Specificity in monosynaptic and disynaptic bulbospinal connections to thoracic motoneurones in the rat

Abstract

The respiratory activity in the intercostal nerves of the rat is unusual, in that motoneurones of both branches of the intercostal nerves, internal and external, are activated during expiration. Here, the pathways involved in that activation were investigated in anaesthetised and in decerebrate rats by cross-correlation and by intracellular spike-triggered averaging from expiratory bulbospinal neurones (EBSNs), with a view to revealing specific connections that could be used in studies of experimental spinal cord injury. Decerebrate preparations, which showed the strongest expiratory activity, were found to be the most suitable for these measurements. Cross-correlations in these preparations showed monosynaptic connections from 16/19 (84%) of EBSNs, but only to internal intercostal nerve motoneurones (24/37, 65% of EBSN/nerve pairs), whereas disynaptic connections were seen for external intercostal nerve motoneurones (4/19, 21% of EBSNs or 7/25, 28% of EBSN/nerve pairs). There was evidence for additional disynaptic connections to internal intercostal nerve motoneurones. Intracellular spike-triggered averaging revealed excitatory postsynaptic potentials, which confirmed these connections. This is believed to be the first report of single descending fibres that participate in two different pathways to two different groups of motoneurones. It is of interest compared with the cat, where only one group of motoneurones is activated during expiration and only one of the pathways has been detected. The specificity of the connections could be valuable in studies of plasticity in pathological situations, but care will be needed in studying connections in such situations, because their strength was found here to be relatively weak.

Key points

• In the rat, unlike in other species, motoneurones of both the internal intercostal nerve and the external intercostal nerve show a phase of excitation in expiration.

• This study investigated the pathways transmitting this excitation from the medulla.

• Direct (monosynaptic) excitation was found from individual expiratory neurones in the medulla to internal intercostal nerve motoneurones, but only indirect (disynaptic) excitation was found from the same neurones to the motoneurones of the external intercostal nerve.

• This is the first demonstration of two separate pathways from individual long descending fibres specific to two different sets of motoneurones.

• This specificity could be useful in studying plasticity or regeneration in thoracic segments in investigations of mechanisms involved in spinal cord injury and repair.

Introduction

Rats and mice have been the species of choice in many areas of neurobiology for at least 20 years, including investigations in the spinal cord and including disease states. However, much of our knowledge of the connectivity between the various cell types or afferent fibres in the spinal cord is based on earlier studies in larger species, in particular the cat. This raises the important question of possible species differences, particularly where one needs to identify the normal targets for a given fibre system. Descending motor systems are a case in point. Numerous studies of spinal cord injury have studied improvements of motor function following experimental treatments; many have measured sprouting or regeneration of descending fibre tracts (often the corticospinal tract). However, for hardly any of the descending tracts in rodents do we know what their actual, normal targets might be.

For instance, the corticospinal tract in the primate is well known to make direct connections to forelimb motoneurones (Lemon & Griffiths, 2005), whereas in the cat, the simplest of such connections are disynaptic, via the C3–C4 propriospinal neurones (Alstermark & Lundberg, 1992). In the rat these propriospinal connections have not been found, although disynaptic responses may be recorded via reticulospinal neurones (Alstermark et al. 2004), and in mice the simplest connections to motoneurones may be trisynaptic (Alstermark & Ogawa, 2004). Here, we consider the respiratory system, this being a system that might be thought more likely to be conserved across species. For the descending connections from the medulla to phrenic motoneurones this may well be the case: in both cats (Monteau & Hilaire, 1991) and in rats (Ellenberger & Feldman, 1988; Lipski et al. 1994; Tian & Duffin, 1996), a strong monosynaptic connection conveying the inspiratory drive to the motoneurones is present. However, many spinal cord injury studies involve thoracic segments, where total cord transections or contusion injuries can be made without compromising the life of the animal. The availability of a set of descending axons traversing the thoracic spinal cord, and whose normal connections were known, would be a valuable asset in such studies, especially if the targets were motoneurones, which are both the most easily identified of neurones and often the easiest to study.

In the cat, the axons from neurones in nucleus retroambiguus form one such descending pathway. Many of these axons extend through the whole length of the spinal cord (Holstege, 1991) and their synapses have been found on motoneurones at several different segmental levels (Boers et al. 2006). Physiologically, a high proportion of these axons, but not all (Boers et al. 2005), are those of expiratory bulbospinal neurones (EBSNs), a set of neurones that are responsible for conveying the expiratory drive to thoracic and upper lumbar motoneurones, as well as commands for a number of other vital motor acts, such as emesis, coughing and defecation (Iscoe, 1998). A substantial part of the expiratory drive has been shown to be transmitted monosynaptically (Cohen et al. 1985; Kirkwood, 1995; Saywell et al. 2007; Road et al. 2013). With a view to exploiting these axons in studies related to spinal cord injury, we have therefore investigated whether such connections are also present in the rat, as might be expected from anatomical tracing experiments (Hardy et al., 1998). Interest in such an interspecies comparison was sharpened by previous observations from this laboratory that the temporal and spatial patterns of activation of intercostal motoneurones in the rat are different from those already known in the cat, dog and human (de Almeida et al. 2010; de Almeida & Kirkwood 2010). We now find that monosynaptic connections from EBSNs to motoneurones of the internal intercostal nerve are indeed present, similar to those in the cat, although apparently somewhat weaker. However, they are accompanied by a second set of connections, deduced to be disynaptic, to a second set of motoneurones, those of the external intercostal nerve. These results therefore provide both an example of homology between the functional connectivity of descending systems in the two species and an example of interspecies heterogeneity.

Methods

The experiments described here are a subset of those already reported by de Almeida et al. (2010). They were performed on adult female Sprague-Dawley rats (Harlan, UK) weighing between 180 and 287 g, as approved by the Ethical Review Process of the Institute of Neurology, in accordance with UK legislation (Animals Scientific Procedures Act, 1986). A range of different anaesthetic regimens were used, which are listed in Table 1, together with the numbers of animals used under each regimen.

Table 1.

Anaesthetic regimens employed

Anaesthetic and dose	No. of experiments (no. of EBSNs) Cross-correlation	No. of experiments (no. of EBSNs) STA
Ketamine only	—	1 (1)
100 mg kg−1, i.p., then supplements, i.v., as required
Ketamine/xylazine	—	3 (5)
Induction, i.p., 100 mg kg−1, 10 mg kg−1 respectively, then supplements of ketamine/xylazine (same proportions), i.v. as required
Fentanyl and ketamine, then α-chloralose	1 (2)	—
Induction, fentanyl, 2 mg kg−1, plus ketamine 50 mg kg−1, i.p., maintained with ketamine and fentanyl i.p., then i.v. as required. For recordings, α-chloralose (20 mg kg−1, i.v.)
Halothane then α-chloralose	3 (4)	2 (2)
Halothane: induction 5%, maintenance 1–2% during surgery, titrated to α-chloralose (30–80 mg kg−1, i.v.) for recording
Ketamine/xylazine then decerebration	3 (11)	—
As above, until decerebration
Halothane then decerebration	3 (8)	5 (7)
Induction 5%, maintenance 1–2% during surgery. Lowered to 0.5% at start of decerebration; turned off once decerebration completed

One animal (halothane, then decerebrate) was used for both cross-correlation and STA. Suppliers: ketamine, Vetalar, Pfizer, UK; xylazine, Rompun, Bayer plc, UK; urethane, Sigma; halothane; Merial Animal health Ltd, UK; fentanyl, Sublimaze, Janssen–Cilag, UK; α-chloralose; α-Chloralose-HBC complex, Sigma (doses cited above are the equivalent doses of α-chloralose; actual doses of the complex were 10× higher).

Surgical procedures

Atropine sulphate (60 μg, i.m., Hameln Pharmaceuticals Ltd, Gloucester, UK) was administered to minimise airway fluid secretion. Rectal temperature was maintained at 37–38°C with a thermostat-controlled heating blanket (Harvard, Edenbridge, UK). The jugular vein was cannulated for the administration of anaesthetic supplements and fluids, and the trachea for mechanical ventilation. The right carotid artery was cannulated for the measurement of blood pressure and bilateral vagotomies were performed. In the decerebrate experiments, the left carotid artery was also prepared, with a loose tie around it. Sodium lactate solution (Hartmann's solution; Baxter Healthcare Ltd, Newbury, UK) was administered periodically (up to 1 ml h⁻¹) and a plasma substitute (Gelofusin; Braun Medical Ltd, Sheffield, UK) was infused as required (both i.v.), to stabilise the animal's fluid balance and to maintain blood pressure within physiological limits. A mean arterial pressure above 80 mmHg was maintained for nearly all of the recordings, although occasionally, towards the end of an experiment, a decline to 60 mmHg was accepted, but only if the pattern of respiratory discharges remained stable (de Almeida et al. 2010).

