BDNF Activates Postsynaptic TrkB Receptors to Induce Endocannabinoid Release and Inhibit Presynaptic Calcium Influx at a Calyx-Type Synapse

Brain-derived neurotropic factor (BDNF) has been shown to play critical roles in neural development, plasticity, and neurodegenerative diseases. The main function of BDNF in the brain is widely accepted to be synaptic regulation. However, how BDNF modulates synaptic transmission, especially the underlying signaling cascades between presynaptic and postsynaptic neurons, remains controversial.


Introduction
Precise and efficient neurotransmission is the basis of neuronal function and plasticity in the CNS (Saheki and De Camilli, 2012;L. G. Wu et al., 2014). Among all neurotrophins, BDNF has attracted much interest for its high expression and potent effects in neural development, functional neural circuit formation, and neurologic diseases (de Jong and Verhage, 2009;Park and Poo, 2013;Choo et al., 2017). BDNF is generally accepted to mainly regulate synaptic function on both excitatory and inhibitory synapses, and one key aspect of the diverse effects of BDNF stems from its complex signaling cascade (Lu et al., 2014). Although many studies have reported BDNF as a potent modulator of synaptic transmission via activation of the tropomyosin receptor kinase B (TrkB) receptor, the underlying signaling cascade is still controversial (Reichardt, 2006;Guo et al., 2018;Lin et al., 2018). A recent study reported that BDNF inhibits synaptic transmission by slowing presynaptic calcium current (ICa) activation, and impairs subsequent exocytosis/endocytosis via activation of the TrkB receptors at a giant calyx-type synapse located in the brainstem (Baydyuk et al., 2015). However, at calyces, the TrkB receptors are expressed not only in the presynaptic nerve terminal, but also in the postsynaptic principal neuron. Whether presynaptic and/or postsynaptic TrkB receptors are involved in the BDNF-induced inhibitory effect is unclear. In addition, the signaling cascade downstream of TrkB activation remains unknown. Interestingly, a previous study reported that BDNF-TrkB signaling in the postsynaptic dendrite leads to a decrease in the probability of presynaptic GABA release on layer 2/3 neocortical inhibitory synapses by rapid mobilization of endocannabinoids (eCBs) into the synaptic cleft (Lemtiri-Chlieh and Levine, 2010). Activation of presynaptic cannabinoid Type 1 receptors (CB1Rs) at the inhibitory synapse makes it necessary to investigate whether BDNF-TrkB signaling at the glutamatergic postsynaptic neuron can also activate the release of eCBs to retrogradely inhibit presynaptic function.
Many studies have demonstrated that eCBs are key activitydependent signals that can modulate synaptic transmission by activating presynaptic CB1Rs (Castillo et al., 2012). For example, strong depolarization of the postsynaptic neuron can lead to reduced synaptic transmission via the release of eCBs (Wilson and Nicoll, 2001). In addition, excessive glutamate release could activate metabotropic glutamate receptors (mGluRs), promoting the synthesis of eCBs to retrogradely regulate neurotransmission (Kushmerick et al., 2004). Evidence also supports mutual interactions between BDNF and eCB signaling. CB1R antagonist has been shown to block BDNF-induced LTP and LTD (Maglio et al., 2018;Pan et al., 2019). Furthermore, CB1R activation upregulates BDNF expression via the PI3K/Akt/mTORC1 pathway (Blázquez et al., 2015), whereas BDNF can induce the release of eCBs at neocortical inhibitory synapses (Lemtiri-Chlieh and Levine, 2010;Zhao and Levine, 2014) and dopamine neurons in the mouse midbrain (Zhong et al., 2015). A recent study showed that the BDNF-induced increase in the mEPSC frequency can be unmasked by blocking eCB signaling at cortical excitatory synapses, suggesting opposing roles of BDNF (Yeh et al., 2017). Whether such opposing roles of BDNF also exist at the glutamatergic calyx-type synapse is not yet known.
CB1Rs are GPCRs and have a well-documented inhibitory effect on adenylyl cyclase (AC) and protein kinase A (PKA) (Childers and Deadwyler, 1996;Castillo et al., 2012). A previous study demonstrated that the AC/PKA signaling pathway can modulate vesicle endocytosis in an activity-dependent manner (Yao and Sakaba, 2012). These findings urged us to investigate whether AC/PKA signaling is involved in BDNF-induced inhibition of synaptic transmission at calyx of Held synapses.
