Activity-Dependent Release of Endogenous BDNF From Mossy Fibers Evokes a TRPC3 Current and Ca²⁺ Elevations in CA3 Pyramidal Neurons

Organotypic slice cultures

Hippocampi were dissected from postnatal day 7–11 Sprague Dawley rats (Harlan, Indianapolis, IN, or Charles River, Wilmington, MA) and cut transversely into ∼500-μm-thick slices using a custom-made wire-slicer fitted with 20-μm-thick Pt wire (Pozzo-Miller et al. 1995). Hippocampal slices were individually plated on Millicell-CM filter inserts (Millipore, Billerica, MA) and cultured in 36°C-5% CO₂, 98% relative humidity incubators (Thermo-Forma, Waltham, MA). Slices were maintained in culture media (Neurobasal-A plus B27, InVitrogen, Carlsbad, CA) containing 20% equine serum for the first 4 days in vitro (div). To avoid the confounding effects of hormones and growth factors in the serum, its concentration was gradually reduced over a period of 48 h starting at 4 div (24 h each in 10 and 5% serum). After a period of 24 h in serum free media (Neurobasal-A plus B27), 7–10 div slices were used for electrophysiology (Tyler and Pozzo-Miller 2001). Some slice cultures remained in serum containing culture media (20% equine serum) as described in the original publications (Gahwiler 1981; Pozzo-Miller et al. 1993; Stoppini et al. 1991; Yamamoto et al. 1989). All procedures followed national and international ethics guidelines and were annually reviewed and approved by the Institutional Animal Care and Use Committee at University of Alabama at Birmingham.

Acute slices

Hippocampi were dissected from postnatal day 45–60 C57BL/6 wild-type mice, placed in ice-cold modified artificial cerebrospinal fluid (ACSF, see following text), and cut transversely into 400-μm-thick slices using a vibrating blade microtome (VT1200S, Leica, Nussloch, Germany). For the dissection and storage of acute slices, the modified ACSF contained (in mM) 87 NaCl, 75 sucrose, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 0.5 CaCl₂, 7 MgCl₂, and 10 glucose (bubbled with 95% O₂-5% CO₂; pH 7.4). Slices were first transferred to this modified ACSF warmed to ∼32°C for 30 min and then kept at room temperature in modified ACSF until use for electrophysiology.

Simultaneous electrophysiology and Ca²⁺ imaging

Individual 7–10 div slice cultures or acute slices were transferred to a recording chamber mounted on a fixed-stage upright microscope (Zeiss Axioskop FS, Oberkochen, Germany) and continuously perfused (2 ml/min) with oxygenated ACSF at room temperature (24°C). For slice cultures, ACSF contained (in mM): 124 NaCl, 2 KCl, 1.24 KH₂PO₄, 1.3 MgSO₄, 17.6 NaHCO₃, 2.5 CaCl₂, and 10 glucose (adjusted to 310–320 mOsm with sucrose); for acute slices, ACSF contained (in mM) 125 NaCl, 5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 2 CaCl₂, 1 MgCl₂, and 25 glucose (adjusted to 300–320 mOsm with sucrose); both ACSF solutions were continuously bubbled with 95% O₂-5% CO₂ (pH 7.4). Superficial pyramidal neurons in CA3 stratum pyramidale were visualized with a water-immersion ×63 objective (0.9 NA) using IR-DIC microscopy (Dodt and Zieglgänsberger 1990). Simultaneous whole cell recording and microfluorometric Ca²⁺ imaging was performed as described (Petrozzino et al. 1995; Pozzo-Miller 2006; Pozzo-Miller et al. 1996). Briefly, unpolished patch pipettes contained (in mM) 120 Cs-gluconate, 17.5 CsCl (or KCl), 10 Na-HEPES, 4 Mg-ATP, 0.4 Na-GTP, 10 Na₂ creatine phosphate, 2 QX314, and 0.2 Na-EGTA (replaced by 0.2 mM fura-2 for Ca²⁺ imaging); 290–300 mOsm; pH 7.2 (resistance: 3–4 MΩ). Membrane currents were recorded in the voltage-clamp mode at a holding potential of −65 mV using an Axopatch-200B amplifier (Molecular Devices, Sunnyvale, CA), filtered at 2 kHz, and digitized at 10 kHz. Recordings were accepted only if access resistance was ≤25 MΩ. Input resistance (R_i) was measured with hyperpolarizing voltage pulses (50 ms, 20 mV), and cells were discarded if any whole cell parameter (i.e., C_m, R_i, R_s) changed by ≥20% during the course of an experiment. A subsample of CA3 pyramidal neurons had R_i of 275.3 ± 77.9 (SD) MΩ and whole cell capacitances of 135.9 ± 30.4 (SD) pF under the preceding recording conditions (n = 31). To reduce the contribution of voltage-gated Ca²⁺ channels during simultaneous whole cell recording and Ca²⁺ imaging, neurons were voltage clamped at +40 mV (Malinow et al. 1994; Perkel et al. 1993; Pozzo-Miller et al. 1996). Fura-2 (InVitrogen) was alternatively excited at 360 and 380 nm using a monochromator (Polychrome-IV, TILL Photonics, Munich, Germany), and its emission (>510 nm) was filtered and detected with a frame-transfer cooled CCD camera (Quantix-57, Photometrics/Roper, Duluth, GA); digital image pairs were acquired every 4 s (50 ms exposures for ∼100 × 200 pixel subarrays). Background-subtracted fluorescence intensity measurements were obtained within regions of interest (ROIs) defined over apical primary and secondary dendrites, as well as somata. The average ratio of 360 and 380 nm fluorescence within each ROI was used as an estimate of intracellular Ca²⁺ concentration as these two parameters are directly proportional to each other (Grynkiewicz et al. 1985). Electrical and optical data were simultaneously acquired on a single G4 Macintosh computer (Apple) running custom-written software (TIWorkBench, written by T. Inoue).

