Astrocytes Impose Postburst Depression of Release Probability at Hippocampal Glutamate Synapses

Results

Hippocampal CA3–CA1 glutamatergic synapses, as a population, exhibit facilitation immediately (tens of milliseconds) after short synaptic burst activation, and augmentation is observed during seconds after longer synaptic bursts (Andersen, 1960; Gustafsson et al., 1989). However, at intermediate intervals (0.1–1 s), burst activation results in depression as illustrated in Figure 1. Even a single conditioning burst consisting of three impulses at 50 Hz gives rise to a prominent synaptic depression of 41 ± 3% (compared to the first fEPSP in the conditioning burst, n = 28, p < 0.0001) 0.5 s after the burst (Fig. 1A). This postburst depression (PBD) of the fEPSP was associated with an increase in the paired-pulse ratio of 14 ± 5% (n = 28, p = 0.004), and with no change of the presynaptic volley (3 ± 1%, n = 28, p = 0.98). The PBD was observed to the same extent using whole-cell recordings from CA1 pyramidal neurons, both when reported using AMPA EPSCs at −80 mV and when reported using NMDA EPSCs recorded at +40 mV (Fig. 1B). The NMDA EPSCs were recorded in the presence of 10 μm NBQX, an AMPA/kainate antagonist, implying that AMPA and kainate receptors are not necessary for the PBD. The PBD of the AMPA EPSC was associated with a comparable decrease in 1/CV² (Fig. 1B). Twenty millimolar BAPTA was included in the pipette solution in these whole-cell recording experiments, indicating that a rise in postsynaptic calcium is not necessary for the generation of the PBD. Based on the equal depression of AMPA and NMDA EPSCs, the increased paired-pulse ratio, and the decreased 1/CV² ratio, we conclude that the PBD is expressed as a depression in release probability.

The PBD was most evidently expressed when there was a delay between conditioning burst and the test stimulus of 0.1–1 s (Fig. 1C). Changing the intraburst frequency in the conditioning burst indicated that frequencies higher than 20–30 Hz produced maximal PBD (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Moreover, changing the number of impulses in the conditioning burst showed that, whereas a single stimulus is insufficient, three impulses at 50 Hz produced maximal PBD (Fig. 1D). These findings indicate that the PBD is saturable and that there is a certain threshold for the induction. It is noteworthy that the conditioning–pause–test protocol resulted in larger depression than did continuous high-frequency activation during the same period. Thus, the 20th fEPSP in a 20 impulse, 50 Hz train was only depressed by 16 ± 10% (of the first fEPSP, n = 5). Together, these results would indicate that the PBD is particularly engaged during theta burst activity. Especially during exploratory behavior and during REM sleep, hippocampal pyramidal neurons typically fire action potentials in short high-frequency (40–100 Hz) bursts with interburst intervals corresponding to the theta frequency (4–10 Hz) (Lisman and Buzsáki, 2008). This theta burst activity is thought to be critical for coding information in the hippocampus (Lisman and Buzsáki, 2008), and it is an efficient mean to induce long-term synaptic plasticity (Larson et al., 1986; Huerta and Lisman, 1996).

In contrast to most other forms of short-term plasticity (Zucker and Regehr, 2002), it was necessary to simultaneously activate a certain minimal number of synapses to induce the PBD. As shown in Figure 1E, a stimulation intensity low enough to only activate a few synapses targeting the CA1 pyramidal neuron recorded from (as indicated by frequent synaptic failures) was not sufficient to induce the PBD. Recruiting more synapses, by increasing the stimulation intensity, revealed a PBD, which did not become larger by further synaptic recruitment, showing that the PBD exhibits a distinct cooperativity threshold. This cooperativity threshold for the PBD contrasts with the lack of such requirement for paired-pulse plasticity, which remained the same at the three different stimulation intensities (1.46 ± 0.04, 1.47 ± 0.03, 1.47 ± 0.03, n = 5). The cooperativity threshold for the PBD is, however, likely low, since stimulation intensities evoking fEPSPs up to 10 times smaller than fEPSPs showing signs of population spike activity were still sufficient to elicit full PBD (n = 5). Since the PBD was unaffected by voltage clamping the membrane potential (at −80 mV and at +40 mV) and by calcium buffering in the postsynaptic neuron, it is unlikely that the postsynaptic neurons are responsible for mediating the cooperativity effect. Two other possibilities, which could account for the cooperativity effect, are spillover of neurotransmitter (Arnth-Jensen et al., 2002) and activity-induced release of gliotransmitters from astrocytes.