In the present study, we used time-resolved capacitance measurements at a giant glutamatergic synapse, the calyx of Held, to investigate the signaling cascade underlying BDNF-induced inhibition of synaptic transmission (Barnes-Davies and Forsythe, 1995;Borst et al., 1995). We found that BDNF selectively activates postsynaptic TrkB receptors, which induces the release of eCBs in a calcium-dependent manner and retrogradely activates presynaptic CB1Rs to induce presynaptic inhibition. These results suggest a different interpretation of the previous study (Baydyuk et al., 2015), and the trans-synaptic signaling cascade of BDNF-TrkB-eCB coupling may provide a comprehensive understanding of neurotrophin-regulated neurotransmission in the CNS.
For postsynaptic recordings, EPSCs were induced by an afferent stimulus via a bipolar electrode placed near the midline of the MNTB. Stimulation pulses were delivered every 10 s (AM2100, A-M Systems), and the voltage was set to 20% above threshold (,10 V). EPSCs were recorded by an EPC-10 amplifier via a pipette (2-3 MV) containing the following (in mM): 125 K-gluconate, 20 KCl, 10 Na 2 -phosphocreatine, 0.3 GTP, 4 Mg-ATP, 10 HEPES, and 0.5 EGTA, pH 7.2 (adjusted with KOH). The series resistance (,10 MV) was compensated by 95% (10 ms lag). For paired-pulse recording, a paired stimulus was applied with an interval of 20 ms to induce two consecutive EPSCs. The paired-pulse ratio (PPR) was calculated as the second EPSC divided by the first EPSC (Liu et al., 2019).
BDNF was purchased from Merck and applied in the extracellular recording solution at a final concentration of 100 ng/ml. WIN55212-2, AM251, forskolin, and K252a were purchased from Sigma Millipore. MDL12330A, KT5720, U73122, and RHC80267 were purchased from Tocris Bioscience. All drugs except BDNF were dissolved in DMSO. The final concentration of DMSO was 0.1% (X. S. Wu et al., 2009).
All of the methods were conducted in accordance with the approved guidelines, and all animal experimental protocols were approved by the Animal Care and Use Committee of Fudan University.
To detect activation of postsynaptic TrkB receptors, parasagittal brain slices (400 mm thick) containing MNTB were prepared and incubated in normal ACSF in the absence or presence of BDNF (100 ng/ml) for 30 min. The target proteins were identified using a mouse antibody against p-Trk (1:50; Santa Cruz Biotechnology) and a rabbit antibody against microtubule-associated protein 2 (MAP2; 1:200; Abcam). DyLight-594 donkey anti-mouse antibody and DyLight-488 donkey anti-rabbit antibody (1:100; Thermo Fisher Scientific) were used as secondary antibodies. ImageJ software (National Institutes of Health) was used to quantify the TrkB activation. The relative fluorescence was calculated as the fluorescence ratio between p-TrkB and MAP2. Data collection and measurements. As described previously (Xue et al., 2012b;Sun et al., 2016), capacitance measurements were made within 10 min after break-in to avoid rundown, and capacitance jumps were measured 250 ms after depolarization to avoid artifacts. The time constant (t ) was obtained from monoexponential or biexponential fitting of the capacitance decay. The initial rate of endocytosis (Rate endo ) was measured 1-2 s after depolarization. The percentage of residual capacitance 15 (DCm 15s %) or 30 s (DCm 30s %) after depolarization was measured to represent capacitance recovery.
For typical drug bath application experiments, BDNF was applied to the recording chamber by a peristaltic pump at least 30 min before recording at room temperature. When BDNF was coapplied with another drug, both the drug and BDNF were delivered to the chamber at least 30 min before recording and continuously present throughout the experiment. For intracellular drug application experiments, drugs were added to the presynaptic or postsynaptic pipette solution before recording. To specifically block the postsynaptic TrkB receptors, 200 nM K252a was included in the postsynaptic pipette solution via a whole-cell configuration before application of BDNF. After bath application of BDNF for 20 min, we obtained paired recordings by applying another presynaptic pipette at the nerve terminal of the same synapse. The same methods were applied to postsynaptic delivery of BAPTA (20 mM).