For the recordings, neuromuscular blockade was used (pancuronium bromide; 0.3 mg h⁻¹ i.v., Hospira UK Ltd). Before neuromuscular blockade, anaesthesia was maintained at a level where the animal showed no more than a minimal withdrawal reflex to a noxious paw pinch. After neuromuscular blockade, depth of anaesthesia was assessed from the recordings of blood pressure and respiratory discharges. Only minimal transient changes of blood pressure, heart rate or respiratory pattern following a similar pinch were allowed before anaesthetic supplements were administered. Particular care was taken to monitor anaesthetic levels with the use of ketamine/xylazine, where frequent supplements were needed (de Almeida et al. 2010). Artificial ventilation was carried out using O₂-enriched air at a rate of 100 cycles min⁻¹. Expired CO₂ was monitored at the trachea, stroke volume adjusted to bring the end-tidal CO₂ fraction to about 4% and then CO₂ was added to the gas mixture to increase the end-tidal CO₂ level to a value adequate to give brisk respiratory discharges in the intercostal nerves, usually around 7–8%.

Animals were supported, prone, by vertebral clamps usually at T6 and T12 and by a strong ligature tied into the lumbar fascia. The head was supported in anaesthetised animals by a plate screwed to the skull and in decerebrate preparations by a custom-made clamp with bilateral spikes screwed into the frontal and premaxillary bones. A laminectomy of 2–4 vertebrae (T7–T10) was made, the dura opened and a pair of stimulating electrodes (cut back tungsten microelectrodes) was placed in the left ventral spinal cord, usually at spinal cord segment T11, for antidromic identification of bulbospinal neurones. A partial occipital craniotomy and a partial laminectomy at C1 were made and the dura was opened. A small patch of pia was removed from the surface of the medulla on the right side, caudal to obex, and a pressure plate was lightly applied to the medullary dorsum to minimise movement.

For a variable number of intercostal spaces on the left side (T5–T11, depending on the experiment), the external and/or the internal intercostal nerves (Smith & Hollyday, 1983; de Almeida et al. 2010) were cut and their central ends were mounted on pairs of platinum wire electrodes for recording or for stimulation. In some of the cross-correlation experiments, an external intercostal nerve from the right side of one or two intercostal spaces (T8 and/or T9) was also prepared. In experiments involving intracellular recordings, bilateral pneumothoraces were made to minimise intrathoracic pressure fluctuations and thus to reduce movement of the cord. An end-expiratory pressure of 2–3 cmH₂O was applied to prevent atelectasis. A paraffin oil pool was constructed from skin flaps, which submerged both the dissected nerves and the exposed spinal cord.

For decerebration, the left carotid artery was tied. A wide craniotomy was made, with ligation of the sagittal sinus. Animals were artificially ventilated and all brain tissue rostral to the superior colliculus was removed by use of a blunt spatula and aspiration. To minimise further bleeding, the exposed brainstem was gently covered with Spongostan (Johnson & Johnson Medical, UK) soaked in saline.

At the end of the experiment, the animal was killed with an anaesthetic overdose, as confirmed by zero blood pressure for 5 min. The particular substance used was usually that used in the earlier part of the experiment (but for decerebrates following halothane, pentobarbital sodium, Lethobarb; Fort Dodge Animal Health, UK).

Recordings and analysis

EBSNs

Glass microelectrodes, tip diameter 2.6–2.8 μm, filled with 3 m NaCl were used, with conventional amplification, to record extracellularly from spontaneously active single units in the right medulla, in a similar region to that reported by Shen and Duffin (2002), centred at about 1.5 mm caudal and lateral from obex and around 1.5 mm deep. Units were identified as bulbospinal by antidromic activation from the spinal cord stimulating electrodes, including a collision test with stimuli at twice axonal threshold and double shocks to confirm that the critical interval in the collision test was not due to soma refractoriness. Orthodromic conduction times to the stimulation site were taken as the critical interval minus 0.5 ms, as in Davies et al. (1985a), and verified by Kirkwood (1995). Conduction velocities were calculated from these, as were conduction times to individual segments for which cross-correlation peaks were detected or excitatory postsynaptic potentials (EPSPs) observed (‘axonal times’, Saywell et al. 2007), using distances from the medulla measured along the cord dorsum postmortem. All times were referred to the rising negative phase of the EBSN spike. The single-unit nature of the EBSN recordings was always confirmed by auto-correlation histograms showing an absence of very short intervals.

Cross-correlation measurements

Simultaneous recordings, typically 30 min in duration, were made of the discharges of each EBSN, together with efferent discharges in up to five internal or external intercostal nerves on the left (internal T8–T11, usually T9 and T10; external T6–T10, often T6 and T9). Frequently one external intercostal nerve on the right was included, sometimes in place of the T6 nerve from the left. Nerve and EBSN recordings were band-pass filtered (300 Hz–3 kHz). Data comprised both an analog recording on magnetic tape and a digital recording of the EBSN and nerve discharges, the latter acquired via a 1401 interface and Spike2 software (Cambridge Electronic Design, Cambridge, UK). Cross-correlation histograms between the times of occurrence of EBSN and efferent nerve spikes of various selected amplitude ranges, together with auto-correlation histograms for the EBSN discharges, were constructed using 0.2 ms bin-widths.

Peaks in correlation histograms were accepted as significant if a single bin exceeded 3.29 (Sears & Stagg, 1976), where m is the baseline count, as in Davies et al. (1985a) and Kirkwood (1995). Single bin counts exceeding this limit (which sometimes occur by chance) were ignored if they fell outside the range of expected latencies (0–10 ms; Davies et al. 1985a). In cases of a sloping baseline, which could occur owing to the variable respiratory phasing of the discharges (for instance a positive slope could occur if the motoneurone firing generally phase-lagged the EBSN discharge), a baseline was fitted by eye and m was taken as the value on that line at the time of a peak (Davies et al. 1985a; Kirkwood, 1995; Vaughan & Kirkwood, 1997). The strength of peaks was assessed by k, the ratio of the highest count within a peak to m (Sears & Stagg, 1976; Davies et al. 1985a; Kirkwood, 1995). The sensitivity of detecting a peak of a given value of k depends on m, and therefore on the length of the recording and the discharge rates of both the EBSN and the efferent motoneurone population, both of which were highly variable between the different nerves and between different runs. To give some meaning to an absence of any peak, only histograms with m≥ 1082 were considered. A baseline count of 1082 will allow the detection of peaks with k≥ 1.1 (cf. Davies et al. 1985b; Kirkwood, 1995).

STA measurements

The aim in these experiments was to quantify the presence of EPSPs from individual EBSNs recorded via intracellular recordings from motoneurones of T9. The experiments were a subgroup of those described by de Almeida and Kirkwood (2010), consisting of motoneurones antidromically identified from either the internal or the external intercostal nerve or ‘unidentified’ motoneurones (for more details, see de Almeida & Kirkwood, 2010). Following the identification of an EBSN, the recording of its discharge was maintained while stable recordings from motoneurones were sought. For each motoneurone, a period of recording that included the motoneurone identification, together with a period with all stimuli switched off (20 s–12 min), was stored on magnetic tape and digitally acquired via the 1401 interface and Spike2 software. The channels recorded consisted of the EBSN recording, an external intercostal nerve for definition of the respiratory cycle (T6 or, for one experiment, T9), a marker pulse for stimulation timing, and a high- and a low-gain version of the intracellular recording (see de Almeida & Kirkwood, 2010). Motoneurones were only accepted for analysis if the membrane potential was more negative than −30 mV near the start of the recording and recordings were usually maintained until the membrane potential deteriorated to an unacceptable level, which could be as positive as −20 mV. For a justification for using low membrane potentials, see de Almeida and Kirkwood (2010). Analysis consisted of performing spike-triggered averaging (STA) for each pair of motoneurones and EBSN, usually using the high-gain motoneurone recording. The procedures and criteria for acceptance of waveforms as representing genuine EPSPs were similar to those in Saywell et al. (2007), including a maximum noise criterion in the averages (10 μV) and the need for repeatability (a similar waveform being seen at the same latency for each time segment of the recording, when the whole recording was split into three successive segments). The minimum number of sweeps accepted for averaging was 850. More details are included in the Results.

In the Results, means are quoted ± SD.

