Preferential formation of strong synapses during re-innervation of guinea-pig sympathetic ganglia

Abstract

1. Re-innervation of partially denervated sympathetic ganglion cells was investigated using intracellular recording in guinea-pig lumbar paravertebral ganglia in vitro. The question addressed was whether the pattern of innervation by strong (suprathreshold) and weak (subthreshold) inputs seen normally was restored during re-innervation.

2. L5 ganglion cells each normally received 3.9 ± 0.2 preganglionic inputs of which 1.2 ± 0.1 were strong. Only 0.9 ± 0.1 inputs arose from the L4 segment, the last of the thoracolumbar outflow, and only 11 % of these were strong.

3. Three to five weeks after cutting the sympathetic chain above the L4 white ramus, each neurone received 2.1 ± 0.1 inputs after sprouting of surviving axons. Nearly 60 % of neurones received a strong input and the normal ratio of weakstrong synapses was restored.

4. Total charge transfer evoked by L4 inputs increased from 8.8 ± 1.4 to 27.6 ± 2.4 pC per neurone after re-innervation, reaching 77 % of that in normal ganglia. This was primarily due to the formation of new strong inputs of normal size.

5. The synaptic events at the new strong synapses and the types of Ca²⁺ channel mediating transmitter release (N-type and channels resistant to specific antagonists) were the same as those at control strong synapses.

6. The data indicate that, following partial denervation, sprouting of surviving preganglionic axons results in the preferential formation of strong synapses with the same characteristics as those in normal ganglia. Thus the pattern of functional transmission by a single strong input to each cell was restored rather than the recovery of the number of synaptic connections.

Cholinergic synapses in mammalian sympathetic ganglia, unlike most central nervous system (CNS) synapses, have two distinct degrees of effectiveness: (1) each ganglion cell receives at least one preganglionic input which is always suprathreshold or ‘strong’ and (2) larger numbers of subthreshold or ‘weak’ inputs converge on the same cell. From in vivo recordings (Skok & Ivanov, 1983; McLachlan et al. 1997), it is clear that strong synapses are responsible for the discharge of postganglionic neurones in paravertebral ganglia, leaving the role of weak synapses as yet unexplained. Yet weak synapses make up the majority of inputs received by each paravertebral ganglion cell.

In rat sympathetic ganglia, the first synapses to appear during development are weak and these progressively increase in quantal content (Hirst & McLachlan, 1984). Strong inputs appear about 2 weeks postnatally and the proportion of cells with a strong input progressively increases thereafter (Hirst & McLachlan, 1984). These observations suggest that during development, strong synapses mature from originally weak synapses. Following denervation of adult ganglion cells, a similar sequence occurs; initial synaptic contacts are weak, but suprathreshold transmission is subsequently re-established (McLachlan, 1974; Njå & Purves, 1977, 1978). In partially denervated superior cervical ganglia (SCG), if only restricted functional subsets of preganglionic axons survive, the surviving axons sprout and form inappropriate connections with postganglionic neurones that they did not previously innervate (Murray & Thompson, 1957; Guth & Bernstein, 1961; Liestol et al. 1987). The capacity of the preganglionic neurones to sprout appears to be large since, even when the number of remaining preganglionic axons is small, the majority of neurones are re-innervated after 3–4 weeks by sprouting of the residual axons (Maehlen & Njå, 1981; Fonnum et al. 1984). Whilst the relative strength of new inputs has not been reported, the finding that after sprouting, mean synaptic amplitude was particularly large in cells with few inputs would be consistent with the preferential formation of strong synapses. This is the question addressed in the present study.

As well as differences in the number of quanta of acetylcholine that are released by each impulse, the types of Ca²⁺ channel that mediate release differ between strong and weak preganglionic terminals (Ireland et al. 1999). At strong synapses, Ca²⁺ entering through N-type channels is responsible for about one third of the postganglionic response and the remainder is resistant to blockade of P-, Q-, L- and T-type channels with selective pharmacological agents. In contrast, at weak synapses, Ca²⁺ entering through N- and P-type channels contributes approximately equally to the postganglionic response with about 30 % of the postganglionic response resistant to selective blockade. Therefore, it is possible that weak and strong synapses are formed by separate populations of preganglionic axons, with the phenotype of an innervating neurone reflected by the types of presynaptic Ca²⁺ channels that it expresses.

In the present study, the determinants of synaptic strength were investigated in L5 ganglia of the guinea-pig lumbar paravertebral chain by examining the re-innervation of ganglion cells by sprouted preganglionic axons derived only from the most caudal white ramus. In the guinea-pig, this arises from the L4 spinal segment. Preganglionic neurones in L4 are located towards the rostral end of the segment, the number present varying markedly between individuals (McLachlan et al. 1985). This limited set of preganglionic inputs was used in experiments designed to test whether or not the predominant pattern of innervation by single strong inputs was restored after removal of the majority of inputs and whether the types of Ca²⁺ channel expressed by the new presynaptic terminals were the same as those at established synapses.

