifts the F_n-P_t relationship upward. However, the quantitative relationships used to generate Fig. 11 indicate that there is a synergistic interaction between Pa_CO2 and the excitatory component of the P_t response. The positive slope, Slp₀, increases by 38% for a change in Pa_CO2 from 30 to 50 mmHg and results in higher F_max values. The negative slope, Slp₁, is not altered by Pa_CO2 and can be seen as a parallel shift if the F_n-P_t relationship (Fig. 11). The negative slope, Slp₂, becomes less negative with increases in Pa_CO2, and the inhibition of the higher P_t range is less effective at higher Pa_CO2 levels.

Physiological significance of the model parameters. Although the empirical model for the F_n-P_t relationship (Fig. 10) is useful in quantifying the neuronal responses to arbitrary inflation patterns, the model parameters themselves have physiological correlates. P_thr0 indicates the threshold pressure at which the PSR activation begins to excite the E neurons, and Slp₀ represents the sensitivity of that reflex component. P_thr1 appears to represent the point where the P_t-related progressive recruitment of the PSRs and/or PSR activity, which mediate the inhibitory component of the E neuronal response to lung inflation, opposes further neuronal excitation, and actually reverses it. It is not known if there exists a separate group or type of PSRs that mediates the inhibitory response. This inflation-mediated excitatory/inhibitory interaction determines F_max and the sensitivity of the inhibitory component, Slp₁, as P_t increases. Because P_thr2 is the P_t value at which the two piecewise linear approximations of the curvilinear inhibitory response component intersect, it represents a point of maximum slope inflection. Slp₂ represents the sensitivity of the second subcomponent of the inhibitory portion of the F_n-P_t relationship.

Correlation of E bulbospinal neuronal with E muscular responses. The various thoracic and abdominal E muscles respond differentially to changes in Pa_CO2 and PSR inputs (1). It is possible that these differences may be accounted for by the variability in the sensitivities of the model parameters, although the behavior of the 41 type D neurons of this study was qualitatively similar. A measure of the degree of parameter variation is given by the 10th and 90th percentile values of Table 1. For example, in 80% of these neurons, the slope of the excitatory portion of the F_n-P_t relationship ranged from 5 to 30 Hz/mmHg and that of the inhibitory portion ranged from 5.2 to 20.4 Hz/mmHg (Table 1, intercept of Slp₀ and Slp₁ at a Pa_CO2 of 40 mmHg). In addition to parameter variability, it is possible that some E muscles do not depend on PSR feedback. For example, during postural changes and use of the rib cage, triangularis sterni E muscles are largely independent of vagal inputs, in contrast to E abdominal muscles, which rely on vagal feedback for this purpose (11).

Perspectives

PSR-mediated inhibition of E bulbospinal neurons appears to play an important role in E airflow control. At end-inspiration, elastic recoil pressure and P_t are greatest, and maximum inhibition of E bulbospinal neuronal activity occurs. As deflation proceeds, disinhibition results in an augmenting E neuronal discharge pattern, which would act to maintain E airflow when elastic recoil pressure is decreasing. Hypercapnia increases E neuronal F_n; however, tidal volume and peak P_t are also increased and PSR-mediated inhibition of F_n would minimize the contribution of E muscles in early expiration. Due to the higher recoil pressure, the lung would empty faster and disinhibition of the Pa_CO2-elevated E neuronal activity would aid in emptying the lung during the later part of the E phase. In addition, as lung volume and P_t decrease below the P_thr1, F_n decreases due to the reduction in the PSR-mediated excitation, effectively braking active expiration and limiting deflation below function residual capacity. Expiratory phase duration is also highly dependent on the E volume trajectory and history via the Hering-Breuer E facilitatory reflex (34), and the discharge pattern of the E bulbospinal neurons would therefore contribute to the control of T_E by altering PSR feedback.

E bulbospinal neurons appear to provide a good neural model in which the central integration of different types of information may be studied, such as those arising from mechanosensory (5, 7), chemosensory (12), and central pattern-generating (6, 15, 19) inputs. The model also allows the study of the neurotransmitters and synaptic mechanisms involved.

In summary, this study characterizes the control of the E neurons in the caudal VRG of dogs by the combination of inputs arising from PSRs and central chemosensory sources. The interaction between these two types of input appears to be mainly additive with regard to the PSR-mediated inhibition and synergistic with regard to the PSR-mediated excitation.