HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by McLachlan, E. M. Articles by Davies, P. J. Search for Related Content PubMed PubMed Citation Articles by McLachlan, E. M. Articles by Davies, P. J.

Analysis of the periodicity of synaptic events in neurones in the superior cervical ganglion of anaesthetized rats

Elspeth M. McLachlan, Heinz-Joachim Häbler *, John Jamieson and Philip J. Davies

MS 8128 Received 14 April 1998; accepted after revision 22 May 1998.

	ABSTRACT

Top Abstract Introduction Methods Results Discussion References

1. The patterns of on-going synaptic events recorded intracellularly in neurones of superior cervical ganglia (SCG)of anaesthetized female rats were analysed by constructing inter-event interval histograms, autocorrelograms, ln-survivor curves and histograms triggered by the arterial pulse wave and by the intercostal EMG.

2. In 11/12 cells with on-going frequencies > 0·5 Hz, one or two inputs were strong (i.e. always suprathreshold). In five cells, action potentials also arose from synaptic potentials with amplitudes close to threshold.

3. Synaptic events in 5/11 neurones tested were phase-related to the arterial pressure wave (i.e. had cardiac rhythmicity, CR).

4. Synaptic events in 9/10 neurones tested (including all with CR) were phase-related to the intercostal EMG and/or their autocorrelograms showed peaks at multiples of the respiratory interval (i.e. had respiratory rhythmicity, RR).

5. The intervals between all synaptic events were exponentially distributed in 8/12 neurones although intervals between single strong events showed peaks related to the respiratory cycle. Bursts occurred only by chance.

6. Event patterns could be simulated by combining events from several respiration-modulated inputs with their timing distributed over nearly half the cycle. From the simulations, the mean number of active preganglionic inputs was estimated to be ~6 with mean discharge frequency ~0·4 Hz.

7. We conclude that, in the spontaneously breathing anaesthetized rat, most preganglionic neurones to the SCG fire with relatively low probability in relation to the respiratory cycle. Rhythms in a postganglionic neurone reflect the activity of its suprathreshold preganglionic inputs.

	INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

The discharge of individual neurones in sympathetic pathways often demonstrates a particular rhythm. Sympathetic postganglionic vasoconstrictor activity in nerves to muscle, viscera and kidneys has pronounced pulse-related rhythmicity in anaesthetized animals (Jänig, 1985; Häbler et al. 1994b; Bartsch et al. 1996), as well as in conscious humans (Delius et al. 1972; Hagbarth et al. 1972; Sundlöf & Wallin, 1978). In contrast, other targets like the skin are supplied by postganglionic neurones mostly lacking cardiac rhythmicity (CR) (Hagbarth et al. 1972; Jänig, 1985; Häbler et al. 1994b) and CR is not a feature of neurones innervating visceral effector tissues (Bahr et al. 1987). Many postganglionic vasoconstrictor neurones also show respiratory rhythmicity (RR) (Boczek-Funcke et al. 1992; Häbler et al. 1993, 1994a, 1996; Johnson & Gilbey, 1994; Macefield & Wallin, 1995), but the activity of neurones supplying non-vascular targets in the cat pelvic organs is not coupled to the respiratory cycle (Boczek-Funcke et al. 1992). The rhythmicity is characterized by a tendency to discharge in one particular phase of the cardiac or respiratory cycles. These rhythms reflect the integrative behaviour of the central neural circuits controlling the sympathetic outflow, particularly to the cardiovascular system.

Superior cervical ganglion (SCG) cells are a functionally heterogeneous population of neurones, as they innervate a wide variety of target tissues. It is estimated that about 50 % of the SCG neurones are vasoconstrictor (to both skin and muscle), about 25 % are secretomotor in the salivary glands in rodents, and the remainder are pilomotor, pupillomotor, or project to the airways or pineal gland (Gibbins, 1991). Because of these different targets, one might expect only subpopulations of SCG neurones to show pulse and/or respiratory rhythmicity in their discharge.

In sympathetic ganglia, there is extensive convergence and divergence from pre- to postganglionic neurones. In rat SCG, for example, it has been shown in vitro that each postganglionic neurone receives input from on average nine preganglionic axons and the ratio of pre- to postganglionic neurones is 1: 20 (Purves et al. 1986). In most paravertebral ganglia, only one or two (rarely three) of these inputs is always suprathreshold ('strong' or dominant) and the others produce subthreshold or 'weak' excitatory synaptic potentials (ESPs) (Skok & Ivanov, 1983; Hirst & McLachlan, 1984; Cassell & McLachlan, 1986). It is not clear whether all these inputs are active in vivo. In recent experiments in which we recorded intracellularly from SCG neurones in anaesthetized rats (McLachlan et al. 1997), it was noted that transmission only rarely occurred by summation of ESPs from convergent preganglionic axons, even during bursts of activity elicited by pinching the skin. It appeared that the discharge of the postganglionic neurones is mostly determined by the discharge of its strong inputs.

It is not clear whether all preganglionic inputs converging on a particular ganglion cell discharge with the same rhythms, although it is known that preganglionic neurones exhibit types of rhythmicity similar to those of postganglionic neurones, at least in the lumbar sympathetic outflow (Jänig, 1985). The simplest explanation for the preservation of well-defined rhythms across the ganglionic synapse is that only the strong inputs transmit their patterned signals. Another explanation would be that all the preganglionic inputs to a given SCG neurone have the same rhythmical firing patterns and these are relatively synchronized, so that both strong inputs and summation of ESPs lead to discharge in relation to cardiac or respiratory cycles.

Here, we have analysed further intracellular recordings of membrane potential from SCG neurones in anaesthetized rats in order to examine the patterning of synaptic inputs. We examined the relation of events to the cardiac and respiratory cycles and also the intervals between synaptic events of different strengths. We sought to ascertain (i) whether synaptic events occurred with particular relationhips to the cardiac and respiratory cycles, (ii) whether bursts of synaptic events arising from convergent preganglionic inputs to a particular neurone were synchronized by these cyclic rhythms, so that summation of ESPs contributed to rhythmic discharge, and (iii) whether inputs of different synaptic strengths to a given neurone showed the same patterns of activity. We used simulations to identify how the observed patterns of synaptic events could result from activity in convergent preganglionic inputs. The simulations enabled us to suggest limits on the patterns of preganglionic activity and the amounts of convergence for each recorded cell.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Experimental procedures have been described in detail previously (McLachlan et al. 1997). Briefly, female Wistar rats were anaesthetized with urethane (1-1·3 g kg^-1 I.P.) or sodium pentobarbitone (45 mg kg^-1 I.P.) and given supplementary doses as necessary (via a catheter in the right saphenous vein) to maintain anaesthesia, the depth of which was considered adequate by the absence of spontaneous fluctuations of blood pressure and withdrawal reflexes to pinching the foot. The majority of data reported here comes from cells recorded in animals under pentobarbitone anaesthesia. The trachea was cannulated and the animals spontaneously breathed air supplemented with O₂. Core temperature was kept relatively low (33-36°C) to enhance cutaneous vasoconstrictor activity. Blood parameters during the recordings from seven animals were (mean ± S.E.M. (range)): arterial P_O2 (Pa,O₂), 124 ± 11 mmHg (71-184 mmHg); arterial P_CO2 (Pa,CO₂), 55 ± 2 mmHg (45-61 mmHg); and pH 7·3 ± 0·0 (7·2-7·3). Respiratory EMG was recorded via two steel needle electrodes stitched into the external intercostal muscles. The animals were killed at the end of the experiment with an overdose of pentobarbitone (120 mg kg^-1 I.V.). These procedures were approved by the Animal Care and Ethics Committee of the University of New South Wales.

The efferent nerve trunks from the left SCG were cut, the SCG with the intact cervical sympathetic trunk was freed from the carotid artery and pinned out on a rubber platform. Intracellular recordings were made from ganglion cells using microelectrodes filled with 0·5 M KCl (resistance 60-100 M) while the preparation was superfused either with warmed paraffin oil or with physiological salt solution, in each case bubbled with 95 % O₂-5 % CO₂. Criteria for impalements have been given elsewhere (McLachlan et al. 1997). Data were recorded using a MacLab/S system and the program Chart (ADInstruments, Castle Hill, NSW) (1 kHz sampling frequency for membrane potential, 200 Hz for EMG and arterial pressure).