METHODS

Young guinea-pigs (140–280 g) of either sex were anaesthetized with a mixture of ketamine (80 mg kg⁻¹; i.p.) and xylazine (14 mg kg⁻¹; i.p.). A midline abdominal incision was made and the gut was displaced from the abdominal cavity, irrigated with warm sterile saline and covered. In each animal, about 2 mm of lumbar sympathetic trunk (LST) between the L3 and L4 was removed (Fig. 1A). In separate experiments a similar lesion was made between the L4 and L5 ganglia (see Results). The gut was then replaced and further irrigated with saline. The wound was closed and the animal allowed to recover. Antibiotic treatment took the form of tetracycline powder applied between the abdominal muscle layer and the skin, and a tetracycline injection immediately following the operation. Procedures were approved by the Animal Ethics and Care Committee of the University of New South Wales.

Figure 1. Methods of partial denervation and graded stimulation.

A, diagram of experimental preparation of L5 paravertebral ganglion. The lumbar sympathetic trunk was cut between L3 and L4 ganglia so that the preganglionic axons arising from L3 and above (dashed lines) degenerated leaving intact only the axons arising from L4. B, EPSPs (a) and EPSCs (b) were recorded at −90 mV holding potential and evoked by graded stimulation of the preganglionic trunk above the L5 ganglion with pulses of 1 ms duration. Three weak synaptic inputs were recruited successively followed by a strong synaptic input. No further responses could be detected at higher stimulus voltages. Each trace is the mean of approximately 10 sweeps. Stimulus artifacts have been truncated. Note that the latency of responses does not usually correlate with their recruitment order. C, relationship between stimulus voltage and EPSC amplitude in the experiment shown in B. Stimulus voltage was increased in steps of approximately 0.2 V. Each point represents a single response.

Experiments were performed on three sets of animals: (1) control experiments were performed on L5 ganglia from unoperated animals. (2) Ganglia from animals 2–3 days after sectioning of the LST above L4 were used to define the population of L4 preganglionic inputs as the preganglionic axons arising rostral to L4 would have degenerated by this time (Fonnum et al. 1984). These are referred to as ‘partially denervated’ ganglia. (3) Ganglia from similarly operated animals 3–5 weeks postoperatively were used to examine sprouting and re-innervation by intact L4 axons. These are referred to as ‘re-innervated’ ganglia.

Synaptic responses were evoked via close-fitting suction electrodes applied to the severed LST immediately rostral to the L5 ganglion. In control ganglia, this activated all preganglionic inputs, whereas in those animals in which the LST had been cut, only L4 axons were activated. In addition, in all experiments, the LST immediately caudal to the L5 ganglion was stimulated. This evoked synaptic responses which would be expected to be due to the activation of collateral branches of preganglionic axons which pass through the L5 ganglion en route to more caudal ganglia.

At the time of each experiment, the animal was killed by deep anaesthesia with pentobarbitone sodium (100 mg kg⁻¹; i.p.) and perfused through the descending aorta with oxygenated physiological salt solution of the following composition (mM): Na⁺, 151; K⁺, 4.7; Ca²⁺, 2.0; Mg²⁺, 1.2; Cl⁻, 144.5; H₂PO₄, 1.3; HCO₃⁻, 16.3; and glucose, 7.8; pH 7.2–7.4. The LST was inspected to confirm the absence of regeneration, and the L5 ganglion was dissected free and pinned in a plastic recording chamber (volume 0.8 ml) perfused with physiological salt solution of the above composition, gassed with 95 % O₂-5 % CO₂, and warmed to 35°C at a flow rate of about 5 ml min⁻¹.

Intracellular recordings were made from L5 ganglion cells using techniques described previously (Hirst & McLachlan, 1984). Excitatory synaptic potentials (EPSPs) were recorded in single-electrode current clamp and excitatory synaptic currents (EPSCs) in single-electrode voltage clamp. Measuring synaptic currents in voltage clamp avoids the influence of non-linear summation and voltage dependent currents that are associated with synaptic potentials, and thus reflects the synaptic conductance change and also, more accurately, the quantity of transmitter released. Strong inputs were identified by their ability always to evoke an action potential at resting membrane potential (RMP) and sometimes also when the membrane was hyperpolarized by up to −40 mV, as well as by the distinctive shape of their postsynaptic potential when the action potential was blocked by hyperpolarization of the membrane (Fig. 2C; see Ireland et al. 1999). Only the later part of the decay of strong synaptic potentials could be fitted with an exponential. Therefore, the time course of decay of strong potentials was measured between 5 and 30 % of peak amplitude. Rise times of synaptic potentials (10–90 % of peak amplitude) were also measured. Recordings were made at −90 mV in order to block active responses (action potentials in current clamp and action currents in voltage clamp; Cassell & McLachlan, 1987). The number of preganglionic axons innervating a neurone was estimated by counting the number of discrete steps in the amplitude of the EPSP and EPSC at −90 mV in response to graded stimulation of the preganglionic nerve (Fig. 1B and C).

