Synaptic Integration of Subquantal Neurotransmission by Colocalized G Protein-Coupled Receptors in Presynaptic Terminals

GABA_B and 5-HT_1B receptors inhibit synaptic transmission from CA1 to subiculum

5-HT_1B and GABA_B receptors have previously been shown to be coexpressed in CA1 axon presynaptic terminals (Boeijinga and Boddeke, 1996; Bonaventure et al., 1998; Lein et al., 2007; Hamid et al., 2014) where both receptors inhibit synaptic transmission, but by different mechanisms. We have previously shown that GABA_B receptors inhibit Ca²⁺ entry, while 5-HT_1B receptors cause Gβγ to act directly at the SNARE complex (Hamid et al., 2014). The amount of inhibition of Ca²⁺ entry by GABA_B receptors in that study was sufficient to account for the effect of the GABA_B receptor agonist baclofen on synaptic transmission. In contrast, the effect of the 5-HT_1B receptor agonist CP93129 was eliminated by treatment of the tissue with botulinum A toxin to remove the c-terminal 9 residues from SNAP-25; and in a further study (Zurawski et al., 2019), genetic manipulation of the c-terminal of SNAP-25 also prevented 5-HT_1B receptor modulation of neurotransmitter release. We also showed in those studies no loss of transmission between soma and axons. This was indicated by Ca²⁺ transients that were unaffected by 5-HT_1B receptors and reduced in amplitude, but not prevented by GABA_B receptors (Hamid et al., 2014).

To investigate the effects of GABA_B and 5-HT_1B receptors on neurotransmitter release, we whole-cell voltage-clamped subicular pyramidal neurons (Fig. 1A) and stimulated CA1 axons with glass-coated tungsten microelectrodes at low intensity to ensure focal stimulation (5-20 µA, 200 µs). Recorded events were of low amplitude (Fig. 1B): ∼10 times the unitary amplitude determined from spontaneous release in TTX shown in earlier work (Hamid et al., 2014), which ensured that responses were monosynaptic. To prevent polysynaptic firing by recurrent excitation of CA1 neurons, the tissue was cut at the CA1-subicular boundary. We recorded responses in bicuculline (5 μm) and 2-amino-5-phosphonopentanoic acid (AP5, 50 μm) to isolate AMPAR-mediated currents. We applied the selective 5-HT_1B receptor agonist CP93129. Effects of a saturating concentration (1 μm) and at a concentration approximately the EC₅₀ (50 nm) are shown (for the concentration response, see Fig. 1Ba,Da). CP93129 reduced the EPSC peak amplitude to 19.6 ± 3.9% at a saturating dose (1 μm, n = 8; EC₅₀ = 0.033 ± 0.04 μm). These EPSCs are also inhibited by the GABA_B receptor agonist baclofen to 7.2 ± 2.3% of control (100 μm baclofen, n = 5, EC₅₀ = 0.35 ± 0.18 μm) (Fig. 1Bb,Db).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Presynaptic 5-HT_1B and GABA_B receptors inhibit evoked transmission from CA1 pyramidal neurons. A, Subicular pyramidal neurons were whole-cell patch-clamped. CA1 pyramidal axons were stimulated with a monopolar, glass-coated tungsten microelectrode. The tissue was cut at the CA1/subiculum boundary to prevent recurrent excitation. Ba, Ten consecutive AMPAR-mediated EPSCs from a cell recorded in bicuculline (5 μm) and D-AP5 (50 μm) in control (black) and in CP93129 (blue). Means of responses in control (black) 50 nm (dark blue) and 1 μm CP93129 (blue) are shown below. Bb, Similar evoked EPSCs were recorded in control (black) and baclofen (10 μm) green and means of EPSCs in control and in baclofen (1 μm, dark green) and 10 μm (green). C, Synaptic responses recorded in control (Ca) after addition of 10 mm CaCl₂ (Cb, black) and with CP93129 (50 nm, purple). In raised extracellular Ca²⁺ (Cc), baclofen 1 μm inhibited the synaptic response as for controls. D, Concentration responses for (Da) CP93129 in control saline (blue circles) and in 10 mm CaCl₂ (purple squares) and (Db) baclofen in control saline (light green circles) and in 10 mm CaCl₂ (dark green square).

We have previously shown that 5-HT_1B receptors target Gβγ to SNARE complexes in CA1 terminals (Hamid et al., 2014), and Gβγ competes with Ca²⁺-synaptotagmin binding to SNARE complexes. In lamprey, this may confer Ca²⁺ dependency on inhibition (Yoon et al., 2007). To test for this in CA1 subicular synapses, we recorded EPSCs in HEPES buffered Ringer (to prevent Ca²⁺ precipitation). Increasing CaCl₂ concentrations (10 mm) enhanced the EPSC (Fig. 1Ca,Cb), but CP93129 now caused significantly less inhibition (Fig. 1Cb,Da; at 50 nm, paired Student's t test, t₍₁₅₎ = 6.00, p = 0.00001, n = 13 and n = 8; and at 400 nm t₍₄₎ = 2.46, p = 0.03, n = 10 and n = 4). The EC₅₀ of CP93129 increased from 33 ± 4 nm to 320 ± 19 nm. In contrast, baclofen (600 nm) inhibition was unaffected by high Ca²⁺ concentration (t₍₁₆₎ = 0.88, p = 0.38; Fig. 1Cc,Db). This latter result implies that 10 mm Ca²⁺ does not cause saturating Ca²⁺ entry with respect to neurotransmitter release, allowing reduced Ca²⁺ entry by baclofen to reduce P_r.

Paired-pulse ratios are modified by GABA_B receptors but not by 5-HT_1B

Colocalized GABA_B and 5-HT_1B receptors act at different targets, Ca²⁺ channels or the SNARE complex, respectively; therefore, their activation may evoke different effects on vesicle fusion. We measured paired-pulse ratios (PPRs) to determine whether these receptors cause inhibition by a change in P_r (Dobrunz and Stevens, 1997). As a control, we raised the extracellular Mg²⁺ concentration (Del Castillo and Katz, 1954) from 1 to 4 mm to reduce presynaptic Ca²⁺ entry and therefore P_r. In 4 mm Mg²⁺, the mean amplitude of the first EPSC was reduced (to 42 ± 3% of control n = 14; Fig. 2A), and the PPR (EPSC2/EPSC1) was increased by a factor of 1.64 ± 0.09, n = 14 (quantified in Fig. 2E, red) demonstrating a Ca²⁺-dependent reduction in P_r.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

P_r is unaffected by 5-HT_1B but reduced by GABA_B receptor activation. A, Paired-pulse stimuli applied to CA1 axons (50 ms intervals). Aa, Means of 10 EPSCs. Extracellular [Mg²⁺] was raised from 1 to 4 mm to reduce Ca²⁺ entry (red) and thus P_r. This reduced the mean amplitude of the first EPSC more than the second. The effect of high Mg²⁺ on the PPR is quantified to show scatter plots of PPRs in control and high Mg²⁺ for each cell and box plots of the same data (Ab). B, A similar experiment but with addition of the GABA_B receptor agonist, baclofen (600 nm). Ba, Mean EPSCs show an increase in PPR quantified in Bb as for Ab. C, A negative control using a glutamate receptor antagonist, kynurenate (200 μm; gray). This gave no change in PPR (Ca, quantified in Cb). D, Similar experiment but with the 5-HT_1B receptor agonist, CP93129 (50 nm; blue). This again gave no change in PPR (Da, quantified in Db). E, Changes in PPRs normalized to pretreatment controls shown as box plots (1.0 = no change, boxes 25% and 75% with means, whiskers to 10% and 90%) for all recordings in the conditions shown in A-D. F, The effect of baclofen concentration versus normalized PPR was plotted. This is concentration-dependent. G, The effect of CP93129 concentration on PPR was plotted. This was not concentration-dependent. ***p < 0.001.