Results

Properties of EBSNs

Recordings from 39 EBSNs are described here, from 20 rats, 25 of the EBSNs contributing data to the cross-correlation results and 15 to the STA results (Table 1). Note that these 39 EBSNs include only those considered tested for connections (see sections below). Their locations are indicated in Fig. 1 and comprise a cluster spanning 1.1–1.9 mm from the midline, 0.9–1.7 mm caudal to obex and 1.05–1.85 below the medullary surface. Although the inevitable variation between animals makes the spread of locations here seem wider than it would be in a single animal (Merrill, 1970), the EBSNs in the rat almost certainly are more scattered than they are in the cat. We can assert this because, while tracking for units, we hardly ever, if at all, observed a dense column of expiratory neurones, which is such an obvious feature in the cat (Merrill 1970; Boers et al. 2005). Nevertheless, through the course of these experiments, EBSNs became easy to recognise, given their characteristic fast firing and incrementing expiratory drive. None of the other expiratory units (e.g. those showing a tonic background discharge or a decrementing pattern) were activated antidromically when tested from the spinal cord electrodes at T11. It should be remembered that all of these recordings were made under moderate hypercapnia.

Figure 1. Positions of EBSNs.

Top, positions projected to a horizontal plane; bottom, projected to a transverse plane. Open circles, units used for cross-correlations; filled circles, units used for STA; triangle, unit used for both.

Most of the EBSNs started to fire around 0.5 s or more into the expiratory phase (Figs 2, 7 and 9). Only two started firing immediately at the termination of inspiration, i.e. in post-inspiration, typically with only one or two spikes, followed by a silent period before the main expiratory burst. For all the EBSNs, the firing rate for each EBSN showed an incrementing pattern. None of the EBSNs was of the step–ramp pattern, as classified by Cohen et al. (1985); all were of the ramp variety.

Figure 2. Example of a set of recordings used for cross-correlation measurements.

Top trace, the firing rate of an EBSN (instantaneous frequency), the discharge of which appears in the second trace. Remaining four traces, efferent discharges in T11 and T10 internal intercostal nerves on the left of the animal, and in the T8 external intercostal nerves on the left and on the right, as indicated. Note that all four nerves show both an expiratory burst (coincident with firing of the EBSN) and a following inspiratory burst.

Figure 7. Example of a recording used for STA.

Simultaneous recording from an EBSN (second trace, with firing frequency shown in the top trace) with an intracellular recording from an internal intercostal nerve motoneurone in T9 (bottom trace) in a decerebrate preparation. The recording from the T6 external intercostal nerve is also included to show the respiratory phase.

Figure 9. Cross-correlation analyses for efferent spikes of different amplitudes.

A, an EBSN recording (instantaneous frequency in the top trace), together with the efferent discharges in the T9 internal intercostal nerve. B, expanded version of the same recording, showing the final part of the first expiration shown in A. Dotted lines (a–e) show the amplitude ranges used for selecting efferent spikes. The level e is also shown in A, for reference. Note that the obvious constancy in the form of the large spikes in the EBSN recording in B shows that the irregular firing frequency evident in A is not due to contamination, but to real irregularity in the EBSN firing. C, auto-correlation histogram for the EBSN, together with cross-correlation histograms calculated separately for five different amplitude ranges for the T9 recording, as indicated, from the smallest efferent spikes (Range 1), up to the largest (Range 5). The histograms are shown scaled according to the baseline count, m, which means that peaks with equal values of k appear with equal amplitudes. The peaks for all five levels, including the peak just visible for Range 1, were all significant. Number of EBSN spikes, 252,047. D, values of k, plotted against the serial order of the amplitude ranges (range 1 involving the smallest spikes) for all 14 of the EBSN/nerve pairs where a similar analysis was possible. Histograms without a significant peak assigned a value of k= 1.0. Symbols: circles, internal intercostal nerves (open circles, the series shown in C); triangles, external intercostal nerves. All of the units were recorded in decerebrates.

For most of the EBSNs, the ramp of activity became particularly steep towards the latter part of expiration, so as to produce a high-frequency burst at a time that usually coincided with the strongest expiratory discharges in the nerves. In some instances, where the strong expiratory bursts in the nerves showed ‘quantal’ variation between respiratory cycles, the EBSN discharge either was absent in the non-expiratory cycles, or lacked the strong burst at the end of expiration (see Figs 2 and 7, also fig. 4A in de Almeida et al. 2010 and fig. 2B in de Almeida & Kirkwood, 2010). The EBSNs illustrated in the figures are typical with regard to their high firing rates. The instantaneous frequency plots make the firing appear very irregular at the high rates, but are nevertheless reasonably accurate (e.g. Fig. 9). Sometimes, the spike amplitude decreased with firing frequency, meaning that a few spikes were missed by the spike discriminator, but the auto-correlation histograms (e.g. Fig. 4A, which applies to the EBSN of Fig. 2) showed that the units were clearly discriminated and the irregularity was not caused by spikes from other units. Modal values of firing frequency (corresponding to the firing rate around end-expiration) were estimated from the auto-correlation histograms as in Davies et al. (1985a) and in Kirkwood (1995) and ranged from 43 to 526 imp s⁻¹. Firing frequencies were fastest for EBSNs in the decerebrates (range 78–526 imp s⁻¹, mean 263 ± 119, n= 25) than in the anaesthetised animals (ranges: ketamine/xylazine or ketamine alone, 90–278 imp s⁻¹, mean 149 ± 62.7, n= 6; α-chloralose, 43–111 imp s⁻¹, mean 79 ± 24.5, n= 8; Fig. 3C). The values were similar whether recorded in cross-correlation or STA experiments. Note that this measure of firing frequency refers to the high activity respiratory cycles if the EBSN was one of those firing fast on only some cycles. Conduction velocities ranged from 13.6 to 43.6 m s⁻¹ (mean 23.8 ± 6.4; Fig. 3A).

Figure 4. Examples of cross-correlation histograms with significant peaks.

Auto-correlation histograms (top traces) for each of two EBSNs, with two cross-correlation histograms from the same EBSN below, for the indicated nerves (T10 Int, T10 internal intercostal nerve; T11 Int, T11 internal intercostal nerve; T9 ext, T9 external intercostal nerve). The EBSNs were recorded in two different decerebrate preparations. Number of EBSN spikes: A, 54,687; B, 144,442. In each instance, the efferent spike amplitude ranges used for the illustrated cross-correlation histograms were the ones which gave the largest signal-to-noise ratio for the histogram peaks (see text).

Figure 3. Characteristics of the EBSNs.

A, conduction velocities. B, conduction times to T10, calculated from the conduction velocities shown in A and for a typical conduction distance of 54 mm. C, modal firing frequencies in late expiration, shown separately for units recorded under two different anaesthetic regimens and after decerebration. Open bars in A, C, units used for cross-correlation; filled bars, units used for STA; grey bars, unit used for both.

Cross-correlation experiments

States of the preparations

For the detection of physiologically significant peaks in EBSN–motoneurone cross-correlation histograms, an appropriately high number of baseline counts per bin must be achieved (Kirkwood, 1979), which therefore requires suitably high expiratory efferent discharge rates in the intercostal nerves. These were only achieved in animals that were decerebrate or under α-chloralose (Table 1). In the rat, varied patterns of respiratory discharges in the intercostal nerves may be seen, according to the anaesthetic protocol and the time within these protocols (de Almeida et al. 2010), the patterns usually including biphasic discharges (i.e. a burst in inspiration and a burst in expiration) in both internal and external intercostal nerves (see also Tian & Duffin, 1996). This is different from the cat or dog, where an expiratory burst is not present in the external intercostal nerve (Sears, 1964a; De Troyer et al. 2005). A typical pattern of nerve discharges is illustrated in Fig. 2. All of the preparations providing data for cross-correlation showed relatively brisk expiratory discharges in at least one external intercostal nerve (in addition to the universal inspiratory burst), although this burst was absent from both of the external intercostal nerves during one of the EBSN recordings. It was also absent from the T6 external intercostal nerve recording for 3/4 of the anaesthetised animals (4/6 EBSNs), but present in all five decerebrate animals tested (T6 not recorded in one) and absent at T6 for only 2/16 EBSN recordings. The state of most these preparations (5/6 EBSN recordings in the four anaesthetised animals and 9/19 of the recordings in the decerebrates) was one where the presumed ‘expiratory oscillator’ was synchronised to the inspiratory oscillator in a one-to-one fashion (de Almeida et al. 2010). For six other EBSNs (one in the anaesthetised group), the pattern included occasional ‘skipping’ of strong expiratory activity (e.g. Figs 2 and 7), and for the remaining five EBSNs, more frequent skipping was seen, e.g. 1 in 2, 1 in 3 or 1 in 4 cycles missed.

