To determine the potential localization of NMDA receptors at the mf terminals of the rodent hippocampus, we performed electron microscopy and post-embedding immunogold labeling in rats using a validated antibody for the obligatory subunit GluN1 (Petralia et al., 1994; Siegel et al., 1994; Takumi et al., 1999; Watanabe et al., 1998). Gold particles were detected in the main body of the postsynaptic density as well as presynaptic mf terminals (Figure 1A–C). GluN1 localized in mf boutons in a relatively high proportion to the active zone, as compared to associational–commissural (ac) synapse in the same CA3 pyramidal neuron (Figure 1D; mf, ~32% presynaptic particles; ac, <10% presynaptic particles; n = 3 animals). Similar quantification for AMPA receptors did not reveal presynaptic localization of these receptors in either mf or ac synapses (Figure 1—figure supplement 1; ~5% presynaptic particles, n = 3 animals). Together, these results provide anatomical evidence for preNMDARs at mf-CA3 synapses.

Presynaptic short-term plasticity, in the form of low-frequency (~1 Hz) facilitation (LFF), is uniquely robust at the mf-CA3 synapse (Nicoll and Schmitz, 2005; Salin et al., 1996). To test a potential involvement of preNMDARs in LFF, we monitored AMPAR-mediated excitatory postsynaptic currents (EPSCs) from CA3 pyramidal neurons in acute rat hippocampal slices. Neurons were held at V_h = −70 mV to minimize postsynaptic NMDAR conductance, and mfs were focally stimulated with a bipolar electrode (theta glass pipette) placed in stratum lucidum ~100 µm from the recorded cell. LFF was induced by stepping the stimulation frequency from 0.1 Hz to 1 Hz for ~2 min in the presence of picrotoxin (100 µM) to block fast inhibitory synaptic transmission, and a low concentration of the AMPAR noncompetitive antagonist LY303070 (0.5 μM) to minimize CA3–CA3 recurrent activity (Kwon and Castillo, 2008). Bath application of the NMDAR irreversible open channel blocker MK-801 (50 μM) significantly reduced LFF (Figure 1E). In addition, the competitive NMDAR antagonists D-APV (100 µM) or R-CPP (50 µM) yielded a comparable reduction of facilitation (Figure 1F). To confirm that these synaptic responses were mediated by mfs, the mGluR2/3 agonist DCG-IV (1 µM) was applied at the end of all recordings (Kamiya et al., 1996). To control for stability, we performed interleaved experiments in the absence of NMDAR antagonists and found that LFF remained unchanged (Figure 1—figure supplement 2A). These findings indicate NMDAR antagonism reduces mf-CA3 short-term plasticity (LFF), suggesting that preNMDARs could contribute to this form of presynaptic plasticity.

The reduction in facilitation of AMPAR transmission could be due to dampening of CA3 recurrent activity by NMDAR antagonism (Henze et al., 2000; Kwon and Castillo, 2008; Nicoll and Schmitz, 2005). To discard this possibility, we repeated our experiments in a much less excitable network in which AMPAR-mediated synaptic transmission was selectively blocked by a high concentration of the noncompetitive antagonist LY303070 (15 μM) and monitored the kainate receptor (KAR)-mediated component of mf synaptic transmission (Castillo et al., 1997; Kwon and Castillo, 2008). In addition, 2 mM MK-801 was included in the intracellular recording solution to block postsynaptic NMDARs (Corlew et al., 2008; Figure 1—figure supplement 3). To further ensure postsynaptic NMDAR blockade, we voltage-clamped the CA3 pyramidal neuron at −70 mV and waited until NMDAR-mediated transmission was eliminated and only KAR-EPSCs remained. Under these recording conditions, bath application of MK-801 (50 μM) also reduced LFF of KAR-mediated transmission (Figure 1G), whereas LFF remained unchanged in interleaved control experiments (Figure 1—figure supplement 2B). At the end of these recordings, 10 μM NBQX was applied to confirm KAR transmission (Figure 1G, Figure 1—figure supplement 2B; Castillo et al., 1997; Kwon and Castillo, 2008). It is therefore unlikely that the reduction of LFF mediated by NMDAR antagonism could be explained by recurrent network activity, suggesting a direct effect on transmitter release.