We inhibited EPSCs with baclofen (600 nm, approximately half-maximal concentration; Fig. 2B), which reduced the first EPSC to 40 ± 6% and enhanced PPRs (by 1.62 ± 0.1, n = 10; quantified in Fig. 2E, green) demonstrating an effect of GABA_B receptors on P_r, consistent with our earlier work showing their effect on presynaptic Ca²⁺ entry (Hamid et al., 2014). The effect on PPR was concentration-dependent (Fig. 2F) with a similar concentration response to the effect on EPSC amplitude (Fig. 1Db).

As a negative control, AMPAR antagonism with half-maximal concentrations of the glutamate receptor antagonist kynurenate (200 μm) showed no effect on PPRs (Fig. 2C). The EPSC2/EPSC1 ratio was 1.13 ± 0.04 (n = 19) of control (quantified Fig. 2E, gray). This is expected for postsynaptic targets. CP93129 inhibits CA1-subicular EPSCs presynaptically (Zurawski et al., 2019). Nevertheless, we now show a half-maximal CP93129 (50 nm) concentration, which reduced the first EPSC to 54 ± 4% of control, did not alter PPRs (PPR in CP93129 was 1.13 ± 0.07 of control; Fig. 2D,E, blue; n = 14). This lack of effect of CP93129 on PPRs was consistent over the full concentration response range of CP93129 on EPSCs (Fig. 2G).

A one-way ANOVA to test effects of raised Mg²⁺ concentration, of baclofen, kynurenate, and of CP93129 on PPRs (data from Fig. 2E), showed significance (ANOVA single factor F_(3,53) = 16.85, p = 8 × 10⁻⁸). Post hoc Tukey HSD analysis revealed no significant difference between effects of raised Mg²⁺ and baclofen (p > 0.1) and none between effects of kynurenate and CP93129 (p > 0.1). However, the effect of CP93129 was significantly different from lowering P_r with raised Mg²⁺ (p < 0.001) or baclofen (p < 0.001). Thus, effects of 5-HT_1B receptors are paradoxical, they are presynaptic but do not alter P_r, whereas GABA_B receptors that are also presynaptic do alter P_r.

5-HT_1B but not GABA_B receptors enhance low-affinity block of AMPA EPSCs

We investigated whether presynaptic receptors alter synaptic cleft glutamate concentrations at postsynaptic AMPARs. Partial block of EPSCs by a low-affinity antagonist is caused by antagonist binding and unbinding. If the antagonist unbinds within the time course of glutamate in the synapse, antagonist bound receptors will then become available for either glutamate or antagonist binding. Therefore, fractional low-affinity antagonism of the AMPAR EPSC will depend on the glutamate concentration (Clements et al., 1992). EPSCs evoked by low glutamate cleft concentrations will be more completely antagonized than those at high concentrations.

To first determine any effect of a reduction in P_r on cleft glutamate concentration, we recorded AMPAR EPSCs. The effect of lowering P_r with high Mg²⁺ was tested against the efficacy of low-affinity kynurenate antagonism. We compared kynurenate efficacy in control (1 mm) and high (4 mm) Mg²⁺ concentrations in the same cell and stimulus intensity because it is not clear that the mean control cleft glutamate concentrations were identical from cell to cell. After obtaining baseline EPSCs, we superfused the slice with kynurenate (200 μm) to obtain a partial block of the EPSC (in the example to 70% of control; Fig. 3Aa, gray). Kynurenate was removed to remeasure control EPSCs. The Mg²⁺ concentration was then increased to 4 mm to reduce P_r (EPSC reduced to 69% of control, Fig. 3Aa, red). Kynurenate (200 μm) was reapplied in 4 mm Mg²⁺ to measure its efficacy at a lower P_r. The EPSC in 4 mm Mg²⁺ was reduced by the same proportion as in control (to 67% of response in 4 mm Mg²⁺; Fig. 3Aa, purple). To compare inhibition by kynurenate in control and 4 mm Mg²⁺ the responses in 4 mm Mg²⁺ alone and combined with kynurenate were scaled to control amplitudes (Fig. 3Ab) showing similar proportionate inhibition. In repeat experiments, the drug sequence was randomized. For all cells, in 1 mm Mg²⁺, kynurenate (200 μm) reduced EPSCs to 44 ± 4% of control; and in 4 mm Mg²⁺, kynurenate similarly reduced the EPSC to a not significantly different 43 ± 4% (n = 10; Fig. 3Ac, purple; paired Student's t test, t₍₉₎ = 0.39, p = 0.7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

5-HT_1B but not GABA_B receptors increase the potency of a low-affinity AMPAR antagonist-mediated block of EPSCs. A, Effect of raised Mg²⁺ on EPSC inhibition by a low-affinity competitive antagonist, kynurenate. Aa, Kynurenate (200 μm; gray) partially inhibited EPSCs, shown as means of 10 traces. After kynurenate wash, 4 mm Mg²⁺ (red), to decrease P_r reduced EPSCs. Kynurenate 200 μm was then combined with high [Mg²⁺] (purple). Ab, For clarity, traces in 4 mm Mg²⁺ scaled to control. Ac, Kynurenate inhibited EPSCs to a similar ratio in control and 4 mm Mg²⁺. Inhibition ratios plotted and overlaid on boxplots (boxes 25% and 75% with means, whiskers to 10% and 90%). Ba, Kynurenate (200 μm; gray) reduced EPSCs. After kynurenate wash, baclofen (600 nm; light green) reduced the EPSC. Kynurenate (200 μm) with baclofen (600 nm; dark green) reduced the EPSC similarly to baclofen alone. Bb, Reponses in baclofen scaled as for Ab. Bc, Data from all cells. Ca, Kynurenate (200 μm; gray) reduced EPSCs. After wash, CP93129 (50 nm; light blue) reduced the EPSC. Kynurenate (200 μm) with 50 nm CP93129 (blue) reduced the EPSC to a greater proportion than CP93129 alone. Cb, Reponses in CP93129 scaled as for Ab. Cc, Data from all cells. D, Noncompetitive AMPAR antagonist GYKI used as control. Da, GYKI (10 μm; orange) reduced EPSCs. After wash, CP93129 (50 nm) inhibited EPSCs (blue). Coapplication of 10 μm GYKI with 50 nm CP93129 (pink) reduced EPSCs as controls. Db, Responses in CP93129 scaled as above. Dc, Data from all cells. E, Box plots overlaid with individual cell data comparing ratio of inhibition by kynurenate or GYKI alone, to inhibition in kynurenate plus high Mg²⁺ (pink), baclofen (green), or CP93129 (blue) or to GYKI and CP93129 (brown). A value of 1 indicates the modulatory agonist had no effect on kynurenate or GYKI potency; higher values show enhanced potency. F, Change in kynurenate inhibition caused by 4 mm Mg²⁺ (data from A) plotted against initial efficacy of 4 mm Mg²⁺ for each neuron. The slope was indistinguishable from zero. G, Similar effect by baclofen on kynurenate inhibition (B) plotted against initial efficacy of baclofen in each neuron. Slope was again indistinguishable from zero. H, Change in kynurenate inhibition by CP93129 (C) plotted against the initial efficacy of CP93129 in each neuron. Slope (−2.3) showed correlation (R² = 0.4). *p < 0.05. ***p < 0.001.

