Interplay of Entorhinal Input and Local Inhibitory Network in the Hippocampus at the Origin of Slow Inhibition in Granule Cells

Abstract

Neuronal activity from the entorhinal cortex propagates through the perforant path (PP) to the molecular layer of the dentate gyrus (DG) where information is filtered and converted into sparse hippocampal code. Nearly simultaneous signaling to both granule cells (GC) and local interneurons (INs) engages network interactions that will modulate input integration and output generation. When triggered, GABA release from interneurons counteracts the glutamatergic signals of PP terminals, scaling down the overall DG activation. Inhibition occurs at fast or slow timescales depending on the activation of ionotropic GABA_A-R or metabotropic GABA_B-R. Although postsynaptic GABA_A and GABA_B-R differ in their location at the synapse, mixed GABA_A/B-R IPSPs can also occur. Here we describe a slow inhibition mechanism in mouse GCs recorded from either sex, mediated by GABA_A/B-R in combination with metabotropic glutamate receptors. Short burst PP stimulation in the gamma frequency range lead to a long-lasting hyperpolarization (LLH) of the GCs with a duration that exceeds GABA_B-R IPSPs. As a result, LLH alters GC firing patterns and the responses to concomitant excitatory signals are also affected. Synaptic recruitment of feedforward inhibition and subsequent GABA release from interneurons, also successfully trigger mixed GABA responses in GCs. Together these results suggest that slow inhibition through LLH leads to reduced excitability of GCs during entorhinal input integration. The implication of LLH in regulation of neuronal excitability suggests it also contributes to the sparse population coding in DG.SIGNIFICANCE STATEMENT Our study describes a long-lasting hyperpolarization (LLH) in hippocampal granule cells. We used whole-cell patch-clamp recordings and an optogenetic approach to characterize this event. LLH is a slow inhibitory mechanism that occurs following the stimulation of the perforant pathway in the molecular layer of the dentate gyrus. We found that it is mediated via postsynaptic ionotropic and metabotropic GABA and metabotropic glutamate receptors. The duration of LLH exceeds previously described IPSPs mediated by any of these receptors. The activation of LLH requires presynaptic gamma frequency bursts and recruitment of the local feedforward inhibition. LLH defines prolonged periods of low excitability of GCs and a restrained neuronal discharge. Our results suggest that LLH can contribute to sparse activation of GCs.

Keywords: dentate gyrus; feedforward inhibition; granule cells.

Figure 1.

EPSPs evoked by PP stimulation are followed by a LLH in dentate GCs. A, Representative current-clamp traces of hyperpolarization following either single pulse (top) or 50 Hz train (V_m = −60 mV) stimulation of the perforant path. B, Firing pattern of a GC (V_m = −45 mV) expressing LLH after PP stimulation with one pulse (black arrow) followed 4 s later by five pulses at 50 Hz (red arrow); inset (bottom) shows a single trial of the same cell and a magnification of the superposition of hyperpolarization events evoked either by a single pulse (black) or a train of 5 pulses (red) at 50 Hz. C, Comparison of the amplitude (mV) and the duration (s) of the hyperpolarization either after a single pulse or a train stimulation of the PP, shows that there is a significantly longer (single: 1.18 ± 0.53 s, train: 2.38 ± 0.74 s, p < 0.0001, n = 12) and bigger (single: 7.17 ± 1.48 mV, train: 14.28 ± 3.78 mV, p < 0.0001, n = 12) hyperpolarization when a train stimulation was applied. D, LLH remains expressed in GCs in the absence of postsynaptic firing (QX 314), as shown in terms of IPSP amplitude (QX 314 single pulse: = 2.858 ± 0.6576 mV; QX 314 train: = 6.44 ± 1.439 mV, p = 0.0103, n = 14), and duration (QX 314 single pulse: 0.285 ± 0.035 s, QX 314 train: 1.173 ± 0.11 s, p < 0.0001, n = 14); left, an example of LLH expression during current-clamp recording of a GC in presence of QX 314 in the patch electrode (5 pulses, 50 Hz). E, An example of the firing pattern of a GC over five trials; traces above correspond to an unstimulated cell, below, stimulated GC with LLH; black arrows indicate single-pulse stimulation and red arrows indicate a five-pulse burst (50 Hz) stimulation of the PP. F, The firing pattern of GCs is altered for an extended period of time (2.12 ± 0.13 s, n = 24) because of LLH expression as revealed by APFDT analysis of the firing pattern of GCs (V_m = −45 mV, 50 ms bins, raw values are normalized to the number of trials per cell); single trials appear in light gray and the mean curve with SEM is in red (running average over 4 points); arrows indicate time of stimulation (black, single pulse; red, train). G, For quantitative summary of APFDT analysis, the total recording time of GCs is split in three time windows as follows (see top scheme, red arrows indicate train stimulation time): time before stimulation (control, 2 s), LLH time (green square overlay, adjusted according to each cell LLH duration), and recovery (2 s). AP frequency in the LLH time window is significantly decreased compared with control and recovery time windows (Ctrl = 0.06 ± 0.004; LLH = 0.01 ± 0.03, Ctrl vs LLH: p < 0.0001, Recovery = 0.08 ± 0.005, Ctrl vs Recovery: p = 0.7565; n = 12 non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.).