Afferent fiber stimulation was performed with an extracellular patch pipette filled with buffered and oxygenated ACSF (∼2 MΩ) positioned within either the granule cell layer or s. lucidum to stimulate the MFs. Square constant-current pulses of 100 μs duration were produced by an isolated stimulator (ISO-Flex, AMPI, Jerusalem, Israel). High-frequency afferent stimulation was delivered as a theta burst pattern consisting of five bursts at 5 Hz, each burst having four pulses at 100 Hz. A subthreshold concentration of the Na⁺ channel blocker TTX (10 nM) was included in the ACSF to reduce membrane excitability and prevent polysynaptic responses, common in organotypic slices. The ACSF contained the GABA_A receptor antagonist picrotoxin (50 μM), while the K⁺ channels coupled to GABA_B receptors were blocked by intracellular Cs⁺. The ACSF also contained the noncompetitive antagonists of AMPAR and NMDAR, GYKI-52466 (20 μM) and MK-801 (20 μM), respectively, as well as the antagonist of group-I mGlu receptors LY-367385 (100 μM) and L-type Ca²⁺ blocker nimodipine (20 μM); additional experiments were performed with the competitive antagonists 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX, 20 μM) and d,l-2-amino-5-phosphonovaleric acid (d,l-APV, 100 μM). Membrane currents observed in the presence of GYKI-52466 were 111.8 ± 12.5 pA (n = 12), whereas currents in the presence of CNQX were 109.0 ± 11.9 (n = 21; P = 0.8803). The current intensity for afferent stimulation (between 10 and 50 μA) was never larger than 10 times that required to evoke maximal AMPAR-mediated fast EPSCs at −65 mV.

Human recombinant mature BDNF (100 μg/ml; Promega; Madison, WI) dissolved in 0.0001% bovine serum albumin was applied from glass pipettes (∼5 MΩ) using a Pressure System IIe (Toohey, Fairfield, NJ), as described (Amaral and Pozzo-Miller 2007a,b). The application pipette was positioned ∼100 μm above the slice and ∼200 μm away from the soma of the CA3 neuron under recording to avoid pressure and mechanical artifacts and aimed at its apical dendrites within s. lucidum (∼150 μm from the soma) against the direction of ACSF perfusion flow. Every pulse of 25–30 s duration at 20–30 psi delivered a maximum of 2 μl of the BDNF solution on top of the slice.

TRPC3 RNA interference

Custom HuSH small hairpin interference RNA (shRNA) constructs (OriGene, Rockville, MD) were designed to target TRPC3 channel subunits (Gene Bank Accession No. J05181). Short hairpin RNAs encoding the following target sequence 5′-GAGTTCAAGAATGACTACAGGAAGCTCTC 3′ were introduced by transfection with the HuSH construct. The same vector also contained the coding sequence for GFP, and its fluorescence enabled identification of transfected neurons before whole cell recording as described (Amaral and Pozzo-Miller 2007a).

