G-protein inhibition of voltage-gated calcium channels can be transiently relieved by repetitive physiological stimuli. Here, we provide evidence that such relief of inhibition contributes to short-term synaptic plasticity in microisland-cultured hippocampal neurons. With G-protein inhibition induced by the GABABreceptor agonist baclofen or the adenosine A1 receptor agonist 2-chloroadenosine, short-term synaptic facilitation emerged during action potential trains. The facilitation decayed with a time constant of ∼100 msec. However, addition of the calcium channel inhibitor Cd2+ at 2–3 μm had no such effect and did not alter baseline synaptic depression. As expected of facilitation from relief of channel inhibition, analysis of miniature EPSCs implicated presynaptic modulation, and elevating presynaptic Ca2+ entry blunted the facilitation. Most telling was the near occlusion of synaptic facilitation after selective blockade of P/Q- but not N-type calcium channels. This was as predicted from experiments using recombinant calcium channels expressed in human embryonic kidney (HEK) 293 cells; we found significantly stronger relief of G-protein inhibition in recombinant P/Q- versus N-type channels during action potential trains. G-protein inhibition in HEK 293 cells was induced via recombinant M2 muscarinic acetylcholine receptors activated by carbachol, an acetylcholine analog. Thus, relief of G-protein inhibition appears to produce a novel form of short-term synaptic facilitation in cultured neurons. Similar short-term synaptic plasticity may be present at a wide variety of synapses, as it could occur during autoreceptor inhibition by glutamate or GABA, heterosynaptic inhibition by GABA, tonic adenosine inhibition, and in many other instances.
- short-term synaptic plasticity
- G-protein inhibition
- calcium channels
- recombinant calcium channels
- cultured neurons
- hippocampal neurons
- HEK 293 cells
Short-term synaptic plasticity may dramatically affect neuronal information transfer (Magleby, 1987;Zucker, 1989; Markram and Tsodyks, 1996; Sejnowski, 1996; Abbott et al., 1997; Tsodyks and Markram, 1997; Dobrunz and Stevens, 1999). Because neurotransmitter release depends supralinearly on presynaptic calcium entry through voltage-gated calcium channels, spike-to-spike changes in calcium currents could contribute substantially to synaptic plasticity. There has to date, however, been little evidence of this class of mechanisms (Borst and Sakmann, 1998; Cuttle et al., 1998;Forsythe et al., 1998). In particular, presynaptic calcium currents in well studied invertebrate neurons appear invariant from stimulus to stimulus (Smith and Zucker, 1980; Charlton et al., 1982; Wright et al., 1996). However, studies of vertebrate preparations provide growing evidence that during repetitive physiological stimuli, calcium channels may manifest progressive increases in current because of relief of G-protein-mediated inhibition. Specifically, in somatic and recombinant channels G-protein-mediated inhibition can be transiently relieved by trains of voltage-clamp action potentials and strong depolarizations (Bean, 1989; Elmslie et al., 1990; Brody et al., 1997; Williams et al., 1997; Park and Dunlap, 1998; Tosetti et al., 1999), although there can be a tonic component of inhibition that is insensitive to depolarization in some preparations (Shapiro and Hille, 1993; Luebke and Dunlap, 1994). Such activity-dependent reversal of calcium channel inhibition could occur in presynaptic calcium channels as well, since at least the initial inhibition of neuronal calcium channels by G-proteins is known to underlie the suppression of neurotransmitter release produced when a wide variety of G-protein-coupled receptors are activated (Toth et al., 1993; Wu and Saggau, 1994, 1995; Dittman and Regehr, 1996; Takahashi et al., 1996). Although several groups have noted that this relief of inhibition potentially could cause a form of short-term synaptic facilitation (Elmslie et al., 1990; Shen and Horn, 1996; Brody et al., 1997), no explicit investigations of this proposal have been performed in synaptic preparations. Here, we tested for such a facilitation in single, cultured rat hippocampal neurons. When grown on glial microislands, the neurons form extensive synaptic connections onto themselves (“autapses”) (Van der Loos and Glaser, 1972), which have properties very similar to those of synapses between neurons (Bekkers and Stevens, 1991; Johnson and Yee, 1995; Goda and Stevens, 1996). Such autapses show large postsynaptic responses and thereby provide a convenient system for studying short-term plasticity, especially under conditions in which neurotransmission is reduced such as during G-protein-mediated inhibition and presynaptic calcium channel blockade.