Figure 2.

LLH is a multicomponent mechanism. A–D, Firing pattern (top) and quantitative summary of APFDT analysis (bottom) of GCs during PP stimulation (5 pulses at 50 Hz) in ACSF and after consecutively adding GABA_A (BIC, 10 μm), GABA_B (CGP 55845, 20 μm) and mGluR II/III (CPPG, 600 μm) receptor inhibitors; LLH time window is adjusted for LLH duration of each cell measured in ACSF in the beginning of each recording; AP frequency is adjusted to number of trials for each cell, n = 12 (average APFDT values: ACSF Ctrl = 0.07 ± 0.004, ACSF LLH = 0.01 ± 0.003, ACSF Recovery = 0.08 ± 0.005; p < 0.0001; BIC Ctrl = 0.07 ± 0.03, BIC LLH = 0.03 ± 0.004, BIC Recovery = 0.07 ± 0.002; p < 0.0001; BIC/CGP 55845 Ctrl = 0.06 ± 0.003, BIC/CGP 55845 LLH = 0.03 ± 0.004, BIC/CGP 55845 Recovery = 0.06 ± 0.003, p < 0.0001; BIC/CGP 55845/CPPG Ctrl = 0.14 ± 0.003, BIC/CGP 55845/CPPG LLH = 0.14 ± 0.007, BIC/CGP 55845/CPPG Recovery = 0.14 ± 0.004, p = 0.7719). E, Top, LLH duration (1.1 ± 0.07 s, n = 12) decreases when GABA_A (0.8 ± 0.03 s, p = 0.0002, n = 12) and GABA_B (0.4 ± 0.04 s, p < 0.0001, n = 9) receptors are inhibited, and becomes negligible when mGluR II-III antagonist (CPPG) is also added to the bath (0.05 ± 0.006 s, p < 0.0001,n = 9). Bottom, LLH duration also decreased after addition of the mGluR2-specific antagonist LY 341495 (0.8 ± 0.04 s, p = 0.0495, n = 10); additionally blocking GABAA-R with BIC (0.8 ± 0.04, p = 0.730, n = 8) and GABA_B-R with CGP 555845 (0.06 ± 0.01 s, p < 0.0001, n = 6) completely abolished LLH. F, Left, AP frequency during LLH is restored to control levels after blockade of GABA_A, GABA_B and mGluR II-III antagonist CPPG (ACSF LLH APF = 15 ± 5% of control; BIC LLH APF = 41 ± 13% of control; BIC/CGP LLH APF = 54 ± 14% of control and BIC/CGP/CPPG LLH APF = 106 ± 18% of control). Right, mGluR2-specific antagonist increases APF in the LLH time window by ∼35% (39 ± 12% of control; LY 341495 Control = 0.06 ± 0.01, LY 341495 LLH = 0.01682 ± 0.006, LY 341495 Recovery = 0.06744 ± 0.009, Ctrl vs LLH, p = 0.0063, n = 10); LY 341495 in combination with GABA_A/B R antagonists (BIC, CGP 55845, respectively) restores APF similar to control levels (ACSF average LLH duration = 0.9 ± 0.06 s, LY 341495/BIC/CGP 55845 average LLH duration = 0.06 ± 0.01 s, p < 0.001, n = 6; LY 341495/BIC/CGP 55845 LLH APFDT = 83% of control, p = 0.1203; n = 6); non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.