We then assayed the efficacy of kynurenate inhibition before and in baclofen which will also alter P_r. In the example, kynurenate (200 μm) reduced the mean EPSC amplitude to 62% of control (Fig. 3Ba, gray). In baclofen (600 nm; Fig. 3Ba, green), the inhibited EPSC was reprobed with kynurenate (200 μm), which now reduced the EPSC to 61% of its amplitude in baclofen alone (Fig. 3Ba, dark green). Responses were again scaled to compare the effect of kynurenate in controls and in baclofen (Fig. 3Bb). In 19 experiments, kynurenate inhibited control responses to 46 ± 3% and the inhibitory effect was slightly but significantly less in baclofen, inhibiting to 51 ± 3% (Fig. 3Bc; green; paired Student's t test, t₍₁₈₎ = 2.19, p = 0.04).

We similarly assayed the efficacy of kynurenate block before and in CP93129, which did not alter P_r. In the example, kynurenate (200 μm) reduced the peak EPSC amplitude to 55% of control (Fig. 3Ca, gray). After removal of kynurenate, EPSC amplitudes were recorded to reconfirm control responses. CP93129 (50 nm) was applied (Fig. 3Ca, light blue), and the CP93129 inhibited EPSC was reprobed with kynurenate (200 μm), which now reduced the EPSC to 38% of its amplitude in CP93129 alone (Fig. 3Ca, blue). Responses were again scaled to directly compare the effect of kynurenate in controls and in CP93129 showing an enhancement in inhibition (Fig. 3Cb). In 17 experiments, kynurenate inhibited control responses to 57 ± 3% and responses in CP93129 to 37 ± 2%. This latter inhibitory effect was significantly greater than the effect of kynurenate alone (Fig. 3Cc; paired Student's t test, t₍₁₆₎ = 5.56, p = 2.1 × 10⁻⁵).

As a negative control for the effect of CP93129 in kynurenate, we replaced the competitive antagonist (kynurenate) with a noncompetitive AMPAR antagonist, GYKI52466 (GYKI). GYKI should inhibit EPSCs similarly, regardless of glutamate concentration because it does not compete with glutamate (Fig. 3D). Mean inhibition by GYKI in control was to 63 ± 4%, and the efficacy of GYKI was slightly but significantly reduced in CP93129 (n = 6, to 68 ± 4% of response in CP93129, paired Student's t test, t₍₅₎ = 3.85, p = 0.011). This result is small but may imply a slight bias in analysis recordings of EPSC amplitudes as their amplitude is reduced to very small values by combined ligands. Recording noise may contribute slightly to the computed result because the EPSC measurement algorithm searched for a smoothed positive maximum. Thus, in very small or results with a failure, there may be a small positive bias in the result. The result may indicate that we have slightly underestimated the effect of 5-HT_1B receptors on cleft glutamate concentrations, but does not change the conclusion. This bias may also underlie the slight but not significant relative reduction in effect of kynurenate seen in baclofen (Fig. 3Bc).

The effects of high Mg²⁺, baclofen, and CP93129 on the efficacy of kynurenate and the effect of CP93129 on the efficacy of GYKI on the inhibition of the synaptic responses were compared with one another by comparing the ratios of effects of antagonist (kynurenate or GYKI) with the same antagonist combined with either 4 mm Mg²⁺, with baclofen, or with CP93129 (Fig. 3E). All four experiments were compared with a one-way ANOVA showing substantial significance (F_(3,48) = 15.88, p = 1.4 × 10⁻⁶). Post hoc Tukey's HSD analysis revealed that the effect of baclofen was not significantly different from the effect of high Mg²⁺ on the efficacy of kynurenate inhibition (p > 0.1). This is consistent with inhibition mediated by a change in P_r of univesicular events at each synapse. In contrast, the effect of CP93129 on kynurenate-mediated inhibition was significantly greater than that of Mg²⁺ (p < 0.005), baclofen (p < 0.001), and of the effect of CP93129 on GYKI inhibition (p < 0.005).

Single raised concentrations of Mg²⁺ (4 mm) or single concentrations of baclofen (600 nm), or CP93129 (50 nm) show varying degrees of inhibition from cell to cell. For example, initial effects of 50 nm CP93129 reduced the EPSC from as little as 0.8 of control to as much as 0.3. If this variance reflects the efficacy of the agonist on cleft glutamate concentration, it will correlate with amplification of agonist inhibition by kynurenate (shown in Fig. 3E). To test this, we compared the efficacy of the agonist alone in each neuron (agonist/control), to the change in agonist efficacy by kynurenate [(kynurenate/control)/(agonist/kynurenate + agonist)]. We first plotted these for effects of high Mg²⁺ (Fig. 3F). There was no correlation between inhibition by 4 mm Mg²⁺ with the subsequent effect of Mg²⁺ on kynurenate inhibition. This is consistent with a Mg²⁺-induced change in P_r at univesicular synapses causing no change in cleft glutamate concentration regardless of the initial efficacy of raised Mg²⁺. Similarly, there is no correlation between the efficacy of baclofen and its effect on subsequent kynurenate block (Fig. 3G). However, there is a strong correlation between the CP93129 efficacy and its amplification of kynurenate inhibition (Fig. 3H), indicating that the greater the effect of presynaptic 5-HT_1B receptors, the lower it drives cleft glutamate concentration and therefore the larger inhibitory effect of kynurenate. We conclude that at CA1 presynaptic terminals changes in P_r do not alter cleft glutamate concentrations, but 5-HT_1B receptors do.

It was important to determine that synaptic responses were not impacted by changes in whole-cell recording access during the experiments. In all experiments, we monitored whole-cell access by applying a small (5 mV, 20 ms) test pulse before each evoked EPSC was recorded. An example is shown of this recording throughout a sequential application of kynurenate, and CP93129; and then the drugs combined simultaneously measuring effects on synaptic responses (Fig. 4A,B), while simultaneously monitoring whole-cell series resistance calculated from double exponentials fitted to the settling of the step current (Fig. 4C). Experiments were terminated and data not used if access resistance changed by >20%. For the examples used for analysis (Fig. 4D), no significant difference was seen for access impedance (ANOVA single factor F_(3,64) = 0.065, p = 0.97) or holding current (F_(3,64) = 0.18, p = 0.91) when comparing between periods of ligand applications.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Stability of recording access. A, Sequential recordings of evoked AMPAR-mediated EPSCs during recording through each drug condition to show recording stability. Individual EPSCs are shown for one recording in which the effects of CP93129 and its amplification by kynurenate were tested. B, Graph of peak amplitudes of all EPSCs recorded. Closed circles represent measured responses corresponding to the EPSCs shown in A. Open circles during wash in or washout of the drugs were not measured. Colors correspond to EPSCs in A. C, Graphs represent whole-cell access impedance (black), and holding current (red) throughout the experiment. Whole-cell access impedances (R_a) were calculated from double exponentials fit to the decays from small depolarizing current steps applied before each EPSC was stimulated (Sigworth, 1983). Example responses to voltage steps at each drug application are shown. Color the same as for A. D, From all cells, histograms represent mean ± SEM for access impedance (positive bars) and holding currents (negative bars) during control, application of kynurenate, CP93129, and kynurenate + CP93129. Colors during ligand applications are the same throughout the figure.

