Schwann cells modulate short-term plasticity of cholinergic autaptic synapses

Results

Microcultures of superior cervical ganglion neurons established functional cholinergic autapses

In the absence of glial cells or their secreted factors a single SCG neuron was able to successfully develop axon-like and dendrite-like processes circumscribed to a microdot of collagen, which we termed a single cell microculture (Fig. 1A). Growth of the dendritic tree was obvious during the first week in culture and during the second and third week numerous branches and ramifications became more common, similar to what has already been described (Furshpan et al. 1986b). The appearance of neurons developed together with non-neuronal cells was very similar to single-cell microcultures (Fig. 1B). The absence of supporting cells did not affect the formation of synapses because the presence of potential autapses along the dendritic network was demonstrated by the positive staining for VAMP-2 (Fig. 1C) and synaptophysin (Fig. 1D) in single cell microcultures (see also Fig. 2 for synapse quantification). In the SCG in vivo, preganglionic fibres tend to establish more axo-dedritic rather than axo-somatic synapses (Forehand, 1985). Similarly, single cell microcultures displayed a larger number of axo-dendritic autapses along their dendritic-like tree compared to the soma (Fig. 1C and D).

The works of Furshpan and coworkers demonstrated that SCG neurons are able to change their phenotype from adrenergic to cholinergic in defined culture conditions. Several factors can induce such change, for example, presence of a feeder layer of myocardiocytes, rat serum or growth factors (Furshpan et al. 1986b). In the present experimental conditions a cholinergic phenotype was induced by CNTF (Saadat et al. 1989). The location of putative cholinergic terminals was demonstrated by staining for the vesicular transporter of acetylcholine (Fig. 1E). Similarly to VAMP-2 and synaptophysin staining, the periphery of the collagen microdot showed a more marked presence of cholinergic terminals if compared to the somatic region. However, the contribution of these peripheral autapses was probably not reflected in electrophysiological recordings due to space-clamp limitations. Only those synapses located in the soma or nearby dendritic tree were likely to provide information (Fig. 1F).

The average resting membrane potential of SCG neurons grown in single cell microcultures was −51 mV (n = 7). After 0.6–1 nA current injections, neurons fired action potentials at frequencies ranging from 0.1 to 20 Hz. The nature of synaptic transmission was studied in experiments such as those depicted in Fig. 1G. Functional cholinergic autapses were revealed by twice challenging neurotransmission: first by hexamethonium, a specific blocker of ganglionar nicotinic receptors, and second by removing calcium from the extracellular medium. Both manoeuvres suppressed autaptic EPSPs reversibly, thus demonstrating the presence of nicotinic autapses. Although SCG neurons in microculture may synthesize other neurotransmitters than acetylcholine (Furshpan et al. 1986a), a complete block of neurotransmission by hexamethonium was always achieved (n = 4). Cholinergic synapses were therefore predominant under our experimental conditions.

What was the nature of non-neuronal cells present in the microcultures? In the superior cervical ganglion, Schwann cells are the most abundant non-neuronal cell type and display a slow growth in culture compared to fibroblasts originating from the perineurium (Freschi, 1982). The SCG also contains satellite cells and small intensely fluorescent (SIF) cells (Baluk, 1995). Although all these cell types could be potentially present in a microculture, the morphology under the phase contrast microscope after 10 days in vitro supported a predominant Schwann cell population. This observation was confirmed by the fact that positive staining for the calcium binding protein S-100B was found in 97% of the non-neuronal cells present in the microcultures (1741 cells evaluated). Typically, all non-neuronal cells observed in a microculture expressed S-100B but this marker was absent from neurons (Fig. 2A). Previous immunohistochemical studies revealed the S-100B marker stains Schwann cells and satellite cells of the superior cervical ganglion (Cocchia & Michetti, 1981) and in addition, transgenic mice expressing fluorescent proteins under the control of the human S-100B promoter show a few Schwann cells enveloping a single postsynaptic neuron (McCann & Lichtman, 2008). On these bases the term ‘Schwann cell’ used in the present study refers to S-100B immunoreactive glial cells, thus considering satellite cells as a type of Schwann cell (Mathey & Armati, 2007). Some glial cells tended to cluster around somas (Fig. 2A, arrows), whilst there were others associated with neuritic processes, further suggesting the presence of a heterogeneous Schwann cell population. The contribution of fibroblasts to the non-neuronal cell population was negligible, as revealed by fibronectin staining (Rohrer & Sommer, 1983). Only a few fibroblast clusters, which did not form part of the microcultures, were observed at the edges of coverslips (data not shown). Thus, microcultures containing non-neuronal cells were reproducing in vitro the interactions between Schwann cells and superior cervical ganglion neurons.