Occurrence of histogram peaks and efferent spike selection

Most EBSNs gave significant correlation peaks for at least one of the nerves, such as those illustrated in Fig. 4, often for more than one nerve, including (for two EBSNs) for all four nerves tested. The occurrence of peaks is summarised in Table 2, an explanation for the numbers in the table being as follows. A given EBSN/nerve pair was considered as tested if a histogram resulted with m≥ 1082 for at least one selection of efferent spikes (by spike amplitude). Most often the discharges of an EBSN were correlated with several selections from a given nerve discharge and a peak was deemed to have been present for a given EBSN/nerve pair if it was significant for one of those tested selections (for the significance criterion, see Methods). This procedure is justified in the Appendix, where we show that, in all instances tested, the strength of the correlation, k, increased with the amplitude of the efferent spikes selected.

Table 2.

Proportion and strength of connections revealed by cross-correlation

	Internal intercostal nerve				External intercostal nerve
	EBSNs	EBSN/nerve pairs			EBSNs	EBSN/nerve pairs
			(Values of k)
Anaesthetised (4 rats)	2/6	2/11	1.126, 1.104 k, means (SD)		None tested

			k, means (SD)				k, means (SD)
			min.	max.			min.	max.
						Left
Decerebrate (6 rats)	16/19	24/37	1.052	1.116	4/19	7/25	1.083	1.111
			(0.018)	(0.051)			(0.015)	(0.014)
						Right
						1/10	1.076

The columns ‘EBSNs’ show the proportion of EBSNs giving a significant histogram peak for any nerve counted as tested, with at least one nerve tested for each EBSN. The columns ‘EBSN/nerve pairs’ show the proportion of EBSN/nerve pairs where a significant histogram peak was seen for any level of efferent spike selection, together with appropriate values of k. For the decerebrates (internal intercostal nerves and external intercostal nerves on the left of the animal), several levels of efferent spike selection were usually tested for each EBSN/nerve pair, so the means (± SD) of the maximum and the minimum values from each EBSN/nerve pair are quoted.

The observed ranges for k were 1.026 to 1.266 (internal intercostal nerve) and 1.060 to 1.122 (external intercostal nerve). To characterise the durations and latencies of histogram peaks for a given EBSN/nerve pair, we chose to measure these parameters for the one histogram from that pair where the peak had the highest signal/noise ratio. The signal/noise ratio was defined as the height of the peak above the baseline, divided by the SD of the baseline counts, where the SD is given by , on the assumption of a Poisson distribution of counts in the baseline (Kirkwood & Sears 1982), i.e. (k− 1)m/= (k− 1) . This took values between 3.3 and 13.6.

Proportions of EBSN/nerve pairs giving histogram peaks

Two results are apparent in Table 2. First, more peaks were detected for the decerebrate than for the anaesthetised preparations (for the internal intercostal nerves, 24/37 EBSN/nerve pairs in the decerebrates, 2/11 in the anaesthetised animals, significantly different, χ², P < 0.05). However, it must be remembered that activity levels, for both the EBSNs and the nerves, and therefore also values of m, were generally higher for the decerebrates than for the anaesthetised animals (for the largest range of spike amplitudes considered for each EBSN/nerve pair (Range 1), m varied between 1654 and 24,077 (median 7157), in decerebrates, but between only 1085 and 6360 (median 2529), in the anaesthetised animals). This difference is also reflected in the fact that no EBSN/external intercostal nerve pairs could be counted as tested in the anaesthetised animals. Second, a higher proportion of EBSN/nerve pairs with internal intercostal nerves gave peaks (24/37) than did those with external intercostal nerves (7/25) (significantly different, χ², P < 0.01). Peaks were seen for nerves at each of the segments tested, in similar proportions (e.g. for internal intercostal nerves in decerebrates: T9, 9/16; T10, 13/18; T11, 2/3).

Next, we wished to interpret the presence of histogram peaks in terms of connections. In particular, in terms of the general question posed in the Introduction, we need to identify which of the histogram peaks represented monosynaptic connections to the motoneurones. This is considered below separately for the two nerves, internal or external intercostal.

EBSN/nerve pairs involving internal intercostal nerves

Both the durations of the histogram peaks and their latencies can be used to identify those representing monosynaptic connections. The first distinction needing to be made is between short duration peaks (‘narrow peaks’), representing a raised probability of motoneurone firing occurring predominantly on the rising phase of single fibre monosynaptic EPSPs (Kirkwood & Sears, 1991), and ‘medium-width’ peaks resulting from other motoneurone inputs synchronised to EBSN used for the correlation (Davies et al. 1985a; Kirkwood, 1995). On the assumption that the most important contributor to the latter type of peaks would be connections from other bulbospinal neurones synchronised with the one recorded, the duration of such peaks should be at least as great as the width of the distribution of EBSN conduction times to the segment concerned. In the cat, for both inspiratory bulbospinal neurones to the phrenic motor nucleus, or for EBSNs to thoracic segments, that critical duration, assessed by its half-width (the width of the peak at half-amplitude) was calculated as 1.1 ms (Davies et al. 1985a; Kirkwood, 1995). Here the distribution of conduction times (Fig. 3B) had a half-width that was slightly longer than this, about 1.6 ms, which is the criterion that can be used here.

The distribution of half-widths of the correlation peaks from EBSN/internal intercostal nerve pairs is shown in Fig. 5A (upper plot). All but two of the values were ≤1.0 ms, the peaks concerned therefore being classified as narrow. The other two peaks had half-widths of 1.8 and 1.7 ms, but we have not classified these two as medium-width, for the following reasons. First, in the cat, the medium-width peaks showed other important properties, which confirmed their origin in presynaptic synchrony (Davies et al. 1985a; Kirkwood, 1995), namely that their latencies were not related to the conduction velocity of the bulbospinal neurone concerned and these latencies were mostly shorter than would be expected for a direct connection. The two peaks here did not have abnormally short latencies, and indeed they were exactly as expected for the direct connection (see below). Moreover, they did not show a slowly rising ‘foot’ to the peak, nor did they have a typically slow rise-time (see examples in above references). Instead, they showed a prolonged tail, including a possible second peak delayed by around 1–1.4 ms. Four other examples could also be identified with a similar time course. Although these were categorised as narrow (half-widths 0.4, 0.6, 0.7 and 0.8 ms), they additionally showed a later, second peak. Examples of these types of peaks are included in Fig. 6A. We therefore suggest that these six peaks represent a combination of a mono- and a disynaptic connection. Note that we have not classified individual histograms in this way, only particular EBSN/nerve pairs where the histograms from several levels of efferent spike selection showed this same feature. Unlike the cat, there were therefore no peaks here classified as medium- width.

Figure 5. Characteristics of cross-correlation histogram peaks.

A, half-widths of the peaks, shown separately for internal or external intercostal nerves, as indicated. B, latencies of the peaks, plotted against the axonal time (see text for definition) as calculated for the EBSN and the segment involved. Filled symbols, peaks from internal intercostal nerves; open symbols, peaks from external intercostal nerves. Filled triangles, peaks which were either double, or which involved relatively long half-widths (see text). Open triangle, external intercostal nerve on the right side of the animal. C, transmission delays for the peaks (difference between latency and axonal times), shown separately for internal or external intercostal nerves, as indicated. The line in B is the line of identity.

Figure 6. Cross-correlation histogram peaks indicating non-monosynaptic connections.

A, cross-correlation histograms from one EBSN with three different nerves, T9 and T10 internal intercostal nerves and T9 external intercostal nerve, as indicated. For both of the peaks from the internal intercostal nerves (a, b), late features were evident. The histogram for the T9 external intercostal nerve (c) is included to emphasise the temporal coincidence of its peak with the late features evident in the peaks in the other two histograms. B, examples of troughs or possible troughs. a, histogram obtained from an EBSN with the T8 external intercostal nerve; b and c, histogram obtained from an EBSN with the T9 external intercostal nerve, shown at two different bin-widths, the usual 0.2 ms (b) and 0.4 ms (c). Note expanded time-scale in A compared with Fig. 4, and contracted time-scales in B. Three different EBSNs, in different animals, all decerebrates. Numbers of EBSN spikes: A, 121,391; Ba, 94,880; Bbc, 94,351.

Confirmation that all of the peaks should represent monosynapic connections came from their latencies, which are plotted against the axonal time of the EBSN concerned (see Methods) in Fig. 5B (filled symbols). These latencies were clearly related to the axonal time, the points forming a line, just a short distance above the line of equality. The difference between the latency and the axonal time, the ‘transmission delay’ (Davies et al. 1985a), varied between 0.63 and 1.71 ms (Fig. 5C, upper plot), with a mean of 1.11 ± 0.32 ms. These data are remarkably like those for the narrow peaks for EBSNs in the cat (Kirkwood, 1995), where the equivalent mean value was 1.24 ms, also taken as an indication of a monosynaptic connection. Note that the six latency values from the peaks that were thought to also include a disynaptic component (filled triangles in Fig. 5B) were entirely unremarkable.

