Synaptic Efficacy Enhanced by Glial Cells in Vitro

See allHide authors and affiliations

Science  12 Sep 1997:
Vol. 277, Issue 5332, pp. 1684-1687
DOI: 10.1126/science.277.5332.1684


In the developing nervous system, glial cells guide axons to their target areas, but it is unknown whether they help neurons to establish functional synaptic connections. The role of glial cells in synapse formation and function was studied in cultures of purified neurons from the rat central nervous system. In glia-free cultures, retinal ganglion cells formed synapses with normal ultrastructure but displayed little spontaneous synaptic activity and high failure rates in evoked synaptic transmission. In cocultures with neuroglia, the frequency and amplitude of spontaneous postsynaptic currents were potentiated by 70-fold and 5-fold, respectively, and fewer transmission failures occurred. Glial cells increased the action potential–independent quantal release by 12-fold without affecting neuronal survival. Thus, developing neurons in culture form inefficient synapses that require glial signals to become fully functional.

Brain development and function depends on glial cells, as they guide the migration of neuronal somata and axons (1), promote the survival and differentiation of neurons (2), and insulate and nourish neurons (3, 4). It is not known, however, whether glial cells also promote the formation and function of synapses, although glial processes ensheath most synapses in the brain (4,5). The recent development of methods to purify (6) and culture a specific type of neuron from the central nervous system (CNS) (7) has allowed us to investigate whether CNS neurons can form functional synapses in the absence of glial cells.

We cultured purified postnatal rat retinal ganglion cells (RGCs) without glial cells under serum-free conditions (Fig.1A) (8) that supported neuronal survival (57 ± 5% of neurons survived after 20 days; mean ± SEM; n = 3), electrical excitability, and the differentiation of axons and dendrites (7). In order to monitor the formation of functional synapses, we recorded spontaneous postsynaptic currents from RGCs (Fig. 1B) (9). After 5 days in culture, 50% of the RGCs tested (15 of 31 cells) showed low levels of synaptic activity with excitatory postsynaptic currents (EPSCs) occurring at a mean frequency of 3 ± 1 min–1 and with a mean peak amplitude of –11 ± 1 pA (Fig.2A). After 20 days of culture, 63% of the RGCs tested (n = 24) displayed spontaneous synaptic activity. The frequency and amplitude of the EPSCs had increased to mean values of 18 ± 7 min–1 and –16 ± 1 pA (n = 15; Fig. 2A), respectively.

Figure 1

Spontaneous synaptic activity in purified RGCs that were cultured for 5 days in defined, serum-free medium in the absence (left) or presence (right) of collicular glia. (A) Hoffmann-modulation contrast micrographs of RGCs. Scale bar, 50 μm. The density of neurons was similar in both cultures. (B) Whole-cell patch-clamp recordings of spontaneous EPSCs from RGCs at a holding potential of –70 mV.

Figure 2

Glial cells potentiated synaptic activity in cultured RGCs. (A) Frequency and mean peak amplitude are shown for spontaneous EPSCs recorded from RGCs that were cultured for 5 (left) or 20 days (right) without (▪) and with (□) collicular glia. Each square represents the synaptic activity in each neuron, tested with the EPSC frequency plotted against the mean EPSC amplitude. (B) The effects of different glial cell types on spontaneous synaptic activity are shown. Astrocytes and oligodendrocytes, but not collicular microglia, potentiated the frequency and amplitude of EPSCs (left). The effect did not require direct contact to neurons, as feeding layers of mixed collicular glia, astrocytes, or oligodendrocytes potentiated synaptic activity in RGCs (right). Before the recordings, RGCs were cultured for 10 to 15 days with different glial cell types in serum-free medium (15). Averaged EPSC frequencies of all neurons from each culture were plotted against the mean EPSC peak amplitudes. Error bars indicate SEM.

To study the effect of neuroglia on synapse formation, we cultured RGCs with glial cells from their target region, the superior colliculus (10). After 5 days of serum-free culture with glial cells, 90% of the RGCs tested (n = 34) showed spontaneous EPSCs. Coculture with collicular glia increased the mean frequency and the mean amplitude of spontaneous EPSCs to 41 ± 12 min–1 and –29 ± 3 pA (n = 30), respectively (Fig. 2A). After 20 days of coculture with glia, every RGC tested (n = 20) showed spontaneous synaptic activity, and the mean EPSC frequency and amplitude were increased to 1265 ± 212 min–1 and –78 ± 8 pA, respectively (n = 20; Fig. 2A). A nearly identical glia-induced enhancement of synaptic activity was observed when RGCs were cocultured with their natural synaptic targets, neurons purified from the superficial layers of the superior colliculus (11, 12). Thus, collicular glia strongly enhanced the frequency and amplitude of spontaneous EPSCs in cultured RGCs.