Figure 2. Summated synaptic responses.

Examples of EPSPs (a) and EPSCs (b) evoked by supramaximal stimulation of all preganglionic inputs synapsing on an L5 ganglion cell in a control ganglion (A), all preganglionic inputs arising from L4 in a partially denervated ganglion (B), and all preganglionic inputs arising from L4 in a ganglion re-innervated after 27 days by sprouting of residual preganglionic axons (C). Dotted lines indicate holding potential of −90 mV and holding currents of −0.4, −0.5, −0.6 nA in A, B and C, respectively. Each record is the mean of 5–10 traces. Time and amplitude scales in A apply throughout. Note the small EPSCs detectable on the peak and falling phase of the strong EPSC in A.

Single inputs and summated responses to supramaximal stimulation were recorded. Strong synaptic inputs that produced EPSCs with amplitudes > 2.0 nA were usually only poorly voltage clamped. This means that the values for summated EPSC amplitude may sometimes have been underestimated. In 12 % of neurones in re-innervated ganglia, active responses in voltage clamp could not be blocked even at −90 mV holding potential and the true amplitude of the EPSC could not be assessed. Due to these omissions, the mean of the summated EPSCs underestimated that of the whole population of preganglionic inputs. Peak synaptic conductance evoked by supramaximal stimuli was calculated as described previously (Ireland et al. 1999). The total charge transfer underlying summated synaptic responses was calculated by integrating the synaptic current. This reflects the total quantity of transmitter released even when the component synaptic responses have different latencies. When only one input is activated, charge transfer is proportional to the peak amplitude of the conductance.

In experiments using Ca²⁺ channel antagonists, only cells with a single strong input were studied (see Ireland et al. 1999). If EPSCs were > 2 nA, voltage-clamp control was improved by reducing the peak amplitude of strong inputs by applying hexamethonium (10⁻⁵-10⁻⁶ m). Stimulation and recording protocols were the same as described previously (Ireland et al. 1999).

ω-Conotoxin GVIA (ω-CTX GVIA, Auspep, Parkville, Australia) and ω-agatoxin IVA (ω-Aga IVA, Peptide Institute, Osaka, Japan) were made up as aliquots which were kept frozen until immediately prior to use. ω-Aga IVA was initially dissolved as a stock solution containing 1 mg ml⁻¹ cytochrome c. The concentration of ω-Aga IVA (120 nM) was chosen so as to block P-type channels completely (Mintz et al. 1992; Randall & Tsien, 1995). As the IC₅₀ for Q-type channels is 90 nM (Randall & Tsien, 1995), some block of Q-type channels might be expected if they were present. Using ω-conotoxin MVIIC, no evidence of Q-type channels was found at these synapses in control ganglia (Ireland et al. 1999). Nifedipine (Sigma) was made up fresh for each experiment as a stock solution (10 mM) in ethanol and care was taken to minimize its exposure to light. Drugs were added by transferring the inlet of the perfusion system into a recirculating system containing 20 ml of a solution containing the stated concentration of drug. Previous experiments had shown that a steady state of drug concentration was achieved in the organ bath within 2–3 min (see Davies et al. 1996)

All values are expressed as means ± standard error of the mean (s.e.m.) except where specified. Differences were tested for statistical significance using Student's t test and χ² tests (Statview, Abacus Software). All reported significant differences have P values < 0.05 unless stated otherwise.

RESULTS

Preganglionic inputs to neurones in L5 ganglia in unoperated (control) animals

Graded stimulation of the LST immediately rostral to the L5 ganglion recruited 1–7 inputs in all neurones in control L5 ganglia (Fig 2A and Fig 3A) (mean, 3.9 ± 0.2 inputs; n = 47). Of these, 2.7 ± 0.2 were weak inputs and 1.2 ± 0.1 were strong inputs (Fig. 4A). This represents a ratio of weak to strong inputs of 2.3 (Fig. 4B). Eighty-two per cent of neurones received at least one strong input and 32 % received more than one strong input. Strong inputs were always recruited at higher voltages than weak inputs. There were no cases in which additional synaptic currents could be detected on the decay phase of the strong EPSC when the stimulating voltage was increased to more than twice threshold for the strong input. EPSCs due to weak inputs were readily detected at longer latencies than EPSCs due to strong inputs despite their lower threshold (Fig. 2A).