Figure 3.

Positive allosteric modulators (PAMs) of mGluR2 (LY 487379) and GABA_B-R (CGP 7930). A, Application of the mGluR2 PAM–LY 487379. Top, An example trace of the firing pattern of a GC (V_m = −45 ± 2 mV) and corresponding graph showing the normalized APFDT in presence of LY 487379 and LY 341495; LLH expression in the presence of LY 487379 (41 ± 6% of control); and of the mGluR2-specific antagonist LY 341495 (LLH APF = 31 ± 11%). Bottom, LLH duration in the presence of mGluR2 PAM (LY 487379 LLH = 0.7 ± 0.07 s, p = 0.3389, n = 9) and mGluR2 antagonist (0.9 ± 0.09 s, p = 0.1764, n = 9). B, Application of GABA_B-R PAM–CGP 7930. Top, An example trace of the firing pattern of a GC (V_m = −45 ± 2 mV) in the presence of CGP 7930 and corresponding graph showing normalized APFDT in CGP 7930 (36 ± 10% of control, n = 11) and CGP 55845 (49 ± 10% of control, n = 7). Bottom, CGP 7930 increased LLH duration = 1.3 ± 0.09 s (p = 0.0005, n = 11) and CGP 55845 decreased LLH duration in a significant manner (0.6 ± 0.07 s, p < 0.001, n = 7). C, Application of mGluR2 PAM LY 487379 and GABA_B-R PAM CGP 7930 and the respective antagonists LY 341495/CGP 55845. Top, An example trace of the firing pattern of a GC (V_m = −45 ± 2 mV) and corresponding APFDT (LY 487379/CGP 7930 LLH APF = 46 ± 12% of control; LY 341495/CGP 55845 LLH APF = 65 ± 13% of control). Bottom, LLH duration is significantly affected by LY 487379/CGP 7930 (LY 487379/CGP 7930 LLH = 1.17 ± 0.14 s, p = 0.0125, n = 11; LY 341495/CGP 55845 LLH = 0.44 ± 0.05 s, p < 0.001, n = 8; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.).

Figure 4.

LLH is an age-dependent mechanism. A, Single-trial current-clamp traces representative for the firing pattern of GCs(V_m = −45 ± 2 mV) of juvenile (P17), young (P23), and adult (P37) mice during PP stimulation (black arrows; 5 pulses at 50 Hz). Inset, Zoom over superposition of LLH in juvenile (black), young (red), and adult (blue) animals to reveal differential expression and impact over AP firing pattern of these cells. B, Comparison of duration (s) and amplitude (mV) of LLH shows significant increases (Durations: juvenile = 0.392 ± 0.067 s, young = 1.2 ± 0.152 s and adults = 1.6 ± 0.119 s, juvenile vs young: p = 0.0006, n = 6; young vs adults: p = 0.1094, n = 6; Amplitudes: juvenile = 3.2 ± 0.9 mV; young = 8.2 ± 0.74 mV and 11.09 ± 0.56 mV in adults, juvenile vs young: p < 0.0010, n = 6; young vs adults: p = 0.0063, n = 6) with aging from juvenile to young mice with slight increases further into adulthood. C, Summary of increases of LLH in terms of duration (s) and amplitude (mV) in GCs from juvenile to adult animals, reaching maximal values (in this subset of GCs maximal amplitude was 16 mV and maximal duration was 6 s) and plateau in older healthy individuals. Average durations (s): P15+ = 0.6 ± 0.06; P20+ = 1.2 ± 0.08; P31+ = 1.8 ± 0.12; 4 months = 1.8 ± 0.11, 7 months = 1.8 ± 0.26; 9 months = 1.6 ± 0.08; Mean differences: Adults (P31+, n = 6) vs 4–9 months (n = 6), p = 0.290; Average amplitudes (absolute, mV): P15+ = 2.3 ± 0.24; P20+ = 9.3 ± 0.39; P31+ = 10.7 ± 0.16; 4 months = 12 ± 0.39; 7 months = 11 ± 0.3; 9m = 11.4 ± 0.29; Mean differences: Adults (P31+) vs 4–9 months, p = 0.0740; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.