Experimental design and statistical analyses. For experiments recording the ICa, capacitance changes, and postsynaptic responses, each group of data were collected from 5 to 13 calyces, which were from 3 to 7 rats of either sex. All data are presented as mean 6 SEM. We used the Kolmogorov-Smirnov test to assess differences in calcium inactivation and paired or unpaired Student's t test to assess differences between two groups. One-way ANOVA with post hoc Bonferroni test was applied for multiple group comparisons. A p value , 0.05 was considered significant. All statistical analyses were performed using MATLAB (2019b, The MathWorks). Details of all statistical analysis are provided in Table 1.

Results
BDNF inhibits apparent EPSC, presynaptic ICa, and release probability First, we investigated how BDNF modulates synaptic transmission by examining the EPSCs of the postsynaptic principal neurons at rat calyces in a whole-cell configuration with 0.5 mM EGTA in the pipette solution. We applied a pair of stimuli with an interval of 20 ms every 10 s near the midline of the MNTB in p8-p10 rats, which can induce two consecutive EPSCs. The PPR was used to evaluate the change in release probability as described above. Recording a baseline for ;10 min, the averaged EPSC amplitude was 3.2 6 0.5 nA (n = 8 calyces). After bath application of BDNF (100 ng/ml), the EPSC gradually decreased and reached a plateau in ;15-20 min (Fig. 1A, top). The EPSC amplitude decreased ;50% (1.6 6 0.3 nA, n = 8; Fig. 1B) and was significantly smaller than the baseline (p = 0.0020, repeatedmeasures one-way ANOVA with Bonferroni post hoc test). The reduced EPSCs recovered to a level similar to baseline after an ;15-30 min washout (99 6 7%; p . 0.9999, repeated-measures one-way ANOVA with Bonferroni post hoc test), showing a reversible inhibitory effect of BDNF (Fig. 1A,B). The PPR at baseline was 0.77 6 0.13 and increased to 1.13 6 0.15 after BDNF application (p = 0.0006, repeated-measures one-way ANOVA with Bonferroni post hoc test; n = 8; Fig. 1A, bottom, B), suggesting a presynaptic mechanism for BDNF-inhibited synaptic transmission (Liu et al., 2019).
A recent study reported that BDNF slows presynaptic ICa activation to inhibit EPSCs at the rat calyx synapse (Baydyuk et al., 2015). Therefore, we applied a similar 20 ms depolarization pulse from À80 to 10 mV (depol 20ms ) and recorded the ICa at the presynaptic nerve terminal of rat calyces (Fig. 1C). We observed an obvious reduction in ICa amplitude after bath application of BDNF (100 ng/ml; Ctrl: 2.1 6 0.1 nA, n = 7; BDNF: 1.4 6 0.1 nA, n = 8; p , 0.0001, unpaired t test; Fig. 1C). However, we did not find a slowing of ICa activation. The 20%-80% rise time for ICa after depol 20ms in control rats was 0.60 6 0.05 ms (n = 7), which is similar to the previous report (Baydyuk et al., 2015). After incubation with BDNF (100 ng/ml) in the extracellular solution for 30 min, the 20%-80% rise time was not significantly changed (0.64 6 0.06 ms, n = 8; p = 0.5619, unpaired t test; Fig. 1C). Next, we plotted the current-voltage curve induced by a 200 ms depolarization pulse from À80 mV to À70, À60, . . . 80 mV with an interval of 30 s (Fig. 1D, left); the ICa in the presence of BDNF was smaller than control at every voltage step, but the peak amplitude did not shift (Ctrl: n = 5; BDNF: n = 6; Fig. 1D, right), confirming that BDNF inhibits ICa amplitude but does not affect its activation. We also examined  Figure 1. BDNF inhibits EPSC, presynaptic ICa, and release probability. A, Top, Sampled paired EPSC recordings in response to 0.1 Hz fiber stimulation at the midline of the trapezoid body; 100 ng/ml BDNF was added to the extracellular solution after obtaining a baseline, and BDNF was washed out after the reduced EPSC amplitudes were stable. Bottom, The corresponding PPR calculated from the paired EPSC amplitudes. B, Top, Sampled EPSC pairs at time points a (baseline, black), b (BDNF treatment, red), and c (washout, blue) from A were overlapped, showing the EPSC changes in response to BDNF application. Bottom, Statistics for EPSC amplitude and PPR (n = 8) from A using a repeated-measures one-way ANOVA with Bonferroni's multiple comparisons test. **p , 0.01. C, Top, Averaged traces of ICa induced by depol 20ms in the