To further support a role of preNMDARs in mf LFF, we took a genetic approach by conditionally removing NMDARs from GCs in Grin1 floxed mice. To this end, an AAV5-CaMKII-Cre-GFP virus was bilaterally injected in the DG to selectively delete Grin1 , whereas AAV5-CaMKII-eGFP was injected in littermates as a control at postnatal days 16–20 in both groups (Figure 2A). Two weeks after surgery, we prepared acute hippocampal slices and examined the efficacy of Grin1 deletion by analyzing NMDAR-mediated transmission in GFP⁺ GCs of Grin1-cKO and control mice. We confirmed that in contrast to control mice, no NMDAR-EPSCs were elicited by electrically stimulating medial perforant-path inputs in Grin1-cKO GCs voltage-clamped at +40 mV in the presence of 100 μM picrotoxin and 10 μM NBQX (Figure 2B). As expected, the NMDAR/AMPAR ratio was significantly reduced in Grin1-cKO mice compared to control (Figure 2C). Only acute slices that exhibited robust GFP fluorescence in the DG were tested for LFF of AMPAR transmission in CA3. We found that LFF was significantly reduced in Grin1-cKOs as compared to controls (Figure 2D), indicating that genetic removal of NMDARs from GCs recapitulated NMDAR antagonism (Figure 1E–G). Grin1 deletion did not affect basal transmitter release as indicated by a comparable paired-pulse ratio to control (Control: 2.5 ± 0.36, n = 13 cells; Grin1 cKO: 2.4 ± 0.31, n = 13 cells; U > 0.5, Mann–Whitney test). Collectively, our findings using two distinct approaches strongly suggest that NMDAR activation in GCs increases LFF of mf-CA3 synaptic transmission.

Bath application of MK-801 could have blocked dendritic NMDARs in GCs and potentially affected transmitter release (Christie and Jahr, 2008; Duguid, 2013). To address this possibility, we repeated our experiments after performing a surgical cut in the granular layer of the DG in order to isolate mf axons from GCs (Figure 3—figure supplement 1A). Under these conditions, MK-801 bath application still reduced LFF (Figure 3A), and LFF was stable in control, acutely transected axons (Figure 3B). In addition, puffing D-APV (2 mM) in stratum lucidum near (~200 µm) the recorded neuron also reduced LFF (Figure 3C), whereas puffing artificial cerebrospinal fluid (ACSF) had no effect (Figure 3D). Lastly, in a set of control experiments, we confirmed that D-APV puffs were sufficient to transiently block NMDAR-mediated transmission in CA3, but not in DG (Figure 3—figure supplement 1B,C). Together, these results support the notion that LFF reduction was due to the blockade of preNMDARs but not somatodendritic NMDARs on GCs.

GCs in vivo typically fire in brief bursts (Diamantaki et al., 2016; GoodSmith et al., 2017; Henze et al., 2002; Pernía-Andrade and Jonas, 2014; Senzai and Buzsáki, 2017). To test whether preNMDARs contribute to synaptic facilitation that occurs during more physiological patterns of activity, mfs were activated with brief bursts (five stimuli, 25 Hz). We first took an optogenetic approach and used a Cre-dependent ChIEF virus to selectively light-activate mf-CA3 synapses in Grin1-cKO and control mice. Thus, animals were injected with a mix of AAV5-CaMKII-CreGFP+AAV DJ-FLEX-ChIEF-tdTomato viruses in the DG (Figure 4A). At least 4 weeks after surgery, acute slices were prepared and burst-induced facilitation of AMPAR-mediated transmission in CA3 was assessed (Figure 4B,C). Burst-induced facilitation, triggered by light stimulation and measured as the ratio of EPSCs elicited by the fifth and first pulse (P5/P1 ratio), was significantly reduced in Grin1-cKO animals as compared to controls. Because these bursts of activity can activate the CA3 network (Henze et al., 2000; Kwon and Castillo, 2008; Nicoll and Schmitz, 2005), we next monitored KAR-EPSCs under conditions of low excitability (as in Figure 1G). MK-801 bath application also reduced burst-induced facilitation, whereas facilitation remained unchanged in naïve slices (Figure 4D,E). In a separate set of experiments, we confirmed the reduction of MK-801 on burst-induced facilitation under more physiological recording conditions (Figure 4—figure supplement 1). Lastly, we tested whether preNMDARs, by facilitating glutamate release during bursting activity, could bring CA3 pyramidal neurons to threshold and trigger postsynaptic action potentials. To test this possibility, we monitored action potentials elicited by KAR-EPSPs (resting membrane potential −70 ± 2 mV) from CA3 pyramidal neurons intracellularly loaded with 2 mM MK-801. Under these recording conditions, MK-801 bath application significantly reduced the mean number of spikes per burst (Figure 4F). No changes in mean spikes per burst were observed in naïve slices over time (Figure 4G). Application of 10 μM NBQX at the end of these experiments confirmed that action potentials were induced by KAR-mediated synaptic responses. Consistent with these observations, MK-801 also reduced the mean number of spikes per burst when AMPAR-mediated action potentials were recorded from CA3 pyramidal neurons (Figure 4—figure supplement 2). In control experiments, we found that intracellular MK-801 effectively blocked postsynaptic NMDAR transmission during burst stimulation (Figure 4—figure supplement 3). Altogether, these results indicate that preNMDARs at mf-CA3 synapses can contribute to information transfer from the DG to CA3.