Figure 3. Number of preganglionic axons converging on L5 ganglion neurones.

Frequency distributions of numbers of all convergent presynaptic inputs on L5 ganglion cells (a), convergent weak inputs (b) and convergent strong inputs (c). A, control ganglia; B, partially denervated ganglia (i.e. L4 inputs only); and C, re-innervated ganglia (L4 inputs after sprouting).

Figure 4. Proportions of weak synapses relative to strong synapses.

A, mean number of weak (□) and strong (▪) inputs in control, partially denervated (L4 axons only), and re-innervated (sprouted L4 axons) ganglia. * Values for weak synapses in both partially denervated and re-innervated ganglia are significantly different from control. † Values for strong synapses in both partially denervated and re-innervated ganglia are significantly different from control. B, ratios of weak to strong synapses in control, partially denervated ganglia (L4 axons only), and ganglia re-innervated by sprouted L4 axons.

The amplitudes of the summated EPSPs and EPSCs in these cells were 40 ± 3 mV (n = 8) and 2.8 ± 0.4 nA (n = 14), respectively. From the summed currents, the total synaptic conductance was calculated to be 31.4 ± 3.9 nS (n = 14). The total charge transfer was 37.8 ± 4.4 pC.

Synaptic responses due to the activation of single strong inputs arising from innervating segments (L1-L4) were obtained in some neurones by partially severing the LST between the stimulating electrode and the L5 ganglion. This eliminated a proportion of the preganglionic inputs to the ganglion and resulted in some cells which had a strong input and no weak inputs. In these cells, EPSPs and EPSCs had amplitudes of 43 ± 2 mV and 3.0 ± 0.3 nA, respectively (n = 25). The rise time of the EPSP was 2.5 ± 0.1 ms, its time constant of decay was 17.1 ± 1.7 ms and the time constant of decay of the EPSC was 7.8 ± 0.3 ms. The peak synaptic conductance of the strong response was calculated to be 32.2 ± 3.1 nS which is similar to the total synaptic conductance. This indicates that the contribution of weak inputs to the total peak synaptic conductance is minor.

Collateral preganglionic inputs to neurones in L5 ganglia in unoperated animals

Graded stimulation of the nerve trunk immediately caudal to the L5 ganglion activated 0–6 synaptic inputs (mean, 3.0 ± 0.1 inputs; n = 125). Of these, 2.3 ± 0.1 were weak and 0.7 ± 0.1 were strong inputs. This represents a ratio of weak to strong inputs of 3.5. Responses were observed in 93 % of neurones.

In order to confirm that synaptic events evoked by stimulation below the ganglion were due solely to activation of collateral branches rather than of axons originating distal to the L5 ganglion, the nerve trunk immediately rostral to the L5 ganglion was severed 2–3 days before. This period allowed the severed preganglionic axons to degenerate (Fonnum et al. 1984). Only one synaptic response was evoked in 47 cells (3 ganglia) by stimulation below the ganglion (none were evoked by stimulating the stump of the severed trunk above the ganglion). This indicates that synaptic responses evoked by stimulation below the L5 ganglion are primarily due to activation of collateral branches of preganglionic axons projecting to more distal ganglia. This was also confirmed in several neurones by colliding responses arising from stimulation rostral and caudal to the ganglion.

Stimulation of the LST caudal to the L5 ganglion also evoked antidromic action potentials in 18 % of cells (26/145). At hyperpolarized potentials, these responses were distinguished from synaptic potentials by their small amplitude and fast rise time (Cassell & McLachlan, 1986). This indicates that about 20 % of postganglionic neurones in the L5 ganglion project more caudally before they leave the paravertebral chain.

Inputs arising from the L4 spinal segment (partially denervated)

Graded stimulation of the LST immediately rostral to the L5 ganglion, 2–3 days after cutting it above L4, would be expected to excite only axons that originally arose from the L4 segment. All ganglia (n = 7) contained some cells that responded, indicating that the L4 white ramus remained intact. L5 neurones received 0–2 synaptic inputs from L4 (mean, 0.9 ± 0.1; n = 104) (Fig 2B and Fig 3B). Only half (47 %) of the cells were innervated. On average, 0.8 ± 0.1 inputs were weak and 0.1 ± 0.03 were strong (Fig. 4A). This gives a ratio of weak to strong inputs of 8.6 (Fig. 4B). Strong inputs were recruited at higher thresholds than weak inputs. Only 9 % of neurones received a strong input, and none received more than one strong input from the L4 segment. Only 8 % of strong inputs and 29 % of weak inputs to L5 ganglion cells arose from the L4 spinal segment. Thus the majority of strong inputs to L5 neurones arise from above L4.