control group (n = 7; black) and in the presence of BDNF (100 ng/ml, n = 8; red). ICa of the BDNF-treated group is scaled for comparison (BDNF-scaled, blue). Dashed lines indicate the 20% and 80% rise time. Bottom, Statistics for ICa amplitude and the 20%-80% rise time induced by depol 20ms using an unpaired Student's t test. **p , 0.01. D, Left, Sampled ICa traces in response to 200 ms depolarization pulses from À80 to À40 mV (black), À10 (green), 0 (purple), 10 (red), and 40 mV (blue) in the control and BDNF-treated groups. Right, Plot of the current-voltage relationship in control (n = 5 for each data point; black) and BDNF-treated calyces (n = 6 for each data point; red). p values were calculated using an unpaired Student's t test. *p , 0.05. E, Plot of ICa inactivation curves in the control (n = 5 for each data point; black) and BDNF-treated calyces (n = 5 for each data point; red). F, Sampled ICa (top) and capacitance changes (bottom) induced by 1 (black), 2 (red), 5 (blue), 10 (green), 20 (yellow), 30 (brown), and 50 ms (purple) depolarization pulses from À80 to 10 mV in the control and BDNF-treated groups. G, Left, The relationship between DCm and the duration of depolarization pulses in the control (n = 6 for each data point; black) and BDNF-treated (n = 5 for each data point; red) groups. Right, Statistics of the release probability measured by the percentage of RRP release induced by a 1 ms depolarization pulse from À80 to 10 mV in the control (n = 6; black) and BDNF-treated groups (n = 5; red) using an unpaired Student's t test. Data were from left. **p , 0.01. n.s., not significant. Detailed statistical information is provided in Table 1.
the inactivation curve of the ICa and found no significant difference between the control and BDNF-treated groups (p = 0.9819, Kolmogorov-Smirnov test; n = 5; Fig. 1E). These results suggest that inhibition of the presynaptic ICa amplitude, not a slowdown of ICa activation, is involved in the BDNF-induced inhibition of EPSC amplitude.
The increased PPR recorded for the EPSC after bath application of BDNF suggests a reduction in release probability (Fig.  1A,B). To address this issue at the presynaptic site, we applied stimulation pulses of various lengths (1, 2, 5, 10, 20, 30, and 50 ms) from À80 to 10 mV to induce vesicle release and determined the changes in the readily releasable pool (RRP) size and release probability in the presence of BDNF as described previously (Xue et al., 2012b). In control rats, depol 20ms induced a capacitance jump of 493 6 35 fF (n = 6), which represents the RRP size (Sun and Wu, 2001). Depolarization pulses of 1, 2, 5, and 10 ms induced 6 6 1%, 23 6 4%, 65 6 4%, and 87 6 5%, respectively, of the capacitance jump induced by depol 20ms measured at the same synapses (n = 6 for each depolarization step; Fig. 1F, left, G, black). No further capacitance increase was observed in the control group when the step duration was increased after 30 ms, which is consistent with previous studies showing that a 10-20 ms depolarization pulse from À80 to 10 mV can deplete the RRP (Sun and Wu, 2001;Xue et al., 2012a). After application of BDNF (100 ng/ml) to the extracellular solution, depol 20ms induced a capacitance increase of 378 6 13 fF (n = 5), which is smaller than the capacitance jump measured in controls (p = 0.0033, unpaired t test). When longer stimulation pulses (30-50 ms) were applied, the capacitance jump increased further, until reaching a similar level as controls (528 6 23 fF after 50 ms depolarization pulse, n = 5; p = 0.9770, unpaired t test; Fig. 1G, red), suggesting that the RRP size is not affected. However, depolarization pulses of 1, 2, 5, and 10 ms induced 4 6 1%, 13 6 2%, 39 6 4%, and 56 6 3%, respectively, of the capacitance jump induced by depol 50ms after incubation with BDNF (n = 5 for each depolarization step, Fig. 1F, right, G, red), which is much smaller than the capacitance jump in controls. These results clearly indicate a reduction in the vesicle release probability, which was calculated the percentage of RRP release (DCm 1ms /RRP) induced by an action potential-like stimulation (depol 1ms , 1 ms depolarization pulse from À80 to 10 mV, the data for depol 1ms were obtained from the same experiments; Fig. 1G) (Sun and Wu, 2001).