PreNMDARs could facilitate glutamate and BDNF release by increasing presynaptic Ca²⁺ rise (Bouvier et al., 2016; Buchanan et al., 2012; Corlew et al., 2008; Park et al., 2014). To test this possibility at mf-CA3 synapses, we combined a conditional knockout strategy with Ca²⁺ imaging using two-photon laser scanning microscopy. We first deleted preNMDARs by injecting AAV5-CaMKII-mCherry-Cre virus in the DG of Grin1 floxed mice, and littermate animals injected with AAV5-CaMKII-mCherry virus served as control (Figure 5A). Two weeks after surgery, we confirmed the efficacy of Grin1 deletion by activating medial perforant-path inputs and monitoring NMDAR/AMPAR ratios in GCs of control and Grin1-cKO animals (Figure 5A). Virtually no NMDAR-EPSCs were detected at V_h = +40 mV in Grin1-cKO animals (Figure 5A). Acute slices that exhibited robust mCherry fluorescence in the DG were used for Ca²⁺ imaging experiments. To maximize our ability to detect preNMDAR-mediated Ca²⁺ signals, we used a recording solution that contained 0 mM Mg²⁺, 4 mM Ca²⁺, and 10 μM D-Serine (Carter and Jahr, 2016). GCs expressing mCherry were patch-loaded with 35 µM Alexa 594 (used as morphological dye) and 200 µM Fluo-5F, and mf axons were imaged and followed toward CA3 until giant boutons (white arrows) were identified (Figure 5B). We found that Ca²⁺ transients (CaTs) elicited by direct current injection in the GC soma (five action potentials, 25 Hz) were significantly smaller in Grin1-cKO animals as compared to control (Figure 5C–E). In addition, NMDAR antagonism with D-APV reduced presynaptic Ca²⁺ rise even under more physiological Mg⁺² concentration in acute rat hippocampal slices (Figure 5—figure supplement 1). Thus, preNMDARs contribute significantly to presynaptic Ca²⁺ rise in mf boutons, and by this means likely facilitates synaptic transmission, although a potential contribution of Ca²⁺ rise-independent effects cannot be discarded.

Lastly, we sought to determine whether direct activation of preNMDARs could drive Ca²⁺ influx in mf giant boutons. To test this possibility, we elicited CaTs by two-photon glutamate uncaging (2PU) on mf boutons of control and Grin1-cKO animals (Figure 6A). As previously described, mCherry GCs were patch-loaded with Alexa 594 and Fluo-5F in a recording solution designed to maximize the detection of preNMDAR-mediated Ca²⁺ signals (as in Figure 5). We first confirmed that glutamate 2PU-induced CaTs in dendritic spine heads of GCs were strongly reduced in Grin1-cKO animals as compared to controls (Figure 6B,C). To verify that reduced Ca²⁺ signals (ΔG/R) were a result of Grin1 deletion and not differences in uncaging laser power, we performed a laser power intensity–response curve and found that Grin1-cKO animals exhibited reduced ΔG/R signals as compared to control regardless of laser power intensity (Figure 6—figure supplement 1). We next measured glutamate 2PU-induced CaTs in mf giant boutons (identified as in Figure 5B) and found that single uncaging pulses were insufficient to drive detectable CaTs in control boutons (Figure 6—figure supplement 2). However, a burst of 2PU stimulation (5 pulses, 25 Hz) induced CaTs in mf boutons of control but not in Grin1-cKO animals (Figure 6D,E). Additionally, CaTs elicited by 2PU stimulation were abolished by D-APV application (Figure 6—figure supplement 3). These findings indicate that brief bursts of glutamate 2PU, a manipulation that mimics endogenous release of glutamate during physiological patterns of activity, induces presynaptic Ca²⁺ influx in mf boutons by activating preNMDARs.