Synaptic activity was recorded with the membrane hyperpolarized by 15 ± 2 mV (6-22 mV) from resting membrane potential (RMP) in order to block the majority of action potentials and so determine the amplitude of the underlying excitatory synaptic potentials (ESPs) (see Figs 1B and 2A). Individual strong inputs could be recognized by the configuration of the after-depolarization/after-hyperpolarization following the action potential and the large and relatively constant amplitude of the underlying ESP if the action potential could be blocked by large hyperpolarizations (see McLachlan et al. 1997). These inputs are referred to as 'strong' throughout this paper. As also described in our earlier paper, there was sometimes a peak of ESPs > 15 mV in amplitude under these recording conditions, most of which would have elicited action potentials at RMP, as predicted from threshold ESP amplitudes at RMP (12 mV) (see McLachlan et al. 1997) and the reversal potential (0 mV, Yawo, 1989). It was not possible to know how many active inputs contributed to this (or any other) peak in the amplitude distribution. For analysis, all action potentials and ESPs > 15 mV amplitude were therefore pooled as 'suprathreshold' responses (including strong responses); ESPs < 15 mV were pooled as 'subthreshold'. Interval analyses were performed for these events both separately and together.

The present analysis is based on twelve SCG cells with relatively high resting activity ( 0·5 Hz) from which impalements were held for 39 ± 16 min (range 7-82 min). The analysis was performed on data over periods of 55 s to 13 min, during which 746 ± 218 (range 56-2414) synaptic events were recorded, including 228 ± 84 (range 5-809) suprathreshold ones. Over these periods, mean arterial blood pressure was constant and > 70 mmHg and the mean and variance of the intervals between synaptic events were constant.

Rhythmicity

The presence of cardiac rhythmicity (CR) in subthreshold events, in suprathreshold events and in all events was determined by pulse-triggered histograms. Synaptic event timing was triggered by the peak of the differential of every second arterial blood pressure wave (see Fig. 2B), so as to reveal the degree of cyclic activity over two consecutive cardiac cycles (see Fig. 2). The differentiation, automatic peak detection and consequent trigger generation was performed using off-line analysis in Chart. Histograms were constructed as a two-cycle window using 4 ms bins. CR was quantified by determining the minimum (min) and maximum (max) number of events in eight consecutive bins (i.e. 32 ms) and deriving the value 100 × [(max - min)/max] averaged over the two cycles (for details see Boczek-Funcke et al. 1991). CR was considered to be present if the histograms of double cardiac cycles revealed that the timing was clearly cyclic, and if the CR ratio was > 40 %. The link to the cardiac cycle was taken to be strong if CR was > 60 % (Häbler et al. 1994a).

The presence of respiratory rhythmicity (RR) in subthreshold events, in suprathreshold events and in all events was recognized, where possible, from peri-EMG time histograms (bin size of 10 ms, see Fig. 4 and Häbler et al. 1996). Using as a marker the midpoint of the intercostal inspiratory EMG burst marked manually in Chart as a trigger, the timing of synaptic events was analysed with respect to every second inspiration (Fig. 2A).

These analyses were applied only to data sets of more than 100 synaptic events.

Analysis of relation between synaptic events

The intervals between synaptic events were measured with 1 ms resolution. The distributions of intervals between subthreshold events, between suprathreshold events and between all synaptic events were compared with those predicted for Poisson processes with the same mean intervals (Cox & Lewis, 1966). Time interval histograms were constructed. Cumulative probability functions on data sets > 250 events were compared with those predicted for a Poisson process using a single-sided Kolmogorov-Smirnov test (StatView®, Abacus Software, Berkeley, CA, USA) with a significance level of 1 % (Zar, 1984) or 0·3 % for n > 1000 when the test is more stringent.

Ln-survivor curves were constructed for each data set and for the corresponding Poisson prediction. Comparing the real and Poisson predicted ln-survivor curves is an established method for detecting an excess of events at short intervals, i.e. bursting (Cox & Lewis, 1966; Bornstein, 1978). The ln-survivor curve is the plot of the natural logarithm of the survivor function (= 1 - the cumulative probability function), R(t) = 1 - i/(N + 1) (i.e. the proportion of intervals t, where N is the number of intervals and i is the number of intervals < t), which is linear for a Poisson process with a negative slope equal to the mean event frequency. The presence of high frequency bursts yields a ln-survivor curve which is concave at short time intervals (see Cox & Lewis, 1966), because of the excess of short intervals, whereas a convex curve indicates that the process is ordered (see Bornstein, 1978). An ordered process has more intervals close to the mean interval than a Poisson process. Deviation below the Poisson prediction at long intervals indicates their relative absence, whereas deviation above the linear prediction implies too many long intervals, i.e. periods of silence or inhibition. (It should be noted that the logarithmic scale exaggerates the deviations at longer intervals.) The fit to the linear prediction was assessed by comparison with the 95 % confidence limits for the Poisson rate parameter (= 1/mean interval). These limits were derived as (A + ) ± (A(A + 2)), where A = 0·5c²/(length of record in seconds) (Cox & Lewis, 1966), where c is the 95 % point of the normal distribution (= 1·96).

Another means that we used to identify rhythmicity was to plot interval autocorrelograms (histograms of intervals between each event and all subsequent events) (see Johnson & Gilbey, 1996). Autocorrelograms (bin widths of 0·1 s) were arbitrarily limited to 5 s.

Simulation of synaptic event patterns

The synaptic events recorded in individual neurones were simulated in a computer model. The distribution of events produced by each simulated 'input' was derived from a random series generator (StatView®). Events resulting from each of several numerically derived inputs were interleaved in time to produce a corresponding sequence of 'synaptic events'. The number of inputs and their pattern of occurrence were varied. The pattern for each input was constructed separately but the events pattern generated by the combination of various inputs was made to match the observed events pattern for each cell. The models were constructed to have the minimum number of inputs consistent with the observed sequence. Event series arising from various numbers of inputs and combinations of different distributions, either unsynchronized or synchronized, were merged over a period that matched the recording period to generate interval histograms with similar numbers of events to those observed, or over about 250 s for theoretical modelling. Normal, exponential and uniform distributions were tested.

Goodness of fit was determined by comparing the shape and size of interval histograms, ln-survivor curves, autocorrelograms, and respiration-triggered histograms with those derived from the recorded data and applying the ² test. The observed ESP amplitude distribution also provided information about the likely number of active preganglionic inputs.

Data are presented as means ± S.E.M. and range unless stated otherwise.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Resting synaptic activity ( 0·5 Hz) was analysed over periods of 4·5 ± 0·9 min to detect periodicity and clustering of events and the data were simulated by combining events arising from multiple inputs using a computer model.

The frequency of on-going synaptic events ranged from 0·5 to 5·4 Hz (mean 2·6 ± 0·5 Hz). Ten neurones were classified as burst-inhibitory (BI) and two as excitatory (E) on the basis of their reflex responses to activation of cutaneous nociceptors (see McLachlan et al. 1997). Neurones recorded in the same animal (see Tables 1 and 2) showed different frequencies and patterns of activity.

The passive properties of neurones in this preparation have been described in detail previously (McLachlan et al. 1997). In this subset, resting membrane potential (RMP) was -48 ± 2 mV (range -38 to -57 mV), input resistance was 103 ± 13 M (range 58-169 M) and the time constant was 12 ± 2 ms (range 4-24 ms). Mean arterial blood pressure (MABP) during analysis periods was 95 ± 3 mmHg (range 76-112 mmHg).

ESP amplitude distributions

At RMP, all neurones discharged action potentials. Action potentials appeared to arise from single synaptic events of suprathreshold amplitude (Fig. 1b), i.e. summation of subthreshold ESPs to threshold was extremely rare. The recordings were made with the membrane hyperpolarized to -63 ± 1 mV (range -58 to -68 mV), thus increasing the ESP amplitude and revealing some ESPs underlying action potentials at RMP (Figs 1 and 2). In 8/12 cells, the distribution of subthreshold ESP amplitudes had one or more peaks (see Figs 3-5 and 7) that were wider than the variance of ESPs arising from a single preganglionic axon (McLachlan, 1975). This implies that ESPs of similar amplitudes arose from several axons so that the events arising from individual inputs could not be separated for analysis.

In five neurones (4 BI, 1 E), one (single) strong input could be clearly identified and was analysed separately, as discussed below. These cells lacked ESPs > 15 mV in amplitude (see Fig. 7), as was the case for two other cells with two strong inputs. In the five other cells (Figs 3A and B and 4A), there was a peak in the ESP histogram around and above threshold (15 mV, see Methods); four of these cells had either one or two strong inputs whereas the remaining cell lacked an active strong input (Fig. 5A). As ESPs > 15 mV would probably have triggered action potentials if the membrane was at RMP (see Methods), these were included with the action potentials in the subsequent analyses (as 'suprathreshold events').