5-HT_1B but not GABA_B receptors differentially inhibit AMPA and NMDA EPSCs

Glutamate activates receptors with varying sensitivities: for example, AMPA, NMDA, and metabotropic glutamate receptors have different affinities for glutamate and different numbers of glutamate binding sites. Consequently, changes in cleft glutamate concentration may differentially alter their activation (Choi et al., 2003; Schwartz et al., 2007; Gerachshenko et al., 2009). We compared 5-HT_1B and GABA_B receptor inhibition on AMPAR- and NMDAR-mediated synaptic responses. AMPAR-mediated EPSCs were recorded as before. We recorded NMDAR-mediated synaptic responses in bicuculline (5 μm), and NBQX (5 μm) at a holding potential of −35 mV. Baclofen (600 nm) equally inhibited AMPA and NMDA components of EPSCs (Fig. 5A; to 49 ± 6%, n = 8, and to 54 ± 4%, n = 6, respectively; t₍₁₂₎ = 0.80, p = 0.44). CP93129 (50 nm) inhibited AMPA EPSCs significantly more than NMDA EPSCs (Fig. 5B; to 54 ± 6%, n = 10, and to 93 ± 4%, n = 6, respectively; t₍₁₄₎ = 7.15, p = 4.9 × 10⁻⁶). Concentration response curves of CP93129 inhibition of AMPA and NMDA EPSCs diverged (Fig. 5C). AMPAR-mediated EPSCs were maximally inhibited to 20 ± 3% of control, whereas NMDAR-mediated EPSCs were significantly less inhibited to 75 ± 6% (paired Student's t test, t₍₇₎ = 6.48, p = 0.00017). We obtained similar results when we recorded NMDA and AMPA EPSCs simultaneously from the same synapses. AMPAR- and NMDAR-mediated responses were recorded simultaneously by holding the membrane potential at positive values and recording events early in the response or later. We measured AMPAR EPSC amplitudes within 2 ms of stimulation, whereas NMDA responses were measured during the decay after 100 ms (Fig. 5D) because these receptors show much slower decay kinetics than AMPAR-mediated responses (Lester et al., 1990). In control conditions with no drugs, NMDAR EPSC amplitudes covary with those of AMPARs (Fig. 5E), indicating that the receptors are activated on the same synapses. In 6 neurons, we plotted the mean amplitudes of AMPAR- and NMDAR-mediated EPSCs with increasing doses of CP93129. AMPAR EPSCs were inhibited to a significantly greater extent than were NMDAR EPSCs at all applied concentrations of CP93129 (Fig. 5F). For the approximate half-maximal concentration (50 nm) t₍₅₎ = 5.61, p = 0.002) and for the maximal concentration (1000 nm) t₍₅₎ = 7.52, p = 0.0007).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Differential inhibition of AMPAR- and NMDAR-mediated EPSCs. A, In separate recordings AMPA-mediated EPSCs were recorded in D-AP5 (Aa) and NMDA-mediated EPSCs in NBQX (Ab). Baclofen (600 nm; green) similarly inhibited both of these responses. B, Similar recordings were made but CP93129 (Ba, 50 nm; blue) inhibited the AMPAR-mediated EPSC, but not the NMDAR-mediated EPSC (Bb, light blue). C, Concentration response effects of CP93129 versus the AMPAR (blue) and NMDAR EPSCs (light blue) show a large difference in efficacy. D, The differential effect of CP93129 on AMPAR and NMDAR responses is found at the same synapses. Compound EPSCs without glutamate receptor antagonists recorded at a holding potential of 50 mV. Early responses (2 ms after stimulus) are dominated by AMPARs (bottom arrow), late responses (100 ms after stimulus) by NMDARs. CP93129 at a saturating dose (1 μm) was applied (blue). E, For individual recordings, sequential response amplitudes of AMPA and NMDA responses were plotted against each other. These responses covaried, indicating that they colocalize to the same synapses. F, Mean amplitudes of AMPAR- and NMDAR-mediated EPSCs during the same stimuli plotted with increasing concentrations of CP93129. AMPAR EPSCs were inhibited to a greater extent than were NMDAR EPSCs. G, CP93129-mediated inhibition of the NMDA component is amplified by the low-affinity competitive antagonist (L-AP5). In control experiments (Ga), 10 and 50 nm CP93129 (blue) had little effect on isolated NMDA-mediated EPSC (recorded at 50 mV in NBQX and bicuculline). L-AP5 (250 μm) partially inhibited this response (Gb, black). This inhibited response was now substantially inhibited by 10 and 50 nm CP94129 (red and pink). H, Quantitation of the data from G, showing inhibition by CP93129 in control (blue, normalized to precontrol recordings) and after treatment with L-AP5 (pink, normalized to recordings in L-AP5 alone.

If the lack of effect of CP93129 on NMDA EPSCs is because it causes reduced cleft glutamate concentrations, then an effect of CP93129 would be revealed by low-affinity NMDAR antagonists that will amplify effects of changing glutamate concentrations at NMDARs (Choi et al., 2000). We recorded evoked pharmacologically isolated NMDAR EPSCs (n = 6) at 50 mV in bicuculline (5 μm) and NBQX (5 μm). Low doses of CP93129 (10 and 50 nm) did not significantly reduce the EPSC in 6 neurons (ANOVA single factor F_(2,12) = 3.80, p = 0.052; Fig. 5Ga,H). In a further 6 neurons, the same experiment was performed in the low-affinity competitive NMDAR antagonist L-5-amino pentanoic acid (L-AP5, 250 μm). The EPSC was reduced to 55 ± 4% of control by L-AP5 (Fig. 5Gb,H). In L-AP5, CP93129 at both 10 and 50 nm CP93129 now inhibited these partially blocked EPSCs to 66 ± 7 and 45 ± 4% of the response in L-AP5 alone. These reductions were significant (ANOVA single factor F_(2,12) = 7.22, p = 0.009). Thus, NMDARs are exposed to a similar 5-HT_1B receptor-mediated reduction in glutamate concentration to that seen by AMPARs.

The effect of 5-HT_1B receptors on glutamate release

It is possible to image glutamate release using genetically engineered glutamate sensors. To achieve this, we virally infected subicular neurons with iGluSnFR using an AAV1 containing iGluSnFR.WPRE.SV40 under a human synapsin promotor (pAAV.hSyn.iGluSnFr.WPRE.SV40 (AAV5) (Marvin et al., 2013) by injecting the viral vector into hippocampi of 22 d rats. In hippocampal slices from these animals, epifluorescence imaging (excitation 470 nm, emission, 520 nm) revealed fluorescent subicular pyramidal neurons. The slices were placed in a custom-designed imaging chamber and a stimulation electrode placed as for electrophysiological recording. The subicular region was then imaged using LLSM (Fig. 6A). The light sheet penetrated to ∼50 µm in the slice, and time-series images of single planes (512 × 512 pixels; 52 × 52 µm) were captured with a single exposure time of 10 ms and a frame rate of 80 Hz. Stimuli were applied to CA1 axons at 5 s intervals to evoke events that appeared stochastically located, varying from stimulus to stimulus across the FOV (Fig. 6B). Repeated stimuli revealed multiple locations of release that were resolved over a sequence of up to 21 individual stimuli. Events were well resolved using LLSM imaging (Fig. 6C,D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Imaging evoked glutamate release with iGluSnFR. A, Imaging of iGluSnFr expressed in subicular neurons was performed under LLSM adapted to allow recording and stimulation of hippocampal slices. The schematic shows the arrangement of preparation and lenses. B, While imaging labeled subicular pyramidal neurons dendrites, repeated stimulation (1 shock at 5 s intervals) caused stochastic fluorescent transients in subiculum. Shown are 3 responses from 2 separate stimuli. C, Evoked transients recorded in one z plane at imaged at 80 Hz expanded from a spot to larger diameter (5-10 µm diameter) objects (time point noted in D, red). D, Transient amplitudes and time courses were calculated from a 5-µm-diameter ROI over the transients. Shown are three control responses at the same location, obtained from 14 stimuli (black). Application of CP93129 (100 nm) reduced the amplitude of the transients that were recorded. Two such responses were recorded in the 14 stimuli in this case (blue). Examples of a failure to evoke a response in control conditions (black) and similarly in CP93129 (100 nm) (blue). E, Example of evoked transient (same location as in C, recorded in CP93129, 100 nm). Fa, CP93129 (100 nm, blue) caused no change in the probability of recording events at the same location over 21 stimuli. In contrast, baclofen (1 μm, green) significantly reduced the probability of recording and event. A two-way ANOVA revealed significance between effects of baclofen and CP93129 (F_(1,16) = 16.1, p = 0.001). Events were determined to be failures if the amplitude after stimulation at that location did not exceed noise. Fb, During paired pulses, overall probability of recording events rose. CP93129 (100 nm, blue) again caused no reduction in probability, whereas baclofen (1 μm, green) did. A two-way ANOVA revealed significance between effects of baclofen and CP93129 (F_(1,32) = 7.12, p = 0.011). G, All evoked event amplitudes were measured and displayed as event/frequency histograms. Under baseline conditions, events were distributed into distinct amplitudes (gray). CP93129 (100 nm, blue) removed this distinction. Top, Histogram example. Bottom, Mean results from five locations normalized to the first baseline event amplitude distinct from noise. Gaussian curves were fitted to baseline data with the constraint that the amplitude and width of non-zero peaks scaled with the number of peaks. This showed a difference in control data between no response and one quantal size. Gaussian fit to the data in CP93129 was not possible. H, Similar experiment with baclofen. Top, Histogram example. Bottom, Average of 5 locations, before and after baclofen (1 μm, green). Baclofen left recorded amplitude of events unaffected but reduced their probability. Mean data in both control (black) and baclofen (green) were fitted with Gaussians with the same constraints as in G. Gaussian fits showing similar quantal peak amplitudes in both conditions. I, Point spread functions (PSFs) of the microscope obtained from a 100-nm-diameter fluorescent bead. Ia, x/y PSF. Ib, x/z PSF of the same bead. The bead was embedded 100 µm deep in 1% agarose. *p < 0.05. ***p < 0.001.