The amplitudes of the summated EPSPs and EPSCs in these cells were 19 ± 2 mV (n = 58) and 0.8 ± 0.1 nA (n = 54), respectively. From the summed currents, the total synaptic conductance was calculated to be 8.3 ± 1.3 nS (n = 54) (Fig. 5Aa). The total charge transfer was 8.8 ± 1.4 pC. The latencies of component responses were similar to those in control ganglia.

Figure 5. Effects of partial denervation on total synaptic conductance and conductance of individual synapses.

A, frequency distributions of total peak synaptic conductance underlying summated EPSCs in partially denervated L5 ganglia (a), and re-innervated ganglia (b). □, neurones with no strong synapses; ▪, neurones with 1 strong synapse; , neurones with ≥ 2 strong synapses. B, frequency distributions of peak synaptic conductance underlying the first recruited weak synaptic currents in partially denervated L5 ganglia (a) and re-innervated ganglia (b) (χ² test, P = 0.62).

As a measure of the strength of individual weak synapses, the amplitude of the first recruited (lowest threshold) weak input in each neurone was also determined. The amplitudes of the EPSP and EPSC evoked at −90 mV at the first recruited weak synapse were 10 ± 1 mV (n = 44) and 0.3 ± 0.03 nA (n = 39), respectively (Fig. 5B a). The amplitudes of these inputs covered a similar range to that observed previously for individual weak inputs to the L5 ganglion in unoperated animals (personal observations). The EPSP and EPSC decayed with time constants of 20.2 ± 1.8 ms (n = 45) and 8.6 ± 0.6 ms (n = 40), respectively, which were not different from those of weak synapses reported previously (Ireland et al. 1999). The corresponding synaptic conductance was 3.2 ± 0.4 nS.

Only 22 % of L5 neurones had responses to graded stimulation immediately caudal to L5. Responses arose from 0–3 inputs (mean, 0.4 ± 0.1; n = 81). Of these, 0.3 ± 0.1 inputs were weak and 0.04 ± 0.02 were strong. This represents a ratio of weak to strong inputs of 8.0, similar to that from stimulating rostral to the same (partially denervated) ganglion. Stimulating caudal to the L5 ganglion activated 42 % of the inputs that were activated from above the ganglion. This indicates that 42 % of L4 axons that synapse in the L5 ganglion give off collateral branches which descend below the L5 ganglion and that the majority (58 %) of descending preganglionic axons projecting caudal to L5 arise above L4. Stimulating caudal to the L5 ganglion also evoked antidromic responses in 24 % of cells (24/102).

Preganglionic inputs 3–5 weeks following partial denervation (re-innervated ganglia)

Three to five weeks after cutting the LST between L3 and L4 ganglia, stimulation above L5 recruited 0–6 preganglionic inputs (mean, 2.1 ± 0.1; n = 221) (Fig 2C and Fig 3C). Of these, 1.5 ± 0.1 were weak and 0.7 ± 0.04 were strong inputs (Fig. 4A). As in normal ganglia, the threshold for activation of strong inputs was always higher than that of weak inputs. Responses were recorded in 88 % of neurones. Thirty per cent of neurones received one weak input; 51 % of neurones received one strong input and only 6 % received two strong inputs (cf. 32 % in control ganglia) (Fig. 3Cc). When the number of cells that received 0, 1 or 2 strong inputs was tested against the expected number based on a random allocation of new strong inputs, there were significantly more cells with a single strong input than would be expected if they formed at random on any cell (P < 0.01, χ² test). This indicates that strong synapses were preferentially formed on neurones that did not already have a strong input arising from L4. From the mean number of strong synapses made on each cell (0.7) and the mean number that originally arose from L4 (0.1), it is evident that 86 % of strong synapses were new.

Formation of new strong synapses was independent of the number of weak inputs innervating a neurone. The mean number of weak synaptic inputs to neurones that received a strong input (1.4 ± 0.1) was the same as that to neurones that lacked one (1.5 ± 0.1, n = 94; P = 0.75). In addition, strong synapses were formed equally readily on neurones with 0–4 weak inputs.

The numbers of both strong and weak inputs were significantly greater in re-innervated ganglia than in partially denervated ganglia (P < 0.0001). In fact, the total numbers of both weak and strong synapses in re-innervated ganglia reached about 55 % of those in control ganglia. This means that the increase in the number of strong synapses was much greater than that of weak synapses, so that the ratio of weak to strong synapses recovered to 2.2 (cf. 2.3 in control, 8.6 at 2–3 days postoperative) (Fig. 4B).