It has been widely observed that short-term synaptic depression (STD) is attenuated by an overall reduction of neurotransmitter release (Lev-Tov and Pinco, 1992; Barnes-Davies and Forsythe, 1995; Isaacson and Hille, 1997; Varela et al., 1997; Brenowitz et al., 1998) such as during G-protein-mediated inhibition. These effects on synaptic efficacy have been attributed to release-dependent mechanisms such as vesicle depletion (Zucker, 1989). By contrast, we have found that G-protein-mediated inhibition reduces STD or converts it to relative facilitation, in a manner that appears distinct from release-dependent mechanisms but consistent with relief of G-protein-mediated inhibition of calcium channels.
MATERIALS AND METHODS
Cell culture. Hippocampal neurons were cultured on glial microislands essentially as reported (Furshpan et al., 1986;Bekkers and Stevens, 1991). Briefly, a 0.15% agarose solution was spread uniformly on glass coverslips; then 2 mg/ml poly-d-lysine plus 3 mg/ml collagen (Cellprime; Collagen Corporation, Palo Alto, CA) in 8.5 mm acetic acid was airbrush (Aztek, Rockford, IL) spattered to form 50- to 750-μm-diameter “islands” of adhesive substrate. Cultured rat astrocytes were plated at a density of 6,000–24,000 cells/ml in Minimal Essential Medium (MEM) with Earle's salts (Life Technologies, Gaithersburg, MD), 10% fetal calf serum, 20 mm glucose, 0.5% N2 supplement (Life Technologies), 0.5% penicillin/streptomycin stock, and phenol red. After 4–6 d at 37°C in a 5% CO2 atmosphere, the astrocytes spread out over the microislands but did not grow on the agarose. Neurons were isolated by trituration of papain-digested, CA1- and CA3-enriched hippocampi from 1- to 2-d-old Sprague Dawley rats. Between 2,000 and 28,000 cells/ml were added to the plates containing astrocytes, and they were cultured in MEM with Earle's salts plus 25 mm HEPES (Life Technologies), 10% horse serum (Life Technologies), 20 mmglucose, 1% N2 supplement, 1 mm sodium pyruvate, 0.5% penicillin/streptomycin stock, and 0.875 mg/ml biotin. The day after plating, 35 μm 5-fluoro-2-deoxyuridine (Sigma, St. Louis, MO) with 75 μm uridine was added. The culture medium was not changed for up to 21 d in vitro.
Electrophysiology. Recordings were made after 7–21 din vitro. Whole-cell patch-clamp pipets with a resistance of 3–4 MΩ were pulled from borosilicate glass and filled with (in mm): 137 Kgluconate, 12 NaCl, 10 HEPES, 4 EGTA, 0.5 CaCl2, 4 MgATP, and 0.3 LiGTP, pH 7.2 with KOH. The standard extracellular solution contained (in mm): 145 NaCl, 5.4 KCl, 2 CaCl2, 1 MgCl2, and 10 HEPES, pH 7.4 with NaOH, adjusted to 310 mOsm with 15–25 mm glucose. For early experiments, the intracellular solution contained 145 Kgluconate and 4 NaCl but was otherwise unchanged. No differences were apparent because of these changes, and results have been pooled. All chemicals were from Sigma except for baclofen, 2-chloroadenosine (2-CADO), and 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzo(f)quinoxaline-7-sulfonamide (NBQX) from Research Biochemicals (Natick, MA), ωCtx-GVIA from Alomone Labs (Jerusalem, Israel), and ωAga-IVA (a gift from Dr. K. Elmslie, Tulane University). A −14 mV junction potential (pipet minus bath) was corrected before sealing. All recordings were made at room temperature (23–25°C). Filtering was at 5 kHz, sampling was at 25 kHz, and the holding potential was −80 mV. Series resistance was typically 6–10 MΩ and compensated 50–85% when possible. EPSCs stabilized in 3–5 min after break-in, and data traces were acquired every 15–30 sec. EPSC integrals followed the same trends as peak EPSCs but showed less variability from sweep to sweep and were more sensitive to the contribution of slightly asynchronous release events (data not shown). EPSC records have been displayed after subtraction of NBQX-insensitive currents (Fig. 1 A), and 1–3 msec around each stimulus has been blanked for clarity. NBQX-insensitive currents were unchanged in baclofen, 2-CADO, or 2–3 μm Cd2+ (data not shown). Rundown or decreases in EPSC amplitude over tens of minutes occurred in some cells but did not significantly affect short-term synaptic plasticity (data not shown). When EPSC amplitudes in two conditions were compared, measurements were bracketed or made within 5–10 min of each other. Minis were recorded without series resistance compensation and were bandpass filtered between 10 and 1000 Hz; 500–800 sweeps of minis were obtained per cell. Mini amplitudes were measured in a semiautomatic manner using custom software in Matlab and confirmed by visual inspection. Cd2+ (2 μm) had no effect on minis; their amplitudes were clearly altered by changing the baseline potential, and NBQX eliminated them entirely (data not shown). In 4 mm Ca2+ plus 100 μm 4-aminopyridine (4-AP) experiments, extracellular Mg2+ was removed, extracellular NBQX was used at 5 μm, and intracellular Na+ was 4 mm. For toxin experiments, 1 mg/ml bovine serum albumin and 0.1 mg/ml cytochrome C were added to the bath. ωCtx-GVIA rapidly and irreversibly reduced initial EPSCs, but ωAga-IVA's effects reversed partially over tens of minutes, so all measurements were obtained within a few minutes of toxin application.