We tested the effect of 5-HT_1B receptor activation on these evoked fluorescence transients by recording events at single sites during repeated stimuli before and after application of CP93129 (100 nm). Frequency of events was calculated by whether an event was recorded on stimulation or not. The probability of recording an event at the same location as in control was unchanged in CP93129 0.20 ± 0.02 versus 0.19 ± 0.03 in control (Fig. 6F; t₍₁₂₎ = 0.17, p = 0.87). Similar recordings were made before and during baclofen (1 μm) application. In these recordings, the probability of recording events was reduced significantly from 0.29 ± 0.07 in control to 0.08 ± 0.05 in baclofen (t₍₈₎ = 4.24, p = 0.0014).

We also measured effects of agonist application on event amplitude. However, event frequencies on single stimuli were low. To increase event frequency, stimuli were applied as paired pulses at 100 ms intervals, and all poststimulation peak fluorescence transient amplitudes were determined as ΔF/F calculated from the ratio of peak response to amplitude immediately before the stimulus after background subtraction of fluorescence adjacent to the ROI. Data were captured for each of every 24 stimuli at 5 release locations located in 3 slices for application of CP93129 and for 5 locations in 4 slices for baclofen. Event amplitudes were plotted as frequency histograms. An example is given for baseline and in CP93129 (Fig. 6G, top). In baseline conditions, events largely fell into either failures, with mean signal similar to background noise, or into a signal centered on a recurring amplitude (quantum). Some larger events were also recorded. CP93129 (100 nm, blue) caused this dual amplitude distribution to be lost. Mean effects of CP93129 were compared for all 5 locations with amplitudes normalized to the mean value of a single quantum in baseline conditions (Fig. 6G, bottom). The distinction between failures and a single quantum value was lost in CP93129, but there was no significant change in the number of observed failures (t₍₄₎ = 1.2, p = 0.15). This is emphasized by plotting a multiple Gaussian curve to the baseline data with the constraints that sequential quantal events amplitudes and widths increased as integer multiples of the quantal amplitude.

Similar experiments were performed in baseline, then baclofen (Fig. 6H, green). In contrast, baclofen (1 μm) significantly increased the incidence of failures (t₍₄₎ = 5.28, p = 0.003), but remaining event amplitudes fell into the same amplitude bins as baseline responses. This is emphasized by multiple Gaussian fits to baseline results and from events in baclofen in which the single event amplitude calculated from these fits was not significantly altered (t₍₄₎ = 1.7, p = 0.16), whereas both the number of failures (t₍₄₎ = 2.9, p = 0.04) and the number of single quantum events (t₍₄₎ = 4.0, p = 0.02) were. We also measured the decay kinetics of the responses in baseline and in both agonists, but no changes were measured. This is likely because of the relatively high affinities of the iGluSnFr variant used such that unbinding of glutamate from the sensor dominates the signal off rate. Indeed, the size of the events compared with the point spread function of the microscope (Fig. 6I), which demonstrate submicron resolution, indicates that we detect glutamate spillover from the synapse as a large part of the signal.

Simulating glutamate release and receptor activation

GABA_B receptors reduce P_r, whereas 5-HT_1B receptors lower cleft glutamate concentrations. The latter may be because of a modification of the fusion pore during neurotransmitter release (Harata et al., 2006; Photowala et al., 2006), changes in MVR (Rudolph et al., 2015), or modulation of the location of release with respect to postsynaptic receptors (Tang et al., 2016; H. Chen et al., 2018; Haas et al., 2018). We created a Monte Carlo simulation of a simple synapse using MCell to probe these possibilities.

The simulated synapse comprised a disk-shaped synaptic cleft, 300 nm in diameter and 20 nm deep, with kinetically modeled AMPARs (Robert and Howe, 2003) and NMDARs (Banke and Traynelis, 2003) placed on the postsynaptic surface (Fig. 7). For each condition, 20 random seeds were simulated, whereby glutamate release was modeled from between one and three vesicles attached to the presynaptic surface with a dynamically opening pore (Movie 1), initially with a 0.4 nm diameter and the length of two lipid bilayers. This is approximately the value proposed for the conductance of foot events in large dense core vesicles (Vardjan et al., 2007). The simulated pore then expanded to the full diameter of the vesicle over a range of time courses from 0.2 to 20 ms. We tested three hypotheses: (1) glutamate release was slowed through a fusion pore; (2) multivesicular vesicle numbers were reduced; and (3) relative locations of release sites and receptors were altered. The parameter files for these simulations are available on our website (https://alford.lab.uic.edu/GPCRs.html). The individual parameters and sources are listed in Table 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Simulating neurotransmitter release and receptor activation. Aa, Simulated 300 nm diameter synapses, 20 nm deep synaptic clefts, and a 1.7 nm path at the edge to restrict fusion. Image is from 500 µs after the start of fusion. To simulate a fusion pore, fusion started from a pore at fixed diameter (0.4 nm for 20 ms) before fully opening (Aa) or pores that fully opened within 200 µs. Simulated receptors were randomly seeded on the postsynaptic surface (receptor colors as for kinetic models, D). Ab, Varying initial fixed pore diameters slowed glutamate release, reduced its cleft concentration (graphed in Ae), and differentially inhibited AMPA (Ac) over NMDA (Ad) receptor-mediated responses (graphed in Af; filled circles represent AMPA; hollow squares represent NMDA). Ba, Simulated MVR pattern. Three simulated vesicles where vesicle number was varied from 3 to 1 with 333-5000 glutamates/vesicle. Bb, The effect on cleft glutamate concentration of reducing vesicle numbers from 3 to 1 when each contained either 1000 glutamates (blue) or 700 glutamates (black). The AMPAR response (Bc) and the NMDAR responses (Bd) are shown below. Graphs represent the AMPAR (Be) and the NMDAR responses (Bf). To each condition, both varying glutamates/vesicle (color-coded) and number of vesicles each point in the colored sections. Ca, The effect of varying the position of the release site with respect to postsynaptic clustered receptors was simulated. Two receptor clusters were seeded with a mean separation of 150 nm and the releasing vesicle positioned over one. Cb, Cleft glutamate concentrations over the clusters (red at release site, blue away from release) and mean cleft concentration (black) after 5000 glutamates were released in 200 µs. Cc, AMPAR responses simulated under the release site (red) and away from the site (blue). Cd, NMDAR responses. Cf-Ch, Similar experiment to Cb-Cd, but with 1500 glutamate molecules to reveal modulation of AMPAR- and NMDAR-mediated responses. Da, Kinetic model for AMPARs and (Db) for NMDARs (parameters in Table 1). Colors are the same as shown for receptors in Aa, Ba, and Ca.