Summated EPSP and EPSC amplitudes in innervated neurones were 33 ± 2 mV (n = 80) and 2.1 ± 0.2 nA (n = 83), respectively. Thus total synaptic conductance was 24.0 ± 2.0 nS (Fig. 5Ab). The total charge transfer was 27.6 ± 2.4 pC which was three times larger than in partially denervated ganglia (P < 0.0001) and 77 % of that in control ganglia. The latencies of component responses were similar to those in control and in denervated ganglia.

The amplitudes of the EPSPs and EPSCs evoked by the first recruited weak synapse to innervated cells were 12 ± 1 mV (n = 79) and 0.3 ± 0.03 nA (n = 72), respectively, and their synaptic conductance was 3.7 ± 0.3 nS. The EPSPs and EPSCs decayed with time constants of 22.1 ± 0.9 ms (n = 79) and 9.7 ± 0.4 ms (n = 71), respectively. None of these values is significantly different from those determined in partially denervated ganglia (Fig. 5B b). If the amplitude of all weak synapses is similarly unchanged, it can be concluded that the increase in the amplitude of summed synaptic responses results from the appearance of new strong inputs, together with an increase in the number of synapses in the ganglion, and not from a generalized increase in the strength of all synapses.

When synaptic responses were evoked in cells that received only a single strong input in re-innervated ganglia, EPSPs recorded in the absence of hexamethonium had amplitudes of 41 ± 3 mV, fast rise times of 3.0 ± 0.3 ms and brief sharp peaks (n = 11). The time constant of the latter part of the EPSP decay phase was 23.5 ± 2.2 ms. EPSC amplitude was 2.3 ± 0.5 nA (n = 11) and the decay phase of the EPSC had a time constant of 8.3 ± 0.9 ms. Synaptic conductance was 25.0 ± 5.4 nS. These values were not significantly different from those of strong synapses recorded in the absence of hexamethonium in control ganglia.

Graded stimulation immediately caudal to the L5 ganglion activated preganglionic inputs on 57 % of neurones. The mean number of inputs was 1.0 ± 0.1 (n = 115). Of these, 0.7 ± 0.1 were weak and 0.3 ± 0.1 were strong, which represents a ratio of weak to strong inputs of 2.6, similar to that from stimulating rostral to the ganglion. Stimulating caudal to the L5 ganglion activated 48 % of the inputs that were activated from above the ganglion, which indicates that 48 % of L4 axons that synapsed in the re-innervated L5 ganglion had collateral branches. This is a similar proportion to that seen in partially denervated ganglia (42 %) and means that, despite the sprouting in the L5 ganglion, there was no change in the number of L4 preganglionic neurones that sent a collateral branch to more caudal ganglia.

Antidromic action potentials were evoked in 24 % of neurones (47/197) by stimulation of the LST caudal to the L5 ganglion. This was not significantly different from that in control (18 %) or in partially denervated ganglia (24 %) (χ² test).

Presynaptic Ca²⁺ channels at single newly formed strong synapses

To test whether newly formed strong synapses in sympathetic ganglia express the same populations of presynaptic Ca²⁺ channels as control strong synapses, single strong inputs in re-innervated ganglia (85 % of which are newly formed) were tested using selective pharmacological blockade of Ca²⁺ channel subtypes (Ireland et al. 1999).

In re-innervated ganglia, non-specific blockade of all Ca²⁺ entry by the replacement of extracellular Ca²⁺ with equimolar Co²⁺ abolished all synaptic responses within 10 min, confirming that transmitter release from sprouted preganglionic terminals is dependent on influx of extracellular Ca²⁺.

In control ganglia, application of the specific N-type Ca²⁺ channel blocker ω-CTX GVIA reduces the amplitude of both weak and strong EPSCs by about 35 % (Ireland et al. 1999). Application of ω-CTX GVIA (100 nM) significantly reduced peak synaptic conductance at single strong synapses in re-innervated ganglia by 55 ± 6 % (n = 8, Fig. 6A and C). This change was significantly larger (P < 0.01) than the reduction in peak synaptic conductance produced by blockade of N-type channels in control ganglia (34 ± 6 %, n = 6; Ireland et al. 1999).

Figure 6. Blockade of N- and P-type Ca²⁺ channels at new strong synapses.