Transfection of human embryonic kidney 293 cells. Human embryonic kidney (HEK) 293 cells were transiently transfected using calcium phosphate precipitation with the following calcium channel clone cDNAs: rat brain α1A (Starr et al., 1991) or human brain α1B (Williams et al., 1992) plus α2δ (Tomlinson et al., 1993) and one of three β subunits: β2a (Perez-Reyes et al., 1992), β3 (Castellano et al., 1993a), or β4 (Castellano et al., 1993b). M2 receptor cDNA (Peralta et al., 1987) was included as well. Methods were otherwise as described (Brody et al., 1997).
Statistical analysis. All p values for pairwise comparisons represent two-tailed, paired t test results or two-tailed, unequal variance t test results as appropriate;p > 0.05 was considered not significant. Error values (±) were SEs. Approximate confidence intervals for fit parameters were calculated using the “locally linear hypothesis” (Pandit and Wu, 1990). t tests between pairs of fit parameters were performed using approximate SDs calculated in the same manner.
Changes in short-term synaptic plasticity after activation of G-protein-coupled receptors
To record postsynaptic currents, we delivered brief voltage-clamp stimuli via a somatic, whole-cell patch pipet (Fig.1 A, top). A propagating action potential was generated, with momentary loss of voltage-clamp control because of large sodium and potassium currents corresponding to initial inward and outward currents, respectively. Subsequently, inward current flowed through AMPA-type glutamate receptors (Fig.1 A, a) that were recorded with good voltage-clamp control. Subtraction of responses obtained with the AMPA-antagonist NBQX (Fig. 1 A, n) isolated synaptic currents (Fig. 1 A, a–n), which were integrated to quantitate EPSCs (Fig. 1 A,Q EPSC). Evidence of good voltage-clamp control during EPSCs included an extrapolated EPSC reversal potential near 0 mV and unchanged short-term synaptic depression when the sizes of the EPSCs were reduced using subsaturating concentrations of NBQX (data not shown) (Bekkers and Stevens, 1991).
We tested for changes in short-term synaptic plasticity attributable to relief of G-protein-mediated inhibition by comparing synaptic responses in the absence and presence of G-protein-coupled receptor agonists. In control conditions, autapses showed STD in response to 50 Hz trains of stimuli, as illustrated by the decrement ofQ EPSC between the first and second responses (Fig. 1 B) and exemplar records (Fig.1 C, left). Population data uniformly confirmed such baseline depression (Fig. 1 D, solid squares). By contrast, the pattern of short-term synaptic plasticity was qualitatively different after application of the GABAB agonist baclofen (10 μm). Initial EPSCs were inhibited by 75 ± 4%, and the depression converted to relative facilitation (Fig.1 C, middle) or appeared markedly reduced (Fig.1 D, open circles), as would be predicted for relief of G-protein-mediated inhibition of presynaptic calcium channels. These effects were not unique to inhibition via GABAB receptors, because nearly identical inhibition and changes in short-term plasticity were seen with 1 μm 2-CADO, which also activates G-proteins via adenosine A1 receptors (data not shown).