EBSN/nerve pairs involving external intercostal nerves

Eight EBSN/nerve pairs showed significant peaks, seven with nerves on the left, one on the right (Table 2). Examples are included in Figs 4B, 6Ac and 10B. These eight peaks were derived from four EBSNs in two animals, and thus each of these two EBSNs gave significant peaks for both of the EBSN/external intercostal nerve pairs tested. The half-widths of the peaks varied from 0.8 to 1.7 ms (Fig. 5A, lower plot). Six of these peaks (half-widths 0.8–1.2 ms) could have been classified as narrow, but we have chosen not to do this, nor to classify the other two as medium-width, because all of the peaks appear to fit better a single population with unique characteristics. Unlike medium-width peaks, their latencies were all greater than expected from a monosynaptic connection and were also clearly related to the EBSN axonal time. Like the points derived from the EBSN/internal intercostal nerve plots, the equivalent measurements here (open symbols in Fig. 5B) also lie on a line parallel to the line of identity, but higher up on the plot. Thus, their transmission delays ranged from 2.16 to 2.93 ms (mean 2.62 ± 0.18 ms, i.e. a mean 1.51 ms longer than the previous values taken as representing a monosynaptic connection and constituting a non-overlapping range; Fig. 5C).

Figure 10. Cross-correlation analyses for efferent spikes of different amplitudes (second method).

In this method, the amplitude ranges were not independent, but represented all spikes with an amplitude greater than a given level. Range 1 corresponded to the lowest level, so always included the largest number of efferent spikes, and therefore the highest value of m. A and B, two different EBSNs in two different animals: the EBSN auto-correlation histogram and the cross-correlation histograms for all four ranges analysed for these two EBSNs with a particular nerve are illustrated for each EBSN. A T10 internal intercostal nerve was used for A and a T9 external intercostal nerve for B. Histograms scaled according to the values of m in each of A and B. All the illustrated peaks were significant. C, values of k plotted against m for all of the 28 EBSN/nerve pairs (22 for the internal intercostal nerves) where this analysis yielded significant peaks for more than one range. The continuous, curved line represents the threshold for values of k to be significant for the given value of m. Note that for a few EBSN/nerve pairs, where significant peaks were seen for at least two of the chosen amplitude ranges, peaks were present for another range, as assessed by eye, but were not significant. The values for these peaks are joined to those of the other ranges by dotted lines. Symbols: small filled circles, pairs including an internal intercostal nerve; triangles, pairs including an external intercostal nerve; open circles and open triangles, measurements from the histograms in A and B, respectively. All the data from decerebrate preparations. Numbers of EBSN spikes: A, 77,156; B, 144,442. Note the log scale for m in C.

From the latency measurements alone one can deduce therefore that these peaks were derived from disynaptic connections. It is then all the more interesting that the four EBSNs that gave the peaks for the external intercostal nerves were the same that gave the six peaks with probable disynaptic components in the internal intercostal nerve correlations (e.g. Fig. 6A). The mean half-width of the eight peaks was 1.20 ± 0.28 ms, compared with 0.64 ± 0.21 ms for the peaks from internal intercostal nerves (except for the two outliers ascribed to the presence of a disynaptic connection). The one peak seen from an external intercostal nerve on the right did not appear different from the other seven.

Histogram troughs

Troughs or possible troughs were detected for six EBSN/nerve pairs (five EBSNs), all involving external intercostal nerves. Two of the troughs, one of these being significant (Fig. 6Ba, k= 0.94) had relatively fast falling phases (i.e. initial slopes), but the others (e.g. Fig. 6Bbc) were harder to quantify, with slower falling phases and apparently longer durations. It was therefore hard to separate these from a sloping baseline, so their significance was not tested, and their existence should be regarded as provisional. The latency of the one significant trough was about 5.4 ms. With an axonal time of 1.19 ms, this therefore had a transmission delay of 4.21 ms, suggesting a trisynaptic linkage. However, there are suggestions that the trough is preceded by a short-duration peak (although not significant), as was the case for some of the other possible troughs. In such a situation, the latency becomes hard to define.

STA experiments

Averages were derived for 40 EBSN/motoneurone pairs. Two of these were rejected because they showed ‘synchrony potentials’ (Saywell et al. 2007), which in these instances comprised slowly rising depolarising waves (rise-times about 4 ms), with latencies shorter than the axonal times (about 0 and 1.5 ms) and amplitudes of 15 and 40 μV. Such potentials are believed to originate in excitation from other presynaptic neurones synchronised with the EBSN providing trigger spikes (analogous to medium-width peaks in the cross-correlations). These potentials could have concealed small EPSPs, so these two pairs were counted as not tested. Thus, there remained 38 EBSN/motoneurone pairs, arising from 37 motoneurones (18 in six anaesthetised animals, 19 in five decerebrates) paired with 15 EBSNs (eight in anaesthetised animals and seven in decerebrates). In the anaesthetised animals, 13 of the pairs were recorded under ketamine/xylazine, three under ketamine alone and the remaining three under α-chloralose. The membrane potential of the motoneurones, measured early in the recording, ranged from −30 to −70 mV (mean 48.3 ± 10.1 mV), being very similar in the anaesthetised and decerebrate animals (mean values 47.0 ± 8.5 and 49.5 ± 11.3 mV, respectively).

The numbers of motoneurones of different types used for averaging are shown in Table 3, along with the EPSPs observed. The first four columns list the motoneurones according to the types of their central respiratory drive potentials (CRDPs, Sears, 1964b), as defined by de Almeida and Kirkwood (2010), namely expiratory (depolarised in expiration), inspiratory (depolarised in inspiration) and biphasic (showing a depolarising wave in both phases). The biphasic group is split into two, type 1, where the larger of the two waves is inspiratory, and type 2, where the larger is expiratory. Also included is a summary of the amplitudes of the expiratory ramp components in the CRDPs for different categories of motoneurones (inspiratory CRDPs excluded), measured as in Saywell et al. (2007). The three unidentified motoneurones in the decerebrate group (all from one experiment) were all recorded at the most rostral locations sampled, and therefore were most probably unidentified because they were motoneurones from the next segment (T8). The one unidentified motoneurone in the anaesthetised group had a large (9 mV) inspiratory CRDP and probably innervated levator costae (de Almeida & Kirkwood, 2010). The numbers of sweeps used varied from 859 to 28,212, with similar ranges for the anaesthetised and decerebrate populations (medians 6035 and 6243, respectively). Figure 7 shows one of the biphasic (type 2) CRDPs, recorded in an internal intercostal nerve motoneurone, together with the discharge from the simultaneously recorded EBSN.

Table 3.

Summary of motoneurones, CRDPs and EPSPs

						Expiratory ramp (mV)
	Exp.	Bi(t1)	Bi(t2)	Insp.	Total	Range	Median
Anaesthetised
Internal intercostal nerve (n= 15)	11	4	—	1	17	0–5	1
No. of EPSPs (amplitude)		1 (13 μV)
External intercostal nerve	1	—	—	—	1	1	1
Unidentified	—	—	—	1	1	—	—
Decerebrate
Internal intercostal nerve	6	1	8	—	15	0–10	4
No. of EPSPs (amplitude)	2 (12, 3 μV)		2 (36, 4 μV)
External intercostal nerve	—	—	1	—	1	4	4
No. of EPSPs (amplitude)			1 (20 μV)
Unidentified	2	—	1	—	3	2–5	3
No. of EPSPs (amplitude)	2 (11, 9 μV)		1 (11 μV)

Each entry in the first four columns lists the number of motoneurones in that category with (below) the number of EPSPs found and their individual amplitudes. Types of CRDP: Exp., expiratory; Bi(t1), biphasic type 1; Bi(t2), biphasic type 2; Insp., inspiratory (see text for definitions). The fifth column lists the total numbers of EBSN/motoneurone pairs tested by STA. The last two columns show the range and median value for the expiratory ramp components of the CRDPs. In the first row, n= 15, instead of 17, because one motoneurone was inspiratory and one motoneurone was tested with two EBSNs.