The effects of 100 nM ω-CTX GVIA (A), and 120 nM ω-Aga IVA (B) at single strong synapses in re-innervated ganglia after sprouting. Both sets of recordings were taken in the presence of hexamethonium (5 μM). a, mean EPSPs in current clamp; b, mean EPSCs from same synapses in voltage clamp; c, residual voltage during voltage clamp. Thin lines indicate control recordings, thick lines indicate recordings taken in the presence of the drugs. In B, the control and ω-Aga IVA traces overlie each other. Dashed lines indicate holding potential (−90 mV) in a and c, and holding current of −0.6 nA in Ab and −0.4 nA in Bb. C, comparison of blocking effects of ω-CTX GVIA, ω-Aga IVA and nifedipine at strong synapses in control ganglia (data from Ireland et al. 1999), and in re-innervated ganglia after sprouting. □, control ganglia; ▪, re-innervated ganglia. Error bars indicate s.e.m.* Significant difference from control, P < 0.01.

In control ganglia, application of the P-type Ca²⁺ channel blocker ω-Aga IVA (40 nM) reduces the amplitude of weak EPSCs by about 40 % but does not affect the amplitude of strong inputs (Ireland et al. 1999). Application of ω-Aga IVA (120 nM) also did not affect peak synaptic conductance at strong synapses in re-innervated ganglia (3 ± 3 %, n = 5, Fig. 6B and C). It is highly unlikely that Q-type Ca²⁺ channels were involved in transmission at strong synapses in re-innervated ganglia, since 120 nM ω-Aga IVA should have blocked at least half of these channels had they been present (EC₅₀ = 90 nM; Randall & Tsien, 1995).

Although L-type Ca²⁺ channels have not been shown to be directly involved in transmitter release at any mature fast junction or synapse so far examined (see Catterall, 1998), they have been reported to be involved in modulation of transmission (e.g. Jensen et al. 1999). They have also been shown to contribute to transmitter release at newly formed neuromuscular junctions in the mouse (Katz et al. 1996) and chick (Gray et al. 1992). However the addition of nifedipine (10 μM) to re-innervated ganglia had no significant effect on transmission at strong synapses (−6 ± 7 %, n = 3, Fig. 6C), as is the case in control ganglia (Ireland et al. 1999).

DISCUSSION

This study has shown that, when L5 paravertebral ganglion cells are re-innervated by sprouting of the preganglionic axons that arise from only the L4 spinal segment, greater numbers of strong inputs were formed than would have occurred by chance. This indicates that one strong preganglionic synapse is preferentially formed on each postganglionic neurone during re-innervation of partially denervated ganglia. The new strong synapses express the same types of presynaptic Ca²⁺ channel as strong synapses in control ganglia.

Normal L5 ganglion cells receive, on average, one strong input and ∼3 weak inputs. Following removal of 80 % of their synapses, the number of both strong and weak synapses in the ganglion increased, but proportionally the increase in the number of strong synapses was about seven times greater than the increase in weak synapses. The preferential formation of strong synapses by L4 preganglionic axons restored the ratio of weak to strong inputs observed in control ganglia. Data from the SCG led to the conclusion that there was a generalized increase in the efficacy of each synapse in the ganglion after sprouting (Maehlen & Njå, 1981). However, the similarity between the strength of the first recruited weak synapse following re-innervation with that in control (Fig. 5) is not consistent with this conclusion. Instead, the new strong synapses largely account for the observed increase in the total charge transfer arising from the L4 segment. The preferential formation of strong synapses also explains the full restoration of the strength of end-organ responses reported after re-innervation of the partially denervated SCG, despite the only partial recovery (to 60 % of control) of the number of inputs to each neurone (Maehlen & Njå, 1981).

It is unlikely that the new strong synapses were formed by an increase in the strength of previously existing weak inputs from L4. This would not be compatible with the observation that neurones which received a strong input after sprouting had, on average, the same number of weak inputs as neurones without a strong input. This observation also indicates that the new strong synapses are formed in addition to, not instead of, weak synapses, implying that strong and weak synapses occupy distinct sites for which they do not compete. Such an idea would be consistent with the concept that each input occupies a different domain of the cell surface, possibly a particular dendrite (Hume & Purves, 1983; Forehand & Purves, 1984; Hirst & McLachlan, 1986; see also Gibbins et al. 1998).