		View larger version [in this window] [in a new window]
Figure 1. On-going synaptic activity in a neurone in the superior cervical ganglion (SCG) of a pentobarbitone-anaesthetized rat Membrane potential recorded at resting membrane potential (RMP, -45 mV) (A) and hyperpolarized to -62 mV (B). One suprathreshold input was always large enough to evoke an action potential (indicated by 1). This 'strong' input elicited action potentials that were followed by an after-depolarization which truncated the after-hyperpolarization (AHP), whereas another input (indicated by 2) was just suprathreshold at RMP, as action potentials were readily blocked by hyperpolarization. Expanded traces to the right (a, b, c) are from the regions shown by dashed lines. Arrow in a indicates threshold for the action potential. Note that, in the group of responses in b, the action potential arises without an inflexion and is followed by an ESP that is abbreviated in time course by the AHP. This ESP probably arose from input 2. Mean frequency of synaptic events 4·7 Hz, MABP 95 mmHg.

		View larger version [in this window] [in a new window]
Figure 2. Relating synaptic activity in a neurone in the SCG to physiological parameters A. recordings of arterial blood pressure (top trace), electromyogram of the external intercostal muscle (second trace), membrane potentials in a neurone (third trace) and transmembrane current (bottom trace). Note that the cell was held at a hyperpolarized membrane potential of about -65 mV. The on-going postsynaptic activity consists of several subthreshold excitatory synaptic potentials (ESPs) of varying amplitude and a strong input (marked by dots). Depolarizing current pulses were passed through the microelectrode at intervals to monitor passive membrane properties. The one shown summed with on-going ESPs so that one of them initiated an action potential (marked by ). Peri-EMG histograms (see Fig. 4) were constructed by counting the occurrence of synaptic events at different times during many respiratory cycles in relation to the midpoint of the EMG burst (indicated by arrow). B, arterial blood pressure waves (top trace), their differential (middle trace) and membrane potential (bottom trace) on an expanded time base. Pulse-related rhythmicity (see Fig. 3) was analysed over two cardiac cycles by adding the occurrence of synaptic events during many cardiac cycles in relation to the peak of the differential of the pressure wave (indicated by arrows).

Cardiac rhythmicity

Heart rate was 5·7 ± 0·2 Hz (range 4·8-7·0 Hz), usually much faster than the frequency of synaptic events. Pulse-triggered histograms of synaptic events (789 ± 113 double cardiac cycles, range 443-1683) showed cardiac modulation to be present in only five of the BI neurones (and not in the one E neurone analysed), with synaptic events having the lowest probability of occurring 0·1-0·15 s after the maximum dP/dt (Fig. 3). The amount of CR ranged from 51 to 77 % (Table 1). Modulation when present always occurred in the suprathreshold events and could be detected also in the subthreshold ones in three cases (cf. Fig. 3). The presence of pulse-related modulation was not correlated with the mean level of arterial blood pressure during recording from different neurones (Table 1).

		View larger version [in this window] [in a new window]
Figure 3. Modulation of synaptic events by the arterial pressure pulse A, amplitude histogram of synaptic responses in a BI neurone with the membrane held at -65 mV (A1) and pulse-triggered histograms of the timing of these responses over two cardiac cycles in relation to the maximum dP/dt of the arterial pressure wave (A2,3). Histograms were constructed separately for synaptic responses >15 mV (A2) and < 15 mV (A3). The former would probably have been suprathreshold at a normal resting membrane potential of -50 mV. Both populations would have arisen from more than one axon. Data were averaged over 488 cycles. Mean frequency of synaptic events 4·8 Hz, mean arterial blood pressure (MABP) 85 mmHg. As in many of the cells, no pulse-related rhythmicity could be detected. B, similar histograms in another BI neurone with inputs that showed pronounced pulse rhythmicity. B1 shows the distribution of response amplitudes with the membrane held at -70 mV. B 2 and 3 are for the timing of supra- (>15 mV) and subthreshold synaptic responses, respectively, with respect to peak dP/dt (averaged over 1990 cycles, mean frequency of synaptic events 3·1 Hz, MABP 103 mmHg). The phase of maximum inhibition occurred 130 ms after peak dP/dt. C 1 shows the distribution of response amplitudes in another BI neurone held at -60 mV. C 2 and 3 show a single suprathreshold and several subthreshold (< 15 mV) synaptic responses, respectively, averaged over 464 cycles. Mean frequency of synaptic events 2·2 Hz, MABP 95 mmHg. The single suprathreshold input (C 2) showed pulse-related rhythmicity whereas the subthreshold inputs did not (C 3). For all amplitude frequency histograms (here and in subsequent figures), the number of action potentials not blocked at the holding potential are shown in the hatched column to the right.

Table 1. Patterning of synaptic events in SCG neurones

Evidence for respiratory rhythmicity (RR)

ln-survivor curve

Cell

Animal

Type

CR All

MABP

All

Supra

Prob (supra)

Sub

Prob (sub)

All

Supra

Sub

1 *

A

BI

77 %

90

Yes

A, H

1·00

A, H

0·25

Cx

Cx

L

2 *

B

BI

51 %

95

Yes

A, H

0·33

R

0·33

L/a-

Cx

L/a

3 *

C

BI

No

90

Yes

A, H, R

0·75

A, H, R

0·15

Cx-

Cx

L

4 *

D

BI

59 %

95

Yes

A, H, R

1·00

H, R

0·80

Cx-

Cx

Cx

5

E

BI

52 %

105

Slight

-

-

H, R

0·33

L/a+

L

L

6

E

BI

No

93

No

No

-

No

-

L/b+

Cx

L/b

7

E

BI

No

76

-

-

-

-

-

L

L

L

8

E

BI

65 %

103

Yes

A, H, R

0·67

R

0·25

L/a+

Cx

L/a

9

F

BI

No

112

Yes

A, R

0·86

A, R

0·67

L+

Cx

L

10

F

BI

No

88

Yes

A, H, R

0·67

A, R

0·67

Cx-

Cx/a

L/b

11 *

G

E

-

85

-

-

-

-

-

L/a

-

L/a

12

F

E

No

102

Slight

A, H

0·15

-

-

L/a+

L

L

Respiratory rhythmicity

EMG-triggered histograms over 201 ± 22 double respiratory cycles (range 134-306) showed evidence of respiratory modulation in 6/8 BI neurones. Respiratory rate was 0·8 ± 0·1 Hz (range 0·6-1·1 Hz); in all animals, the respiratory interval was relatively constant during the recordings (S.D. 0·06 ± 0·02 s). A peak at the respiratory period was also evident in the autocorrelograms of synaptic events for five of these cells. In the two other BI cells and one E cell, for which EMG-triggered histograms were not available, the autocorrelograms also had a peak at the respiratory period (as indicated by the respiratory oscillations in blood pressure). We refer here to the presence of either of these periodicities as respiratory rhythmicity (RR), which was present in 9/10 neurones tested (Table 1). The modulation usually consisted of a relative inhibition of synaptic events during inspiration (Fig. 4A and B) although there was an inspiratory peak in one cell and also for some subthreshold inputs. Most commonly, peak activation occurred during the first part of expiration (Fig. 4A) (see Häbler et al. 1994b).

There were differences in the degree of modulation for inputs of different amplitudes, i.e. suprathreshold inputs tended to show greater modulation than subthreshold ones. As for CR, the phase relationships for the two types of input were not always identical (see Fig. 4A). All cells showing CR also had RR (Table 1).

		View larger version [in this window] [in a new window]
Figure 4. Modulation of synaptic events during the respiratory cycle A, amplitude histogram of synaptic responses in a BI neurone (A1, membrane held at -68 mV) and peri-EMG histograms showing the timing of these responses over two respiratory cycles (A2 and 3). The average smoothed and rectified intercostal EMG is shown for reference (A2, B 2). Synaptic responses >15 mV (A2) and < 15 mV (A3). Frequency of synaptic events 5·4 Hz, MABP 112 mmHg. Average of 306 cycles. As in most cells, marked respiratory rhythmicity was present in both kinds of inputs, with maximum inhibition in inspiration. Note slightly earlier peak activity in subthreshold responses. B, response amplitude histogram in another BI cell (B 1) held at -65 mV, showing respiratory rhythmicity of intervals in both a single suprathreshold (B 2) and >1 subthreshold (B3) responses. Mean frequency of synaptic events 0·8 Hz, MABP 90 mmHg. Average of 190 cycles. This neurone showed peak activation soon after inspiration.