Alternative hypotheses to relief of G-protein-mediated inhibition of calcium channels
In addition to the possibility of dynamic changes in Ca2+ influx through calcium channels, a number of more traditional explanations could account for the effects of baclofen on synaptic plasticity. The most prominent among these involves release-dependent short-term depression mechanisms such as vesicle depletion, postsynaptic receptor desensitization, and autoreceptor inhibition by glutamate (Zucker, 1989), which may underlie short-term depression in many synapses. For example, if vesicle depletion were occurring, the baclofen-induced reduction in overall presynaptic Ca2+ influx through calcium channels would decrease fractional depletion of releasable vesicles per action potential, thereby diminishing short-term synaptic depression. To exclude this possibility, we titrated the calcium channel blocker Cd2+ into the bath to mimic the initial reduction of release probability produced by baclofen (Fig. 1 B). Like G-protein-mediated inhibition, Cd2+ in effect reduces channel open probability. Unlike G-protein-mediated inhibition, however, Cd2+ blockade appears to be voltage independent over a physiological voltage range (Leonard et al., 1987) and remains invariant with repetitive stimulation of calcium channels (see Discussion). Despite a substantial reduction in transmitter release by Cd2+, the short-term synaptic depression was unaffected in almost all cells (Fig. 1 C, right, D, gray triangles). In rare recordings (3 of 29), there was somewhat less depression in Cd2+. However, baclofen or 2-CADO always produced even greater changes in short-term plasticity than did Cd2+ (data not shown). These results excluded a contribution of release-dependent mechanisms to the changes in synaptic plasticity produced by baclofen. Instead, our unpublished results suggest that baseline depression in this preparation may reflect action potential propagation failure at axonal branch points (Parnas, 1972; Hatt and Smith, 1976; Macagno et al., 1987; Streit et al., 1992; Wall, 1995; Debanne et al., 1997).
A remaining conventional explanation for the baclofen-induced changes in plasticity was that facilitation caused by residual calcium bound to presynaptic effectors was potentiated with lowered calcium entry (Rahamimoff, 1968; Stanley, 1986). This possibility seemed unlikely because our pipet solution contained a high concentration of EGTA, which would chelate residual Ca2+ (Hochner et al., 1991; Atluri and Regehr, 1996). It appeared that this EGTA effectively diffused to the presynaptic terminals, because after whole-cell access was obtained, the rate of spontaneous miniature EPSCs was initially high but declined over 3–5 min. Furthermore, bath-applied EGTA AM had no effect on short-term synaptic plasticity (data not shown). Most importantly, facilitation from residual calcium was ruled out by the invariance of short-term synaptic plasticity with the addition of Cd2+ (Fig.1 D).
Having excluded major known forms of synaptic plasticity that could have produced our results, we next considered novel mechanisms other than relief of G-protein-mediated inhibition. One possibility was that baclofen and 2-CADO attenuated the release-independent short-term depression mechanism present at baseline. To explore this hypothesis, we examined short-term plasticity as the duration (d) between pairs of stimuli was varied (Fig.2). Recovery from depression in control proceeded in two rising exponential phases with time constants of ∼6 and ∼5000 msec (Brody and Yue, unpublished observations). Mennerick and Zorumski (1995) have observed a similar time course previously. Most importantly, in control, there was little change in short-term plasticity for interstimulus intervals between 20 and 420 msec (Fig. 2 B, solid squares). In baclofen, however, the second EPSC was larger than the first EPSC at short intervals, but smaller at longer intervals (Fig. 2 B, dashed line, open circles). Thus there appeared to be a distinct component of facilitation in baclofen that decayed rapidly (τ = 96 msec), leaving the subsequent recovery from depression indistinguishable from that of the control over the remaining period in which comparisons were made (d = 170–420 msec). These results would be difficult to reconcile with simple attenuation of depression. They fit well, however, with reestablishment of G-protein-mediated inhibition after its relief by the first stimulus (Lopez and Brown, 1991; Zamponi and Snutch, 1998), all superimposed on unchanged baseline depression.
To exclude novel postsynaptic mechanisms underlying the baclofen-induced facilitation, we analyzed miniature EPSCs (minis), which arise from the asynchronous release of single synaptic vesicles. Changes in postsynaptic receptor sensitivity and dendritic attenuation that might produce synaptic plasticity would be apparent as changes in mini amplitudes (Wyllie et al., 1994; Isaac et al., 1996; Carroll et al., 1998). During baseline periods (Fig.3 A), baclofen did not change either the mean or distribution of mini peak amplitudes (Table1; Fig. 3 B; Kolmogorov–Smirnov test, p > 0.2). A particular advantage of single-neuron cultures is that all minis arise from the same presynaptic cell as evoked EPSCs. In other preparations, minis may reflect release events at numerous synapses outside of the subset that contributes to evoked EPSCs. Therefore, if there were a change in the amplitude of minis originating from this subset, the change might be poorly resolved in the amplitude histogram for minis arising from all synapses. Instead, in single-neuron cultures, the lack of change in mini amplitudes is sufficient to exclude novel postsynaptic mechanisms as the basis for the baclofen effect on evoked EPSCs.