BDNF induces presynaptic inhibition via activation of postsynaptic TrkB receptor and presynaptic CB1R
BDNF regulates synaptic functions via the activation of TrkB receptors, the major BDNF receptors in the CNS (Reichardt, 2006). We examined the location of TrkB receptors at the calyx of Held synapse by immunostaining. The antibody against Bassoon (BSN), a presynaptic cytomatrix protein selectively localized at the active zone of the nerve terminal (tom Dieck et al., 1998), overlapped partially with the staining of the antibody against TrkB receptors ( Fig. 2A). Staining of TrkB receptors was also observed at the postsynaptic principal neurons, indicating that the TrkB receptors have both presynaptic and postsynaptic expression. We asked whether presynaptic and/or postsynaptic TrkB receptors are involved in the BDNF-induced inhibition of synaptic transmission. A previous study demonstrated that BDNF can act on postsynaptic TrkB receptors to activate the release of eCBs, which would then diffuse retrogradely to reduce the presynaptic release probability of neurotransmitters at cortical inhibitory synapses (Lemtiri-Chlieh and Levine, 2010). In the glutamatergic calyx-type synapse, eCBs are synthesized postsynaptically, and the CB1Rs are specifically localized at the presynaptic nerve terminal (Kushmerick et al., 2004;Zou and Kumar, 2018) (Fig. 2B). Therefore, the postsynaptic expression of TrkB receptors, together with the presynaptic localization of CB1Rs, prompted reinvestigation of the previous study to examine the possibility of BDNF-TrkB-CB1R-induced inhibition of synaptic transmission at calyces (Baydyuk et al., 2015).
Second, we investigated whether activation of TrkB receptors was involved in BDNF-induced presynaptic inhibition. We applied ANA-12 (500 nM), a potent and selective TrkB receptor antagonist (Montalbano et al., 2013), with BDNF (100 ng/ml) in the extracellular solution. After incubating for 30 min, the BDNF-induced inhibition of calcium influx, exocytosis, and endocytosis after depol 20ms was abolished, confirming that the effect of BDNF was mediated by the activation of TrkB receptors (Fig. 2D). Incubation with only ANA-12 did not affect the presynaptic ICa and exocytosis/endocytosis (Fig. 2D).
We further investigated the involvement of presynaptic and/ or postsynaptic TrkB receptors by specifically blocking the postsynaptic TrkB receptors. We applied K252a (200 nM Administration of K252a in the postsynaptic pipette solution did not affect the presynaptic ICa and vesicle exocytosis/endocytosis (Fig. 2E,F, black). Therefore, we concluded that BDNF induces presynaptic inhibition via activation of postsynaptic TrkB receptors at calyces.

BDNF inhibits rapid endocytosis via the eCB signaling pathway
After exocytosis, the fused presynaptic membrane is retrieved via endocytosis to maintain efficient synaptic transmission and normal morphology of the presynaptic nerve terminal (L. G. Wu et al., 2014). Two forms of endocytosis are commonly observed at calyces: clathrin-and dynamin-dependent slow endocytosis, and clathrin-independent dynamin-dependent rapid endocytosis (W. Wu et al., 2005;X. S. Wu et al., 2009). We examined whether BDNF also affects rapid endocytosis.
The presynaptic inhibition of ICa and vesicle exocytosis/ endocytosis demonstrated above critically depend on the activation of postsynaptic TrkB receptors. Therefore, we verified TrkB activation in MNTB principal neurons using immunohistochemistry. We chose MAP2 as the marker for postsynaptic neurons and SC-8058 as the phospho-TrkB marker. Minimal phospho-TrkB receptor was observed in the control group (except some staining of blood vessels; Fig. 4, left). In the BDNF-treated group, the phospho-TrkB receptors dramatically increased at the postsynaptic neurons (Ctrl: 34 6 1%, n = 15; BDNF: 92 6 7%, n = 11; p , 0.0001, unpaired t test; Fig. 4), confirming the activation of postsynaptic TrkB receptors.