In normal ganglia, it is not clear whether a single preganglionic neurone forms both weak and strong synapses. The fact that weak and strong synapses express distinctly different populations of presynaptic Ca²⁺ channels raises the possibility that a single preganglionic axon only forms one type of synapse. The formation of strong synapses on more than one ganglion cell by a single preganglionic axon has been demonstrated (Purves & Wigston, 1983). Thus there may be ‘strong’ and ‘weak’ preganglionic neurones. In re-innervated ganglia, newly formed strong synapses did not express P-type presynaptic Ca²⁺ channels, in this regard resembling strong synapses but not weak synapses in control ganglia. If it is the case that preganglionic neurones in normal ganglia are phenotypically ‘weak’ or ‘strong’, then this result would suggest that the new strong synapses arose from a sevenfold expansion in the number of synaptic connections made by a small number of surviving ‘strong’ phenotype preganglionic neurones. The alternative is that an individual preganglionic neurone can form both types of synapse, in which case the strength of a synapse and the makeup of its presynaptic Ca²⁺ population might rather be determined by local signals from the postganglionic neurone. If this were the case, any of the surviving preganglionic neurones could potentially have contributed to the increase in the number of strong synapses. It is also possible that all newly sprouted synapses, whether strong or weak, do not express P-type presynaptic Ca²⁺ channels.

The reduction in synaptic responses in re-innervated ganglia by 53 % after blockade of N-type channels with ω-CTX GVIA was significantly greater than occurs in control ganglia where only 30 % of the response is blocked. This may reflect the fact that most of the strong synapses resulted from recent sprouting. At developing CNS synapses, the proportion of N-type channels is initially high after synaptogenesis but declines as the synapse matures (Scholz & Miller, 1995; Verderio et al. 1995; Iwasaki & Takahashi, 1998). This possibility could be tested by examining the sensitivity of strong synapses to ω-CTX GVIA after longer survival times.

After re-innervation by sprouting L4 axons, the L5 ganglion contained 55 % of the number of inputs present in the normal ganglion but they transferred 77 % of the charge normally transferred to each ganglion cell during supramaximal stimulation. Of the L5 ganglion cells, 12 % remained denervated and 41 % did not receive a strong input. In addition, only 6 % of neurones were innervated by two or more strong synapses after sprouting. Further, the number of weak synapses did not exceed four per neurone which is less than in the normal ganglion. The failure of the L4 axons to fully restore the number of synapses present in the ganglion prior to partial denervation is not surprising as the number of L4 preganglionic neurones is only ∼200, which is 14 % of the total number of axons projecting to the L5 ganglion (McLachlan et al. 1985). Nevertheless, this small population of L4 preganglionic neurones sprouted to form twice the number of weak synapses and seven times the number of strong synapses that they originally made, indicating a large capacity for terminal expansion. Strong preganglionic inputs to postganglionic neurones are analogous to the motor endplates on skeletal muscle fibres. When a skeletal muscle is partially denervated, motor axons can increase the number of muscle fibres that they supply by two to fifteen times (Thompson & Jansen, 1977; Brown & Ironton, 1978; Luff et al. 1988). It is clear that preganglionic neurones can expand their peripheral fields to a comparable extent. Further, the presence of multiple strong synapses on a single postganglionic cell is uncommon in re-innervated ganglia. This is consistent with the postganglionic neurone being responsible either for inducing the preferential sprouting of axons that form strong synapses or for determining the phenotype of preganglionic terminals which synapse on it and for limiting the number of strong synapses that are formed on each neurone.

During re-innervation, new connections were formed between preganglionic axons from L4 and cells that did not originally receive any inputs from L4. When the cat SCG was partially denervated by removing all but a functionally restricted subset of preganglionic axons (Murray & Thompson, 1957; Guth & Bernstein, 1961; Liestol et al. 1987), inappropriate connections were formed during sprouting of these preganglionic inputs. When the original axons regenerated into the ganglion, these connections were displaced so that correct functional pathways were re-established (Guth & Bernstein, 1961). In the present study, it is possible that the limitation to re-innervation by L4 axons resulted from a mismatch of functional connections rather than their capacity for divergence. However, as functionally inappropriate connections form easily in the SCG, this is unlikely. Further, unlike the functionally heterogeneous neurone population in the SCG, the great majority of neurones in L5 ganglion project to the hindlimb, contain neuropeptide Y and are probably vasoconstrictors (McLachlan & Llewellyn-Smith, 1986) so that most or all of the new strong connections may have been functionally appropriate.

In conclusion, this study has demonstrated that, following the loss of most synapses and virtually all strong synapses after partial denervation, surviving preganglionic axons preferentially form new strong (suprathreshold) synapses during re-innervation. These are formed independently of weak synapses, probably at spatially separate sites on the neurone. The characteristics of the new strong synapses, in particular the types of presynaptic Ca²⁺ channel that are responsible for transmitter release are the same as those of established strong synapses. Together, these observations suggest that the strong synapses do not arise from upregulation of existing weak synapses, but from either sprouting of axons with a strong phenotype or the induction of new strong synapses by the postganglionic neurone. Since postganglionic neurones are predominantly driven by strong synapses in vivo (see Introduction), the preferential formation of strong, rather than weak, synapses ensures that functional transmission across the ganglion is restored.