Previous work implicated preNMDARs in the release of BDNF at corticostriatal synapses following burst stimulation and presynaptic Ca²⁺ elevations (Park et al., 2014). Given the high expression levels of BDNF in mfs (Conner et al., 1997; Yan et al., 1997), we examined the potential role for preNMDARs in BDNF release from mf terminals. To this end, a Cre-dependent BDNF reporter (BDNF-pHluorin) was injected in Grin1-floxed and control animals. Littermate mice were injected with a mix of AAV5-CaMKII-mCherry-Cre + AAV-DJ-DIO-BDNF-pHluorin viruses in the DG (Figure 7A). At least 4 weeks after surgery, acute slices were prepared for two-photon laser microscopy to image mf boutons. After acquiring a stable baseline of BDNF-pHluorin signals, mfs were repetitively activated (see Materials and methods) (Figure 7B). BDNF-pHluorin signals were analyzed by measuring ΔF/F, where ΔF/F reductions indicate BDNF release (Park et al., 2014). We found that GluN1-deficient mf boutons showed a significant (~50%) impairment in BDNF release as compared to control (Figure 7C–D). Furthermore, using a more physiological pattern of burst stimulation, GluN1-lacking mf boutons still displayed altered BDNF release as compared to control (Figure 7—figure supplement 1). Taken together, our results suggest preNMDARs contribute significantly to BDNF release during repetitive or burst stimulation of mf synapses.

In addition to providing a major excitatory input to the hippocampus proper, mf axons also synapse onto excitatory hilar MCs and inhibitory neurons in CA3 (Amaral et al., 2007; Henze et al., 2000; Lawrence and McBain, 2003). To test whether preNMDARs could also play a role at these synapses, we visually patched MCs and INs in acute rat hippocampal slices, loaded them with 35 µM Alexa 594 (Figure 8A) and 2 mM MK-801, and monitored AMPAR-EPSCs (V_h = −70 mV). Unlike mf-CA3 synapses, mf synapses onto CA3 INs in stratum lucidum do not express LFF, but can undergo burst-induced facilitation or depression (Toth et al., 2000). We found that MK-801 bath application had no effect on burst-induced facilitation or depression (Figure 8B), suggesting preNMDARs do not play a role at mf-IN synapses in CA3. Mf inputs onto hilar MCs undergo robust activity-dependent facilitation (Lysetskiy et al., 2005). Similar to mf-CA3 synapses, we found that MK-801 reduced LFF (Figure 8C). Stability experiments of mf transmission at CA3 INs or hilar MCs showed no significant differences (Figure 8—figure supplement 1). Taken together, our findings demonstrate that preNMDARs facilitate mf transmission onto excitatory neurons, but not onto inhibitory INs.

A monoclonal antibody against GluN1 (clone 54.1 MAB363) was obtained from Millipore (Germany), and its specificity was characterized previously (Siegel et al., 1994). An affinity-purified polyclonal rabbit anti-GluA1-4 (pan-AMPA), corresponding to aa 724–781 of rat, was used and characterized previously (Nusser et al., 1998).

Immunohistochemical reactions at the electron microscopic level were carried out using the post-embedding immunogold method as described earlier (Lujan et al., 1996). Briefly, animals (n = 3 rats) were anesthetized by intraperitoneal injection of ketamine-xylazine 1: 1 (0.1 mL/kg b.w.) and transcardially perfused with ice-cold fixative containing 4% paraformaldehyde, 0.1% glutaraldehyde, and 15% saturated picric acid solution in 0.1 M phosphate buffer (PB) for 15 min. Vibratome sections 500 μm thick were placed into 1 M sucrose solution in 0.1 M PB for 2 hr before they were slammed on a Leica EM CPC apparatus. Samples were dehydrated in methanol at −80°C and embedded by freeze-substitution (Leica EM AFS2) in Lowicryl HM 20 (Electron Microscopy Science, Hatfield, PA), followed by polymerization with UV light. Then, ultrathin 80-nm-thick sections from Lowicryl-embedded blocks of the hippocampus were picked up on coated nickel grids and incubated on drops of a blocking solution consisting of 2% human serum albumin in 0.05 M TBS and 0.03% Triton X-100. The grids were incubated with GluN1 or pan-AMPA antibodies (10 μg/mL in 0.05 M TBS and 0.03% Triton X-100 with 2% human serum albumin) at 28°C overnight. The grids were incubated on drops of goat anti-rabbit IgG conjugated to 10 nm colloidal gold particles (Nanoprobes Inc) in 2% human serum albumin and 0.5% polyethylene glycol in 0.05 M TBS and 0.03% Triton X-100. The grids were then washed in TBS and counterstained for electron microscopy with 1% aqueous uranyl acetate followed by Reynolds’s lead citrate. Ultrastructural analyses were performed in a JEOL-1010 electron microscope.