Particle-mediated gene transfer

Hippocampal slices were transfected with a HuSH expression plasmid (OriGene) encoding green fluorescent protein (GFP) and the TRPC3-specific shRNA sequence. A custom-modified Helios gene gun (Bio-Rad; Hercules, CA) was used to perform the biolistic transfection following established protocols (Alonso et al. 2004; Lo et al. 1994). Briefly, the TRPC3-specific shRNA expression vector was precipitated onto 1.6 μm colloidal gold and this mixture was coated onto Tefzel tubing using 0.06 mg/ml polyvinylpyrrolidone. Slices were bombarded using He pressure at 100 psi from a distance of 15 mm. For negative control experiments, gene transfer was performed as in the preceding text using a random shRNA or only enhanced yellow fluorescent protein, except the plasmid encoding eYFP was precipitated onto 1.6 μm colloidal gold at a ratio of 50 μg of DNA to 25 mg of gold.

Drugs

Some compounds were dissolved in DMSO (0.01%) and others directly into the ACSF or intracellular solution; vehicle controls using 0.01% DMSO were routinely performed. When compounds were included in the intracellular solution (e.g., k-252b), a diffusion period of 5–10 min was allowed to elapse before the start of the experiment. All of the chemicals used for these experiments were obtained from Sigma (St. Louis, MO), Calbiochem (San Diego, CA), or Tocris (Ellisville, MO).

Coefficient of variation analysis

The coefficient of variation (CV²) of evoked AMPAR- and NMDAR-mediated EPSCs was calculated after subtraction of the recording noise variance (σ²): [(σ)² – (σ of noise)²]/mean², as described (Bender et al. 2009). The fractional change in CV² (r = CV² _{before TBS}/CV² _{after TBS}) was plotted against a synaptic gain factor (f = EPSC amplitude _{after TBS}/EPSC amplitude _{before TBS}), as described (Faber and Korn 1991; Shen and Horn 1996).

Statistical analysis

Averages of multiple measurements are presented as means ±SE; mean ± SD was used to calculate the coefficient of variation (CV). Data were statistically analyzed using unpaired Student's t-test, ANOVA test, and pair-wise correlation and linear regression using the Prism software package (GraphPad Software, San Diego, CA). Probability values <0.05 were considered statistically significant (i.e., P < 0.05, <5% probability that the observations are due to chance). When different than this cutoff value, the actual P values ≤4 decimal points are given in results (rather than just the statement “greater than” or “less than”) to provide readers with more detailed information regarding the outcome of the statistical analyses. Compromise power analyses were performed to determine the statistical power given the number of observations, sample means and SDs, using G*Power (Erdfelder et al. 1996). These power analyses yielded values of statistical Power (1 − β; where β is the type-II error) larger than 0.95 (i.e., 95% confidence of accepting the null hypothesis when is true).

Activity-dependent release of BDNF from MFs evokes MF-I_BDNF, a slow membrane current in CA3 pyramidal neurons

We recently showed that stimulation of afferent Schaffer collaterals activated TRPC-like nonselective cationic currents in CA1 pyramidal neurons by means of the release of endogenous—native—BDNF (Amaral and Pozzo-Miller 2007a). We now focus on the MF pathway in CA3 because these fibers contain the highest levels of BDNF in the brain (Conner et al. 1997; Danzer and McNamara 2004). In the following experiments, the identity of MF-mediated responses was routinely confirmed using the group-II mGluR agonist DCG-IV (Toth et al. 2000). Indeed, EPSC amplitudes decreased significantly after bath application of 1 μM DCG-IV (76.00 ± 5.42 vs. 204.52 ± 5.59 pA, n = 5; P < 0.001; Fig. 1A). TBS has been shown to be an effective stimulation pattern to release GFP-tagged BDNF (Hartmann et al. 2001) or endogenous BDNF from cultured neurons (Balkowiec and Katz 2002). Consistently, stimulation of MFs in CA3 s. lucidum of 7–10 div organotypic slice cultures with a single TBS train evoked slowly activating inward currents in CA3 pyramidal neurons voltage clamped at −65 mV in the presence of ionotropic and metabotropic glutamate and GABA receptor antagonists (20 μM GYKI-52466 or 20 μM CNQX for AMPARs; 20 μM MK-801 or 50 μM d,l-APV for NMDARs; 100 μM LY-367385 for mGluR1/5; 50 μM picrotoxin for GABA_ARs; Cs⁺ intracellular solution to block GABA_B-coupled K⁺ channels; 10 nM TTX to reduce polysynaptic activity). The average amplitude of MF-induced I_BDNF was 109.02 ± 11.89 pA, and they peaked 1.47 ± 0.23 s following the onset of afferent stimulation (n = 21; Fig. 1B). MF-I_BDNF observed in the presence of GYKI-52466 was 111.8 ± 12.5 pA (n = 12), whereas MF-I_BDNF in the presence of CNQX was 109.0 ± 11.9 (n = 21; P = 0.8803), demonstrating that kainate receptors did not contribute to the measured responses. The amplitude of MF-I_BDNF was very stable, lacking any significant “rundown” during recording times ≤90 min from whole cell access (Fig. 1B). Similar responses were observed in acute slices from 45- to 60-day-old C57BL/6 wild-type mice (117.83 ± 11.85 pA, n = 9; P = 0.6469 compared with MF-I_BDNF from organotypic cultures from rat hippocampal slices; Supplemental Fig. S1A).¹