In order to evaluate the influence of glial cells on synaptogenesis, microcultures were co-stained with the postsynaptic and presynaptic markers, PSD-93 and VAMP-2, respectively (see Methods for details). As shown in Fig. 2B, axo-somatic synapses were rare after a few days in culture but showed a marked increase when neurons reached 10–15 days in vitro. Development of SCG neurons in culture is associated with an increase of their somatic surface (Johnson et al. 1980; see also Fig. 2A). Immediately after plating, neuron size was ∼10 μm, and after 2 weeks in culture neurons of single cell microcultures increased their surface diameter to 24 ± 2 μm (mean ± s.e.m., n = 16). Neurons developed in the presence of Schwann cells showed a similar growth to a size of 26 ± 2 μm (mean ± s.e.m., n = 16). On these bases, the densitiy of axo-somatic contacts was comparable in single cells and microcultures developed in a glial environment (Fig. 2C). Altogether, these results showed that Schwann cells neither promoted an increase of somatic surface nor enhanced the number of axo-somatic synapses. Axo-dendritic synapses were not evaluated because: (i) the complexity of the dendritic-like tree was very variable among microcultures, (ii) axo-dendritic synapses tended to form clusters, and (iii) most of the axo-dendritic synapses did not contribute to electrophysiological responses due to space clamp limitations.

To evoke synaptic transmission, the soma of a single microcultured SCG neuron was depolarized during 2 ms from a holding voltage of −60 mV to 0 mV. In 2 mm extracellular calcium concentration ([Ca²⁺]_o) autaptic responses ranging from 0.4 to 4 nA were observed immediately after the sodium current peaked (Fig. 2D). The nature of evoked EPSCs was nicotinic because addition of 100 μm hexamethonium at the end of the experiment always abolished them (n = 6, data not shown). In agreement with synaptic immunostainings, autaptic currents were never observed in microcultures developed for less than 10 days in vitro. In those cultures the sodium current was not followed by an autaptic response (Fig. 2D), showing a lack of mature cholinergic synapses.

Spontaneous miniature EPSCs in microcultures established in the presence and absence of Schwann cells

Changing extracellular calcium concentration is a well established method to modify the release probability of synaptic terminals (Dodge & Rahamimoff, 1967). However, this manoeuvre could also have an effect at the postsynaptic level because SCG neurons express nicotinic receptors highly permeable to calcium (Trouslard et al. 1993). To investigate this possibility, the effect of [Ca²⁺]_o on spontaneous miniature EPSCs (mEPSCs) was analysed by exposing neurons to 1 mm, 2 mm and 4 mm [Ca²⁺]_o (Fig. 3).

The size of mEPSCs was variable, showing amplitudes ranging from 20 to 140 pA (Fig. 3A and B). Both, the intrinsic variability of postsynaptic responses and cable filtering properties of dendrite-like processes were likely to contribute to this wide distribution (Bekkers & Stevens, 1996). On average, the amplitude of mEPSCs was little affected by raising [Ca²⁺]_o from 1 mm to 4 mm, but their duration was increased at the higher [Ca²⁺]_o tested. The decay phase of mEPSCs was well fitted by a single exponential whose time constant rose gradually from 1 mm [Ca²⁺]_o to 4 mm [Ca²⁺]_o (Fig. 3C). As a result, mEPSCs obtained in high [Ca²⁺]_o carried more charges into the cell (Fig. 3D). Calculation of mEPSCs integrals was therefore influenced by [Ca²⁺]_o, whilst measurement of their amplitudes was relatively insensitive (Fig. 3E).