Animal handling followed an approved protocol by the Albert Einstein College of Medicine Institutional Animal Care and Use Committee in accordance with the National Institute of Health guidelines. Acute rat hippocampal slices (400 µm thick) were obtained from Sprague-Dawley rats, from postnatal day 17 (P17) to P28 of either sex. For procedures regarding transgenic mouse slice preparation, see below. The hippocampi were isolated and cut using a VT1200s microslicer (Leica Microsystems Co.) in a solution containing (in mM): 215 sucrose, 2.5 KCl, 26 NaHCO₃, 1.6 NaH₂PO₄, 1 CaCl₂, 4 MgCl₂, 4 MgSO₄, and 20 glucose. Acute slices were placed in a chamber containing a 1:1 mix of sucrose cutting solution and normal extracellular ACSF recording solution containing (in mM): 124 NaCl, 2.5 KCl, 26 NaHCO₃, 1 NaH₂PO₄, 2.5 CaCl₂, 1.3 MgSO₄, and 10 glucose incubated in a warm-water bath at 33–34°C. The chamber was brought to room temperature for at least 15 min post-sectioning, and the 1:1 sucrose-ACSF solution was replaced by ACSF. All solutions were equilibrated with 95% O₂ and 5% CO₂ (pH 7.4). Slices were allowed to recover for at least 45 min in the ACSF solution before recording. For physiological Mg⁺² and Ca⁺² experiments, ACSF solutions were adjusted to (in mM): 1.2 MgSO₄ and 1.2 CaCl₂, and temperature was maintained at 35 ± 0.1°C in the submersion-type recording chamber heated by a temperature controller (TC-344B Dual Automatic Temperature Controller, Warner Instruments).

Electrophysiological recordings were performed at 26.0 ± 0.1°C (unless otherwise stated) in a submersion-type recording chamber perfused at 2 mL/min with normal ACSF supplemented with the GABA_A receptor antagonist picrotoxin (100 µM) and the selective AMPA receptor (AMPAR) antagonist LY303070 at a low concentration (0.5 µM) to minimize CA3-CA3 recurrent activity, or at a high concentration (15 µM) to isolate KAR-EPSCs and KAR-EPSPs to assess monosynaptic mf transmission. Whole-cell recordings were made from CA3 pyramidal cells voltage-clamped at −70 mV using patch-type pipette electrodes (3–4 mΩ) containing (in mM): 131 cesium gluconate, 8 NaCl, 1 CaCl₂, 10 EGTA, 10 glucose, 10 HEPES, and 2 MK-801 pH 7.25 (280–285 mOsm) unless specified otherwise. KOH was used to adjust pH. Series resistance (8–15 MΩ) was monitored throughout all experiments with a −5 mV, 80 ms voltage step, and cells that exhibited a series resistance change (>20%) were excluded from analysis. A stimulating bipolar electrode (theta glass, Warner Instruments) was filled with ACSF and placed in stratum lucidum to selectively activate mfs using a DS2A Isolated Voltage Stimulator (Digitimer Ltd.) with a 100 µs pulse width duration. AMPAR-EPSCs were recorded for a baseline period of 2 min, and LFF was induced by stepping the stimulation frequency from 0.1 to 1 Hz for 2 min. Facilitation was measured by taking a ratio of the mean EPSC during the steady-state, LFF period of activity and the 2-min baseline (EPSC_1Hz/EPSC_0.1Hz) before and after bath application of NMDAR antagonists.