TrkB-IgG, a chimeric recombinant protein that fuses together the ligand-binding domain of the TrkB receptor with the Fc domain of human IgG (Shelton et al. 1995), was used as an extracellular scavenger to implicate endogenous BDNF binding to membrane receptors in MF-I_BDNF activation. Consistent with observations in CA1 (Amaral and Pozzo-Miller 2007a), TrkB-IgG (1 μg/ml, equivalent to 3.8 nM of the receptor dimer) significantly reduced MF-I_BDNF amplitude (20.49 ± 6.4 pA, n = 10; P < 0.001 vs. controls; Fig. 1B). As controls, a nonspecific human IgG (1 μg/ml) did not affect MF-I_BDNF (147.26 ± 39.25 pA, n = 7; P > 0.05 vs. controls; Fig. 1B). In addition, increasing stimulus intensities evoked larger currents (Fig. 1C), suggesting an activity-dependent process.

Vesicular Zn²⁺ released during MF stimulation does not contribute to MF-I_BDNF

At glutamatergic synapses, neurotransmitter vesicles oftentimes store and release Zn²⁺ alongside their glutamate content, thereby increasing the extracellular concentration of Zn²⁺ (Frederickson et al. 2005). Nowhere is this more apparent than in CA3 s. lucidum, the brain region with the highest vesicular Zn²⁺ content. In addition to its action on NMDA receptors (Mayer and Vyklicky 1989; Mayer et al. 1989), Zn²⁺ also modulates LTP induction at MF-CA3 synapses: removal of free Zn²⁺ with Ca-EDTA prevented LTP induction, and exogenous addition of Zn²⁺ potentiated MF-CA3 extracellular field EPSPs (Li et al. 2001). Because synaptically released Zn²⁺ was recently proposed to trans-activate TrkB receptors in CA3 (Huang et al. 2008), we experimentally altered its extracellular concentration to determine whether MF-I_BDNF is in any way influenced by vesicular Zn²⁺ released during TBS of MFs.

Addition of 1.5 mM of the membrane impermeable Zn²⁺ chelator Ca-EDTA did not affect MF-I_BDNF (123.66 ± 13.81 pA, n = 18; P > 0.05 compared with control responses). Similar results were obtained using 2.5 mM Ca-EDTA (124.51 ± 18.91 pA, n = 34, P > 0.05 compared with controls; Fig. 2A). Considering that the Kd of EDTA for Zn²⁺ is 10¹⁶ (Bers 1982; Vogt et al. 2000), these concentrations of Ca-EDTA are 2 orders of magnitude higher than the apparent affinity constant of Ca-EDTA for Zn²⁺ (3–4 nM). On a practical note, further attempts to increase the concentration of Ca-EDTA (e.g., to 10 mM) (as in Huang et al. 2008; Li et al. 2001) yielded very unstable whole cell recordings, likely due to excessive buffering of divalent cations that reduced membrane charge screening. Screening and binding of negative charges on the extracellular surface of plasma membranes by divalent cations contributes to the electric field across the membrane, and thus is critical for normal membrane excitability (Hille 2001); in fact, Zn²⁺ itself exerts very potent charge screening effects (Blaustein and Goldman 1968; Hille et al. 1975).