Furthermore, baclofen did not alter average mini amplitudes during a time window stretching from 20 to 100 msec after an evoked EPSC (Table1; Fig. 3 C), despite the persistence of baclofen-induced facilitation throughout this period (Fig. 2 B). Amplitudes did increase with time after the evoked EPSCs (∼20% in 80 msec), but the increases were indistinguishable between control and baclofen (Fig. 3 C). This increase could in no way account for the facilitation that occurred only in baclofen and decayed over the same time interval (Fig. 2 B). Thus, these results ruled out a major postsynaptic component as the basis of the facilitation in baclofen.
A final class of alternative mechanisms we examined concerned the possibility that baclofen unmasked activity-dependent but Ca2+-independent changes in maximal release probability, downstream of calcium entry. If such a mechanism accounted for the facilitation in baclofen, then this facilitation would be insensitive to raising overall presynaptic calcium entry. Alternatively, if relief of calcium channel inhibition explains the facilitation, it should be blunted with saturation of the Dodge–Rahamimoff relation between calcium influx and vesicle release (Dodge and Rahamimoff, 1967; Reid et al., 1998). In this case, proportionate spike-to-spike increases in presynaptic calcium currents would enhance neurotransmitter release to a lesser extent when absolute calcium current magnitudes are larger. To test this, we increased presynaptic calcium entry mainly (see Materials and Methods) by raising extracellular Ca2+ from 2 to 4 mm and adding 100 μm 4-AP to broaden the presynaptic action potentials (Fig.4 A,4Ca/4-AP) (Llinas et al., 1976; Wheeler et al., 1996;Hjelmstad et al., 1997). In 4Ca/4-AP, initial EPSCs were on average 3.47-fold larger than in control (Fig. 4 B), and baclofen, 2-CADO, and Cd2+ all inhibited the EPSCs less effectively (compare Fig. 4 Bwith Fig. 1 D, left). Most importantly, in4Ca/4-AP, baclofen or 2-CADO caused only small changes in short-term plasticity (Fig. 4 C), thereby excluding progressive changes in maximal release probability as the mechanism of facilitation. These results support other mechanisms, including relief of G-protein-mediated inhibition, presynaptic action potential changes, and spike-to-spike increases in the calcium affinity of the release apparatus.
Blockade of N- versus P/Q-type channels results in predicted differential effects on short-term synaptic plasticity
To generate a highly discriminating test of whether relief of G-protein-mediated inhibition underlies the facilitation, we hoped to exploit a clear intrinsic difference between the two types of presynaptic calcium channels that trigger transmitter release (Takahashi and Momiyama, 1993; Wheeler et al., 1994; Reuter, 1995; Reid et al., 1997). One such distinction was apparent between recombinant P/Q-type (α1A) and N-type (α1B) channels expressed in HEK 293 cells along with M2 muscarinic acetylcholine receptors (Fig.5). M2 receptor activation by carbachol, an acetylcholine analog, inhibited P/Q- and N-type channels to comparable extents (27 ± 5 and 36 ± 8%, respectively). However, 50 Hz trains of voltage-clamp action potential waveforms (APWs) relieved more of the inhibition of P/Q-type (Fig.5 A,C, left, D) than of N-type (Fig.5 B,C, right, D) channels. Voltage pulses to +100 mV relieved inhibition nearly completely in both channel types (92% for N-type; 95% for P/Q-type). Thus the differing behavior of channel types during trains reflected distinctions in the kinetics of relief and not in the voltage-dependent nature of the inhibition. Furthermore, these distinctions between P/Q- and N-type channels were not unique to the particular combination of auxiliary subunits used in these experiments. The stronger relief of G-protein-mediated inhibition shown by P/Q-type channels was qualitatively preserved when either the β1b or β3 subunit was substituted for the β2a subunit used in the experiments shown in Figure 5 (data not shown), after taking into account the overall changes in baseline inactivation properties seen with these other subunits (Patil et al., 1998). Assuming that the contrasting features of M2 receptor-mediated inhibition of P/Q- and N-type channels in HEK 293 cells would hold true for baclofen-induced channel inhibition in hippocampal neurons, we made the following specific prediction: if relief of G-protein-mediated inhibition produces the facilitation seen in baclofen, then the facilitation should be larger for transmission mediated exclusively by P/Q-type channels than for transmission mediated by N-type channels.