To qualify for analysis, mf responses met three criteria: (1) The 20–80% rise time of the AMPAR-EPSC was less than 1 ms, (2) LFF was greater than 150%, (3) the AMPAR-EPSC displayed at least 70% sensitivity to the group 2/3 mGluR agonist, DCG-IV (1 µM). Isolated KAR-EPSCs were elicited by 2 pulses with a 5 ms inter-stimulus interval for LFF experiments. Baseline measurements were acquired at least 10 min after ‘break-in’ to achieve optimal intracellular blockade of postsynaptic NMDARs by MK-801 (2 mM) in the patch-pipette. To transect mf axons in acute slices, a 45° ophthalmic knife (Alcon Surgical) was used to make a diagonal cut across the hilus from the dorsal to ventral blades of the DG, and the subregion CA3b was targeted for patch-clamp recordings. For D-APV (2 mM) puff experiments, a puffer device (Toohey Company) was set to deliver two to three puffs of 100 ms duration at 3–4 psi during the 2 min of LFF activity. The puffer pipette was placed at least 200 µm away from the recording site, and both the puff pipette and hippocampal slice were positioned to follow the direction of the laminar perfusion flow in a low profile, submersion-type chamber (RC-26GLP, Warner Instruments). Burst-induced facilitation was elicited by 5 pulses at 25 Hz with a 0.03 Hz inter-trial interval for a baseline period of 10 min. Facilitation was measured by calculating the ratio of the mean KAR-EPSC peak of the fifth pulse to the first pulse (P5/P1) before and after bath application of MK-801 (50 µM). To study KAR induced action potentials, CA3 pyramidal cells were whole-cell patch-clamped with internal solution containing (in mM): 112 potassium gluconate, 17 KCl, 0.04 CaCl₂, 0.1 EGTA, 10 HEPES, 10 NaCl, 2 MgATP, 0.2 Na₃GTP, and 2 MK-801, pH 7.2 (280–285 mOsm). Current-clamped CA3 cells were held at −70 mV during burst stimulation of mfs (5 pulses at 25 Hz) to monitor evoked action potentials. Spike transfer was measured by quantifying mean number of spikes/burst for a 10 min period before and after bath application of MK-801 (50 µM). Robust sensitivity to the AMPAR/KAR selective antagonist NBQX (10 µM) confirmed KAR-EPSC responses. Similarly, CA3 pyramidal cells were kept in current-clamp mode for AMPAR-mediated action potential monitoring in the presence of LY303070 (0.5 µM) and picrotoxin (100 µM). AMPAR-mediated mf action potentials were confirmed by blockade of responses following application of DCG-IV (1 µM). Both hilar MCs and CA3 INs were visually patched-loaded with Alexa 594 (35 µM), and morphological identity was confirmed by two-photon laser microscopy at the end of experiments. MCs were voltage-clamped at −70 mV, and a bipolar electrode was placed in the DG to activate mf inputs. The data analysis and inclusion criteria used for mf experiments (described above) was also implemented for MC recordings. CA3 INs were voltage-clamped at −70 mV and burst stimulated, facilitation was assessed as previously mentioned. Both facilitating and depressing mf responses were included for analysis given the diversity of mf to CA3 IN transmission (Toth et al., 2000). Whole-cell voltage and current-clamp recordings were performed using an Axon MultiClamp 700B amplifier (Molecular Devices). Signals were filtered at 2 kHz and digitized at 5 kHz. Stimulation and acquisition were controlled with custom software (Igor Pro 7).

Grin1-floxed littermate mice of either sex (P16-20) were injected with 1 μL of AAV5-CaMKII-eGFP, AAV5-CaMKII-CreGFP, AAV5-CaMKII-mCherry, or AAV5-CaMKII-mCherry-Cre viruses at a rate of 0.12 μL/min at coordinates (−1.9 mm A/P, 1.1 mm M/L, 2.4 mm D/V) targeting the DG using a stereotaxic apparatus (Kopf Instruments). Two weeks post-surgery, mice were sacrificed for electrophysiology or Ca²⁺ imaging experiments. Mice were transcardially perfused with 20 mL of cold NMDG solution containing (in mM): 93 NMDG, 2.5 KCl, 1.25 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 25 glucose, 5 sodium ascorbate, 2 Thiourea, 3 sodium pyruvate, 10 MgCl₂, 0.5 CaCl₂, brought to pH 7.35 with HCl. The hippocampi were isolated and cut using a VT1200s microslicer in cold NMDG solution. Acute mouse slices were placed in an incubation chamber containing normal ACSF solution that was kept in a warm-water bath at 33–34°C. All solutions were equilibrated with 95% O₂ and 5% CO₂ (pH 7.4). Post-sectioning, slices recovered at room temperature for at least 45 min prior to experiments. For NMDAR/AMPAR ratios, GCs were patch-clamped with the cesium internal solution previously mentioned containing either Alexa 594 (35 µM) for GFP⁺ cells (laser tuned to 830 nm/910 nm, respectively) or Alexa 488 (35 µM) for mCherry⁺ cells (laser tuned to 910 nm/780 nm, respectively). AMPAR-EPSCs were recorded at −65 mV in the presence of picrotoxin (100 µM) by placing a bipolar electrode near the medial perforant path and delivering a 100 μs pulse width duration using an Isoflex stimulating unit. AMPAR-EPSCs were acquired for at least 5 min followed by bath application of NBQX (10 µM) to isolate NMDAR-EPSCs. GCs were brought to +40 mV to alleviate magnesium block and record optimal NMDAR-EPSCs. NMDAR/AMPAR ratios were measured by taking the mean NMDAR-EPSC/AMPAR-EPSC for a 5 min period of each component. Only acute mouse slices with optimal GFP and mCherry reporter fluorescence (i.e. robust expression, ≥75% of DG fluorescence) were used for electrophysiology, and Ca²⁺ and BDNF imaging experiments. Grin1-floxed animals (The Jackson Laboratory) were kindly provided by Dr. Michael Higley (Yale University).