Analysis of relation between synaptic events

Interval histograms usually peaked at short intervals, even when the number of events was small. The interval histograms of all synaptic events could be fitted by single decaying exponentials in 5/9 neurones (P > 0·01, Kolmogorov-Smirnov, Table 1, see Figs 10C 1 and 11C 1), including one E neurone. In two of the cases that did not fit an exponential, the histograms were positively skewed but with fewer short intervals than an exponential distribution (Fig. 5A; see Johnson & Purves, 1983) and, in the other two, a peak was present between 0·3 and 2 s (Fig. 5B). This interval peak bore no relation to the cardiac period (0·18 ± 0·01 s), but was related to the respiratory period (1·2 ± 0·3 s) (see below). The longest interval in each cell was 5 ± 1 s (range 1-12 s). In seven cells, there were intervals > 2·5 s (see Fig. 7), i.e. there were relatively silent periods.

		View larger version [in this window] [in a new window]
Figure 5. Distributions of amplitudes and intervals between synaptic responses A1, amplitude histogram of synaptic responses in a BI cell with the membrane held at -55 mV. A2, interval histogram of all synaptic events has a peak at small values but the fit to an exponential is poor (P = 0·001). Mean frequency of synaptic events 4·0 Hz, MABP 88 mmHg. B 1, amplitude histogram of synaptic responses in a BI cell (with a single strong input) with the membrane held at -60 mV. B 2, interval histogram for all synaptic events has a broad peak between 0·5 and 1 s. Mean frequency of synaptic events 2·4 Hz, MABP 90 mmHg.

The presence of clustering at short intervals was tested by constructing ln-survivor curves (see Methods). In six BI neurones and both E neurones, the plots were close to linear (e.g. Fig. 6A, Table 1); these plots indicated that the intervals were exponentially distributed, even in two cases in which the Kolmogorov-Smirnov test could not be applied. In two of the BI cells (Cells 2 and 7 in Table 1), the curves for the subthreshold inputs became slightly concave between intervals of 100-800 ms; this implies clustering of ESPs but the intervals were too long to permit summation. For the other neurones, the curves were convex (Fig. 6B and C) and this was associated with peaks in the interval distributions (e.g. Figs 7A2 and C 2).

It was notable, however, that the curves that were approximately linear had shapes which at several points lay very close to the 95 % confidence limits. At short intervals (< 50 ms), the curve usually lay just above the linear prediction, implying fewer brief intervals than would be expected if they occurred randomly. This mismatch might be explained by the failure to detect some small amplitude ESPs superimposed on the falling phase of much larger ones, but might also result from recurrent inhibition (Lebedev et al. 1980). We had, however, no difficulty in detecting large events occurring only 1-2 ms after smaller ones. At longer intervals on the ln-survivor plot, the last few points often lay close to or even above the upper confidence limit (Fig. 6A, Table 1). The simplest explanation for this deviation is that there were occasional periods of inhibition.

		View larger version [in this window] [in a new window]
Figure 6. Tests for clustering of synaptic events A, the ln-survivor function for all synaptic events lies very close to the 95 % confidence limits for the linear prediction for a random and independent process. Note the deviations above the line at long intervals. B, in another neurone, the ln-survivor function for all synaptic events is convex indicating that the data are not random. Same cell as Fig. 5B. C, in a neurone with a single strong input, the convex ln-survivor function for the intervals between events arising from this input shows an inflexion at intervals close to the respiratory interval (1·2 s). Same cell as Fig. 4B.

Synaptic event intervals in cells with single strong inputs

In five cells, the activity of a single strong input was identified independently. The frequency of these strong inputs ranged from 0·04 to 1·3 Hz (mean 0·60 ± 0·24 Hz) but one was not analysed (an E cell in which only five action potentials occurred during the recording period of 105 s, Cell 11 in Table 2).

Of the BI cells (Cells 1-4 in Tables 1 and 2), three had only small amplitude subthreshold events clearly distinct from the suprathreshold responses (Fig. 7A1-C 1, cf. e.g. Fig. 5A). In only one cell did the interval histogram for all events resemble a single exponential (Fig. 7B 2) and, in three cases, the ln-survivor curve for all events was convex (Table 1). There were only rare intervals < 250 ms between strong responses (a total of 3 intervals in 2/4 cells). In all cases, there was a large peak in the interval histogram between 0·8 and 2·8 s and an excess of long intervals (Fig. 7A3-C 3).

There was no evidence of a peak at the cardiac period (180-200 ms) in the interval distribution of any of the single strong inputs although three of the cells with strong inputs showed CR (Table 1); this is consistent with the rate of firing of individual preganglionic axons (generally < 2 Hz, Gilbey et al. 1986; Lewis & Coote, 1995; Bartsch et al. 1997) being very much lower than the heart rate. All of the single strong inputs discharged with a modal interval related to the respiratory period (Fig. 7; see Gilbey et al. 1986) but respiratory rhythmicity in the subthreshold inputs was generally obscured by activity in other convergent active inputs (see simulations below). Ln-survivor curves for single strong inputs were all convex and their autocorrelograms all showed peaks at the respiratory interval.

The subthreshold events in these cells occurred at 0·9 ± 0·4 Hz (n = 5, range 0·3-2·3 Hz). In two cells, the intervals between ESPs were widely distributed, including some very long intervals (see Fig. 7A4). In only one case (Fig. 7B 4) did the interval distribution resemble an exponential and, in another (Fig. 7C 4), there was a large peak, as for the strong inputs. Overall, the mean interval between subthreshold events (1·8 ± 0·7 s, range 0·4-3·9 s, n = 5) was not significantly different from that between the strong ones (1·7 ± 0·5 s, range 0·8-2·8 s, BI cells, n = 4) but, because the activity arose from multiple inputs, the mean frequency of individual subthreshold inputs must have been lower.

		View larger version [in this window] [in a new window]
Figure 7. Interval histograms for three BI neurones in which a single strong input could be analysed independently A1, amplitude histogram of synaptic responses with the membrane held at -65 mV. A2, interval histogram of these responses shows a frequency peak at about 1·2 s. Separate interval histograms for the strong input (A3) and for all subthreshold events (A4) are shown below. The relative distribution of respiratory periods is shown as hatched columns. The strong input (mean frequency 0·58 Hz) had a peak of intervals around the respiratory period and some longer intervals close to multiples of the respiratory period. Subthreshold inputs occurred infrequently (at 0·26 Hz) so that the activity of the strong input dominates the overall histogram. RR but not CR was present in the strong input. Mean cardiac period 154 ms. Same cell as Fig. 4B. B1, amplitude histogram with the membrane potential held at -60 mV. B 2, interval histogram of these responses was similar to a Poisson prediction except for the presence of some intervals >1·5 s. Separate interval histogram for the strong input (B 3) with low frequency discharge (0·35 Hz) reveals a peak at about three times the respiratory period (hatched columns). B 4, subthreshold events occurred at a higher frequency (1·85 Hz); these data do not fit an exponential (P < 0·01, Kolmogorov-Smirnov), having a deficit of very short and an excess of long intervals. Both CR (55 % in the strong input) and RR were present. Mean cardiac period 159 ms. Same cell as Fig. 3C. C 1, amplitude histogram with the membrane potential held at -60 mV. Interval histogram of synaptic responses (C 2) showing a peak at about 0·7 s. Separate interval histograms for the strong input (C 3) and for all subthreshold events (C 4) are shown below. The strong input (mean frequency 0·63 Hz) discharged with greatest probability around the respiratory period (hatched columns). Subthreshold inputs occurred more frequently (at 1·1 Hz) predominantly at about half the respiratory period. All inputs had CR = 59 % and RR. Mean cardiac period 204 ms, mean frequency of synaptic events 1·8 Hz, MABP 95 mmHg.

Synaptic event intervals in cells with two or more suprathreshold inputs

In 7/12 cells, more than one suprathreshold event could be distinguished (see Table 2). The average frequencies of supra- and subthreshold events in these cells were 1·1 ± 0·4 Hz (range 0·1-2·9 Hz) and 1·9 ± 0·4 Hz (range 0·5-3·3 Hz), respectively. In four cells with more than fifty suprathreshold events occurring within 6 ± 2 min (range 3-13 min), there was a hump in the distribution of intervals between suprathreshold events at 0·3-1 s but many short intervals (< 250 ms) were always present. The histograms for the intervals between multiple suprathreshold inputs all took the form of the histogram shown in Fig. 7C 4 and, for the intervals between subthreshold inputs, like those in Fig. 7B 4, i.e. they were exponential (n = 7, Kolmogorov- Smirnov). Most ln-survivor curves for events in these neurones (including one E cell) were close to linear (Table 1). In 4/6 BI neurones, the suprathreshold events showed convex ln-survivor curves and there was evidence of RR in the autocorrelogram and/or EMG-triggered histogram.