To investigate whether the presence of Schwann cells in the microculture could modifiy neurotransmission in autaptic synapses we first analysed mEPSCs. In this type of microcultures, and as described for single-cell ones, the rise in [Ca²⁺]_o increased the decay phase of mEPSCs (Fig. 4A and B). As a result, mEPSCs found at high [Ca²⁺]_o carried more charges into the cell (Fig. 4C), thus reflecting the high permeability of nicotinic receptors to Ca²⁺. In terms of amplitude, no differences existed between mEPSCs obtained in single cell microcultures and in cultures developed in the presence of Schwann cells (Fig. 4D). The analysis of mEPSCs obtained in 2 mm [Ca²⁺]_o showed that their temporal profile was not modified by the presence of glial cells. The rise time from 20% to 80% was 1.5 ms (n = 102) and 1.3 ms (n = 101) for single cell and glial microcultures, respectively. In addition, there was no obvious correlation between mEPSC amplitudes and their decay time constant in both culture conditions (Fig. 4E). Altogether, Fig. 4A–E shows that the individual characteristics of mEPSCs did not change in the presence of Schwann cells, but they appeared at a frequency almost an order of magnitude higher when neurons developed in the glial environment (Fig. 4F; compare to Fig. 3A). Schwann cells increased synaptic activity from ∼0.2 to ∼1.5 mEPSCs s⁻¹. Such enhancement of spontaneous neurotransmission was not associated with a larger number of active synapses because Schwann cells did not promote synaptogenesis of axo-somatic synapses (Fig. 2C). Hence, Schwann cells had a direct effect on the spontaneous activity of nicotinic autapses, but what was their effect on evoked neurotransmission?

Neurotransmission in cholinergic autapses studied at low frequencies of stimulation

Long periods of recording allowed the diffusion of the intracellular solution from the patch pipette to neuritic processes. This phenomenon was seen by the addition of 0.2 mm tetramethyl-rhodamine labelled 3 kDa dextran to the internal solution. Staining of processes was obvious after 28 min of dialysis. Maximum dye accumulation was at dendrites located in the proximity of the soma but labelling of ∼1 μm diameter axon-like processes was also observed (see Supplementary Fig. 1). To test whether the presence of dye along the axon was reflecting a washout of synaptic proteins, rundown for autaptic responses was tested in recordings lasting more than 30 min (n = 6). In this set of neurons series resistance was maintained around initial levels for more than 50 min. Although the amplitude of EPSCs was decreased by 25% at the end of the recording period, little rundown was observed in the charge carried by autaptic responses (Fig. 5). This observation confirmed that Schwann cells were not required for the robustness of the experimental preparation.

Long lasting recordings offered the possibility to investigate the effect of recombinant proteins able to interfere with presynaptic function. When the recombinant light chain of tetanus toxin was diluted to 2 μm in the internal patch solution, it blocked neurotransmission. Success, however, was only achieved in small microdots and it developed after ∼25 min of recording (Fig. 5). The molecular weight of the light chain of tetanus neurotoxin is 50 kDa, more than 10 times larger than the fluorescently labelled dextran used to evidence neuritic processes. This difference probably caused a slower diffusion through long, tortuous axon-like processes. Obtaining a successful delivery of recombinant proteins to presynaptic terminals therefore required the use of small microdots, usually not bigger than ∼100 μm diameter.

Long lasting recordings where responses were evoked at stimulation frequencies ranging from 0.1 to 1 Hz did not show obvious plasticity features. On average, the amplitude of EPSCs showed little variation during the initial 20–25 min of recording. During this period, R_s showed little change, since it increased from an average value of 14 MΩ to 17 MΩ (n = 28). However, only a few neurons were recorded after this time window, because significant increases of the initial series resistance developed.

Synaptic plasticity of single cell microcultures evoked by high frequency stimulation

Experiments in vivo have shown that preganglionic fibres innervating the SCG ganglion fire at a maximal frequency of 10–40 Hz (Birks & Isacoff, 1988; Huang & Cohen, 2000). Trains of stimuli delivered at frequencies up to 20 Hz evoked action potentials without showing failures in SCG single cell microcultures (data not shown), and a frequency of 14 Hz was chosen to test for use-dependent plasticity mechanisms. Depression was consistently evoked when [Ca²⁺]_o was above 1 mm but was absent below this concentration (Fig. 6A). The amplitude of autaptic responses in 0.5 mm and 2 mm [Ca²⁺]_o was 715 ± 34 pA (n = 5) and 1872 ± 119 pA (n = 12), respectively, meaning that the release probability at which the synapse was operating was key to determining the presence of synaptic depression during high frequency stimulation.

In 2 mm [Ca²⁺]_o depression reached a steady state value after four to six depolarizations delivered at 14 Hz (Fig. 6A). By using cumulative plots of EPSC amplitudes it was possible to fit a line to the steady-state phase of synaptic depression and extrapolate it to time 0 (Fig. 6B). Assuming the refilling of the readily releasable pool of vesicles (RRP) took place uniformly throughout stimulation, the intercept on the Y-axis reported an estimate of the RRP size (see Supplementary Fig. 2 for details of the procedure). This type of calculation has been previously performed at the calyx of Held synapse (Schneggenburger et al. 1999), hippocampal synapses in microculture (Otsu et al. 2004) and neuromuscular junctions (Millar et al. 2002). The main constraint on performing this calculation was that it could only be applied if synaptic depression developed. Therefore, it was not a valid approach to establish the size of the RRP in conditions of low release probability, for example at 0.5 mm [Ca²⁺]_o (Fig. 6A and B).