To simulate a slowly opening fusion pore, the vesicle (containing 5000 glutamates) was attached to the presynaptic membrane through a pore (Fig. 7A). The simulated pore expanded to the diameter of the vesicle, but arrested at a diameter of 0.4, 0.5, 0.75, 1.0, 1.5, or 2 nm for 20 ms. Receptors were randomly seeded on the postsynaptic membrane, and their states represented by colors in kinetic models (Fig. 7D). Arresting the fusion pore opening at set diameters slowed release and reduced peak cleft glutamate concentrations (Fig. 7Ab) compared with full expansion of the pore to the diameter of the vesicle within 200 µs, which represents the default state. This slower glutamate release reduced simulated AMPAR, but not NMDAR, responses (Fig. 7Ac,Ad) (Movie 2). Both the peak simulated glutamate concentration and peak AMPAR activation were reduced with smaller pore sizes, whereas NMDAR responses remained unchanged (Fig. 7Ae,Af).

MVR was simulated by varying numbers of simulated fusing vesicles from 1 to 3 (Fig. 7B). One vesicle was placed at the center and two offset by 75 nm (Fig. 7Ba). Fusion was simulated by pores fully opening in 200 µs. Vesicle content varied from 333 to 5000 glutamates because at high glutamate content the AMPAR response saturated with more than one vesicle, preventing any modulation. Varying vesicle numbers altered cleft glutamate concentration and modulated AMPAR and NMDAR responses nearly equally (Fig. 7Bb-Bd). The effect on both NMDARs and AMPARs was modulated by the total number of released glutamates rather than vesicle number (Fig. 7Be-Bf).

To vary vesicle fusion position compared with receptors, receptors were seeded within two proscribed areas of the postsynaptic density: one under the fusing vesicle and the other a mean distance of 150 nm away (Fig. 7Ca). Release of 5000 glutamates was simulated with complete fusion in 200 µs causing peak glutamate concentrations that differed at the two sites. Little modulation of AMPAR or NMDAR responses occurred (Fig. 7Cc,Cd). Earlier work simulating vesicles displaced from the release site reported larger effects with fewer released glutamates (Haas et al., 2018). Consequently, the simulation was repeated with 1500 glutamates (Fig. 7Cf-Ch). The more distant clusters of AMPAR and NMDAR responses were reduced to 80% and 83% of the amplitude under the release site, respectively. Thus, the response is sensitive to relative locations of vesicle fusion and receptor location, but this effect is small compared with that of experimental 5-HT_1B or GABA_B receptor activation and does not recapitulate the difference in effect between AMPAR and NMDAR responses. Overall, changing fusion pore dilation best mimicked the physiological effect of 5-HT_1B receptor activation.

Simulating the amplification of inhibition by kynurenate

Experimental effects of the low-affinity antagonist, kynurenate, on AMPAR EPSCs indicate that 5-HT_1B receptors, but not GABA_B receptors, lower the concentration of glutamate in the synaptic cleft. To determine whether kynurenate might discriminate between mechanisms by which cleft glutamate concentration is changed based on receptor effects, we simulated the action of kynurenate using a modification of the kinetic model of the AMPAR.

The kinetic model assumed that kynurenate can substitute for glutamate at all states of the receptor but prevented receptor channel opening (Fig. 8Aa). Simulations were performed with kynurenate (240 μm) present in the synaptic cleft. This reduced the simulated AMPAR-mediated response to 55% of control (Fig. 8Ab). The simulations were then repeated with the synaptic vesicle fusion pore diameter arrested for 20 ms at diameters from 0.4 to 1.5 nm. The effect of kynurenate was enhanced when the pore size was restricted (Fig. 8B). The effect of simulated kynurenate was also tested against MVR. Modifying the number of fused vesicles was only effective at a reduced number of glutamates present in each vesicle; thus, these simulations were performed from with three, two, or one vesicles each containing 700 glutamates as for Figure 7Bb-Bd. Kynurenate also enhanced the effect of reduced vesicle number on the simulated AMPAR response (Fig. 8C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Simulating cleft glutamate modulation and kynurenate inhibition. Aa, Kinetic model of AMPAR activation by glutamate and block by kynurenate. Colors and lowest tier of 3D model are as for Figure 7Da. Vertical elements of the 3D representation show kynurenate occupying receptor bound and desensitized states (parameters in Table 1). Ab, Simulated full fusion in 200 µs with 5000 glutamates (black, control) and the effect of 240 μm kynurenate (gray). Ba, The effect of kynurenate (gray) on control simulated responses (black) and on the response when the vesicle fusion pore was restricted to 0.4 nm (light purple without, purple with kynurenate). Bb, Full fusion responses and 0.4 nm pore responses scaled to emphasize effect of kynurenate is amplified by the 0.4 nm pore. Ca, Simulation as numbers of vesicle were varied from 3 to 1, each with 700 glutamates as in Fig. 7Bb-Bd. Effect of kynurenate (gray) on 3 vesicle response (black), and on 1 vesicle response (light orange) plus kynurenate (orange). Cb, Responses scaled to compare effect of kynurenate on the 3 vesicle and 1 vesicle responses. D, Comparison of magnification of kynurenate inhibition by baclofen (green) and CP93129 (blue) from experimental data (Fig. 3E) to the effect on kynurenate inhibition in simulations of reducing vesicle numbers (orange) or changing the initial fixed vesicle pore diameter (purple). E, Simulation data (from D) plotted against the effect of reducing vesicle number (orange) or fusion pore size (purple) on AMPAR responses. Green line indicates the fit to the similar effect of kynurenate on variations in experimental baclofen inhibition (from Fig. 3G). Blue line similarly indicates experimental CP93129 variation (from Fig. 3H).

Effects of both these manipulations were compared with the experimental effects of GABA_B and 5-HT_1B receptor activation on AMPA-mediated synaptic responses. Box plots (from Fig. 3E) showing CP93129 and baclofen effects on kynurenate-mediated antagonism are presented alongside simulated kynurenate inhibition following a change in number of fusing vesicles (normalized to the maximum, 3) and following arrest of the fusion pore at between 0.4 and 1.5 nm diameters normalized to full fusion (defined as complete opening of the fusion pore the vesicle diameter in 200 µs). Both of these manipulations enhanced the efficacy of kynurenate (Fig. 8D).

As for the experimental effects of agonists (Fig. 3E,H), simulations of variation in inhibition of the AMPA response, either by varying the number of multivesicular fusing vesicles at one active zone (Fig. 8D, orange) or varying the vesicle pore diameter (Fig. 8D, purple) correlated with the enhancing effect of kynurenate on both manipulations (Fig. 8E). Both approaches recapitulated experimental effects of CP93129 but not of baclofen (Fig. 8E). From this combination of experimental effects of low-affinity antagonist, and its simulation, the experimental data can be explained by a reduction synaptic cleft glutamate. Results with kynurenate alone, however, do not discriminate between the effect on release, whether change in fusion pore diameter or change in fusing vesicle numbers that causes this change in concentration.