Table 3 also includes the numbers and amplitudes of EPSPs detected. Only one EPSP was seen out of 17 tests in the anaesthetised animals (inspiratory motoneurones excluded, one expiratory motoneurone tested with two EBSNs), but 8/19 were seen for the decerebrate group (a significant difference, exact test, P= 0.014). Four out of seven of the EBSNs gave rise to these EPSPs, in 4/5 of the decerebrate animals. The EPSP amplitudes were generally small, the largest being 36 μV (Fig. 8Aa, averaged from the EBSN/motoneurone pair in Fig. 7). The mean EPSP amplitude was 13.1 ± 9.4 μV (n= 9), or 13.25 ± 9.9 μV (n= 8) for the decerebrates alone. Two other EPSPs are included in Fig. 8A, from an expiratory internal intercostal nerve motoneurone (Fig. 8Ab) and from a biphasic (type 2) external intercostal nerve motoneurone (Fig. 8Ac), together with the auto-correlation histograms of the three EBSNs. Note that in Fig. 8Ab the periodicity in the auto-correlation histogram is reflected in the averaged synaptic waveform, just as was the case in the cross-correlations (e.g. Fig. 4B).

Figure 8. STA EPSPs and their characteristics.

A (left), three EPSPs, averaged in three motoneurones of different types, in three different animals. The averaged trigger spike is shown below each EPSP and the auto-correlation histograms are shown on the left for the same recordings. a, internal intercostal nerve motoneurone (biphasic, type 2, illustrated in Fig. 7); b, internal intercostal nerve motoneurone (expiratory); c, external intercostal nerve motoneurone (biphasic, type 2). Numbers of EBSN spikes: a, 15,985; b, 20,015; c, 1825. B, segmental delays for all the EPSPs detected (symbols), compared with the transmission delays from the cross-correlation analyses (histogram replotted from Fig. 5C, two histograms combined, with external intercostal nerve data now represented by filled columns). Open circles, internal intercostal nerve notoneurones, expiratory; open up triangles, internal intercostal nerve notoneurones, biphasic; open down triangles; unidentified motoneurones; filled circle, external intercostal nerve motoneurone. C, amplitude of EPSPs plotted against the amplitude of the expiratory ramp in the CRDP of the motoneurone. Circles, decerebrates; diamond, anaesthetised animal. The line is the regression line for the values from the decerebrates (slope not significant).

The latencies of the EPSPs are shown as segmental delays in Fig. 8B, compared with the transmission delays from the cross-correlation measurements. The mean for the internal intercostal nerve motoneurones, considered together with the unidentified ones (most likely to be internal intercostal nerve motoneurones from T8), was 0.91 ± 0.47 ms and the value for the one external intercostal nerve motoneurone was 2.07 ms. These values were therefore 0.20 and 0.46 ms shorter than the mean values for transmission delays. The rise-times (10–90%) for the EPSPs for six of the internal intercostal nerve or unidentified motoneurones were between 0.5 and 0.7 ms with the other two being 1.4 and 1.8 ms. That for the one external intercostal nerve motoneurone was 1.0 ms. Finally, it seemed possible that the EPSP amplitude might be related to the expiratory ramp amplitude as in Saywell et al. (2007), so these were plotted against each other (Fig. 8C). Although there appears to be a positive correlation, the slope of the regression line fitted to the eight points for the decerebrates was not significant (P= 0.114). Note also that the mean amplitude of the expiratory ramps (decerebrate preparations only) for those motoneurones that showed an EPSP (3.6 ± 1.6 mV, n= 8) was similar to that for those that did not show an EPSP (4.2 ± 2.1 mV, n= 11).

Discussion

There are two important results from this study. First, the EBSNs make widespread monosynaptic excitatory connections to internal intercostal nerve motoneurones in the rat. Second, although the expiratory discharges seen in the external intercostal nerve of the rat can sometimes be as strong as those in the internal intercostal nerve, the EBSNs do not show a monosynaptic connection to this group of motoneurones, but instead a disynaptic excitatory connection can be found. We believe that this is the first time that two different connections have been demonstrated to be made selectively to two anatomically defined sets of motoneurones from the same single-fibre inputs. There are also a number of other new observations with regard to the EBSNs, which will be dealt with first

EBSN properties

The present study is the first to report that EBSNs in the rat project as far as T10 or T11. Although we did not formally investigate what proportion did so (e.g. by testing whether those which projected into the cervical spinal cord also projected to T11), we suspect that nearly all EBSNs did so. Our reason for this conclusion is the observation that all of the neurones that we recorded that showed a strong, phasic, incrementing expiratory discharge were antidromically activated from T11.

A second new observation is that we found that the mean conduction velocity of the EBSN axons was 23.8 ± 6.4 m s⁻¹. This is much faster than previously reported by Saether et al. (1987, 13.2 ± 3.2 m s⁻¹) or by Shen and Duffin (2002, 7.4 ± 0.4 m s⁻¹). This is almost certainly because a more accurate method was used here. This method is based on the orthodromic conduction time and was verified by Kirkwood (1995), having also been used by Davies et al. (1985a). The values of conduction velocity measured by Davies et al. (1985a) were similarly faster than those measured by other authors, who, like both Saether et al. (1987) and Shen and Duffin (2002), used single-point antidromic conduction times from the cervical cord, but they agreed with values measured by Dick and Berger (1985) who used a two-point stimulation method.

Finally, it is worth noting the very fast firing frequencies attained by the EBSNs, particularly in the decerebrates. The mean for this group, 263 imp s⁻¹, is at least 2.5 times faster than the figures for the cat from our laboratory (Kirkwood, 1995; Ford et al. 2000; Saywell et al. 2007), although it should also be noted that the frequencies under anaesthesia here were lower, particularly under α-chloralose (79 imp s⁻¹). This is close to the values from the cat, which also were determined under anaesthesia. However, judging by the illustrated examples, EBSNs in decerebrate cats also show relatively modest firing frequencies (Cohen et al. 1985). It is remarkable that some of the EBSNs here fired in bursts of 400 or 500 imp s⁻¹ for several hundred milliseconds every respiratory cycle, for many hours. Moreover, their synapses transmitted apparently successfully at this rate (e.g. the unit of Fig. 9A–C (Appendix) or that of Fig. 8Ab, although the EPSP for the latter was very small). Note, however, that EBSNs are not unique in this: Duffin and van Alphen (1995) observed that their inspiratory equivalents in the rat, inspiratory bulbospinal neurones (defined in this case by their connections to phrenic motoneurones), had mean firing rates of 190 ± 71 imp s⁻¹ under pentobarbital anaesthesia and at near eupnoeic levels of CO₂, which compares with 52 ± 27 imp s⁻¹ for the cat (Long & Duffin, 1984).

The monosynaptic connections

There seems little doubt, based on both the time courses and the latencies of the individual histogram peaks for the internal intercostal nerve discharges, that these peaks represent monosynaptic connections, just as in the cat. The logic for this conclusion is set out under the Results. The only peaks that were questionable in this regard are the few with relatively long half-widths or those which were accompanied by a second peak (Fig. 6A). For these, again as previously argued, the presence of an additional disynaptic component is a better explanation. It is of interest that there was a general absence of pre-synaptic synchronisation, as deduced from the paucity of medium-width peaks in the histograms, and despite the expectation that in decerebrates, as compared with anaesthetised cats, the increased liveliness of the preparations might make such effects more prominent. Of course, very weak medium-width components in the observed peaks cannot be excluded, but the general absence of these effects might be explained by the generally weaker connections seen in the rat (see below), together with the probable existence of fewer EBSNs, which were the cells hypothesised by Kirkwood (1995) to be the source of the effect. The firing of some pairs of medullary expiratory neurones in the rat have been shown to be synchronised (Shen & Duffin, 2002), but not specifically the EBSNs.

From a methodological point of view, it is of interest that the widest of the narrow peaks here (ignoring the two peaks assumed to involve later components) had a half-width of 1 ms. This corresponds to the very similar observation that the half-widths of the narrow peaks derived from correlations between the discharges of inspiratory bulbospinal neurones and thoracic motoneurones by Davies et al. (1985a) were no greater than 1.1 ms. Thus, the new data reinforce the view that this half-width upper limit is a good descriptor of peaks derived from monosynaptic connections to motoneurones, which is coincidently the same as the theoretical lower limit for medium-width peaks derived from the conduction times both of the cat inspiratory bulbospinal neurones to the phrenic nucleus (Davies et al. 1985a) and of the cat EBSNs to thoracic levels (Kirkwood, 1995).

A concomitant observation is the absence of monosynaptic connections to external intercostal nerve motoneurones. Is this as secure as the observation of the presence of the connections to those of the internal intercostal nerve? The sample was smaller (fewer histograms reached the threshold level for m and the values of m were generally lower than for the internal intercostal nerve), but nevertheless the sensitivity was great enough to reveal the disynaptic peaks for several of the EBSNs. We are therefore confident that the difference between the two nerves is clear enough to be taken as good evidence for specificity in the connections.