Postsynaptic release of eCBs is required for the BDNFinduced presynaptic inhibition
Having shown that BDNF acts on the postsynaptic TrkB receptors and retrogradely induces presynaptic inhibition of the ICa and exocytosis/endocytosis via activation of the presynaptic CB1Rs, we investigated the involvement of postsynaptic eCB synthesis. The eCBs have been shown to be released from postsynaptic neurons via Ca 21 -dependent mechanisms (Kushmerick et al., 2004). We recorded the EPSCs with 20 mM BAPTA in the postsynaptic pipette solution to examine whether the BDNFinduced inhibition of synaptic transmission could be abolished. After 30-60 min bath application of BDNF (100 ng/ml), the EPSC amplitude was not changed (baseline: 5.1 6 0.8 nA, BDNF: 5.1 6 0.8 nA; p = 0.8688, paired t test; n = 5; Fig. 5A,B), confirming that the release of eCBs is required for BDNF-induced inhibitory effect.

C D B A
Rate endo : 2436 22 fF/s, p . 0.9999; DCm 30s %: 16 6 5%, p . 0.9999; one-way ANOVA with Bonferroni post hoc test for all four groups; n = 7; Fig. 6C, blue, D). We performed similar experiments with application of 30 mM RHC80267, a DGL inhibitor (Lemtiri-Chlieh and Levine, 2010), in the presence of BDNF (100 ng/ml) and obtained similar results as control (Fig. 6A,C, green; Table 1). Application of only U73122 or RHC80267 did not induce any significant change from control (Fig. 6B,D). These results further confirm that postsynaptic eCB synthesis is required for BDNF-induced presynaptic inhibition.
The AC/PKA signaling pathway is involved in CB1R-induced presynaptic inhibition Many studies have suggested that activation of CB1R modulates synaptic transmission by reducing neurotransmitter release at both excitatory and inhibitory synapses (Heifets and Castillo, 2009;Kano et al., 2009;Ohno-Shosaku et al., 2012). However, how CB1R activation leads to the inhibition of synaptic transmission remains elusive (de Jong and Verhage, 2009). A previous study demonstrated that the AC/PKA signaling pathway modulates endocytosis in response to strong stimulation (Yao and Sakaba, 2012). Therefore, we investigated whether AC/PKA are also involved in BDNF-TrkB-CB1R-induced presynaptic inhibition.

Discussion
BDNF and eCBs are widely expressed neuromodulators that play crucial roles in various neuronal functions, plasticity, and physiological processes. In the present study, we found that BDNF inhibits presynaptic calcium influx and vesicle exocytosis/endocytosis via the BDNF-TrkB-CB1R signaling pathway. BDNF selectively activates postsynaptic TrkB receptors to evoke the release of eCBs via the PLCg /DGL pathway that retrogradely activates presynaptic CB1Rs, leading to the suppression of downstream AC/PKA signaling. Our study suggests a new mechanism of BDNF-induced inhibition of synaptic transmission at the calyx synapse (Fig. 9). Figure 6. Postsynaptic release of eCBs is required for the inhibitory effects of BDNF. A, Averaged presynaptic ICa (top) and Cm (bottom) induced by depol 20ms in the control (DMSO bath , 0.1% DMSO in the extracellular solution, black), BDNF treatment (BDNF, identical to Fig. 2C), BDNF treatment in the presence of U73122 (U731221BDNF, blue), and BDNF treatment in the presence of RHC80267 (RHC1BDNF, green) groups. B, Statistics for DCm, ICa, Rate endo , and DCm 15s % with different treatments in extracellular solution (DMSO bath , n = 7; BDNF, identical to Fig. 2C, n = 8; U731221BDNF, n = 7; RHC1BDNF, n = 6; U73122, n = 6; RHC, n = 5). The p values were calculated using a one-way ANOVA with Bonferroni's multiple comparisons test. **p , 0.01. C, D, Similar to A, B, except the stimulation was depol 20msx10 (DMSO bath , n = 5; BDNF, identical to Fig. 3A, n = 13; U731221BDNF, n = 7; RHC1BDNF, n = 6; U73122, n = 5; RHC, n = 6). *p , 0.05. **p , 0.01. Detailed statistical information is provided in Table 1. BDNF induces a retrograde eCB signaling pathway How does activation of postsynaptic TrkB receptors lead to inhibition of presynaptic calcium influx and exocytosis/endocytosis? In the present study, we report the involvement of the eCB signaling pathway in BDNF-induced inhibition of synaptic transmission at calyces. Inhibition of the presynaptic CB1R or disruption of postsynaptic eCB synthesis abolishes the BDNFinduced presynaptic inhibition, and exogenous cannabinoid agonist WIN55212-2 can mimic the inhibitory effect of BDNF, demonstrating BDNF-induced retrograde presynaptic inhibition.