Modulation of the time course of the AMPAR-mediated response

Simulating the slowing of the fusion pore opening slowed the rise and decay times of the simulated AMPAR response (Fig. 7Ac) demonstrated by replotting the inhibited response scaled to the peak of the control (Fig. 9A). In simulated data, the arrested pore diameter (0.4 nm) increased the time course (τ) of the rise from 0.09 to 0.35 ms and of the decay from 0.97 to 1.59 ms measured by single exponentials fitted to the rise and decay. Simulated kinetic properties of AMPAR opening are more rapid than those experimentally obtained. This may be because they are neither affected by dendritic filtering nor by asynchronous vesicle fusion. Nevertheless, we investigated effects of agonist on the kinetics of synaptic responses.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Effect of agonist on AMPA-mediated EPSC kinetics. A, Data from simulating full fusion and a 0.4 nm pore scaled to the same amplitude to show the slowing of the rise and decay times of the resultant AMPAR-mediated response. Ba, Experimental data of control responses (mean of 10 sequential EPSCs, black) and responses in the same neuron in CP93129 (50 nm, blue). Red curves are single exponentials fitted to 10%-90% of the rise and of the decay phases. Bb, Response amplitudes scaled to emphasize the change in rates of the response in CP93129. Bc, Box plots and data points for the rise time in control (gray) and in CP93129 (blue). Bd, For decay times. C, Simulated AMPA-mediated EPSCs scaled to show the kinetic effect of reduced number of responses to mimic a change of P_r at univesicular synapses on simulated kinetics (right) or reduced number of vesicles (from 3 to 1) at multivesicular synapses (left). Da, Experimental control responses (black) and in the same neuron in baclofen (600 nm, green). Red curves indicate single exponentials fitted to rise and decay phases as in B. Db, Response amplitudes scaled as in B. Dc, Box plots and data points for the rise time in control (gray) and in baclofen (green). Dd, Similarly for decay times. **p < 0.01.

Responses from 11 neurons were recorded in which EPSCs showed monotonic decays (Fig. 9B). CP93129 (50 nm) increased τ's of rise and decay phases, visible when the response in CP93129 was scaled to the control amplitude (Fig. 9Bb). Single exponential fits to the 10%-90% rise times showed a significant increase from a mean of 1.4 to 2.0 ms (t₍₁₀₎ = 3.06, p = 0.006) as did the decay times from 7.3 to 9.4 ms (t₍₁₀₎ = 3.54, p = 0.003) (Fig. 9B).

Simulations of reduction of P_r between synapses is trivial, a reduction in numbers of sites; consequently, kinetic properties of the simulated EPSC are unchanged (Fig. 9C, left). Alternatively, a simulated reduction in P_r at a multivesicular synapse shows a slight decrease in decay rate (Fig. 9C, right). We compared this to effects of baclofen (1 μm), which reduces P_r, in 10 neurons, again in which the decay was monotonic. Baclofen had no significant effect on mean rise time (control = 1.56 ± 0.12 ms, baclofen = 1.44 ± 0.16 ms; t₍₉₎ = 0.83, p = 0.21) or decay (control = 7.92 ± 0.47 ms, baclofen = 8.11 ± 0.41 ms; t₍₉₎ = 0.59, p = 0.30) (Fig. 9D). These results emphasize that 5-HT_1B and GABA_B receptors on the same presynaptic terminals mediate inhibition by different mechanisms.

Repetitive stimulation and train-dependent responses

At the same terminals, GABA_B receptors inhibit Ca²⁺ entry, whereas 5-HT_1B receptors cause Gβγ to interfere with Ca²⁺-synaptotagmin SNARE complex interactions (Hamid et al., 2014). This implies selective effects of receptor activation, but also synergy following activation of both receptors (Zurawski et al., 2019). As presynaptic Ca²⁺ accumulates during stimulus trains (Yoon et al., 2007; Hamid et al., 2014) that induce long-term plasticity (O'Keefe, 1990), it may modify competition for interaction with the SNARE complex between Gβγ and synaptotagmin or other Ca²⁺ sensors, such as Doc2 (Groffen et al., 2006) or synaptotagmin 7 (Jackman et al., 2016; Jackman and Regehr, 2017).

GABA_B receptors inhibit train-dependent Ca²⁺ accumulation. To show this, we recorded Ca²⁺ transients in CA1 pyramidal neuron terminals by filling the axons with Ca²⁺-sensitive dyes. CA1 neurons were whole-cell recorded under current clamp with a red dye (Alexa-594) to determine the axon location and a Ca²⁺-sensitive dye (Fluo 5F) to record action potential evoked Ca²⁺ transients. We have previously shown that Fluo 5F allows reliable recording of Ca²⁺ signals (Hamid et al., 2014, 2019). Individual presynaptic varicosities in the subiculum were imaged using line scanning (500 Hz) and neurons stimulated with 2 ms depolarizing current pulse trains (5 stimuli at 20 ms intervals) to evoke action potentials (Fig. 10A, top). Line scanning resolved Ca²⁺ transients over accumulating Ca²⁺ during the stimulus train (Fig. 10Aa, bottom). Baclofen (600 nm) reduced the Ca²⁺ transient following each stimulus in the train (Fig. 10Ac, green squares). The last transient was reduced to 0.46 ± 0.07 of control (t₍₃₎ = 2.94, p = 0.049). CP93129 (100 nm) had no effect on the train-evoked Ca²⁺ transients (Fig. 10Ac, blue circles). The last transient in the train in CP9319 was 1.18 ± 0.11 of control (t₍₃₎ = 1.55, p = 0.13). The absolute value of Ca²⁺ surrounding synapses may be lower than used in these recordings (Borst, 2010), although the values of P_r that we obtained from iGluSnFr experiments were similar, but the overall finding that 5-HT_1B receptor effects are both Ca²⁺ and train dependent holds true.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Effect of repetitive stimulation on inhibition. A, Line scans through single CA1 neuron presynaptic varicosities imaged from neurons filled with Ca²⁺ dye (Fluo5F) from whole-cell patch recordings. Aa, A train of 5 action potentials (top) were triggered via the patch pipette to evoke a train of fluorescence transients in the varicosity (middle) that were quantified (bottom). Ab, Baclofen (600 nm) did not prevent evoked events but reduced the Ca²⁺ transient amplitude following each action potential. Ac, Means of peak amplitudes in baclofen divided by pre-agonist control (green squares) of the peak of each transient in the train, from 4 varicosities. In contrast, in a further 4 varicosities, CP93129 (1 μm, blue circles) had no effect on transient amplitudes throughout trains. B, Train of stimuli (4 shocks, 50 Hz) applied to CA1 axons to evoke AMPA EPSCs recorded in subicular neurons (in D-AP5 and bicuculline at a holding potential of –60 mV). Baclofen (1 μm) inhibited EPSCs throughout the train of stimuli. C, A similar train stimulus applied for isolated AMPAR-mediated EPSCs in which box plots and raw data are as for B. CP93129 (400 nm, blue) inhibited the first response in the train, but the EPSCs recovered to control amplitudes by the fourth stimulus. D, A similar train stimulus applied in a preparation superfused with NBQX and bicuculline to isolate the NMDAR EPSCs. EPSCs were recorded at –35 mV to eliminate Mg²⁺ block of the NMDAR. CP93129 (400 nm) had little effect on any of the responses in the stimulus train. Box plots and raw data as for B.