To test this key prediction, we compared the effects of baclofen during selective blockade of N-type channels with those during blockade of P/Q-type channels. Synaptic transmission via P/Q-type channels was isolated with ωCtx-GVIA, a peptide toxin that selectively blocks N-type channels (Dunlap et al., 1995; Mori et al., 1996). During steady-state ωCtx-GVIA blockade (Fig.6 A), baclofen still inhibited transmission by 70 ± 4% (n = 7), and short-term depression was reduced or converted to facilitation (Fig.6 A,C, left, D), as in the absence of toxins (Fig. 1). In separate cells, N-type channel-mediated transmission was isolated with ωAga-IVA, a peptide inhibitor of both P- and Q-type channels (Dunlap et al., 1995; Mori et al., 1996). After ωAga-IVA blockade (Fig. 6 B), baclofen inhibited neurotransmission by an additional 73 ± 4% (n = 6). In contrast to the results with isolated P/Q-type channels (Fig.6 A), baclofen caused little change in short-term plasticity when only N-type channels triggered transmission (Fig.6 B,C, right, D). The distinction between the effects of baclofen with the two toxins was actually greater than might be expected from our studies of recombinant channels, even after factoring in a standard third to fourth power relationship between Ca2+ current and EPSCs. Quantitative differences between the extents of facilitation in the cultured neurons and HEK 293 cells could reflect differences in the nature of the G-proteins or in the parameters of action potentials. These parameters may change during repetitive presynaptic firing but were held fixed in experiments with recombinant channels. Overall, the remarkable agreement of these results with our prediction strongly supported the hypothesis that relief of G-protein-mediated inhibition of calcium channels underlies the short-term synaptic facilitation observed in the presence of baclofen.
The experiments summarized in Figure 6 merit an important technical clarification. To conserve on expenditure of toxins, we stopped toxin application before baclofen exposure (e.g., Fig.6 A,B). Nonetheless, the baclofen experiments can be interpreted as if steady-state toxin blockade were still in effect, because these experiments were always performed within an ∼5 min window shortly after toxin application, and control experiments showed almost no reversal of blockade over this time span. In such 5 min windows, there was −0.1 ± 0.96% recovery of pretoxin currents after cessation of ωCtx-GVIA (n = 4 cells) and −1.89 ± 2.3% recovery after ωAga-IVA (n = 3 cells). Hence, the conclusions based on results in Figure 6 are no different from those that would be drawn using continuous toxin application.
This report is the first that specifically implicates a contribution of relief of G-protein-mediated inhibition to synaptic plasticity. This form of relative facilitation is distinct from the mechanisms of short-term facilitation that have been described in other preparations. Presynaptic calcium channels in the calyx of Held exhibit a Ca2+ entry-dependent form of facilitation that contributes to short-term synaptic plasticity (Borst and Sakmann, 1998; Cuttle et al., 1998), but this effect may involve the action of Ca2+-calmodulin (Lee et al., 1999) and does not require G-protein activation. Others have shown that STD can be attenuated by an overall reduction of neurotransmitter release, yielding a relative facilitation (Lev-Tov and Pinco, 1992;Barnes-Davies and Forsythe, 1995; Isaacson and Hille, 1997; Varela et al., 1997). However, these effects on STD were attributable to release-dependent mechanisms such as vesicle depletion (Zucker, 1989). At an avian calyceal synapse, such attenuation of STD can in fact result in frank facilitation of synaptic responses during high-frequency stimulation, with concomitant increases in the likelihood of firing postsynaptic action potentials (Brenowitz et al., 1998). By contrast, in hippocampal autapses, we have found that G-protein-mediated inhibition reduces STD or converts it to relative facilitation, in a manner that is consistent with relief of G-protein-mediated inhibition of calcium channels but not with Ca2+-dependent facilitation or attenuation of release-dependent depression.