Next, we determined whether glial cells could promote synaptic activity without contacting RGCs. Treatment of RGC cultures with glia-conditioned medium for 10 to 15 days (13) increased the frequency and the amplitude of spontaneous EPSCs to 425 ± 84 min–1 and –31 ± 2 pA (n = 43), respectively, indicating that a soluble signal from glial cells increased synaptic activity. The glial effect was not mimicked by glutamine (8), combinations of various peptide trophic factors, or components of the extracellular matrix, because culturing RGCs for 10 to 15 days with these components did not change the level of synaptic activity compared with control cultures (14).

Because our preparation of collicular glia contained a mixture of cell types, we next examined which type of neuroglia increased the synaptic activity by culturing RGCs for 10 to 15 days with purified astrocytes, oligodendrocytes, or microglial cells (15). Both, astrocytes and oligodendrocytes purified from rat optic nerve, increased the frequency and amplitude of spontaneous EPSCs in cultures with (astrocytes, n = 20 neurons; oligodendrocytes, n = 18) or without (astrocytes,n = 20; oligodendrocytes, n = 24, Fig.2B) direct contact. Collicular microglia did not increase synaptic activity (n = 16). Thus, the ability to potentiate synaptic activity was specific to macroglial cells.

Glial cells did not enhance the synaptic activity by improving neuronal survival. The percentage of RGCs surviving when cultured above feeding layers of collicular glia or treated with glia-conditioned medium for 10 to 15 days was 107 ± 5% (n = 6) of the survival rate in glia-free cultures. Furthermore, if glial cells were added to RGCs that were cultured for 10 days without glia, at a time when neuronal survival was stabilized (12), the frequency and amplitude of EPSCs were raised within 3 to 4 days to a similar level (386 ± 91 min–1 and –50 ± 6 pA,n = 24, two cultures) as in neuron-glia cocultures of comparable age. This confirmed that glial cells increased the synaptic activity independently from neuronal survival.

Glial cells may have affected neuronal excitability, because in cocultures spontaneous EPSCs often occurred in bursts (Fig.3A). After 20 days of culture with neuroglia, all neurons tested (n = 20) showed bursts of EPSCs (mean frequency 13 ± 1 min−1). These bursts were due to action potential–evoked transmitter release, because action potential–independent release occurs randomly (16). To test whether neuroglia increased synaptic activity by enhancing neuronal excitability, we blocked action potentials with tetrodotoxin (TTX) and recorded miniature EPSCs (mEPSCs) (17). After 12 to 15 days of culture, neuroglia potentiated the mean frequency of mEPSCs by 12-fold from 8 ± 1 min–1(n = 17) in glia-free cultures to 98 ± 45 min–1 (n = 15) and increased their amplitudes significantly (P < 0.001; Kolmogorov-Smirnow test; Fig. 3B). The increase in the frequency and amplitude of miniature postsynaptic currents indicated that glial cells acted directly on synapses (18).

Figure 3

(A) Bursts of spontaneous synaptic activity occurred frequently in RGCs that were cocultured with glial cells (right) but never in glia-free cultures (left). (B) Glial cells increased the frequency and amplitude of action potential–independent mEPSCs in 12- to 15-day-old cultures (17). Cumulative frequency distribution of mEPSC frequencies (left) in RGC cultures without (thick line) and with (thin line) collicular glia. Amplitude histograms (right) of mEPSCs (filled bars) and recording noise (open bars) in the absence (upper panel) and presence (lower panel) of collicular glia. Cumul. frequency, cumulative frequency; Rel. Freq., relative frequency. (C) Electron micrographs of synapses formed by RGCs that were cultured for 20 days in the absence (left) and presence (right) of glial cells. Scale bars, 0.3 μm.