Consequently, we hypothesized that modifying Ca²⁺ entry with GABA_B receptors will alter the efficacy of 5-HT_1B receptors during repetitive activity to provide a form of meta-modulation and synaptic integration. To address this, we first examined effects of stimulus trains (4 shocks, 50 Hz) on inhibition of synaptic responses by CP93129 and baclofen. AMPA EPSCs were isolated in bicuculline and D-AP5. Baclofen inhibited AMPA EPSCs (measured from EPSC initial slopes) throughout these trains with no significant difference in inhibition between responses within the train other than the paired-pulse enhancement (see Fig. 2) (Fig. 10B, ANOVA single factor F_(3,20) = 28.2, p = 0.18). In CP93129 (100 nm), initial slopes of the first responses were reduced to 42 ± 3% of control, while the fourth only to 78 ± 9% (Fig. 10C, single factor ANOVA determined whether CP93129 differentially altered EPSCs in the train; this was significant, ANOVA single factor F_(3,40) = 28.2, p = 5.7 × 10⁻¹⁰). Post hoc Tukey HSD analysis demonstrated the third and fourth EPSCs were inhibited significantly less than the first (p < 0.001). No significant inhibition of NMDA EPSCs isolated in bicuculine (5 μm) and NBQX (5 μm) was recorded for any stimulus (Fig. 10D, ANOVA single factor F_(3,16) = 0.36, p = 0.78).

GPCRs filter the final output of the hippocampus

GABA_B receptors provide fairly uniform inhibition of synaptic output during stimulus trains after the second stimulus (Fig. 10B), when paired-pulse facilitation associated with changes in P_r is increased. By contrast, 5-HT_1B receptor activation reveals a complex modification of synaptic outputs from the hippocampus. NMDAR-mediated EPSCs are barely affected, although in vivo activation of the NMDA responses would require depolarization, which can be provided by synaptic transmission via AMPARs. AMPAR responses are blocked at the start but not the end of short stimulus trains. If 5-HT_1B receptor activation merely inhibited AMPAR responses, then synaptically evoked depolarization would be blocked and NMDARs would remain blocked by Mg²⁺. Therefore, we first determined how 5-HT_1B receptors modify information flow between CA1 and subicular neurons during activity trains in which AMPAR-mediated depolarization triggers NMDAR-mediated responses (Herron et al., 1986).

We stimulated CA1 axons with short stimulus trains as earlier but recorded from subicular neurons under current-clamp conditions. Stimulation intensity was adjusted (up to 20 µA) to ensure that the first stimulus of the burst invariably caused an action potential in the recorded postsynaptic cell (Fig. 11Aa). We found a complex mode of inhibition, mediated by CP93129. After application of the agonist, the first EPSP is significantly inhibited, leading to a dose-dependent delay in depolarization and consequent firing of an action potential (Fig. 11A–D), but CP93129 did not prevent action potential firing during the train, even at saturating concentrations. The mean number of action potentials following the first stimulus in 4 cells was 1.3 ± 0.2 in control, but reduced to 0.2 ± 0.1 by a half-maximal concentration of CP93129 (50 nm, p < 0.05, Fig. 11Ea). However, firing recovered by the second stimulus, in which the control stimulus evoked 1.1 ± 0.4 action potentials and stimuli in CP93129, evoked 1.1 ± 0.4 (Fig. 11Eb), while by the third stimulus, evoked action potentials were enhanced by CP93129 (from 0.4 ± 0.2 to 0.7 ± 0.3, increase in number from first to third was significant, p < 0.01). Thus, while we see a concentration-dependent CP93129-evoked delay of action potentials (Fig. 11A-D). The mean area under the depolarization is barely altered (86 ± 5% of control at 1 μm), even when the first AMPAR-mediated response is substantially inhibited (to 10 ± 2%, Fig. 11F). This reflects the frequency-dependent recovery from 5-HT_1B receptor-mediated inhibition and lack of a CP93129 effect on NMDAR-mediated EPSCs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

5-HT_1B receptors filter synaptic output of the hippocampus, but these filter properties are modified by GABA_B receptors. A, 5-HT_1B receptors suppress activity at the start of a train but not total activity. A stimulus train (5 shocks, 50 Hz) was applied to CA1 axons, EPSPs recorded from subicular neurons. Stimulus intensity was adjusted to ensure the first train stimulus caused an action potential. Aa, Example control trace. Ab, Histogram shows mean number of action potentials following each stimulus from 10 sequential trains from this cell. Ba-Da, Recordings of the cell in A after increasing concentrations of CP3129 (CP). Bb-Db, Action potential counts from each stimulus in the recordings. E, Mean action potential counts from each stimulus from 4 cells (Ea) in control and (Eb) in 50 nm CP93129. F, CP93129 inhibits the first EPSP but not total excitation during trains. From 4 cells, a comparison of dose dependence of CP93129 inhibition of the mean EPSP initial slope on stimulus 1 and mean area under the depolarization for all stimuli (data normalized to control). G, GABA_B receptor activation modifies the filtering effect of 5-HT_1B receptors. Ga, As shown in Figure 10, baclofen (green) 1 μm inhibits EPSPs uniformly throughout the stimulus train. CP93129 (400 nm purple) in the presence of baclofen now also inhibits uniformly throughout the train. Examples are mean responses from 10 sequential trains in one cell. Gb, Box plots and single data show effects in 7 neurons of adding CP93129 (400 nm) to baclofen on responses from each of the 4 stimuli in the train. Gc, The effect of CP93129 on the last EPSC in the train is proportional to the initial inhibitory effect of baclofen. Ha, The modulating effect of baclofen 1 μm on CP93129 inhibition is seen in current-clamp recordings. Baclofen (green) inhibited EPSPs throughout the train. Hb, In the same recording in baclofen, the stimulation intensity was increased (green). CP93129 was added at 50 nm (light blue) and 400 nm (dark blue). EPSPs in baclofen were now uniformly inhibited throughout the stimulus train. Hc, Graph of amplitudes normalized to response in 1 μm baclofen for the first EPSP in the train and for the total area under the depolarization in increasing CP3129 concentrations.

Baclofen inhibits EPSCs throughout trains (Fig. 10B), and has previously been shown to act by inhibiting Ca²⁺ entry (Hamid et al., 2014). This implies that it can modify Ca²⁺-synaptotagmin competition with Gβγ at SNARE complexes and consequently meta-modulate effects in CP93129 which causes Gβγ competition with Ca-synaptotagmin at SNARE complexes (Yoon et al., 2007). We investigated combined receptor activation on presynaptic modulation. Baclofen (1 μm) inhibited AMPAR EPSCs uniformly during stimulus trains (Fig. 11Ga, green). However, in baclofen, CP93129 now also inhibits EPSCs throughout the train (Fig. 11Ga,Gb, purple). This meta-modulation of CP93129 inhibition by baclofen depends on the initial efficacy of baclofen. Plotting initial effects of baclofen alone against CP93129 inhibition of the last EPSC in the train in baclofen reveals a linear relationship (Fig. 11Gc). This is consistent with GABA_B receptor inhibition of Ca²⁺ channels reducing Ca²⁺ that causes evoked release at each stimulus and for Ca²⁺-synaptotagmin competition with 5-HT_1B receptor released Gβγ at SNARE complexes.

We tested this effect of baclofen on CP93129 filtering of EPSP trains. We did not evaluate specific effects of baclofen alone because polysynaptic inhibition is also altered by baclofen, and these current-clamp experiments included no GABA_A antagonists. Nevertheless, recordings in baclofen provided a starting condition to test effects of CP93129 during trains after partial inhibition of presynaptic Ca²⁺ entry by baclofen (Fig. 11Ha). CP93129 in baclofen reduced the area under the depolarization throughout the train (Fig. 11Hb), such that in contrast to the effects of CP93129 alone, in baclofen CP93129 equally inhibited the first EPSP slope and the area under the depolarization during the train. Changes in area and initial slope are not significantly different at 50 or 400 nm CP93129 (Fig. 11H; t₍₇₎ = 1.45, p = 0.19 and t₍₃₎ = 1.39, p = 0.252, respectively). Thus, different receptors with different effectors inhibit release with different temporal outcomes through train evoked responses, but they also interact such that one receptor (GABA_B) alters the effect of a second (5-HT_1B) both qualitatively and quantitatively causing an entirely novel form of presynaptic integration.