Table 2. Estimated numbers and discharge frequency of inputs active under resting conditions

Simulation of synaptic events arising from convergent inputs

We used the information about discharge patterns of strong inputs as the basis for mimicking the behaviour of several preganglionic inputs. Patterns of event intervals were simulated by generating interval series for multiple 'inputs' and merging these to reproduce the overall interval histogram, ln-survivor curve, etc., observed experimentally. Because several preganglionic inputs with on-going activity converge onto each ganglion cell, the overall distribution of synaptic intervals depends not only on the pattern of discharge of individual inputs but also on the number of active inputs.

Two types of input were modelled - those whose occurrence was synchronized to a 'respiratory cycle' and those that were not synchronized.

Unsynchronized inputs

Inputs that occurred without relation to a cyclic event were taken to occur at either exponentially, uniformly or normally distributed intervals. The mean interval and variance for each input and the number of inputs were modified.

Combinations of inputs with exponentially distributed intervals always gave an interval distribution that was also exponential, and combinations of other unsynchronized distributions also tended to be exponential as the number of inputs increased. Two or more inputs with uniform interval distributions produced an interval histogram that was essentially exponential (Fig. 8A2 and 3). Multiple unsynchronized inputs with normally distributed intervals also produced a skewed interval distribution but an exponential distribution resulted only when seven or more inputs were combined.

Ln-survivor curves for exponentially distributed intervals were always linear, whereas uniformly (Fig. 8A4) and normally distributed ones yielded convex plots. When increasing numbers of inputs were combined, the ln-survivor curves progressively became more linear (Fig. 8A4). Autocorrelograms for multiple inputs were also flat. These models indicate that multiple inputs of any distribution converge to give intervals that are distributed exponentially. However, the absence of a peak in the corresponding autocorrelogram at the respiratory interval is not consistent with much of the recorded data. Observed inputs without RR were simulated by either exponential or uniform interval distributions (see below) and not by unsynchronized normally distributed intervals (for which there is no justification).

		View larger version [in this window] [in a new window]
Figure 8. Simulation of interval histograms and ln-survivor functions for events arising from convergent inputs A1, distribution of intervals between approximately 200 events during 240 s resulting when a single input has a uniform interval distribution (mean 1·25 s) providing a range of intervals similar to those observed experimentally. The distribution changes when 2 such inputs occurring independently of each other are combined (A2), and when 6 such inputs are combined (A3). The distributions in A2 and A3 are both good fits to exponentials (P > 0·01, Kolmogorov-Smirnov). A4 shows the ln-survivor curves (continuous lines) for the same event series, together with the linear predictions for random events with the same mean interval (dashed lines). B 1, the distribution of intervals for a single input, between 200 events with normally distributed delay with respect to the respiratory marker (S.D. 0·2 s) (see text) and occurring every respiratory cycle. The respiratory period is normally distributed with a mean of 1·25 ± 0·05 s (S.D.) (B 2). When events from 2 (B 3) and 6 (B 4) such synchronized inputs (i.e. with the same delay and variance) are combined, the resulting interval distributions have two peaks, one at very short intervals and the other just below the respiratory interval. The ln-survivor curves (B 5) for these event patterns show sharp inflexions never seen in the experimental data. The vertical dashed line in B shows the mean respiratory interval.

Inputs synchronized to the respiratory cycle

Simulations were also generated with each 'input' occurring with a variable delay relative to an arbitrary marker during the 'respiratory cycle', that was itself slightly variable in duration. Three sources of variation in the 'respiratory-modulated' model were introduced.

(i) The distribution of respiratory intervals was taken to vary in a manner similar to those recorded from the intercostal EMG (i.e. normally distributed with mean interval 1·25 ± 0·05 s) (Fig. 8B 2).

(ii) The timing of an input with respect to the respiratory marker was taken to be distributed normally with a standard deviation of 0·2 s, so as to mimic the data for single strong inputs. Inputs occurred at different mean delays with respect to the respiratory marker.

(iii) The frequency of occurrence was modified by changing the probability that a particular 'input' occurred during every respiratory interval, because only a few of the preganglionic inputs fired in the majority of respiratory cycles (Gilbey et al. 1986).

When normally distributed events originating from multiple inputs were synchronized to occur with the same mean delay and a probability of 1 (i.e with every respiratory cycle), the distribution of intervals had two peaks, one at very short intervals and the other a little below the mean respiratory interval (Fig. 8B). The ln-survivor curves were markedly concave at short intervals and convex at long ones; this effect was exaggerated as the numbers of inputs increased (Fig. 8B 5).

Figure 9A shows the interval histogram for a single input with variable delay after the respiratory marker (S.D. 0·2 s) coupled to the respiratory cycle with a probability of 1 (yielding intervals of 1·25 ± 0·28 s). The pattern when this was coupled to the respiratory cycle with a probability of 0·67 is shown in Fig. 9B. When two such inputs were combined, with a mean delay separated by 0·1 s, the interval histogram had a peak at short intervals, a broad peak centred just below the respiratory interval, and small peaks near multiples of the respiratory interval (Fig. 9C). As the number of inputs with mean delays within 0·1 s of each other was increased, the peak at short intervals became exaggerated and a broad small peak just below the respiratory interval remained (see Fig. 9D, cf. Fig. 8B 4). However, when the difference in mean delay between two inputs was longer (0·5 s), the combined interval histogram had a broad peak well below the respiratory interval (Fig. 9E). This is similar to that observed in some cells with identified RR (e.g. Fig. 7C). If the mean delays of multiple inputs were distributed evenly over 0·5 s, the interval histogram approximated an exponential (Fig. 9F).

		View larger version [in this window] [in a new window]
Figure 9. Simulation of events occurring with various relationships to a respiratory cycle A, distribution of event intervals resulting when events in a single input have a normally distributed delay and probability = 1 with relation to a variable respiratory cycle (with the interval distribution shown in Fig. 8B 2). This distribution has S.D. of 0·25 s (the combined variance of delays and respiratory intervals). B, as for A except that the probability of occurrence during a respiratory interval is 0·67, resulting in some events occurring at multiples of the respiratory interval. C-F, interval histograms arising from multiple inputs with event patterns as in B. C, event distribution for two inputs with difference in mean delay of 0·1 s results in a histogram with two peaks. D, combining six inputs with mean delays spaced evenly within 0·1 s markedly increases the peak at short intervals. E, distribution for two inputs with difference in mean delay of 0·5 s results in an event histogram with one peak well below the respiratory interval. F, when the mean relative delays of six inputs are spaced evenly over 0·5 s, the distribution approximates an exponential (dashed curve). Ln-survivor (ln-S) curves are shown above each plot (thick traces) together with the prediction for a Poisson process with the same mean interval (dashed lines). The curves tend to be convex with inflexions around peaks in the interval histograms. When multiple inputs are combined, the curve becomes linear (F). G-J, autocorrelograms for the event patterns generated for C-F. Rhythmicity is evident for the data in C, D and F, but is reduced when the mean delays are more widely distributed (E). The vertical arrows indicate the mean respiratory interval.

Ln-survivor curves

Ln-survivor curves were derived for the simulated interval sequences with respiratory modulation (Fig. 9, upper curves). The ln-survivor curve for a single input discharging regularly with every respiratory cycle was convex (Fig. 9A). If mean delays of several inputs were restricted to a 0·1 s period of the respiratory interval, the ln-survivor plot characteristically fell initially below and subsequently above the linear prediction and this curvature increased with the number of inputs (Fig. 9D, cf. Fig. 8B 5). In contrast, if the mean phase relationships of several inputs were spaced over 0·5 s, or if the variance of the delays of two or three inputs was increased, the plot tended to flatten out (Fig. 9F). Overall, the greater the number of inputs that were merged (no matter what their event pattern), the closer the plot for the combined sequence came to fitting within the 95 % confidence limits for a straight line.

Autocorrelograms

Autocorrelograms derived for the respiration modulated models (Fig. 9G -J) exhibited cyclic variation with the respiratory period. Greater modulation was seen with larger numbers of modulated inputs (cf. Fig. 9G and H) whereas greater variation in their relative timing tended to obscure it (cf. Fig. 9G and I, and H and J). An initial trough was always present in the autocorrelogram (corresponding to the interval histogram) when there was one or only a few inputs (Fig. 9I). Increasing variance in the timing of a single input caused the cycling to diminish markedly after the first peak.

Thus the models that incorporated inputs that were rather widely distributed in relation to the respiratory cycle with a probability usually < 1 provided interval distributions, ln-survivor curves and autocorrelograms that took the form of most of those derived from the recorded data.