Grin1 floxed and control mice of either sexes (P17–P20) were injected with a 1:2 mix of AAV5-CaMKII-CreGFP/AAV-DJ-FLEX-ChIEF-tdTomato viruses targeting the DG, using the same coordinates described above. At least 4 weeks post-surgery, acute hippocampal slices were prepared as previously described, and slices showing optimal GFP and tdTomato expression were used for electrophysiology experiments. Mf optical burst stimulation was elicited by using a Coherent 473 nm laser (4–8 mW) delivering 5 pulses at 25 Hz with a 1–2 ms pulse width duration. Facilitation was measured by taking a ratio of the mean AMPAR-EPSC peak of the fifth pulse to the first pulse (P5/P1) in control and Grin1-cKO animals.

mCherry⁺ GCs were patch-loaded with an internal solution containing in (mM): 130 KMeSO₄, 10 HEPES, 4 MgCl₂, 4 Na₂ATP, 0.4 NaGTP, 10 sodium phosphocreatine, 0.035 Alexa 594 (red morphological dye), and 0.2 Fluo-5F (green calcium indicator), 280–285 mOsm. KOH was used to adjust pH. GCs near the hilar border were avoided and GCs that exhibited adult-born GC electrophysiological properties were excluded from analysis. GCs were kept in voltage clamp configuration at −50 mV for at least 1 hr to allow the diffusion of dyes to mf boutons. Recordings were obtained in ACSF solution containing (in mM): 124 NaCl, 2.5 KCl, 26 NaHCO₃, 1 NaH₂PO₄, 4 CaCl₂, 0 MgSO₄, 10 glucose, 0.01 NBQX, 0.1 picrotoxin, and 0.01 D-serine. Using an Ultima 2P laser scanning microscope (Bruker Corp) equipped with an Insight Deep See laser (Spectra Physics) tuned to 830 nm, the ‘red’ photomultiplier tube (PMT) was turned on and with minimal pockel power the red signal was used to identify the mf axon. With 512 × 512 pixel resolution, mf axons were followed for at least 200 µm from the DG toward CA3, until bouton structures were morphologically identified and measured (>3 μm in diameter). GCs were switched to current-clamp mode held at −70 mV and 1 ms current injections were used to elicit a burst of 5 action potentials at 25 Hz. Using line scan analysis software (PrairieView 5.4, Bruker Corp.), a line was drawn across the diameter of the bouton at a magnification of at least 16×. The ‘green’ PMT channel was turned on, and 1000 line scans were acquired in a 2 s period. Action potential induction was delayed for 400 ms to collect a baseline fluorescence time period. Calcium transients (CaTs) were acquired with a 1 min inter-trial-interval and analyzed using the ΔG/R calculation: (G − G₀)/R. CaTs from control animals were compared to Grin1-cKO by taking the mean peak ΔG/R value for a 30 ms period of the fifth action potential. In similar fashion, CaT signals in acute rat hippocampal slices were acquired and tested for sensitivity to D-APV (100 µM) while adjusting ACSF MgS0₄ concentration to 1.3 mM and CaCl₂ to 2.5 mM in the absence of NBQX.