Strength of the monosynaptic connections

Again comparisons should be made with the cat (Kirkwood, 1995), first with regard to the cross-correlation measurements, using the data from the decerebrates in the present study. A similar proportion of EBSNs gave connections to the internal intercostal nerve motoneurones (here, 16/19, 84%; cat, 23/27, 85%) and a similar proportion of EBSN/nerve pairs revealed connections (here, 24/37, 65%; cat 24/33, 73% or 35/58, 60%, excluding significant medium-width peaks and according to the threshold level of m chosen). However, the two species differ in the amplitudes of the peaks, here in the range k= 1.052–1.116, but in the cat a mean of 1.16. We have quoted a range of values here for the rat because of the almost universally observed dependence here of k on the spike amplitude range chosen (see Appendix). This itself could have two origins, first a mixture of alpha and gamma motoneurones for the smaller spike amplitude ranges (assuming weaker histogram peaks for the gammas) and, second, within the alpha range, a dependence of the connections on the efferent spike amplitude similar to that shown for inspiratory bulbospinal neurones (IBSNs) to phrenic motoneurones (Davies et al. 1985b). The problem of identifying the gamma motoneurone spikes is not only that there is a lack of suitable criteria as compared with the cat (a variety of indirect identifying features were used by Sears, 1964a), but also that an overlap of spike sizes between alphas and gammas is to be expected in the rat, as their conduction velocities are known to overlap (Andrew & Part, 1972; Andrew et al. 1978). We therefore suggest (arbitrarily) that a figure for k in the middle of the range, say k= 1.08, should be chosen to represent the connection strength for all the alphas. In that case, the monosynaptic connections in the rat from each EBSN could be said to be around half as strong as those in the cat.

The question of then estimating the total monosynaptic input to internal intercostal nerve motoneurones involves more assumptions. Experience of tracking with a microelectrode in the rat medulla leads to the view that there are far fewer EBSNs in the rat than in the cat, probably no more than 100 on each side. For the cat, the number assumed by Kirkwood (1995) and by Saywell et al. (2007) was 400. This difference is partly compensated for by the fact that the EBSNs were firing 2.5 times faster in the rat than in the cat. Taking these two together leads to another factor (2.5/4, i.e. 0.625) in comparing rat and cat, leading to a view that the overall strength of the monosynaptic input is likely to be around one-third of that in the cat.

The above assessment was made with respect to the decerebrate preparations. In the anaesthetised animals, significantly fewer connections were detected, but an important factor here is the much lower level of activity (fewer efferent spikes and more slowly firing EBSNs), reflected in the lower values of m. This, for statistical reasons, means that fewer histogram peaks would be expected to be detected. Furthermore, the lower level of efferent activity almost certainly also means that the larger motoneurones, whose larger efferent spikes were represented in the strongest of the correlation peaks, were not recruited in this state. Thus, the detection of fewer connections from the cross-correlations cannot be taken as evidence in itself for weaker connections operating in the anaesthetised animals, but it does emphasise the particular requirements that must be met to detect the connections by this technique.

STA vs. cross-correlation and functional relevance

First note that the segmental delays of the EPSPs (Fig. 8B), being on average a little shorter than the comparable transmission delays from the cross-correlations, were consistent with their representing monosynaptic connections to the internal intercostal nerve motoneurones or unidentified motoneurones, together with a disynaptic connection (represented by one EPSP) to an external intercostal nerve motoneurone. To that extent, the STA measurements confirm the cross-correlation results.

With regard to the strength of connections, we should compare the two types of measurements in the decerebrates. Nearly all of the EPSPs were detected in decerebrates, as were most of the cross-correlation peaks. Following the arguments of Kirkwood and Sears (1991) (also see Kirkwood, 1995; Saywell et al. 2007), a cross-correlation peak with k= 1.5 may be taken as equivalent to an EPSP of amplitude 100 μV. The value of k to be used from the present experiments should be that for all the nerves tested. Therefore, the k value assumed above (1.08) gives a value for (k– 1) of 0.08, which must be multiplied by the connection frequency to give 0.08 × 16/19 = 0.0674, on the assumption of k= 1.0 for flat histograms. This therefore predicts an equivalent EPSP amplitude of 100 μV × 0.0674/0.5 = 13.5 μV. The observed mean EPSP amplitude (taking zero amplitude for the absence of an EPSP) was 13.25 × 8/19 = 5.6 μV. These two numbers (differing by a factor of 2.4) are as close as might be expected, given all the assumptions and sampling issues involved. The observed EPSPS can thus be said to be confirmatory of the connections revealed in the cross-correlations.

It is also worth calculating what this might represent functionally, i.e. how much average depolarisation might be expected from this connection. For this we have used the same calculation as in Saywell et al. (2007) for the cat (see their p. 779) and an EPSP amplitude given by the mean of the two values quoted above (9.6 μV). We have not estimated the half-widths of the EPSPs here, because most of the averages were quite noisy, but we have retained the mean value used by Saywell et al. (2007), and also an assumption that there is an input to the motoneurones via crossed collaterals of EBSN axons in the contralateral spinal cord that is of the same proportion as in the cat. These numbers then predict that the amplitude of the expiratory ramp that these monosynaptic connections would produce is as little as 0.82 mV. The observed mean amplitude for the expiratory ramp was 3.9 mV (for all 19 motoneurones in the decerebrates). Thus, we conclude that most of the depolarisation in these ramps is likely to come from sources other than the monosynaptic input. This is then consistent with the lack of overall correlation between EPSP amplitude and expiratory ramp. The apparent relationship in Fig. 8C is probably fortuitous. With regard to the monosynaptic input from EBSNs to internal intercostal motoneurones in the rat, we are therefore in a very similar position as was Davies et al. (1985b) with regard to the monosynaptic input from IBSNs to external intercostal motoneurones in the cat. With appropriate methods, the connection is easy to demonstrate, but it contributes a relatively weak overall input. However, Davies et al. (1985b) also pointed out that even weak inputs may have major effects at threshold. Moreover, since then, it has also become increasingly clear that under some circumstances inputs as weak as this may be greatly amplified by intrinsic motoneurone properties (see Enríquez Denton et al. 2012 for examples and references).

There is one more issue to consider with regard to this connection, which is that we found significantly fewer EPSPs in the anaesthetised than in the decerebrate animals. This is of interest because the equivalent EPSPs in the cat were hypothesised by Saywell et al. (2007) to be very sensitive to barbiturate anaesthesia. Clearly a similar hypothesis, that these synapses were particularly sensitive to anaesthetics, could also explain the observation here. However, although the difference was significant, it must be said that the sample was small, and the hypothesis really needs a more robust test. Nevertheless, the calculation above concerning the small proportion of the expiratory motoneurone input routed via this pathway emphasises that, even if transmission at these synapses was depressed by anaesthesia, the weaker expiratory output from the motoneurones in the anaesthetised animals cannot be directly related to this. Of course, a depression of other synapses from these same axons on the interneurones presumed to transmit the main part of the expiratory drive, coupled with the lower firing rates of the EBSNs, would provide an ample explanation. In this context, note that the expiratory ramps in the anaesthetised animals were generally smaller than were those in the decerebrates (Table 3).

The disynaptic connections

Our identification of the cross-correlation peaks for the external intercostal nerves as representing a disynaptic connection is based solely on their latencies. However, we believe this is a sufficient justification. Both the obvious relationship between latency and EBSN conduction velocity (Fig. 5B, open symbols) and the absence of any very short latencies makes pre-synaptic synchrony a very unlikely explanation. These two observations, coupled with the distribution of segmental delays, which were longer than the segmental delays from the internal intercostal nerves, and with a non-overlapping distribution, make the identification certain. Also, the actual value of the difference in segmental delays between the two populations (1.51 ms) is appropriate to one extra synapse. Perhaps this might be even more appropriate if it were closer to the monosynaptic value (1.11 ms), but such an equality should only be expected if the interneurones concerned were assumed to be strictly segmental. In fact, we do not know in which segment they lie: they might be situated in a more rostral segment, with slightly slower conducting axons, or in a more caudal one, with ascending axons, or they might be situated contralaterally. Indeed, once we had observed the first example of disynaptic connections, it was the strong possibility of contralaterally projecting interneurones (Schmid et al. 1993; Saywell et al. 2011) that made us include in our measurements an external intercostal nerve on the right side. The idea here was to look for the connections from the EBSNs via a set of interneurones ipsilateral to the descending axons, and which had connections to contralateral motoneurones. The result (one such connection observed) gave only minimal support to this possibility.