Consistent with the lack of effect of extracellular Zn²⁺ on MF-evoked responses, membrane currents activated by application of recombinant BDNF to apical dendrites of CA3 pyramidal neurons (in the presence of TTX) were not significantly affected by the addition of 2.5 mM Ca-EDTA (control I_BDNF 83.67 ± 19.37 pA, n = 27, vs. I_BDNF in Ca-EDTA 79.61 ± 19.67 pA, n = 16; P > 0.05; Fig. 2B). We also tested whether I_BDNF was affected by increasing extracellular Zn²⁺. I_BDNF evoked in the presence of 200 μM ZnCl₂ was not significantly different from controls (84.19 ± 17.01 pA, n = 11, vs. 83.67 ± 19.37 pA, n = 27; P > 0.05; Fig. 2B). Because recombinant BDNF applied to CA3 s. lucidum did not increase mEPSC frequency in CA3 pyramidal neurons (M. D. Amaral, Y. Li, T. Inoue, G. Calfa, L. Pozzo-Miller, unpublished data), these observations rule out potential artifacts arising from activity-dependent release of vesicular Zn²⁺ from MFs. Taken together, these results demonstrate that MF-I_BDNF is not sensitive to changes in the concentration of extracellular Zn²⁺.

TRPC3 channels are required for MF-I_BDNF

To directly demonstrate the requirement for TRPC3 channels in the activation of MF-I_BDNF as we did for recombinant BDNF-induced responses in CA1 (Amaral and Pozzo-Miller 2007a), we used a shRNA that was specifically designed to knockdown TRPC3 channel expression. HuSH constructs encoding TRPC3 shRNA (and GFP used for cell identification) were coated onto gold particles and then biolistically transfected into slice cultures (Alonso et al. 2004; Lo et al. 1994). After 48 h of transfection, GFP or eYFP-expressing CA3 pyramidal neurons were identified by epifluorescence microscopy and the presence of a gold particle (Alonso et al. 2004; Amaral and Pozzo-Miller 2007a).

Consistent with our observations in CA1 neurons, TRPC3 knockdown completely prevented the activation of MF-I_BDNF in CA3 pyramidal neurons (27.63 ± 2.92 pA, n = 42, vs. untransfected CA3 neurons within the same slice 316.3 ± 41 pA, n = 12; P < 0.001; Fig. 2C). However, CA3 neurons transfected with a cDNA plasmid encoding eYFP alone or a random shRNA sequence did not affect MF-I_BDNF amplitude (311.88 ± 33.34 pA, n = 14; P > 0.05 compared with untransfected CA3 neurons in the same slice; Fig. 2C). These results demonstrate that MF-I_BDNF in CA3 neurons is mediated by nonselective cationic channels containing TRPC3 subunits.

Activity-dependent release of BDNF from MFs evokes dendritic Ca²⁺ signals in voltage clamped CA3 pyramidal neurons

Elevations in the intracellular concentration of Ca²⁺ ions were the first-described immediate actions of BDNF (Berninger et al. 1993). Activated neurotrophin Trk receptors trigger the formation of IP₃ by PLCγ-mediated hydrolysis of phosphatidylinositol bisphosphate (PIP₂), which then activates IP₃ receptors of intracellular Ca²⁺ stores (Segal and Greenberg 1996). Consistently, I_BDNF evoked by recombinant BDNF in hippocampal CA1 neurons is associated with dendritic transient Ca²⁺ elevations (Amaral and Pozzo-Miller 2007b). Here we performed simultaneous whole cell recording and microfluorometric imaging in CA3 pyramidal neurons filled with the Ca²⁺ indicator fura-2 (100 μM) and voltage clamped at +40 mV to reduce the contribution of voltage-gated Ca²⁺ channels (Perkel et al. 1993; Pozzo-Miller et al. 1996). In the presence of our antagonist cocktail (which also included 20 μM of the L-type Ca²⁺ channels blocker nimodipine), TBS of MFs within CA3 s. lucidum evoked outward membrane currents followed by slowly developing Ca²⁺ elevations in apical dendrites (Fig. 3A, left). The 360/380 nm ratio of fura-2 signals had average peak amplitudes of 0.6 ± 0.06 (n = 33, CV = 0.54; Fig. 3B). To determine whether these Ca²⁺ responses were a result of activity-dependent BDNF release, afferent MFs were stimulated after 30 min incubation with the extracellular BDNF scavenger TrkB-IgG (Fig. 3A, right). Under these conditions, there were no significant changes in the fura-2 ratio (0.05 ± 0.01, n = 14, CV = 0.93) when compared with control responses (P < 0.001; Fig. 3B).