C B A D
AC and PKA are widely reported to be involved in G-protein-activated presynaptic inhibition (Chevaleyre et al., 2007;Castillo et al., 2012). Here, we report that inhibition of AC/PKA leads to suppression of calcium influx and vesicle exocytosis/ endocytosis at calyces. Activation of AC by 50 mM forskolin in the bath solution fully abolished the WIN-induced presynaptic  Figure 8. The AC/PKA signaling pathway is involved in the inhibitory effects of BDNF. A, Averaged presynaptic ICa (top) and Cm (bottom) induced by depol 20ms in the control (DMSO pre , 0.1% DMSO in the presynaptic pipette solution, black), WIN55212-2 treatment (WIN, red), WIN55212-2 treatment in the presence of MDL 12330A (MDL1WIN, blue), and WIN55212-2 treatment in the presence of KT 5720 (KT1WIN, green) groups. B, Statistics for DCm, ICa, Rate endo , and DCm 15s % from different treatments in extracellular solution (DMSO pre , n = 6; WIN, n = 5; MDL, n = 8; KT, n = 7; MDL1WIN, n = 9; KT1WIN, n = 5; Forskolin1WIN, n = 5; Forskolin, n = 7). p values were calculated using a one-way ANOVA with Bonferroni's multiple comparisons test. *p , 0.05. **p , 0.01. C, D, Similar to A, B, except the stimulation was depol 20msx10 (DMSO pre , n = 6; WIN, n = 6; MDL, n = 8; KT, n = 9; MDL1WIN, n = 8; KT1WIN, n = 6; Forskolin1WIN, n = 6; Forskolin, n = 9). *p , 0.05. **p , 0.01. E, Left, Averaged ICa induced by depol 20msx10 in the control (Ctrl 2Ca , identical to DMSO pre in Fig. 8C, n = 6; black), MDL 12330A treatment with 3.5 mM extracellular calcium (MDL 3.5Ca , n = 6; red), and KT 5720 treatment with 3.5 mM extracellular calcium (KT 3.5Ca , n = 6; blue) groups. Right, Statistics for QICa using a one-way ANOVA with Bonferroni's multiple comparisons test. F, Left, Averaged Cm induced by depol 20msx10 from E. Right, Statistics for DCm, Rate endo , and DCm 30s % using a one-way ANOVA with Bonferroni's multiple comparisons test. *p , 0.05. **p , 0.01. n.s., not significant. Detailed statistical information is provided in Table 1. inhibition ( Fig. 8A-D). Previous studies have reported that activation of AC by forskolin potentiates synaptic transmission in many different preparations by different mechanisms (Kaneko and Takahashi, 2004;Cheung et al., 2006;Yao and Sakaba, 2010;Renner et al., 2017). However, the forskolin-induced potentiation of EPSC has been shown to occur independent of calcium and RRP size (Ariel et al., 2012). A detailed study at calyx synapses also demonstrated that increased cAMP concentration leads to a large increase in release probability and much smaller increases in RRP size (Yao and Sakaba, 2010), which is consistent with forskolin not affecting the calcium influx and vesicle exocytosis/endocytosis. Together, our results indicate that the BDNF-induced reduction in calcium influx and vesicle exocytosis/endocytosis is mediated by the retrograde eCB signaling pathway.