Modelling the behaviour of inputs to individual recorded neurones

The simplest combinations of inputs that produced event patterns matching those recorded were derived (see Table 2). In each case, we varied the probabilities, delays and variances (in that order) by small increments from the values used in the models above until a match was obtained (P > 0·05, ² test). In most cases, it was only necessary to vary the probabilities.

The single strong inputs all showed RR, characterized by cyclic autocorrelograms, interval histograms with a peak at multiples of the respiratory interval (Fig. 7) and bimodal EMG-triggered histograms. All of these could be simulated by selecting different probabilities of firing in relation to the respiratory cycle (see Table 1). In one case (Cell 4, Fig. 7C), the strong input was unusual in that it fired twice in about 20 % of respiratory cycles, showing two interval peaks separated by 0·9 s. This could not be simulated using the simple respiratory model. However, it could be mimicked if it was assumed that the (preganglionic) neurone from which the strong input arose received two (descending) inputs with variable delays separated by 0·9 s and probabilities of 0·93 and 0·27, respectively. The pattern for another strong input (Cell 2, Fig. 7B) with a low average probability of occurrence of synaptic events with each respiratory cycle (about 0·33) could be better simulated if the probabilities that the intervals were equivalent to 2, 3 or 4 cycles were taken to be 0·1, 0·8 and 0·1, respectively. In three of these neurones with one strong input, to mimic the behaviour of the subthreshold inputs required a lower probability in relation to respiration than the strong input and, in the remaining case, could best be simulated if they occurred randomly. These examples demonstrate that the simple simulations using one probability do not provide unique explanations of the behaviour of preganglionic inputs.

In all other cells, in which more than one suprathreshold input was identified, it was usually possible to mimic the shape of the observed interval histograms and the linear and convex ln-survivor curves (see Table 1) by assuming some respiratory modulation in all inputs. The probability in relation to a respiratory cycle had to be different for each cell and, again, often had to be higher for suprathreshold than for subthreshold events. Overall this difference in the link to respiration was significant (suprathreshold, 0·75 ± 0·09, n = 7; subthreshold, 0·45 ± 0·10, n = 7, P = 0·03, paired t test; see Table 1). The observed autocorrelograms rarely showed as much respiratory modulation as the simulated ones. For one neurone, there was clearly no evidence of a cycle in the EMG-triggered histogram and models using exponentially or uniformly distributed intervals for some or all of the inputs were able to mimic the observed event patterns (Cell 6 in Table 2).

Two examples are shown in Figs 10 and 11 which represent the extremes of the patterns analysed. Both had a relatively high frequency of events. Strongly respiratory-modulated activity and an exponential interval histogram in one cell (Cell 9, Fig. 10A1 and C 1) was simulated by combining the activity of three suprathreshold inputs and six subthreshold ones, all with RR (see legend for details). In this case, as in two others, the delays of the subthreshold inputs after the respiratory marker had to be clustered around two values rather than evenly distributed, because the rhythmicity showed an inspiratory peak as well as an expiratory one. This suggests that different inputs to the same cell can discharge in different phases of respiration. In contrast, in another cell in which respiratory modulation could not be detected (Cell 6, Fig. 11), a similar exponential interval histogram could be mimicked by combining three suprathreshold inputs and six subthreshold ones firing asynchronously (with an exponential distribution of intervals). In this case, a relative lack of short intervals was reproduced by simulating the effect of the after-hyperpolarization (AHP) in the preganglionic neurones on their discharge. This was achieved by filtering events from the exponential interval distribution of all inputs by reducing the probability of an event occurring after each event by an exponentially declining factor with time constant of 0·3 s. Another possibility is that the inputs to this neurone were linked to respiration but with their relative timing spread much more widely than we found necessary to simulate the behaviour of all other cells.

		View larger version [in this window] [in a new window]
Figure 10. Simulation of recorded data in a neurone with marked respiratory modulation A, observed (A1) and simulated (A2) peri-EMG histograms for all synaptic events. B, amplitude distribution of all synaptic events, with the membrane held at -68 mV. Mean frequency of events 5·4 Hz, MABP 112 mmHg (same cell as in Fig. 2). C, interval histograms for all events combined (C 1), for suprathreshold events (C 3) and for subthreshold events (C 5) showing observed data (open columns) together with those simulated (filled columns). The data show a good fit to an exponential (dashed curve, P = 0·078, Kolmogorov-Smirnov). To simulate the observed data, it was necessary to combine the event patterns for three suprathreshold inputs (example shown in C 2), each with a delay with S.D. of 0·2 s and respiratory modulation with a probability of 0·86; their mean delays with respect to the cycle were 0·7, 0·85 and 1·2 s, respectively. These were combined with six subthreshold inputs (example in C 4) each with the same variance in delay as the suprathreshold inputs and respiratory modulation with a probability of 0·67; four of the inputs had mean delays of 0·6 s with respect to the cycle whereas the delays of the other two were both 1 s. These combinations simulated the suprathreshold (C 3) and subthreshold events (C 5). Ln-survivor (ln-S) curves are shown above the interval distributions (thin lines, observed; thick lines, simulated), together with the prediction for a Poisson process with the same mean interval (dashed lines). D, autocorrelograms derived from the observed (D 1) and simulated (D 2) event patterns show cyclic variation with peaks at the respiratory interval (1·25 s). The downward arrow (in B) shows 15 mV amplitude cut-off between suprathreshold and subthreshold events. The upward arrows indicate the mean respiratory interval.

Numbers of active preganglionic inputs

Bearing in mind the limitations of our simple approach and the lack of an optimization procedure, the simulations (in conjunction with the synaptic response amplitude histograms and the configuration of the action potentials for each cell) allowed us to estimate the number of supra- and subthreshold inputs from which the observed synaptic events arose (Table 2). Overall the estimated number of active synaptic inputs was 6·3 ± 0·7, with 1·8 ± 0·2 suprathreshold inputs and 4·3 ± 0·6 subthreshold ones. It should be noted that some of the suprathreshold events did not arise from strong inputs, i.e. their amplitude was only just above threshold and with quantal variation they would on occasions have failed to initiate an action potential. From these estimates and the observed mean frequencies of occurrence, the derived mean firing frequency of individual suprathreshold inputs (0·48 ± 0·12 Hz) was found to be similar to that of subthreshold inputs (0·35 ± 0·05 Hz, P = 0·48, Wilcoxon signed rank test).

		View larger version [in this window] [in a new window]
Figure 11. Simulation of recorded data in a neurone that lacked respiratory rhythmicity A, observed (A1) and simulated (A2) peri-EMG histograms show no modulation. B, amplitude distribution for all synaptic events recorded at -65 mV (same cell as in Fig. 3A). Mean frequency of events 4·8 Hz, MABP 97 mmHg. C, interval histograms as for Fig. 10. All inputs were assumed to occur randomly, i.e. with exponential interval distributions, but with a simulated 'after-hyperpolarization' (see text) that followed each randomly timed event. Combination of three such asynchronous inputs was necessary to simulate the suprathreshold interval histogram (C 3) and six low frequency ones to simulate the subthreshold interval histogram (C 5). The overall output (C 1) shows a good fit to an exponential (dashed line, P = 0·026, Kolmogorov-Smirnov). D, autocorrelograms show no modulation. Conventions as in Fig. 10.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The present analysis of the discharge patterns of convergent preganglionic inputs to a sample of rat SCG neurones selected on the basis of their high rates of on-going activity (> 0·5 Hz) shows that (i) respiratory, and less often, cardiac rhythmicity could be identified in both suprathreshold and subthreshold inputs, (ii) clusters or bursts of synaptic events occurred no more often than expected by chance, and (iii) although most inputs showed evidence of respiratory modulation, different inputs to the same cell did not necessarily have the same discharge pattern. In addition, unless the frequency of discharge of the subthreshold inputs is many times higher than that recorded under the conditions of these experiments, summation is not significant for the generation of postganglionic discharge. The data suggest that rhythmic patterning in a postganglionic axon will reflect that present in one, or a few, suprathreshold preganglionic inputs. Further, from attempts to simulate the observed event patterns, it appears that the number of convergent active inputs is probably less than the number of connections they receive, consistent with the absence of on-going activity in a proportion of preganglionic neurones (Coote et al. 1981; Gilbey et al. 1986; Zhou & Gilbey, 1992).