Since astrocyte function and morphology develop during an extended postnatal period (Konietzko and Muller, 1994; Bushong et al., 2004; Schools et al., 2006; Andersson et al., 2007), activity-induced release of gliotransmitters from astrocytes would be expected to increase with development. Spillover, on the other hand, would rather be expected to be favored early in development when astrocyte processes around synapses and glutamate uptake are poorly developed (Piet et al., 2004; Diamond, 2005). The developmental profile of the PBD (Fig. 2A), emerging at postnatal day 10 and being fully developed at postnatal day 20, would thus be more consistent with activity-induced release of gliotransmitters from astrocytes than with spillover. To more directly test whether the PBD required some form of astrocyte signaling, we elicited the PBD before and after the application of the glial metabolic inhibitor FAc (Fonnum et al., 1997; Andersson et al., 2007). Consistent with a critical role for astrocyte signaling, FAc completely blocked the PBD (Fig. 2B). We have previously shown that endothelin-1 closes astrocyte gap junctions and inhibits an astrocyte-mediated transient heterosynaptic depression (Blomstrand et al., 2004; Andersson et al., 2007). However, endothelin-1 did not affect the magnitude of the PBD (Fig. 2B), indicating that signaling through astrocyte gap junction is not required for this homosynaptic form of astrocyte-mediated synaptic depression.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Developmental profile and astrocyte involvement in the PBD. A, Developmental profile of the PBD. Results from rats of different postnatal days were binned into 11 groups (n = 5–15 per group), and a sigmoid curve was fit to the data points. B, The PBD is blocked by the glial metabolic inhibitor FAc (10 mm FAc) and by BAPTA (50 mm) infused into the gap junction-coupled astrocyte network, but not by the gap junction blocker endothelin-1 (Et-1, 1 μm). The control PBD before FAc was 42 ± 5% (n = 11) and was reduced to −9 ± 6% (**p < 0.005) after FAc. The control PBD before BAPTA (43 ± 3%, n = 5) was recorded from a region in the slice without BAPTA-filled astrocytes and was compared with the PBD recorded from a region in the same slice with BAPTA-filled astrocytes (−4 ± 3%, n = 5, ***p < 0.001), as schematically indicated on top (scale bar indicates 50 μm). S, Stimulating electrode; E1, E2, extracellular recording electrodes.

To test whether a rise in astrocyte calcium is a necessary step in the generation of PBD, we patched gap junction-coupled astrocytes in the stratum radiatum with a pipette solution containing the calcium chelator BAPTA and Lucifer yellow. To allow diffusion of BAPTA to a large number of gap junction-coupled astrocytes in 20 min (Serrano et al., 2006), we used a high concentration of BAPTA (50 mm). The diffusion of Lucifer yellow was used to estimate the diffusion of the calcium chelator in the astrocytic network as Lucifer yellow is of comparable molecular size to BAPTA (Lucifer yellow MW: 457 to BAPTA MW: 476). Both of these are well below the 1 kDa size limit for gap junction pores (Giaume et al., 2010). When Lucifer yellow and BAPTA had diffused to gap junction-coupled astrocytes for 20 min, the patch pipette was withdrawn and an extracellular recording pipette was placed in close vicinity (within 20 μm) to the originally patched astrocyte. Under these conditions, the PBD could not be elicited (Fig. 2B), although normal PBD was elicited when recorded from another region without BAPTA/Lucifer yellow-filled astrocytes in the CA1 str. radiatum of the same slice. A lower concentration of BAPTA (10 mm) allowed to diffuse for 20 min did not fully block, but substantially inhibited the PBD to 27 ± 9% of control (n = 3, p = 0.008). Notably, paired-pulse plasticity was unaffected by the presence of a high concentration of BAPTA in the astrocytes (102 ± 5%, n = 5, p = 0.6). Experiments using the inhibitor of the sarcoplasmic reticulum calcium pump, cyclopiazonic acid (10 μm), indicated that release of calcium from internal stores is involved in the generation of the PBD. In the presence of cyclopiazonic acid, the PBD was decreased to 63 ± 12% of control (n = 5, p = 0.036).

The blockade of PBD by chelating astrocyte calcium was also observed in another set of experiments in which we first recorded a control PBD and then patched an astrocyte in the vicinity of the extracellular recording pipette with a solution containing 50 mm BAPTA and Lucifer yellow. Twenty minutes after having patched the astrocyte, we again applied the PBD protocol. As illustrated in Figure 3A, the PBD was completely blocked by filling the astrocyte network with 50 mm BAPTA. As a control, we repeated the experiment without patching an astrocyte, simply positioning the patch pipette into the slice for 20 min. In these control experiments, the PBD was unaffected (Fig. 3B). Furthermore, in three experiments we patched uncoupled astrocytes with nonpassive electrophysiological properties. Patching such astrocytes did not affect the PBD (Fig. 3C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Astrocyte calcium signaling is critically involved in the generation of the PBD. The effect on the PBD is evaluated as relative change of control PBD in three different situations: “Network,” “Slice,” and “Uncoupled cell.” A, “Network”: BAPTA (50 mm) and Lucifer yellow were infused for 20 min via the patch pipette into a gap junction-coupled astrocyte exhibiting passive electrophysiological properties (scale bar indicates 20 μm). Control PBD before BAPTA infusion (45 ± 5%, n = 5) was significantly different from the PBD 20 min after BAPTA infusion (**p < 0.005). B, “Slice”: a BAPTA-containing patch pipette was simply placed into the slice for 20 min. Control PBD before placement of BAPTA-containing pipette in the slice (41 ± 7%, n = 5) was not significantly different from the PBD 20 min after (p > 0.05). C, “Uncoupled cell”: BAPTA infused for 20 min via the patch pipette into an uncoupled cell with nonpassive electrophysiological properties. Control PBD before patching (47 ± 4%, n = 3) was not significantly different from the PBD 20 min after patching (p > 0.05). Electrophysiological properties were tested using voltage steps of 10 mV from −100 mV to +50 mV (calibration: 50 ms, 100 pA). Top parts of the panels show schematic drawings depicting placement of extracellular recording electrode (E), intracellular recording electrode (I), and stimulation electrode (S) in the three different situations.

To examine whether continuous train stimulation, as opposed to burst stimulation, elicits this astrocyte-mediated depression, we examined the responses to trains of 20 impulses at 0.2 Hz, 2 Hz, and 50 Hz, in the region of the slice in which 50 mm BAPTA was infused into a gap junction-coupled astrocyte, as well as in a control region of the same slice (compare Fig. 2C). Despite inhibiting the PBD, these trains were largely unaffected by strongly buffering intracellular astrocyte calcium (supplemental Fig. 2, available at www.jneurosci.org as supplemental material).