TBS of MFs causes a BDNF-dependent depression of excitatory postsynaptic currents in CA3 pyramidal neurons

Considering the well-established role of BDNF on hippocampal synaptic plasticity (Bramham and Messaoudi 2005; Poo 2001), we monitored the amplitude of evoked EPSCs recorded from CA3 pyramidal neurons before and after a single brief TBS of afferent MFs. Pharmacologically isolated EPSCs mediated by AMPAR were significantly reduced after TBS of MFs (144.5 ± 26.8 pA vs. 38.52 ± 14.3 pA after 40 min, n = 12; P < 0.001; Fig. 4A); these responses were subsequently blocked by the AMPAR antagonist CNQX (20 μM). Similarly, pharmacologically isolated EPSCs mediated by NMDAR were significantly reduced after TBS of MFs (188.6 ± 13.5 vs. 79.8 ± 10.0 pA after 40 min, n = 30; P < 0.001; Fig. 4B); these responses were subsequently blocked by the NMDAR antagonist d,l-APV (50 μM). This enduring depression of NMDAR-mediated responses unlikely reflects the much short-lived Ca²⁺-dependent inactivation of NMDARs (Legendre et al. 1993) because it developed even if the monitoring of NMDAR-mediated EPSCs was resumed after a 5 min interval from the theta burst stimulus, long enough for the TBS-induced Ca²⁺ transient to be dissipated by buffering, sequestration, and extrusion. Note that TBS of MFs also induced a brief posttetanic potentiation (PTP), which is evident when afferent stimulation was resumed immediately after TBS (see Supplemental Fig. S1B).

To obtain the ratio of NMDAR/AMPAR EPSCs in the same neuron, 40–50 consecutive synaptic responses in control ACSF were first evoked at V_hold = −60 mV for AMPAR-EPSCs and then another EPSC series at +40 mV for NMDAR-EPSCs. Consistent with the pharmacological approach in different cells, the amplitude of AMPAR-EPSCs and of NMDAR-EPSCs were also significantly reduced after MF TBS (94.4 ± 7.6 vs. 55.7 ± 5.1 pA after 20 min, n = 4; P < 0.001 for AMPAR-EPSCs; 176.2 ± 25.2 vs. 122.5 ± 11.4 pA after 20 min, n = 4; P < 0.05 for NMDAR-EPSCs), while the NMDAR/AMPAR ratio was unchanged (0.5 ± 0.06 vs. 0.4 ± 0.08 after 30 min, n = 5; P = 0.3; Fig. 4C). Such parallel changes in the amplitude of both AMPAR-EPSCs and NMDAR-EPSCs with a constant NMDAR/AMPAR ratio strongly suggest a presynaptic expression mechanism.

Changes in presynaptic function are usually accompanied by changes in the coefficient of variation (CV²) of EPSC amplitudes. Consistent with a decrease in presynaptic transmitter release (e.g., Choi and Lovinger 1997; Shen and Horn 1996), the CV² of pharmacologically isolated AMPAR-EPSCs was significantly increased after TBS of MFs (n = 8, P = 0.0039 Wilcoxon signed-rank test); a similar increase was observed for the CV² of pharmacologically isolated NMDAR-EPSCs after MF TBS (n = 14, P < 0.0001 Wilcoxon signed-rank test). Also consistent with a presynaptic change (Faber and Korn 1991; Shen and Horn 1996), the plot of the fractional change in CV² (r) versus the synaptic gain factor (f) showed a significant correlation with all points to the right of the diagonal (r = f; AMPAR-EPSCs linear regression slope = 0.40, significantly different from unity P = 0.0246, n = 8; NMDAR-EPSCs linear regression slope = 0.35, significantly different from unity P = 0.0012, n = 14; Fig. 4D).

The role of activity-dependent release of endogenous BDNF on this enduring depression of NMDAR-mediated responses was tested using the scavenger TrkB-IgG. Indeed, the amplitude of NMDAR-mediated EPSCs remained unaffected after TBS in the presence of 1 μg/ml of TrkB-IgG in the recording ACSF (165.7 ± 12.8 vs. 142.3 ± 15.3 pA after 20 min, n = 9; P = 0.29; Fig. 5A). Consistent with a requirement of TrkB receptors, intracellular application of the membrane impermeable analog k-252b (200 nM in 0.01% DMSO) (Amaral and Pozzo-Miller 2007a; Knusel and Hefti 1992) also prevented TBS-induced depression of NMDAR-mediated EPSCs (185 ± 9.7 vs. 168.9 ± 7 pA after 20 min, n = 13; P = 0.1763; Fig. 5B). These observations suggest that an enduring depression of glutamatergic synaptic transmission at MF-CA3 synapses is a physiological consequence of activity-dependent release of endogenous BDNF from MFs.