Synaptic activity showed respiratory modulation in most BI neurones, but only half of them had cardiac rhythmicity (CR). The presence or absence of CR might be associated with the function of sympathetic neurones in the periphery. Vasoconstrictor neurones supplying skeletal muscle are more likely to exhibit CR than those supplying skin (Delius et al. 1972; Jänig, 1985; Häbler et al. 1994a). However only about 50 % of rat SCG neurones are vasoconstrictor. If only those directed at muscle show a pronounced CR, the proportion of our sample (50 % of BI neurones) is not unexpected. The idea that BI neurones are vasoconstrictor remains speculative.

The respiratory modulation observed here was similar in pattern to that in hindlimb vasoconstrictor neurones (Darnall & Guyenet, 1990; Häbler et al. 1993), vasoconstrictor neurones supplying the submandibular gland (Bartsch et al. 1996) and preganglionic neurones projecting to the SCG (Häbler et al. 1996), consisting of inhibition during inspiration and activation during expiration in the majority of cases (see also Gilbey et al. 1986). The respiratory patterns can be generated by a combination of at least two mechanisms: (i) central coupling between respiratory neurones and sympathetic premotor neurones in the rostral ventrolateral medulla (McAllen, 1987; Haselton & Guyenet, 1989) and (ii) a cyclic reflex through baroreceptor afferents related to the ventilation-induced blood pressure waves (Häbler et al. 1996). Since the animals were breathing spontaneously in our experiments, it is not possible to know which mechanism was dominant. The widespread occurrence of RR indicates that this is probably a property of the activity in several functionally distinct pathways.

We simulated the effect of several inputs with various discharge patterns converging on a single postganglionic neurone and showed that the overall interval histograms of synaptic events tended towards an exponential distribution as the number of convergent inputs increased, no matter what their discharge pattern. An exponential distribution resulted with fewer inputs if (i) the variance of the relative timing of each input to the respiratory cycle was large, or (ii) some inputs were unsynchronized with respect to respiration. However, the greater the number of convergent inputs, the less can be deduced about the pattern of discharge of each input. An essentially similar conclusion was made for the distribution of miniature endplate potentials arising from a number of release sites (Fatt & Katz, 1952). Conversely, when the interval distribution shows a peak, only a few convergent inputs can have been active (see also Johnson & Purves, 1983).

Incorporation of a respiratory rhythm in all simulated inputs reproduced the observed interval histograms, ln-survivor curves and autocorrelograms in the majority of cells. Respiratory modulation yielded peaks to the interval histograms, convexity to the ln-survivor curves and cycling in the autocorrelogram. RR and, when present, CR were detected in both sub- and suprathreshold events (Table 1), including all single strong inputs. However, RR was not detected in at least some subthreshold inputs to cells with respiratory-linked suprathreshold inputs. This might mean that some inputs had unsynchronized discharge patterns (e.g. Fig. 7A4). In a similar study of activity in rabbit ciliary ganglion cells (Johnson & Purves, 1983), in most of which retinal illumination modified the frequency of synaptic events, 30 % of cells also had at least some inputs that responded to auditory or noxious stimuli. Thus not all the inputs to a postganglionic neurone necessarily respond alike.

The discharge of each input to an SCG cell reflects the discharge of a single preganglionic neurone that integrates many small amplitude fast and slow excitatory and inhibitory potentials (McLachlan & Hirst, 1980; Dembowsky et al. 1985), as do most other CNS neurones. In order for the preganglionic discharge to be synchronized with the respiratory cycle, a summed excitation must reach the spinal neurone from the respiratory centre during one part of the cycle. In the present experiments, the probability of a preganglionic input discharging with each respiratory cycle was often less than 1, implying that the excitatory drive was not very powerful and that preganglionic neurones differ in their excitability. The analysis indicated no tendency for the inputs to discharge in close synchrony, confirming our earlier subjective impression that bursts were rare (see McLachlan et al. 1997). In fact, the relative timing of several 'respiratory-modulated' inputs had to be spread over some 40 % of the respiratory interval to simulate the observed inter-event interval histograms. Overall, it must be concluded that, even allowing for differences in conduction velocity, both of descending pathways (conduction delay 20-150 ms, Morrison & Reis, 1991) and of preganglionic axons (conduction delay 40- 100 ms, P. Davies and D. Ireland, unpublished observations), the timing of events in relation to respiration is not very precise. This implies that the respiratory drive to a pool of preganglionic neurones is not constrained to a brief time period consistent with a barrage of fast excitatory potentials arising from descending pathways but might more easily be explained if the underlying membrane events are prolonged.

Postganglionic neurones differ markedly from CNS ones in that they are predominantly activated by only a few of the active convergent inputs, in particular the strong ones. Our subjective identification of single strong inputs based on the configuration of their action potentials was confirmed by the absence of short intervals in the distributions of intervals between strong responses. When more than one suprathreshold response could be recognized, there were always short intervals between suprathreshold events (Figs 8B and 9C and E). Amongst the present sample of neurones selected for their relatively high levels of synaptic activity, cells with few active inputs had only one active strong input, whereas those with many active inputs had two strong inputs as well as others with ESPs large enough on occasions to trigger action potentials. Because of the paucity of summation, the probability that these large ESPs would evoke discharge would depend not only on quantal variation but also on whether or not they occurred during the AHP of a preceding action potential (Fig. 1b; see also McLachlan et al. 1997). However, if the frequency of discharge of these particular inputs increased so that their ESP amplitudes facilitated (McLachlan, 1975), their likelihood to transmit impulses would increase.

When multiple suprathreshold inputs converge, the postganglionic discharge pattern does not necessarily reflect the summed behaviour of the individual preganglionic neurones. An action potential arising from a just-threshold input might be blocked by a preceding AHP or it might be followed by one arising from a strong input (which would not be blocked by the preceding AHP). This means that, if two postganglionic action potentials occur < 200 ms apart, they might have originated either from two strong inputs, from a just-threshold input immediately preceding a strong one or, but very rarely in our recordings, from a single preganglionic neurone discharging twice. Such doublets were uncommon during on-going activity in the present experiments, although they have been recorded in other pathways (Johnson & Gilbey, 1996).

The mean number of active inputs (6) derived for this sample of SCG cells with high levels of activity is lower than the mean number of inputs (9) defined by the steps in the ESPs evoked by graded electrical stimulation (Purves et al. 1986). The presence of inactive inputs is consistent with the failure to detect activity in 50 % of identified preganglionic neurones in the anaesthetized rat spinal cord (Gilbey et al. 1986; Lewis & Coote, 1995). Another contributing factor may be that small amplitude subthreshold ESPs often have low quantal contents resulting from a low release probability (McLachlan, 1975) so that the discharge of some preganglionic inputs usually fails to release transmitter.

Our limited data suggest that whether or not rhythms (either cardiac and respiratory) are present in all preganglionic inputs to a given cell may not be functionally important. If the suprathreshold inputs show a rhythm, the centrally generated pattern will appear postganglionically. In the rabbit (Skok & Ivanov, 1983), respiratory modulation was thought to synchronize the inputs and increase the probability that subthreshold inputs would summate to discharge the postganglionic cell. Our analysis appears to have ruled out this mechanism in the spontaneously breathing anaesthetized rat. The bursts recorded in multifibre sympathetic activity (see Häbler et al. 1994a) are less likely to arise from the activity of groups of neurones that fire synchronously with every cycle than from the low frequency discharge of many pathways, each showing cyclic modulation with low probability.

In summary, both cardiac and respiratory rhythmicity were detected in the on-going synaptic responses recorded in postganglionic neurones in the SCG in the spontaneously breathing anaesthetized rat. The rates of firing of individual inputs were low and there was little evidence that synchronization of convergent preganglionic inputs by centrally generated rhythmicity is responsible for respiratory modulation of the discharge of a single postganglionic neurone. It seems that the brainstem rhythms appear in the activity of postganglionic axons only if they are present in the strong inputs which are the major determinant of action potential discharge. However, the convergence of two strong preganglionic inputs with respiratory rhythmicity might lead to postganglionic discharge patterns with interspike intervals briefer than the respiratory interval, as reported by others (Johnson & Gilbey, 1996). Thus convergence of strong inputs within ganglia can play an important role in determining the pattern of discharge in some sympathetic pathways.