Have there been previously reported hints of a mechanism such as that observed here? At the same avian synapses, conversion of paired-pulse depression to facilitation by baclofen inhibition could have resulted in part from relief of G-protein-mediated inhibition, because comparable Cd2+-mediated inhibition occluded depression but produced less facilitation (Otis and Trussell, 1996). Also, in rat hippocampal slice cultures, adenosine inhibited transmission and increased paired-pulse facilitation. Although lowering [Ca2+]o and increasing [Mg2+]oinhibited transmission more than did adenosine, this maneuver increased paired-pulse facilitation less than did adenosine (Debanne et al., 1996). Finally, in frog sympathetic ganglia, synaptic depression caused by acetylcholine autoinhibition was prominent during 5 Hz stimulation, but not at 20 Hz (Shen and Horn, 1996). The authors proposed that 20 Hz stimulation could have relieved the inhibition of presynaptic calcium channels, whereas 5 Hz action potentials did not. One important difference between some previous results and ours is that the baseline depression in cultured hippocampal neurons was not detectably release dependent. This unusual finding has been investigated in a separate set of experiments, which supports axonal branch-point failure (Parnas, 1972; Hatt and Smith, 1976; Macagno et al., 1987; Streit et al., 1992;Wall, 1995; Debanne et al., 1997) as the underlying mechanism for the short-term synaptic depression (Brody and Yue, unpublished observations). Regardless of the mechanism, the release-independent nature of the baseline depression clarified the interpretation of changes in short-term plasticity during G-protein-mediated inhibition of calcium channels.
Another key aspect of our approach concerns the use of Cd2+ to test for release-dependent mechanisms by reducing presynaptic Ca2+influx. Both G-protein-mediated inhibition and Cd2+ blockade of calcium channels in effect reduce channel open probability without altering unitary current amplitude; Cd2+ at low doses produces a millisecond flickering block of calcium channels (Lansman et al., 1986), whereas G-protein-mediated inhibition of neuronal calcium channels slows the activation of channels (Carabelli et al., 1996;Patil et al., 1996). An advantageous distinction between G-protein-mediated inhibition and Cd2+blockade is that the former can be relieved during repetitive activity (e.g., Fig. 5 A) but Cd2+blockade remains invariant during analogous trains of APWs (Fig.7). Here, the patterns of short-term changes in recombinant P/Q- or N-type currents did not change after robust blockade by Cd2+. The invariance of Cd2+ blockade during repetitive activity probably arises from the relative voltage independence of Cd2+ blockade of neuronal channels over a physiological range of voltages (Leonard et al., 1987). When voltage-dependent unblock of Cd2+ in neuronal calcium channels has been observed (Thevenod and Jones, 1992), it was not apparent for voltages up to +40 mV and manifested clearly only with stronger (unphysiological) depolarization.
Changing extracellular calcium and/or magnesium concentrations to test for release-dependent mechanisms may not be equivalent to Cd2+ blockade. Altering Ca2+ or Mg2+alters unitary channel conductance (Church and Stanley, 1996), which may have distinct effects on calcium concentrations in presynaptic microdomains. Also, changes in these divalent cations may alter membrane excitability by surface-charge-screening effects (Frankenhaeuser and Hodgkin, 1957), which appears to affect various ion channels inhomogeneously (Green and Andersen, 1991). The low doses of Cd2+ required to block calcium channels do not produce detectable surface-charge effects, as gauged by the voltage required for either half-activation of calcium channels (Leonard et al., 1987) or half-inactivation of sodium channels (Hanck and Sheets, 1992). Furthermore, the micromolar concentrations of Cd2+ used in this study had no measurable effect on action potentials measured in hippocampal somata, with half-width durations of 1.42 ± 0.16 msec in control versus 1.39 ± 0.14 msec in 2 μmCd2+ (n = 5 cells;p > 0.40). These distinctions are relevant to the present results, because changing Ca2+and/or Mg2+ does affect short-term synaptic plasticity in this preparation (Mennerick and Zorumski, 1995), but in a complex manner that is inconsistent with the presence of release-dependent depression (Brody and Yue, unpublished observations). Thus, the use of Cd2+ allowed us to exclude unambiguously several potential alternative explanations for the effects of baclofen on short-term synaptic plasticity.