For glutamate uncaging experiments, GCs that were mCherry⁺ were patch-loaded using the internal solution previously described, and a small volume (12 mL) of recirculated ACSF solution containing (in mM): 124 NaCl, 2.5 KCl, 26 NaHCO₃, 1 NaH₂PO₄, 4 CaCl₂, 0 MgSO₄, 10 glucose, 2.5 MNI-glutamate, 0.01 NBQX, 0.1 picrotoxin, and 0.01 D-serine. A MaiTai HP laser (Spectra Physics) was tuned to 720 nm to optimally uncage glutamate and elicit CaTs in GC dendritic spines. Following the measurement of CaTs in GC spines, mf boutons were identified and to mimic bursting activity, five uncaging pulses (1 ms duration) were delivered at 25 Hz. The acquired CaTs in spines and boutons were analyzed using the ΔG/R calculation in control and Grin1-cKO animals. In a subset of control boutons, D-APV (100 µM) was applied to detect CaT sensitivity to NMDAR antagonism.

Grin1 floxed and control mice of both sexes (P16-20) were injected with a 1:2 mix of AAV5-CaMKII-mCherryCre/AAV-DJ-DIO-BDNF-phluorin viruses targeting the DG using the same coordinates as above. At least 4 weeks post-surgery, acute hippocampal slices were prepared as previously described, and slices showing optimal GFP and mCherry expression were taken for imaging sessions. For stimulation, a monopolar micropipette electrode was placed in the stratum lucidum at least 250 µm away from the imaging site. The Insight Deep See laser (Spectra Physics) was tuned to 880 nm, and the imaging site was selected by the appearance of fibers and bouton structures in the stratum lucidum. Using 512 × 512 pixel resolution identified boutons measuring at least 3 μm in diameter were selected as a region of interest (ROI) magnified to 4–6×, and a baseline acquisition of 100 consecutive images at 1 Hz using T-series software (PrairieView 5.4, Bruker Corp.) was acquired (Park et al., 2014). Following the baseline acquisition, a repetitive stimulation consisting of 125 pulses at 25 Hz was delivered 2×, triggering an acquisition of 200 consecutive images at 1 Hz. The fluorescence intensity of the bouton ROI was measured using ImageJ software to calculate ΔF/F of the BDNF-pHluorin signal. To verify reactivity of the ROI, an isosmotic solution of NH₄Cl (50 mM) was added at the end of the imaging session as previously reported (Park et al., 2014). The same experimental and analysis procedure was implemented to measure BDNF release triggered by mf burst stimulation consisting of 5 pulses at 100 Hz, 50×, every 0.5 s.

AAV5-CaMKII-eGFP and AAV5-CaMKII-CreGFP viruses were acquired from UPenn Vector Core. AAV5-CaMKII-mCherry and AAV5-CaMKII-mCherry-Cre were obtained from UNC Chapel Hill Vector Core. The AAV-DJ-FLEX-ChIEF-tdTomato and AAV-DJ-DIO-BDNF-phluorin viruses were custom ordered and obtained from UNC Chapel Hill Vector Core. The DNA of the ChIEF virus was a generous gift from Dr. Pascal Kaeser (Harvard University), and the DNA of the BDNF-pHluorin was kindly provided by Dr. Hyungju Park (Korea Brain Research Institute).

Picrotoxin and all chemicals used to prepare cutting, recording, and internal solutions were acquired from MilliporeSigma. All NMDAR antagonists (D-APV, MK-801, R-CPP), NMDAR agonist (D-serine), the group 2/3 mGluR agonist (DCG-IV), and MNI-glutamate for uncaging experiments were purchased from Tocris Bioscience. D-APV was also acquired from the NIMH Chemical Synthesis Drug Program. NBQX was purchased from Cayman Chemical Company. The noncompetitive AMPAR selective antagonist LY303070 was custom ordered from ABX Chemical Company. Alexa 594 morphological dye, Alexa 488, and the Ca²⁺ indicator Fluo-5F (Invitrogen) were purchased from ThermoFisher Scientific.

All data points from experiments were tested for normality using a Shapiro–Wilk test (p-value < 5% for a normal distribution). Statistical significance was determined if p-value < 0.05. Experiments with a normal distribution and an N > 7 cells were tested for statistical significance with a paired Student’s t-test. Experiments with N < 7 cells or skewed distributions were tested for statistical significance using a paired Wilcoxon signed-rank sum test. For experiments comparing control and Grin1-cKO animals, statistical significance was determined using unpaired t-test and Mann–Whitney test (U < 0.05). All statistical tests were performed using Origin Pro 9 (Origin Lab). Experimenters were blind to the identity of the virus injected in transgenic Grin1 floxed mice during the acquisition of data in CA3 electrophysiology and two-photon imaging. However, data analysis could not be performed blind in those experiments in which NMDAR/AMPAR ratios in GCs were examined in order to assess the efficiency of the cKO.