The half-widths of the histogram peaks for the external intercostal nerves were longer than those for the internal intercostal nerves, as appropriate for the extra temporal dispersion involved in a disynaptic connection. With a mean of 1.2 ms, these half-widths were a little shorter than could have been predicted either from observed single-fibre disynaptic PSPs (Watt et al. 1976; Takakusakai et al. 1989) or from the model used by Vaughan and Kirkwood (1997), and were also shorter than the few peaks seen by Tian and Duffin (1996) in cross-correlations between IBSNs and external intercostal motoneurones, also ascribed to disynaptic excitation. They were nevertheless within the range predicted from Watt et al. (1976) (PSP rise times of 0.6–2.3 ms). Moreover, there is an additional observation here that could account for a reduced duration of these peaks. This is the possible presence of inhibition. A few examples of inhibition (troughs) were seen in the correlation histograms, all of these being in histograms from external intercostal nerves. Some of the troughs that were observed (Fig. 6B) were not very distinct, but it is possible that this was because they were mostly a result of simultaneous excitation and inhibition. The trough with the sharpest onset had a long latency, appropriate for a trisynaptic connection, but it could easily be hypothesised that near-simultaneous disynaptic excitation and inhibition could readily lead to such a resulting waveform. Although hypothetical, this idea is attractive because of the simultaneous (on a broader time scale) excitation and inhibition seen in internal intercostal nerve motoneurones from some of the same motoneurones used in this study (de Almeida & Kirkwood, 2010). Such an effect was not recognised by de Almeida and Kirkwood (2010) in recordings from external intercostal nerve motoneurones (not surprisingly, because they were few in number), but is quite likely to be present, simply by comparison with the internal intercostal nerve motoneurones. Both of these groups of motoneurones showed CRDPs with biphasic excitation.

We cannot assess the functional significance of the strength of the disynaptic connections revealed by cross-correlation, as the extent of the divergence to and convergence from the interneurones involved is unknown, and the influence of these factors on the cross-correlations is equally obscure. For reference, however, the strength of correlation peaks seen here between EBSNs and external intercostal nerves (present for 7/25 pairs, with k between 1.083 and 1.111) is rather stronger than were the likely disynaptic peaks reported by Tian and Duffin (1996) between inspiratory bulbospinal neurones and (mostly) external intercostal nerves, which were present for 6/30 pairs with a mean value of k of 1.06.

Finally, we should consider the evidence for disynaptic excitation that was also apparent between some EBSNs and internal intercostal nerve motoneurones. This evidence is not as firm as that for the external intercostal nerve group, because the simultaneous monosynaptic excitation precluded a clearly defined latency. However, with only a minority of the excitation resulting from the monosynaptic path to these motoneurones, and with the existence of a disynaptic connection to the external intercostal nerve motoneurones, the result is not surprising. Moreover, as far as can be judged, the extra peaks appearing in the correlation histograms from the internal intercostal nerves occurred at latencies very similar to those for the external intercostal nerves (Fig. 6A). In the end, perhaps the most convincing evidence that these extra peaks do represent a disynaptic link is that all of them occurred for the same four EBSNs, in two animals, that showed the disynaptic excitation of the external intercostal nerve motoneurones. Note that we are not claiming that it is the same interneurones that are responsible for the excitation of both groups of motoneurones (although it might be), but rather that it is the state of the preparation at the particular times concerned that allowed this pathway to show up in the cross-correlations. This is the same argument as used by Kirkwood et al. (1982), when noting the correspondence (according to the states of the preparations) between the presumed oligosynaptic excitation of external intercostal nerve motoneurones from homonymous muscle spindles and the duration of simultaneously recorded motoneurone synchronisation.

Wider implications

To our knowledge this is the first report of central single-axon inputs giving selectively a monosynaptic input to one group of motoneurones and a disynaptic input to another group. This selectivity is of interest because it gives support to the idea that the expiratory excitation of the two groups can be independently controlled. This is consistent with the view we previously put forward (de Almeida & Kirkwood, 2010), with regard to the simultaneous excitation and inhibition seen in internal intercostal motoneurones, that the inhibition most likely represented part of the spinal mechanisms which allow reconfiguration of the spinal cord circuits for different motor acts. In a similar way, here, the disynaptic excitation of external intercostal nerve motoneurones might represent activity pertaining to a quite different motor act than does the monosynaptic excitation of the internal intercostal group.

With regard to one of our initial aims, which was to define a long descending pathway to motoneurones that was suitable for exploitation in experimental spinal cord injury, the present demonstration of a pathway with two specific sets of connections to thoracic motoneurones might seem ideal. This would allow investigations into changes in the specificity of the connections under conditions of plasticity or regeneration. However, one difficulty remains: the relative weakness of the pathways demonstrated by the methods chosen (which are probably the only ones suitable for investigating the specificity). If these pathways are to be investigated under other conditions, then care will need to be taken to ensure either sufficiently strong motoneurone discharges (for cross-correlations) or a large enough sample of intracellularly recorded motoneurones (for STA). A considerable advantage could be gained if the interneurones that are assumed here to transmit most of the drive to either set of motoneurones could be identified.

Glossary

CRDP

central respiratory drive potential

EBSN

expiratory bulbospinal neurone

EPSP

excitatory postsynaptic potential

IBSN

inspiratory bulbospinal neurone

STA

spike-triggered averaging

Appendix

Relationship between cross-correlation strength and amplitude of efferent spikes

Our intention in this study was to measure connections to alpha motoneurones, which will generally be those giving the larger spikes in the recordings. However, there is no reliable criterion for separating these from the smaller gamma motoneurones. In the cat, Kirkwood (1995) used a criterion based on the observations of Sears (1964a), derived from intercostal ‘nerve filament’ recordings, that the gamma motoneurones tended to fire throughout the respiratory cycle. Given the different patterns of nerve discharges in the rat, this criterion could not be relied upon, so an empirical approach was adopted, which was to analyse successively different spike amplitude ranges. Spike amplitudes were selected via the Spike2 software after acquiring the medullary recording as an analog signal. Choosing several independent (multi-unit) spike amplitude ranges, as in Davies et al. (1985b), was limited by the decision here to consider only correlation histograms with baseline counts of m≥ 1082. This relatively severe criterion was chosen because of the relatively weak histogram peaks observed (values of k often <1.1). For the 25 pairs of EBSNs and internal intercostal nerves that are listed in Table 2 as showing significant peaks (all in decerebrates), it was possible to choose separate spike amplitude ranges in 12 of these, where each range alone gave a significant peak (Fig. 9). For the seven pairs of EBSNs and external intercostal nerve recordings on the left side where significant peaks were detected, separate spike amplitude ranges that gave significant peaks were similarly possible for two pairs, and on the right side only 1/10 pairs gave a peak, with separate ranges not being possible for this one. In all these 14 pairs where peaks were detected for separate ranges, the values of k increased with the amplitude of efferent spikes selected, as shown in the summary plot (Fig. 9D).

An alternative procedure was also adopted, where successive amplitude ranges that comprised all the spikes above a given threshold were considered, the threshold being successively raised. The threshold level was varied so that m was successively reduced by a factor of about 2, the final value (involving the largest efferent spikes) being between 2000 and 1082. For internal nerve discharges in the decerebrates, 2–6 ranges were involved. In the anaesthetised animals, mostly 2–3 ranges, but for four pairs enough counts were available for only one range. For the external nerve discharges (all in decerebrates), mostly 2–5 ranges, and only one range in two instances. If a histogram peak was present it was usually seen for all or most of the ranges (two examples are shown in Fig. 10) and in all instances k increased with the threshold amplitude, consistent with the increase in k seen for the separate amplitude ranges. Indeed, most often the peaks would not have been detectable if k did not increase with spike amplitude, because m determines the significance threshold and m decreases with the number of efferent spikes. Figure 10C makes this explicit by plotting k for all of the peaks detected for more than one range against m, with the significance criterion also plotted for comparison. An increase in k as m decreased was seen for all of the plots and is directly equivalent to the increase in k as spike amplitude increases.

The universality of this result, that (1) k always increased with the increased bias towards large amplitude spikes, and (2) for most EBSN/nerve pairs, significant peaks were seen for most of the ranges selected, is an important feature justifying the consideration of multiple correlations from each EBSN. Without these two features, the statistical significance of the observations could have been compromised. A problem remains in how to characterise the strength of the correlations (see Discussion), but Table 2 includes an empirical description, where summary figures are given for the lowest and the highest values of k for each EBSN.

Additional information

Competing interests

None.

Author contributions

Both authors contributed to all aspects of this study.

Funding

The work was supported by the International Spinal Research Trust and the Brain Research Trust. A.T.R.deA. held an MRC studentship.