	REFERENCES

Top Abstract Introduction Methods Results Discussion References

Bahr, R., Bartel, B., Blumberg, H. & Jänig, W. (1987). Functional characterization of preganglionic neurons projecting in the lumbar splanchnic nerves: neurons regulating motility. Journal of the Autonomic Nervous System 15, 109-130.
Bartsch, T., Häbler, H.-J. & Jänig, W. (1996). Functional properties of postganglionic sympathetic neurones supplying the submandibular gland in the anaesthetized rat. Neuroscience Letters 214, 143-146	[Medline]
Bartsch, T., Häbler, H.-J. & Jänig, W. (1997). Reflex patterns of preganglionic sympathetic fibres projecting to the superior cervical ganglion in the rat. Pflügers Archiv 433, suppl. 6, R74.
Boczek-Funcke, A., Häbler, H.-J., Jänig, W. & Michaelis, M. (1991). Rapid phasic baroreceptor inhibition of the activity in sympathetic preganglionic neurones does not change throughout the respiratory cycle. Journal of the Autonomic Nervous System 34, 185-194	[Medline]
Boczek-Funcke, A., Häbler, H.-J., Jänig, W. & Michaelis, M. (1992). Respiratory modulation of the activity in sympathetic neurones supplying muscle, skin and pelvic organs in the cat. The Journal of Physiology 449, 333-361	[Abstract]
Bornstein, J. C. (1978). Spontaneous multiquantal release at synapses in guinea-pig hypogastric ganglia: evidence that release can occur in bursts. The Journal of Physiology 282, 375-398	[Abstract]
Cassell, J. F. & McLachlan, E. M. (1986). The effect of a transient outward current (IA) on synaptic potentials in sympathetic ganglion cells of the guinea-pig. The Journal of Physiology 374, 273-288	[Abstract]
Coote, J. H., Macleod, V. H., Fleetwood-Walker, S. & Gilbey, M. P. (1981). The response of individual sympathetic preganglionic neurones to microelectrophoretically applied endogenous monoamines. Brain Research 215, 135-145	[Medline]
Cox, D. R. & Lewis, P. A. W. (1966). The Statistical Analysis of Series of Events. Methuen, London.
Darnall, R. A. & Guyenet, P. (1990). Respiratory modulation of pre- and postganglionic lumbar vasomotor sympathetic neurons in the rat. Neuroscience Letters 119, 148-152	[Medline]
Delius, W., Hagbarth, K. E., Hongell, A. & Wallin, B. G. (1972). General characteristics of sympathetic activity in human muscle nerves. Acta Physiologica Scandinavica 84, 65-81	[Medline]
Dembowsky, K., Czachurski, J. & Seller, H. (1985). An intracellular study of the synaptic input to sympathetic preganglionic neurones of the third thoracic segment of the cat. Journal of the Autonomic Nervous System 13, 201-244	[Medline]
Fatt, P. & Katz, B. (1952). Spontaneous subthreshold activity at motor nerve endings. The Journal of Physiology 117, 109-128.
Gibbins, I. L. (1991). Vasomotor, pilomotor and secretomotor neurons distinguished by size and neuropeptide content in superior cervical ganglia of mice. Journal of the Autonomic Nervous System 34, 171-183	[Medline]
Gilbey, M. P., Numao, Y. & Spyer, K. M. (1986). Discharge patterns of cervical sympathetic preganglionic neurones related to the central respiratory drive in the rat. The Journal of Physiology 378, 253-265	[Abstract]
Häbler, H.-J., Bartsch, T. & Jänig, W. (1996). Two distinct mechanisms generate the respiratory modulation in fiber activity of the rat cervical sympathetic trunk. Journal of the Autonomic Nervous System 61, 116-122	[Medline]
Häbler, H.-J., Jänig, W., Krummel, M. & Peters, O. A. (1993). Respiratory modulation of the activity in postganglionic neurones supplying skeletal muscle and skin of the rat hindlimb. Journal of Neurophysiology 70, 920-930	[Medline]
Häbler, H.-J., Jänig, W., Krummel, M. & Peters, O. A. (1994a). Reflex patterns in postganglionic neurons supplying skin and skeletal muscle of the rat hindlimb. Journal of Neurophysiology 72, 2222-2236	[Medline]
Häbler, H.-J., Jänig, W. & Michaelis, M. (1994b). Respiratory modulation in the activity of sympathetic neurones. Progress in Neurobiology 43, 567-606	[Medline]
Hagbarth, K.-E., Hallin, R. G., Hongell, R. G., Torebjörk, H. E. & Wallin, B. G. (1972). General characteristics of sympathetic activity in human skin nerves. Acta Physiologica Scandinavica 84, 164-176	[Medline]
Haselton, J. R. & Guyenet, P. G. (1989). Central respiratory modulation of medullary sympathoexcitatory neurons in rat. American Journal of Physiology 256, R739-750	[Medline]
Hirst, G. D. S. & McLachlan, E. M. (1984). Post-natal development of ganglia in the lower lumbar sympathetic chain of the rat. The Journal of Physiology 349, 119-134	[Abstract]
Jänig, W. (1985). Organization of the lumbar sympathetic outflow to skeletal muscle and skin of the cat hindlimb and tail. Reviews of Physiology, Biochemistry and Pharmacology 102, 119-213.	[Medline]
Johnson, C. D. & Gilbey, M. P. (1994). Sympathetic activity recorded from the rat caudal ventral artery in vivo. Journal of Physiology 476, 437-442.	[Abstract]
Johnson, C. D. & Gilbey, M. P. (1996). On the dominant rhythm in the discharges of single postganglionic sympathetic neurones innervating the rat tail artery. The Journal of Physiology 497, 241-259	[Abstract]
Johnson, D. A. & Purves, D. (1983). Tonic and reflex synaptic activity recorded in ciliary ganglion cells of anaesthetized rabbits. The Journal of Physiology 339, 599-613	[Abstract]
Lebedev, V. P., Petrov, V. I. & Skobelev, V. A. (1980). Do sympathetic preganglionic neurones have a recurrent inhibitory mechanism? Pflügers Archiv 383, 91-97	[Medline]
Lewis, D. I. & Coote, J. H. (1995). Chemical mediators of spinal inhibition of rat sympathetic neurones on stimulation in the nucleus tractus solitarii. The Journal of Physiology 486, 483-494	[Abstract]
McAllen, R. M. (1987). Central respiratory modulation of subretrofacial bulbospinal neurones in the cat. The Journal of Physiology 388, 533-545	[Abstract]
Macefield, V. G. & Wallin, B. G. (1995). Modulation of muscle sympathetic activity during spontaneous and artificial ventilation and apnoea in human. Journal of the Autonomic Nervous System 53, 137-147	[Medline]
McLachlan, E. M. (1975). An analysis of the release of acetylcholine from preganglionic nerve terminals. The Journal of Physiology 245, 447-466.	[Abstract]
McLachlan, E. M., Davies, P. J., Häbler, H.-J. & Jamieson, J. (1997). On-going and reflex synaptic events in rat superior cervical ganglion cells. The Journal of Physiology 501, 165-182	[Abstract]
McLachlan, E. M. & Hirst, G. D. S. (1980). Some properties of preganglionic neurons in the upper thoracic spinal cord of the cat. Journal of Neurophysiology 43, 1251-1265	[Medline]
Morrison, S. F. & Reis, D. J. (1991). Responses of sympathetic preganglionic neurons to rostral ventrolateral medullary stimulation. American Journal of Physiology 261, R1247-1256	[Medline]
Purves, D., Rubin, E., Snider, W. D. & Lichtman, J. (1986). Relation of animal size to convergence, divergence and neuronal number in peripheral sympathetic pathways. Journal of Neuroscience 6, 158-163	[Abstract]
Skok, V. I. & Ivanov, A. Y. (1983). What is the ongoing activity of sympathetic neurons? Journal of the Autonomic Nervous System 7, 263-270	[Medline]
Sundlöf, G. & Wallin, B. G. (1978). Human muscle nerve activity at rest. Relationship to blood pressure and age. The Journal of Physiology 274, 621-637	[Medline]
Yawo, H. (1989). Rectification of synaptic and acetylcholine currents in the mouse submandibular ganglion cells. The Journal of Physiology 417, 307-322	[Abstract]
Zar, J. H. (1984). Biostatistical Analysis. Prentice Hall, New Jersey.
Zhou, S.-Y. & Gilbey, M. P. (1992). Respiratory-related activity of lower thoracic and upper lumbar sympathetic preganglionic neurones in the rat. The Journal of Physiology 451, 631-642	[Abstract]

Acknowledgements

This work was supported by the National Health & Medical Research Council of Australia. H.-J. H.'s participation was supported by the Humboldt Foundation. We are very grateful to Ansgar Boczek-Funcke for providing the program for quantification of cardiac rhythmicity and to James Brock and Wilfrid Jänig for their helpful comments on the manuscript.

Corresponding author

E. M. McLachlan: Prince of Wales Medical Research Institute, High Street, Randwick, NSW 2031, Australia.

Email: e.mclachlan{at}unsw.edu.au

This article has been cited by other articles: