Unmasking group III metabotropic glutamate autoreceptor function at excitatory synapses in the rat CNS


  • Brian Billups,

    1. Department of Cell Physiology and Pharmacology, University of Leicester, PO Box 138, University Road, Leicester LE1 9HN, UK
    Search for more papers by this author
  • Bruce P. Graham,

    1. Department of Computing Science and Mathematics, University of Stirling, Stirling FK9 4LA, Scotland, UK
    Search for more papers by this author
  • Adrian Y. C. Wong,

    1. Department of Cell Physiology and Pharmacology, University of Leicester, PO Box 138, University Road, Leicester LE1 9HN, UK
    Search for more papers by this author
  • Ian D. Forsythe

    1. Department of Cell Physiology and Pharmacology, University of Leicester, PO Box 138, University Road, Leicester LE1 9HN, UK
    2. MRC Toxicology Unit, University of Leicester, PO Box 138, University Road, Leicester LE1 9HN, UK
    Search for more papers by this author

Corresponding author I. D. Forsythe: MRC Toxicology Unit, University of Leicester, Leicester LE1 9HN, UK.  Email: idf@le.ac.uk


Presynaptic group III metabotropic glutamate receptor (mGluR) activation by exogenous agonists (such as l-2-amino-4-phosphonobutyrate (l-AP4)) potently inhibit transmitter release, but their autoreceptor function has been questioned because endogenous activation during high-frequency stimulation appears to have little impact on synaptic amplitude. We resolve this ambiguity by studying endogenous activation of mGluRs during trains of high-frequency synaptic stimuli at the calyx of Held. In vitro whole-cell patch recordings were made from medial nucleus of the trapezoid body (MNTB) neurones during 1 s excitatory postsynaptic current (EPSC) trains delivered at 200 Hz and at 37°C. The group III mGluR antagonist (R,S)-cyclopropyl-4-phosphonophenylglycine (CPPG, 300 μm) had no effect on EPSC short-term depression, but accelerated subsequent recovery time course (τ: 4.6 ± 0.8 s to 2.4 ± 0.4 s, P= 0.02), and decreased paired pulse ratio from 1.18 ± 0.06 to 0.97 ± 0.03 (P= 0.01), indicating that mGluR activation reduced release probability (P). Modelling autoreceptor activation during repetitive stimulation revealed that as P declines, the readily releasable pool size (N) increases so that the net EPSC (NP) is unchanged and short-term depression proceeds with the same overall time course as in the absence of autoreceptor activation. Thus, autoreceptor action on the synaptic response is masked but the synapse is now in a different state (lower P, higher N). While vesicle replenishment clearly underlies much of the recovery from short-term depression, our results show that the recovery time course of P also contributes to the reduced response amplitude for 1–2 s. The results show that passive equilibration between N and P masks autoreceptor modulation of the EPSC and suggests that mGluR autoreceptors function to change the synaptic state and distribute metabolic demand, rather than to depress synaptic amplitude.

Early evidence for presynaptic glutamate receptors arose from the observation that the phosphonic derivative of glutamate, l-2-amino-4-phosphonobutyrate (l-AP4) depressed excitatory transmission (Koerner & Cotman, 1981; Davies & Watkins, 1982). Presynaptic depression was also induced by local application of glutamate (Forsythe & Clements, 1990), and metabotropic glutamate receptors (mGluR) were implicated by the observation that the specific agonist, trans-1-amino-1,3-cyclopentanedicarboxylic acid (trans-ACPD), depressed both excitatory (Baskys & Malenka, 1991) and inhibitory synapses (Desai & Conn, 1991). There are eight members of the G-protein-coupled mGluR family which are related to GABAB receptors, having a bilobed agonist-binding domain, dimeric structure and broad expression in the CNS (Anwyl, 1999; Schoepp, 2001). Although both group II (mGluR2 and 3) and group III (mGluR4, 6, 7 and 8) receptors are associated with presynaptic actions, attention here is focused on those mGluRs (4, 7 and 8) that are specifically activated by l-AP4 and serve as autoreceptors at many glutamatergic synapses in the central nervous system.

One difficulty in establishing the physiological relevance of presynaptic mGluRs is that by definition, autoreceptor activation is through endogenous release, requiring repetitive stimulation. Hence their action is superimposed on other short-term modulatory phenomena (e.g. facilitation and depression). Application of exogenous agonist (such as l-AP4) potently inhibits transmitter release and confirms the potential for presynaptic activation (Forsythe & Clements, 1990), but provides little information of their physiological activation. On the other hand, studies using antagonists to block endogenous activation during trains of stimuli show little change in synaptic amplitude at several l-AP4-sensitive synapses, including the calyx of Held (von Gersdorff et al. 1997), nucleus tractus solitarius (Chen et al. 2002) and the parallel fibre–Purkinje cell synapse in the cerebellum (Lorez et al. 2003), implying limited physiological significance in modulating synaptic transmission. However here we explore an alternative explanation, namely that the minimal changes in excitatory postsynaptic current (EPSC) amplitude are due to compensatory mechanisms which mask mGluR autoreceptor effects on transmitter release.

We have investigated the role of endogenous mGluR autoreceptor activation in modulating short-term plasticity at the calyx of Held synapse using whole-cell patch clamp of the postsynaptic medial nucleus of the trapezoid body (MNTB) neurone during orthodromic stimulation of the presynaptic axon and terminal at physiological frequencies (200 Hz) and temperatures. The calyx of Held is an excitatory glutamatergic synapse generating a large EPSC mediated by postsynaptic AMPA and NMDA receptors (Forsythe and Barnes Davies, 1993), in addition to group III metabotropic receptors (Elezgarai et al. 1999; Renden et al. 2003) which are expressed on the presynaptic terminal. Application of the specific group III mGluR agonist l-AP4, reduces neurotransmitter release (Barnes-Davies & Forsythe, 1995) through a direct G-protein βγ subunit inhibition of calcium channels (Herlitze et al. 1996; Takahashi et al. 1996). During repetitive stimulation, the EPSC exhibits short-term depression caused by vesicle depletion, reduced release probability and AMPA receptor desensitization (Schneggenburger et al. 1999; Scheuss et al. 2002; Wong et al. 2003).

Under conditions which minimized postsynaptic desensitization, we found that group III mGluRs were activated during trains of synaptic stimuli and caused a cumulative reduction in release probability, evident as a switch from paired-pulse depression to paired-pulse facilitation following the stimulus train. Endogenous mGluR activation was masked when measuring EPSC amplitude during stimulus trains, but on recovery mGluR activation was exhibited as a slowed rate of recovery from synaptic depression. Our modelling suggests a simple explanation for this functional masking during the repetitive stimulation: namely that modulation (in this case by mGluR) of release probability (P) has a passive reciprocal influence on readily releasable pool (N) size (i.e. as P declines, N increases) hence the response amplitude (NP) will converge to the same product as under control conditions.

Short-term plasticity has important implications for communication between central neurones (Zucker & Regehr, 2002), and our results suggest that mGluR autoreceptor action allows short-term depression to proceed in a coherent manner, preserving information within a stimulus train. The mGluR autoreceptors do not depress the response amplitude during repetitive stimulation (maintaining the rate and magnitude of short-term depression), but could serve to redistribute release across a broader pool of release sites during repetitive stimulation.


Ten- to 14-day-old Lister hooded rats were killed by decapitation in accordance with the UK Animals (Scientific Procedures) Act 1986. Transverse brainstem slices (200 μm thick) were prepared as previously described (Wong et al. 2003). The slicing medium was maintained at around 0°C and contained (mm): 250 sucrose; 2.5 KCl; 10 glucose; 1.25 NaH2PO4; 26 NaHCO3; 4 MgCl2; 0.1 CaCl2 and 0.5 ascorbate (pH 7.4 when gassed with 95% O2, 5% CO2). The control aCSF for recording contained (mm): 125 NaCl; 2.5 KCl; 10 glucose; 1.25 NaH2PO4; 26 NaHCO3; 1 MgCl2; 2 CaCl2; 3 myo-inositol; 0.5 ascorbic acid, 2 Na-pyruvate, 2 kynurenate, 0.04 d(–)2-amino-5-phosphonopentanoic acid (AP5), 0.01 MK801, 0.01 bicuculline and 0.001 strychnine (pH 7.4 when gassed with 95% O2, 5% CO2). Under these conditions NMDA, GABAA, and glycine receptors were fully blocked and the evoked AMPA receptor-mediated responses were partially blocked by kynurenate (86%) in order to minimize saturation and desensitization (Wong et al. 2003).

Whole-cell patch clamp recordings were made from visually identified MNTB neurones with an Axopatch 200B amplifier, filtered at 10 kHz and sampled at 20 kHz. Currents were recorded with pCLAMP8 (Axon Instruments). Pipette open tip resistances were 4–6 MΩ, whole-cell access resistances were <20 MΩ and series resistance was compensated >70% with a 10 μs lag time. Experiments were performed at physiological temperature (35–37°C). The intracellular solution contained (mm) 110 CsCl; 40 Hepes; 10 TEA-Cl; 12 Na2-phosphocreatine; 1 EGTA and 2 QX314 (pH adjusted to 7.3 with CsOH). Presynaptic axons were activated (2–8 V and 0.2 ms) by a DS2A isolated stimulator (Digitimer, Welwyn Garden City, UK) and bipolar platinum electrode placed at the midline across the slice. Synaptic connections were detected by loading MNTB neurones with fura-2AM and imaging the resultant postsynaptic calcium rise (Billups et al. 2002). Conditioning trains (1 s, 200Hz) were evoked in the presynaptic axons at intervals of 30 s, and the resultant EPSC trains recorded under voltage clamp at a holding potential of –70 mV. To follow recovery time course, a test EPSC or a short train of 50 EPSCs (200 Hz) was delivered following each conditioning train with varying intervals of up to 10 s. Five repetitions were measured for each interval and each curve required seven intervals, thus requiring over 40 min of stable recording time for a control, drug application and test recovery curve.

A simple model of vesicle exocytosis was implemented in Microsoft Excel. The vesicle pool was refilled to a maximum value of 4000 vesicles with a single exponential time constant, τ. Another model with activity-dependent vesicle recycling (Graham et al. 2004) was extended to include short-term facilitation, and mGluR activation. In the model, vesicles in a releasable pool of size n may release with probability p on the arrival of a presynaptic stimulus (eqn (1)) to give an EPSC amplitude proportional to np (eqn (9)). Vesicles in this releasable pool are replenished at a constant background rate (time constant τn) and at an elevated rate by a fraction of reserve vesicles na (eqn (3)) for a short period (time constant τa) following each presynaptic stimulus (eqn (2)). The finite-sized reserve pool nr may deplete following prolonged stimulation (eqn (4)). Short-term facilitation is modelled as an instantaneous increase in release probability p following a stimulus, which then decays back to a baseline value pb with time constant τp (eqns (5) and (6)) (Markram et al. 1998). The effect of activation of presynaptic mGluRs is modelled as a decrease in this baseline release probability pb in proportion to the amount of release following a stimulus (eqn (7)). This, in turn, affects the vesicle release probability p as it recovers to an increasingly small value (given by pb) on recovery from facilitation. The baseline probability recovers with a single, slow time constant τm to its initial value (eqn (8)). This model was implemented in Matlab. The equations governing this model are:

display math(1)
display math(2)
display math(3)
display math(4)
display math(5)
display math(6)
display math(7)
display math(8)
display math(9)

The time, s, is a presynaptic spike time; s+ is just after release; t is a time before the next spike. Here n is the relative size of the releasable pool, i.e. n(0) = 1. It is replenished with background time constant, τn= 2.8 s and an activity-dependent time constant, τa= 60 ms, to a maximum fraction na(0) = 0.6, from a finite reserve pool, nr (which describes the relative size of the pool, i.e. nr(0) = 1). The initial size of the reserve pool is nrp= 17 vesicles (number of vesicles available to replenish each release site), and it is assumed not to replenish during the course of an experiment. Release, p= 0.21 (initially), is facilitated by Δp= 0.12 on each spike, and recovers with time constant τp= 5 ms. Activation of mGluRs depresses baseline release, pb, by up to Δpm= 0.014 on each spike, which recovers with time constant τm= 7 s. Parameter values were chosen using a combination of least-squares optimization to the depression data, and hand-fitting to the recovery and paired-pulse data. Each component of the model makes an identifiable contribution to the match of the model with the experimental data. Vesicle recycle rates and release probability are comparable to experimental estimates (von Gersdorff et al. 1997; Sakaba & Neher, 2001), as are the time course of its facilitation. The size of the reserve pool is in accord with the number of undocked vesicles seen in close proximity to release sites (Satzler et al. 2002; Nicol & Walmsley, 2002).

l-(+)-2-amino-4-phosphonobutyric acid (l-AP4), CPPG, bicuculline, strychnine, AP5, MK801 and QX314 were obtained from Tocris Cookson. All other chemicals were obtained from Sigma. Data and graphs are expressed as the mean ±s.e.m. and statistical significance (P < 0.05) was tested with paired two-tailed t tests. On presentation of double exponential fits, the relative magnitude (%) for the first component is indicated in brackets.


Inhibition of mGluRs has no effect on synaptic depression during high frequency trains

To investigate the endogenous activation of presynaptic metabotropic glutamate receptors at the calyx of Held synapse, postsynaptic MNTB neurones in rat brainstem slices were whole-cell voltage clamped. AMPA receptor-mediated EPSCs were elicited by electrical stimulation of the trapezoid body fibres and recorded in the presence of 2 mm kynurenate to minimize receptor saturation and desensitization (Wong et al. 2003). Presynaptic group III mGluRs were activated by the specific agonist l-AP4 (50 μm) causing a reduction in the EPSC amplitude to 21 ± 3% of control (n= 3) which was reversed by 300 μm CPPG (Fig. 1A and B). Since the affinity of l-AP4 for group III mGluRs is an order of magnitude higher than that of glutamate, this CPPG dose would also antagonize endogenous activation. Activation of postsynaptic mGluRs was prevented by the exclusion of ATP and GTP from the postsynaptic patch pipette.

Figure 1.

Presynaptic group III mGluRs depress the calyx EPSC
EPSCs were recorded from MNTB neurones under whole-cell voltage clamp. A, the stimulus artefact (asterisk) precedes the EPSC (arrow). The control EPSC (black trace) was reduced by the application of 50 μml-AP4 (grey trace) and reversed by 300 μm CPPG. B, EPSC magnitude plotted every 10 s during l-AP4 and CPPG bath application. C and D, bath application of CPPG (grey trace) alone had no effect on EPSC magnitude or time course. E, summary data: the asterisk indicates a significant reduction in the EPSC by l-AP4 (n= 3, P= 0.02). F, EPSC trains at 200 Hz were elicited for 1 s (stimulus artefacts have been erased for clarity). G, the same data on a faster time-scale, 1 s 200 Hz trains, control (black trace) and following CPPG (grey trace). The EPSCs are almost identical in control and CPPG. H, each EPSC from the 200 Hz train is shown normalized to the magnitude of the first EPSC in the train. Data from five cells are averaged and standard error bars are plotted on every 5th point.

Persistent or resting activation of the presynaptic mGluRs by low concentrations of extracellular glutamate (Losonczy et al. 2003) can be excluded, since CPPG had no effect on unitary EPSC amplitude (Fig. 1C and D; EPSC was 104 ± 5% of control; n= 5, P= 0.4). The 10–90% rise time and single exponential decay time constant were also unaffected, being 0.22 ± 0.01 and 0.68 ± 0.04 ms in control and 0.22 ± 0.01 and 0.74 ± 0.06 ms in CPPG, respectively (n= 5, P= 0.6 and 0.1).

The calyx of Held/MNTB synapse can transmit at very high frequencies under physiological conditions (Wu & Kelly, 1993; Taschenberger & von Gersdorff, 2000; Kopp-Scheinpflug et al. 2003), and presynaptic mGluRs may be activated only during high-frequency trains, with glutamate accumulating in and around the synaptic cleft (Scanziani et al. 1997). To test this hypothesis we studied the effects of CPPG on EPSC trains evoked at 200 Hz for 1 s, at 37°C (Fig. 1F). The resulting EPSCs rapidly depressed to a steady-state amplitude, with the depression rate being well fitted by a double exponential function having time constants of 24 ± 4 ms (91%) and 700 ± 80 ms (n= 5); at the end of the train, EPSCs were 14 ± 3% of their initial magnitude. CPPG (300 μm) had no effect on the magnitude of depression (13 ± 2%; n= 5, P= 0.7) or the depression rate (21 ± 5 ms (91%) and 520 ± 110 ms; n= 5, P= 0.1 and 0.3; Fig. 1G and H; grey traces). These data indicate either that the presynaptic mGluRs are not activated (as previously supposed) or that following mGluR activation secondary changes mask the autoreceptor action.

Recovery from synaptic depression is slowed by presynaptic mGluR activation

Short-term depression during repetitive stimulation appeared unchanged by CPPG; however, the rate of recovery following the train was enhanced, suggesting that endogenous mGluR activation is of functional significance. Recovery from synaptic depression was examined by eliciting a single EPSC at varying time intervals after the end of a 1 s conditioning train (Fig. 2A). These test EPSCs were normalized to the amplitude of the first EPSC in the conditioning train, and plotted as a function of time (Fig. 2B). Application of CPPG (300 μm) enhanced this recovery, with recovery curves being fitted with a double exponential of time constants: 31 ± 13 ms (13%) and 4.6 ± 0.8 s under control conditions, and 67 ± 25 ms (26%) and 2.4 ± 0.4 s following CPPG application. The dominant recovery component, with the slower time constant, was accelerated nearly two-fold to 192% of control (n= 5 P= 0.02) by blocking mGluRs. The magnitude of the recovery for test EPSCs measured 2 s after the end of the conditioning train was increased from 52 ± 4% to 73 ± 6% by CPPG (n= 5 P < 0.01; Fig. 2C and D). Hence, although mGluR activation had no effect on EPSC magnitude during the conditioning train, the recovery curves prove that glutamate release during the train does nevertheless activate mGluRs and influence synaptic signalling.

Figure 2.

Block of mGluRs accelerates recovery from synaptic depression
A, the time course of recovery from depression was assessed by eliciting a single EPSC at varying time intervals (Δt) following a conditioning train. Unitary EPSCs are overlaid to illustrate the recovery time course. B, the magnitude of the unitary EPSC is normalized to the first EPSC of the train and plotted against the recovery time interval for control (black symbols) and following CPPG (300 μm, grey symbols). Recovery time intervals of 0.05, 0.1, 0.2, 0.5, 1, 2, 5 and 10 s were used. Data from five cells were averaged and fitted with double exponential functions. C, sample data from the 2 s recovery time point in control (black trace) and following CPPG (grey trace). The first few EPSCs in the 200 Hz train are shown, which are unaffected by CPPG (as in Fig. 1G). The recovery EPSC is shown, significantly enhanced by CPPG. D, the recovery 2 s after the train was assessed every 4 min and is plotted as a function of time following CPPG application. Data from five cells were averaged.

mGluR activation reduces release probability during synaptic stimulation

Since presynaptic mGluR activation suppresses calcium currents (Takahashi et al. 1996) and reduces release probability, one explanation for the masking of mGluR action during short-term depression is a simple inverse compensatory change in release probability and vesicle availability. This hypothesis was tested by examining the paired-pulse ratio (PPR) of two closely spaced stimuli following the conditioning train, with an increased release probability being indicated by decreased PPR. The PPR of two EPSCs, 5 ms apart, elicited 1 s after the 200 Hz train is shown in Fig. 3A. Activation of mGluRs under control conditions caused paired-pulse facilitation, whereas following mGluR antagonism, EPSC magnitude was enhanced and paired-pulse depression was observed. Averaged data (Fig. 3B and C) shows a PPR of 1.18 ± 0.06 under control conditions and 0.97 ± 0.03 (n= 5, P= 0.01) in CPPG. The PPR at the start of the 200Hz train was unaffected by CPPG, being 0.90 ± 0.06 in control and 0.83 ± 0.07 in CPPG (n= 5, P= 0.06). Following the train, CPPG increased EPSC magnitude, which, combined with the decrease in the PPR, is consistent with a reduction in release probability mediated by the presynaptic mGluRs.

Figure 3.

Inhibition of mGluRs increases release probability
Changes in release probability following the train were assessed using paired pulses 1 s after the train. A, two EPSCs, 5 ms apart, are shown and normalized to the magnitude of the first EPSC, in the absence (black traces) and presence (grey traces) of 300 μm CPPG. The broken lines indicate the change from paired-pulse facilitation to paired-pulse depression following CPPG application. B, the paired-pulse ratio following the train is plotted as a function of time. Data are averaged from five cells. C, The paired-pulse ratio is significantly decreased following CPPG application (asterisk indicates P= 0.01).

An alternative method for assessing release probability is to use the cumulative amplitude of EPSCs in a train to estimate the size of the readily releasable vesicle pool. The y intercept of a line fitted to the steady-state portion of the curve gives an indication of the size of this pool. The proportion of the pool released by the first EPSC of the train can then be calculated (Schneggenburger et al. 1999). Trains at 200 Hz were delivered for 1 s, followed 1 s later by another 200 Hz train (Fig. 4A). The cumulative EPSC amplitudes for the first and second train are shown in Fig. 4B, in the presence and absence of CPPG. For the first train, the data points for the control and CPPG curves overlap completely, as expected since CPPG has no effect on the magnitude of the EPSCs in the initial train (Fig. 1H).

Figure 4.

mGluR activation reduces release probability during recovery from short-term depression
A, 200 Hz, 1 s trains were elicited for 1 s, followed 1 s later by a second train (200 Hz, 0.25 s). The resulting EPSCs are shown with stimulus artefacts removed for clarity. B, the cumulative amplitudes of the first 50 EPSCs in the first train are shown in the presence and absence of CPPG, with the control and CPPG data points overlapping. A straight line was fitted by linear regression to the last 15 EPSCs on the control graph, and back-extrapolated to time 0 to estimate the releasable pool size. The cumulative amplitudes of the EPSCs in the second train are also shown in control (black traces) and in the presence of CPPG (grey traces). Straight lines were also fitted to the last 15 EPSCs to estimate the pool size and release probability. All data are averages from five cells. C, The pool size and release probability are shown for the second train in the presence (grey) and absence (black) of CPPG. D, the time between the two trains was increased and the calculated release probability plotted as a function of the time interval, in the presence of CPPG (grey trace) and in control, with mGluRs being activated (black trace). Asterisks indicate statistical significance in C and D (P < 0.05).

The line fitted through the later portion of this curve intercepts the y axis at 4.8 ± 0.4 nA in control (black line) and 4.6 ± 0.4 nA in CPPG (grey line, n= 5, P= 0.45). This corresponds to a pool size of 879 vesicles assuming a miniature EPSC amplitude of 39 pA (Schneggenburger et al. 1999) and an 86% inhibition by 2 mm kynurenate (Wong et al. 2003). The first EPSC in the train releases 23 ± 3% of the pool in control conditions, and this is unaffected by CPPG (25 ± 4%, n= 5, P= 0.30).

Following a 1 s recovery period, mGluR block has a significant effect on the cumulative amplitude plots. Under control conditions the y intercept for the data from the second train is 2.4 ± 0.2 nA, and the proportion released with the first EPSC of that train is 0.16 ± 0.01. While the intercept value is unchanged by CPPG (2.8 ± 0.3, n= 5, P= 0.06; Fig. 4C), the proportion released is significantly raised to 0.23 ± 0.02 (n= 5, P= 0.01), indicating that mGluR activation lowers release probability under control conditions.

The recovery time course for the release probability after the initial train was assessed by measuring the cumulative amplitude plots of a second train elicited at different time intervals following the end of the first train. This analysis is complicated by the fact that at the shorter time intervals (<1 s) the releasable pool size cannot be estimated from the y intercept because the number of vesicles recycled between each EPSC is large compared to the number of vesicles released. Nevertheless, at longer time intervals, the time course of the mGluR-mediated change in release probability assessed by this method (Fig. 4D) matches the time course of the differences in the recovery curves (Fig. 2B), with significant differences at 1 and 2 s time intervals, suggesting that under control conditions the influence of mGluR activation on transmitter release lasts for at least 2 s.

N versus P: changes in release probability need not alter the evoked EPSC size

It seems surprising that EPSC depression during a high-frequency train is unchanged upon mGluR activation (Fig. 1H), even though release probability is reduced. However, using a simple model of vesicle depletion (Liley & North, 1953; Betz, 1970) it can be demonstrated that changes in release probability are not necessarily manifested as changes in evoked response amplitude. In this model, each action potential releases a proportion (P) of a single pool of vesicles to generate an EPSC. The pool is continually refilled in an exponential manner, with time constant τ (Fig. 5A). Using constant P values and refilling τ of 0.25 and 125 ms, respectively (Fig. 5B grey traces), the model predicts that the EPSC rapidly depresses to 14% of its initial value, as observed experimentally (Fig. 1H). To mimic mGluR activation, P was gradually reduced throughout the train from 0.25 to 0.15, in proportion to the EPSC size (Fig. 5B black traces) and compared to the situation where P is kept constant. The depression curve was virtually unaffected by a gradual change in P, with steady-state depression being 13% of the initial EPSC magnitude (indicated by the arrow on the middle panel). Comparing releasable pool size for the two conditions reveals that as P declined, fewer vesicles were released and the vesicle pool size showed a corresponding increase. Throughout the train, a smaller proportion of a larger pool is released and the resulting EPSC is unchanged. The size of the pool had changed by the end of the train from 14 to 21% of the initial value, an increase of 50% (Fig. 5B, right panel).

Figure 5.

A simple model reveals that depression is insensitive to changes in release probability
A, the presynaptic terminal contains a single pool of vesicles, a proportion (P) of which is released by each action potential, and it is replenished by a first-order process with an exponential time constant (τ). The resulting EPSC magnitude is proportional to the number of vesicles released. B, the model is driven with a 200 Hz 1 s train and P is kept constant (grey traces). The normalized EPSC and pool size decline to a steady state. P is changed to mimic presynaptic mGluR activation (black traces), with little effect on EPSC magnitude (arrow, middle panel), but causing an increase in pool size (right panel). C, if τ is changed by a similar proportion (black line), significant changes in the EPSC are observed (arrow, middle panel).

Reducing the pool refilling rate, τ, by a similar proportion (from 125 to 75 ms; Fig. 5C black traces) resulted in an initially similar depression rate, but then the EPSC amplitude slowly increased by approximately 50% towards the end of the train (arrow, middle panel), confirming that the degree of synaptic depression during a train is ultimately governed by the refilling rate, τ, not by P (Markram & Tsodyks, 1996; Abbott et al. 1997; Brenowitz & Trussell, 2001).

This simple model was extended to include short-term facilitation and activity-dependent vesicle recycling (Graham et al. 2004). The changes in P used to calculate the EPSC magnitudes are shown in Fig. 6 for the period during the train (Fig. 6A, left panel) and the recovery phase following the train (Fig. 6B, left panel). The data are plotted with activation of mGluR receptors (black traces) and without mGluR receptor activation (grey traces). Data points from Fig. 1H and Fig. 2B are plotted on top of the EPSC depression and recovery curves (Fig. 6A and B, middle panels), and show a good fit to the model. Similar to the previous model (Fig. 5), examination of the vesicle pool size during the train reveals an increase that passively compensates for the decrease in P (Fig. 6A, right panel). The modelled paired-pulse ratio of two pulses 1 s after the train was 1.11 with the decrease in P, and 0.95 without the decrease (Fig. 6B, right panel). This result is consistent with the effects of CPPG on the paired-pulse ratio observed in Fig. 3, and supports the hypothesis that mGluR autoreceptors are of functional relevance at these excitatory synapses.

Figure 6.

A more complex model fits the synaptic depression and EPSC recovery following a train
Adding facilitation and activity-dependent vesicle recycling to the model (see Methods) allows assessment of recovery following a conditioning train. A, the model is driven by a 200 Hz, 1 s train with mGluR activation mimicked by an activity-dependent decrease in release proportion (P; black traces) or with mGluR activation absent (grey traces). The resulting EPSC depression curve (black and grey overlapping lines, middle panel) fits well with sample data points from Fig. 1H. The pool size during the train is increased by mGluR activation (right panel). B, P recovers with a time constant of 7 s following the train (black trace, left panel). Data from Fig. 2B are plotted and fitted with the EPSC recovery from the model (middle panel). The paired-pulse ratio of two EPSCs 1 s following the train is increased by the lowered P (right panel).


It is well established that activation of presynaptic group III mGluRs mediates depression of transmitter release at many excitatory and inhibitory synapses, but several recent reports suggest little endogenous activation at excitatory synapses. Our results accord with these findings in that group III mGluR antagonists had no effect on the rate or magnitude of EPSC depression even using high-frequency trains and at physiological temperatures. However our data also clearly show that endogenous mGluR activation changes fundamental aspects of short-term plasticity, reducing the release probability, raising PPR, and slowing the rate of recovery from synaptic depression. Simple models of transmission at the calyx of Held demonstrate that although the overall levels of synaptic depression may be unchanged by mGluR autoreceptors, the raised vesicle pool size and lowered release probability cause a fundamental change in synaptic state. Our results suggest that the physiological role of mGluR autoreceptors is achieved with minimal change in EPSC amplitude during trains, maintaining the coherent decline of the EPSC during a train (short-term depression), and suggesting that their function is not to further depress the EPSC but to redistribute release across a larger pool of active sites.

An important aspect of this study was to explore presynaptic changes under high throughput conditions, while minimizing the confounding problems of simultaneous postsynaptic changes. In order to minimize postsynaptic AMPA receptor desensitization during trains of stimuli, we used a low-affinity glutamate receptor antagonist, kynurenate. Two millimolar kynurenate reduced EPSC magnitude by around 85%, reducing clamping errors and preventing receptor saturation and desensitization by virtue of its fast unbinding rate (Wong et al. 2003). This enables physiologically relevant, high-frequency stimulation to be used to trigger endogenous mGluR activation. One disadvantage of using kynurenate is that the reduced EPSC magnitude increases the variance relative to quantal amplitude and prevents the use of fluctuation analysis to determine vesicle pool size or release probability (Clements & Silver, 2000). Activation of postsynaptic mGluRs or other second messenger pathways was minimized in our experiments by dialysis with an internal patch solution to which no ATP or GTP were added. The lack of effect of CPPG on the initial EPSC of each train is also consistent with no postsynaptic effects under our recording conditions, and argues against tonic activation at the calyx.

mGluR activation does not influence synaptic depression

Despite the changes in release probability that are evident from the recovery curve and PPR following the synaptic train, the onset rate and degree of depression during the train are remarkably constant, with and without mGluR activation (Fig. 1H). The model of vesicle exocytosis and recycling (Fig. 4AC) is of the simplest form that can account for the depression of exocytosis to a steady-state amplitude (Liley & North, 1953; Betz, 1970). A well-documented flaw is that this model greatly overestimates the recycle rate required to fit the magnitude of synaptic depression and the rate of recovery from depression. However, we have also employed other models (which include short-term facilitation and activity-dependent recycling) which can match the EPSC depression magnitude and recovery rate (Trommershauser et al. 2003; Graham et al. 2004) and give qualitatively similar change in N and P, clearly illustrating that the depression magnitude at the end of the train depends on the recycling rate (τ) rather than the release probability. This indicates that changes in release probability are not necessarily evident in EPSC depression curves, and that changes in the EPSC recovery curve can be explained without changes in rates of vesicle recycling. Our results also suggest that the slow onset of the autoreceptor response contributes to the masking process, since rapid changes in P would be detectable as a step change in EPSC amplitude, and this was not observed experimentally. The data also imply that changes in recycling are an unlikely consequence of mGluR activation, since such changes affected EPSC amplitudes and were not masked (Fig. 5C) by passive re-equilibration of N and P. However this differs from exogenous activation of presynaptic GABAB receptors, which do affect vesicle recruitment (Sakaba & Neher, 2003) at the calyx of Held.