BDNF inhibits endocytosis via calcium-dependent and -independent pathways
The previous study reported that BDNF inhibits slow and rapid endocytosis via a calcium-independent pathway because BDNF does not reduce the QICa on mild or intense stimulation (Baydyuk et al., 2015). However, in our study, we found that BDNF inhibits calcium influx and exocytosis/endocytosis on either depol 20ms or depol 20msx10 (Figs. 2C,D, 3A,B). When we increased the extracellular calcium concentration in the BDNFtreated group or decreased the calcium concentration in the control group to induce similar amounts of calcium influx and vesicle exocytosis in the two groups, endocytosis was still partially inhibited because of direct modulation of endocytosis by the AC/PKA pathway. Therefore, our findings suggest that BDNF inhibits presynaptic endocytosis in both calcium-dependent and -independent ways.
Physiologic implication of the BDNF-TrkB-eCB signaling cascade Many studies have shown that BDNF can facilitate the efficacy of excitatory synapses by altering either presynaptic neurotransmitter release (Carmignoto et al., 1997;Jovanovic et al., 2000) or the magnitude of postsynaptic responses (Alder et al., 2005) in brain slices or cultured neurons. However, several studies have shown that BDNF may play a different role in the brainstem (Balkowiec et al., 2000;Clark et al., 2011). For example, in the brainstem nucleus tractus solitarius slice, BDNF can reduce the amplitude of mEPSC, the evoked EPSC, and the action potential discharge, indicating reduced intrinsic neuronal excitability (Clark et al., 2011).
BDNF/eCB-induced inhibition of neurotransmission may serve as negative feedback to provide activity-dependent neuroprotection from excitotoxicity. Depolarization-induced suppression, a strong depolarization of postsynaptic neurons leading to reduced synaptic transmission via the release of eCBs, has been interpreted as an efficient means of neuronal protection (Wilson and Nicoll, 2001). Excessive glutamate release has also been shown to promote the synthesis of eCBs to avoid hyperexcitability (Kushmerick et al., 2004). BDNF/eCB signaling may also exert neuroprotective effects on neurodegenerative diseases, such as Huntington's disease. Delivery and overexpression of BDNF or activation of CB1R protect the striatal neurons from excitotoxicity, reduce motor disorders, and prevent the loss of medium spiny neurons (Kells et al., 2008;Blázquez et al., 2011;Connor et al., 2016;Aymerich et al., 2018).
A recent study reported that a reduction in BDNF expression impairs synaptic transmission at the calyx of Held (Jang et al., 2019). However, an increase in BDNF may modulate synaptic transmission via the activation of different signaling cascades, including the inhibition of synaptic transmission via eCB signaling shown here and in other studies (Lemtiri-Chlieh and Levine, 2010;Zhao and Levine, 2014;Zhong et al., 2015). Furthermore, increased BDNF expression has been observed in many physiological or pathologic conditions. For example, a protective mechanism of the CB1R-dependent increase in BDNF expression has been reported in mice with kainateinduced seizures (Marsicano et al., 2003). At calyces, the basal neuronal firing can be increased to .600-800 Hz on stimulation (Von Gersdorff and Borst, 2002;Hermann et al., 2007), which may increase the BDNF level in an activity-dependent manner (Y. J. Wu et al., 2004;Singer et al., 2014) to induce inhibitory neuroprotection from excitotoxicity. A recent study showed that, in the lower part of the auditory system in the brain, BDNF may improve the signal-to-noise ratio and sound sensitivity by increasing the inhibitory strength of neurons at hearing onset. A significant increase in central noise can be observed after auditory nerve injury (Chumak et al., 2016).
In conclusion, we examined the presynaptic mechanisms and signaling cascades of BDNF-induced inhibition of synaptic transmission at a glutamatergic central synapse. By uncovering the detailed mechanisms underlying how BDNF/TrkB couples with the eCB signaling pathway to modulate synaptic transmission, our study may provide a comprehensive understanding of how BDNF and eCBs associate in an overlapping set of neurologic diseases. Figure 9. Schematic of the proposed signaling pathway for BDNF-inhibited synaptic transmission. BDNF activates postsynaptic TrkB receptors to induce eCB release via the PLCg /DGL pathway. eCBs retrogradely bind to presynaptic CB1Rs and lead to suppression of the AC/PKA signaling pathway, finally inhibiting presynaptic calcium influx and exocytosis/endocytosis. VGCC, Voltage-gated calcium channel; PIP 2 , phosphatidylinositol 4,5-bisphosphate; DAG, diacylglycerol.