Glial cells can potentiate the frequency of quantal synaptic release by enhancing the number or the efficacy of synapses. In order to test whether RGCs formed synapses in the absence of glial cells, we studied 16- to 21-day-old RGC cultures with the electron microscope and identified synapses on the basis of standard ultrastructural criteria (19). RGCs formed numerous synaptic contacts when cultured without glial cells and, on average, 2.3 ± 0.3 (n= 3 cultures) more synapses in the presence of glial cells (20). We observed no apparent differences in the synaptic ultrastructure in glia-free cultures and in cocultures (Fig. 3C). In addition, immunofluorescence staining of glia-free and of cocultures demonstrated that RGCs expressed the synaptic vesicle proteins synapsin I, synaptophysin, and synaptotagmin, as judged by punctate-like staining of axons (12). Because RGCs formed synaptic contacts in glia-free cultures, the low level of synaptic activity indicated that synapses are inefficient in the absence of glial cells. The twofold increase in the synapse number cannot account for the 12-fold increase in quantal release, suggesting that glial cells enhanced the synaptic efficacy.

To examine whether glial cells affected synaptic efficacy, we studied evoked synaptic transmission in glia-free and in neuron-glia cocultures (21). In 10- to 15-day-old glia-free cultures, extracellular stimulation at 1 Hz evoked EPSCs in only 22% of the RGCs tested (n = 94). In cocultures with glial cells, every RGC (n = 25) showed EPSCs upon extracellular stimulation. In cocultures, the charge transfer of EPSCs evoked at 1 Hz was, on average, four times larger (mean 189 ± 42 pF; n = 25) than in glia free cultures (44 ± 6 pF; n = 24). Within a stimulus train, some stimuli failed to induce postsynaptic responses (Fig. 4). In glia-free cultures, the percentage of failures at a stimulation frequency of 1 Hz averaged at 22 ± 3% (n = 24 neurons). When RGCs were cocultured with glial cells, however, the failure rate was reduced to 2 ± 1% (n = 25). In both cultures, the failure rate increased with higher stimulation frequencies. In the absence of glial cells, however, this frequency dependence was greater than in the presence of glial cells (Fig. 4A), and at every stimulation frequency tested, the mean failure rates in cocultures were lower than in glia-free cultures (Fig. 4B). If we raised the stimulation frequency from 1 Hz to 5 Hz and 10 Hz, the normalized failure rate (21) increased, on average, by 53 ± 4% and by 83 ± 4%, respectively, in glia-free cultures (n = 24), but only by 14 ± 4% and by 44 ± 7%, respectively, in cocultures (n = 24). The differences in failure rates at higher stimulation frequencies cannot be explained by the twofold increase in the number of release sites in cocultures (22). In principle, stimulation failures could also be due to unreliable induction or conduction of action potentials. This was unlikely, however, as current-clamp recordings from RGCs (21) in glia-free culture (n = 12) showed that antidromic stimulation induced action potentials at every frequency tested. Furthermore, it has been shown previously that failures in synaptic transmission in CNS neurons are caused by the probabilistic nature of transmitter release (23). Thus, the observation of frequent stimulation failures in glia-free cultures indicated that the synapses have low efficacies. The higher reliability of synaptic transmission in cocultures indicated that glial cells enhanced the synaptic efficacy.

Figure 4

Glial cells decreased the failure rate of evoked synaptic transmission. (A) In glia-free cultures, the failure rate increased steeply with the stimulation frequency, whereas in the presence of glial cells, synapses could sustain high transmission frequencies. EPSCs were evoked by extracellular stimulation in RGCs cultured for 15 days in the absence (upper panel) or presence (lower panel) of collicular glia. For each RGC, four stimulus trains were applied at the indicated frequencies. The failure rates are indicated. (B) Glial cells decreased the failure rate at every stimulation frequency tested. The failure rates averaged from all neurons tested were plotted against the stimulation frequency. Error bars indicate SEM. RGCs were cultured for 10 to 15 days without (▪) and with (□) collicular glia. The inset shows an EPSC evoked by extracellular stimulation (left trace) and a stimulation failure (right trace). The membrane potential was held at –70 mV; scale bars indicate 60 pA and 3 ms.

In summary, our findings show that the efficient function of developing CNS synapses in vitro depends critically on glial cells, and they raise the question of whether glial cells also regulate synapse function in vivo: Newly formed synapses that are not yet ensheathed by glial cells may be inefficient. Moreover, the efficacy of adult synapses may also depend on their intimate partnership with glial cells (24).

  • * Present address: Synapse Formation and Function Group, Max-Delbrück-Center for Molecular Medicine, Robert-Rössle-Strasse 10, 13122 Berlin-Buch, Germany.

  • To whom correspondence should be addressed. E-mail: fpfrieg{at}


View Abstract

Navigate This Article