Although baclofen did not show discernible postsynaptic effects (Fig.3; Table 1), it did reduce the rate of spontaneous EPSCs from 1.62 Hz in control to 0.67 Hz in baclofen (Table 1). In other preparations, presynaptic inhibition has similar effects (Scanziani et al., 1992;Dittman and Regehr, 1996), which have been hypothesized to involve an inhibitory effect on the release machinery downstream of Ca2+ entry. Although this effect merits future investigation, it is unlikely to contribute to the relative facilitation produced by baclofen. First, a tonic decrease in overall release probability, such as might be inferred from the decrease in mini frequency, would not be expected to give rise to facilitation in the autapse preparation, because of the evidence favoring release-independent short-term depression (i.e., Fig. 1, the experiments with Cd2+). Second, effects on the release machinery downstream of Ca2+entry could not account for the near occlusion of baclofen-induced facilitation after blockade of P/Q-type channels (Fig. 6).
One final alternate mechanism that we have not considered is the potential interaction of baclofen with the rab protein pathway (Simons and Zerial, 1993; Fischer von Mollard et al., 1994). Rab proteins are small GTP-binding proteins that may catalyze GTP hydrolysis in the course of facilitating vesicle docking and/or priming (Sudhof, 1995). Activation of G-protein-coupled receptors by baclofen could in principle decrease local GTP concentrations and thereby reduce vesicle docking and/or repriming. Although this could conceivably affect short-term synaptic plasticity, G-protein-coupled receptor activation should, if anything, enhance short-term depression. Instead we found experimentally that baclofen had just the opposite effect, shifting short-term plasticity toward facilitation. Furthermore, the involvement of rab proteins could in no way explain the differential effects on short-term responses produced by blockade of N- versus P/Q-type channels. Hence, it seems unlikely that changes in rab protein cycling underlie the baclofen-induced changes in synaptic plasticity.
The form of short-term synaptic plasticity described here could have important implications for neuronal information processing. G-protein-mediated presynaptic inhibition may not cause an absolute quieting of the synapse but rather a selective damping of low-frequency activity. Information carried in high-frequency bursts of action potentials (Connors and Gutnick, 1990; Gray and McCormick, 1996;Lisman, 1997) may still be transferred reliably, because such bursts could partially reverse the inhibition. This could be a widespread phenomenon, because autosynaptic and heterosynaptic inhibition by a variety of neurotransmitters, as well as tonic adenosine receptor activation, act via G-protein-mediated inhibition of presynaptic calcium channels (Anwyl, 1991; Jones and Elmslie, 1997). For example, the diffusion of GABA from its site of release to presynaptic terminals on glutamate-releasing synapses in CNS slices has been shown to heterosynaptically inhibit excitatory transmission via activation of GABAB receptors (Isaacson et al., 1993; Dittman and Regehr, 1997). Such inhibition could significantly alter short-term plasticity at the affected excitatory synapses, at least in part due to relief of G-protein inhibition. Likewise, presynaptic G-protein-coupled autoreceptors can be activated by transmitter released from the same synapse, inhibiting further transmitter release (Deisz and Prince 1989;Scanziani et al., 1997). Relief of G-protein inhibition could limit the magnitude of this sort of negative feedback during high-frequency firing. Additionally, tonic G-protein inhibition may occur because of the presence of extracellular adenosine in slice preparations (Wu and Saggau, 1994; Masino and Dunwiddie, 1999). Therefore, relief of G-protein inhibition may underlie a portion of the baseline short-term plasticity at some synapses. Thus, this frequency-selective mechanism may affect many aspects of synaptic function throughout the nervous system. Inhibition by adenosine has been implicated in the regulation of arousal and wakefulness in vivo(Porkka-Heiskanen et al., 1997), and it will be interesting to investigate the role of possible concomitant changes in short-term synaptic plasticity in more intact systems.
This work was supported by National Institutes of Health and National Science Foundation grants to D.T.Y. and a Medical Scientist Training Program fellowship to D.L.B. We thank C. Boyer and C. F. Stevens for the microisland culture protocol; A. Ghosh, C. Jahr, and C. Zorumski for initial advice on neuronal cell culture; SIBIA Neuroscience (human α1B), T. Snutch, E. Perez-Reyes, and E. Peralta for clones; K. Elmslie for the generous gift of the ωAga-IVA; and C. Aizenmann, H. Colecraft, D. DiGregorio, J. Dittmann, L. Jones, D. Linden, and P. Fuchs for helpful discussions and comments on this manuscript.
Correspondence should be addressed to Dr. David T. Yue, The Johns Hopkins University School of Medicine, Department of Biomedical Engineering, Program in Molecular and Cellular Physiology, 720 Rutland Avenue, Baltimore, MD 21205. E-mail:.