Recovery from depression is slowed

The rate of recovery from depression is significantly slowed by mGluR activation, such that 1 s after the conditioning train the EPSC is reduced by 32% following mGluR activation (Fig. 2B). The modelling shows that during synaptic depression, mGluR-mediated changes in release probability (P) are passively compensated by reciprocal increases in the pool size (N), such that the product NP remains constant, supporting the view that a shift from higher P to larger N underlies the apparent increase in recovery rate after mGluR block. Following depression, the compensatory change of N rapidly dissipates while P remains depressed, reducing the EPSC magnitude and therefore slowing the EPSC recovery time course. Our activity-dependent model demonstrates this slowed recovery (Fig. 6B), as does a more complex model involving two pools of vesicles with different release probabilities (Trommershauser et al. 2003), if release from both pools is depressed by mGluR activation (results not shown).

Physiological importance

Metabotropic glutamate receptors have a broad role in regulating many ion channels including voltage-gated calcium channels, mobilization of intracellular calcium and modulation of transmitter release (Anwyl, 1999; Schoepp, 2001). It is well established that activation of presynaptic group III mGluRs mediates depression of transmitter release at many excitatory and inhibitory synapses, but several recent reports suggest little endogenous function at the former. There is broad expression of group II and III mGluRs in the neocortex, hippocampus, thalamus, olfactory bulb, cerebellum, brainstem and spinal cord. In synaptic terminals of the hippocampus; the low-affinity mGluR7a is located within the presynaptic active zones of glutamatergic terminals (Dalezios et al. 2002), while mGluR2 is located more remotely, requiring glutamate spillover for physiological activation (Scanziani et al. 1997). There is both pre- and postsynaptic expression of mGluR4a in the hippocampus (Bradley et al. 1999), and this variant is also located at the calyx of Held (Elezgarai et al. 1999). The role of presynaptic receptors is not simply to depress transmitter release. There is evidence for activity-dependent activation of mGluRs (Scanziani et al. 1997; Iserhot et al. 2004) and persistent activation of presynaptic mGluRs in the hippocampus (Losonczy et al. 2003), although we found no evidence for tonic activation in the MNTB.

The activation of presynaptic receptors and their physiological role are dependent on receptor location (within the cleft or elsewhere), spillover, glutamate transporter activities and the recent history of synaptic activity. Previous studies of presynaptic GABAB receptors show potent depression of transmitter release at many excitatory synapses (as heteroreceptors), but can lead to synaptic facilitation during short stimulus trains (Brenowitz et al. 1998; Brenowitz & Trussell, 2001), through preservation of the vesicle pool and secondary effects such reduced desensitization. Within the cerebellar glomeluli, glutamate spillover will inhibit GABA-mediated Golgi synapses, and GABA spillover will suppress excitatory transmission (Mitchell & Silver, 2000a,b). Autoreceptor activation at an excitatory synapse is inherently difficult to study since by definition their activation is in conjunction with other activity-dependent phenomenon. Our approach differs from most previous studies where function has usually been examined following application of specific agonists or antagonists. While such approaches prove the receptors are present (and clearly indicates the potential for modulation, particularly by heteroreceptors), the simple interpretation that activation of presynaptic receptors suppresses transmitter release is not always valid. Counter-intuitively our data suggest that endogenous activation of mGluR autoreceptors does not suppress synaptic amplitude during trains, but does nevertheless change the underlying state of the synapse. Our experimental use of endogenous activation is crucial for understanding the physiological significance of autoreceptors.

Synaptic depression is a form of short-term synaptic plasticity that plays an important role in the information coding of neuronal networks (Markram & Tsodyks, 1996; Abbott et al. 1997; Tsodyks & Markram, 1997). In chick auditory brainstem short-term depression provides an adaptive mechanism compensating for sound volume within the interaural timing pathway (Cook et al. 2003). The fact that the rate of EPSC depression is uncorrupted by mGluR activation reinforces the idea that significant physiological information is contained and preserved on transmission across this synapse (presumably related to decoding of sound source location).

Reduction of the presynaptic calcium current by mGluR activation (Takahashi et al. 1996) will help counteract the depletion of extracellular calcium caused by synaptic transmission (Borst & Sakmann, 1999). In addition, a significant proportion of the energy used by the presynaptic terminal to maintain transmission is needed to remove internal calcium (23%; Attwell & Laughlin, 2001). Reducing the calcium influx will help maintain the presynaptic ATP supply. Finally, as release probability is reduced, the compensatory increase in vesicle pool size would ensure subsequent maintenance of higher frequency activity, thus sustaining the dynamic range of this relay in the auditory pathway.



This work was supported by the Wellcome Trust and the Royal Society.

Author's present address

A. Y. C. Wong: Department of Pharmacology and Therapeutics, McGill University, 3655 Promenade Sir-William Osler, Montreal, Quebec, Canada H3A 1Y6.