The RRP is heterogeneous among synapses and its definition greatly depends on the method used (Murthy et al. 1997). For example, the measurement of the RRP size by using a train of action potentials in hippocampal synapses reported different estimations from another well-established method, the application of a hypertonic sucrose solution (Rosenmund & Stevens, 1996; Moulder & Mennerick, 2005). We defined the RRP size using plots of cumulative EPSC amplitudes, so that it only took into consideration the contribution of vesicles released synchronously with stimulation. The participation of some vesicles from the RRP in asynchronous release has been recently revealed (Stevens & Williams, 2007). Their contribution can be taken into account by analysing EPSC charges instead of currents. This approach, however, was not performed. Nicotinic receptors displayed a high permeability to calcium, and changes in [Ca²⁺]_o affected both the probability of release and the amount of charge carried by EPSCs (Figs 3 and 4). In contrast, the amplitude of mEPSCs was constant under variable [Ca²⁺]_o. Hence, this parameter provided a measurement of the presynaptic effects of changes in [Ca²⁺]_o, with little postsynaptic contribution.

Typically, in 4 mm [Ca²⁺]_o, the RRP was exhausted at the fourth action potential of the train, while in 1 mm [Ca²⁺]_o, at least six stimuli were required (Fig. 6C and D). Steady state depression was always achieved faster under high release probability conditions. In terms of the fraction of the RRP released by the first action potential of the train, a rise from 1 mm to 4 mm [Ca²⁺]_o at most doubled the amount of vesicles mobilized (Fig. 6E). When [Ca²⁺]_o was 1 mm, the arrival of an action potential released approximately a third of the RRP but in 4 mm [Ca²⁺]_o, almost two-thirds were released. This change in the kinetics of depression, however, did not affect the total number of vesicles mobilized from the RRP. The average of seven different neurons successively trialled with 1 mm, 2 mm and 4 mm [Ca²⁺]_o showed that the estimates for the RRP size were independent of release probability at which synapses were operating (Fig. 6E). Remarkably, the size of the RRP was not modified in the presence of Schwann cells (Fig. 7A).

Determinants of paired pulse plasticity in cholinergic autapses

Paired pulse stimuli are a simple approach to probe synaptic plasticity. Paired pulse facilitation (PPF) is defined by a paired pulse ratio (PPR) larger than 1, whilst paired pulse depression (PPD) is defined by a PPR < 1. The red and blue traces in the example of Fig. 6C show that a change from PPF in 1 mm [Ca²⁺]_o to PPD in 4 mm [Ca²⁺]_o took place if the two initial stimuli of the trains of action potentials were considered. As shown in Fig. 6E, in 1 mm [Ca²⁺]_o the first action potential of a train of stimuli released a variable range of vesicles, from 25% to 55% of the whole RRP, whilst in 4 mm [Ca²⁺]_o this range rose from 40% to 70%. The trains of action potentials were therefore used to define the relationship between the PPR for a 70 ms interval and the fraction of the RRP released by the first action potential of the train (R_f) in single cell microcultures. An experimentally determined power function described the relationship:

In general terms, it could be considered that R_f was inversely related to PPR (Fig. 7B, open symbols). The relationship between PPR and R_f in microcultures developed in the presence of Schwann cells also followed well the above function (Fig. 7B, dots). The analysis of trains of stimuli also allowed the study of whether (i) there was a relation between PPR and the RRP size and (ii) PPR was affected by postsynaptic mechanisms.

A direct relation between PPR and the average RRP size was not found in single cell microcultures, this meaning that the number of vesicles available for release was unrelated to short-term plasticity phenomena (Fig. 7C). Microcultures developed in the presence of Schwann cells also did not show an obvious relationship (Fig. 7D). Saturation of postsynaptic receptors can induce a process of short-term depression (Zucker & Regehr, 2002). Hence, there was the possibility that the size of postsynaptic responses was related to the type of plasticity displayed, i.e. the larger EPSCs were associated with a more marked depression. As depicted in Fig. 7E, there was not an obvious relation between PPR and synaptic potency, measured as the amplitude of the first EPSC of the train. Thus, the observed changes in PPR in single cell microcultures were essentially of presynaptic origin. A similar result was obtained in microcultures developed in the presence of Schwann cells (Fig. 7F).

Effect of Schwann cells on the short-term plasticity displayed by cholinergic autapses

Schwann cells did not modify the RRP size (Fig. 7A), but instead altered PPR in conditions of high release probability. Comparison between Fig. 7C and D illustrates this observation. PPR was decreased in glial microcultures but the RRP size was unaffected. To further investigate this aspect, paired pulse stimuli were applied in successive trials in single cell microcultures (Fig. 8A) and compared to neurons microcultured in the presence of glial cells (Fig. 8B). As expected (Figs 6 and 7), PPD was consistently evoked when [Ca²⁺]_o was above 2 mm and pulse intervals were shorter than 0.5 s (Fig. 8C). However, neuronal development together with Schwann cells significantly affected depression. In this condition, PPD was longer lasting and was still present at a 1 s interval (Fig. 8D). When the [Ca²⁺]_o of the bathing medium was lowered to 0.5 mm, neurons changed PPD to PPF. Facilitation was obvious for pulse intervals up to 100 ms and was comparable between the two types of microcultures (Fig. 8E and F). The time course of PPF was exponential and showed a time constant of ∼0.5 s in both cases (Fig. 8G and H). The temporal profile of PPD was also exponential but about 10 times faster than PPF. This observation, however, was valid only in single cell microcultures because Schwann cells prolonged the process of depression. In this condition the time course of PPD was comparable to PPF (Fig. 8H).

Paired pulse depression was evoked in single cell microcultures when [Ca²⁺]_o was both 2 mm (Fig. 9A) and 4 mm (Fig. 9B). In this type of culture the time course of PPD followed time constants of 41 ms and 83 ms in 2 mm and 4 mm [Ca²⁺]_o, respectively (Fig. 9A and B, right). On average, the presence of Schwann cells significantly slowed down recovery from depression. In this condition, the process became bi-exponential (Fig. 9A and B, grey traces). In 2 mm [Ca²⁺]_o PPD took place with fast and slow time constants of 14 ms and 2 s, respectively, while in 4 mm [Ca²⁺]_o this process was even slower, providing time constants of 0.4 s and 31 s. Therefore, Schwann cells were modulating the degree of short-term depression. In addition to the fact that short-term plasticity phenomena are considered essentially presynaptic processes (Zucker & Regehr, 2002), three pieces of experimental evidence favoured a specific effect of Schwann cells at the presynaptic level: (i) synaptic potency was not significantly affected by the presence of glial cells (Fig. 9A and B, left, see also Fig. 10), (ii) PPR was unrelated to the amplitude of the first EPSC (Fig. 7E and F), and (iii) mEPSCs, which reflect basic properties of postsynaptic nicotinic receptors, did not change their characteristics in the presence of glial cells. As a result, it was concluded that Schwann cells were enhancing the short-term depression displayed by cholinergic autapses by acting at the presynaptic level.

Decrease of [Ca²⁺]_o below 2 mm [Ca²⁺]_o shifted the tendency of paired pulse stimuli to evoke depression. For example, for a 100 ms interval, in 4 mm [Ca²⁺]_o all cells tested evoked PPD, in 2 mm [Ca²⁺]_o the proportion was reduced to 90% of the neurons and in 1 mm [Ca²⁺]_o only 5 out 10 single cell microcultures displayed depression (Figs 9 and 10, left). Nevertheless, the net effect in 1 mm [Ca²⁺]_o was PPF (Fig. 10A), this meaning that the degree of facilitation was larger than depression in this condition. When neurons were bathed in 0.5 mm [Ca²⁺]_o, only facilitation was evoked (Fig. 10B). PPF followed an exponential time course and was longer lasting in 0.5 mm [Ca²⁺]_o (τ = 0. 8 s) than in 1 mm [Ca²⁺]_o (τ = 0.41 s). In contrast to the results found for depression, microcultures developed in the presence of Schwann cells showed a comparable facilitation time course to single cell microcultures (Fig. 10), providing time constants of 0.72 s and 0.28 s in 0.5 mm and 1 mm [Ca²⁺]_o, respectively. The longest time constants were thus associated with the lowest release probability condition tested, where fewer vesicles were mobilized from the RRP (Fig. 6E). Altogether, the present results support a scenario where autapses switch between PPF and PPD within a continuous process, which is inversely proportional to release probability. Schwann cells exert a specific effect on synaptic plasticity, enhancing short-term depression but having no effect on facilitation.