Figure 5.

LLH is frequency-dependent. A, Representative current-clamp single-trial trace of the firing pattern of a GC (V_m = −45 mV) when a train of 5 pulses (black arrows) was delivered to the PP at frequencies of 1 Hz (top) or 50 Hz (bottom). B, Summary graph showing differences in LLH amplitudes (mV) and durations (s) as the stimulation frequency varies from 1 to 200 Hz. Average amplitudes (absolute values, n = 8): 1 Hz = 2.53 ± 0.29; 2 Hz = 2.25 ± 0.38; 5 Hz = 3.19 ± 0.44; 10 Hz = 2.98 ± 0.46; 20 Hz = 3.6 ± 0.59; 50 Hz = 6.27 ± 0.77; 100 Hz = 5 ± 1; 200 Hz = 4.84 ± 1.08; Mean differences of amplitudes: 1 vs 50 Hz, p < 0.0001; 50 vs 100 Hz, p = 0.0560; Average durations (s): 1 Hz = 0.4 ± 0.05; 2 Hz = 0.64 ± 0.15; 5 Hz = 0.8 ± 0.12; 10 Hz = 0.67 ± 0.12; 20 Hz = 0.9 ± 0.15; 50 Hz = 1.1 ± 0.09; 100 Hz = 1.1 ± 0.26; 200 Hz = 0.9 ± 0.16; Mean differences of durations: 1 vs 50 Hz, p < 0.0001; 50 vs 100 Hz, p = 0.9711. C, Peristimulus time histogram of action potential firing over 15 trials during PP stimulation at 1 and 50 Hz (black arrows; red dotted line indicates start of stimulation). D, AP frequency distribution (50 ms bins) of GCs during PP stimulation at 1 Hz (n = 12) and 50 Hz (n = 22) showing higher efficiency in LLH induction when PP was stimulated with higher frequency train (1 Hz: 0.47 ± 0.069 s vs 50 Hz: 2.1 ± 0.226 s); single trials appear in light gray, mean distribution curve with SEM in red; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.

Figure 6.

LLH is activity-dependent. A, Representative current-clamp single-trial trace of the firing pattern of a GC (V_m = −45 mV) when PP is stimulated (black arrows) with either 2 pulses (top) or 10 pulses (bottom) at 50 Hz. B, Summary graph showing increases of LLH duration (s) and amplitude (mV) while increasing number of pulses in the stimulation train from 2 to 20. Mean durations (s), n = 11: 2 pulses = 1.1 ± 0.11; 3 pulses = 1.1 ± 0.14; 5 pulses = 1.45 ± 0.12; 10 pulses = 1.5 ± 0.19; 20 pulses = 1.6 ± 0.27; Mean differences of durations: 2 pulses vs 10 pulses, p = 0.0437; Mean amplitudes (absolute, mV): 2 pulses = 6.19 ± 1.04; 3 pulses = 6.9 ± 1.3; 5 pulses = 9.4 ± 1.1; 10 pulses = 9.3 ± 1.6; 20 pulses = 9.1 ± 1.5; Mean differences of amplitudes: 2 pulses vs 10 pulses, p = 0.1750. C, Peristimulus time histogram of firing of GCs over 10 trials, reveal that increasing the number of pulses in a fast train (50 Hz) from 2 (top) to 10 (bottom), increases the efficiency of the PP-induced hyperpolarization. D, AP frequency distribution of GCs (2 pulses, 1.7 ± 0.25 s, n = 20, 50 ms bin, normalized to number of trials per cell, single trials appear in light gray, average curve and SEM in red) reflecting a longer silence period when the PP received more pulses (10 pulses, 3 ± 0.32 s, n = 15; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.).

Figure 7.

LLH is induced through synaptic activation following specific PP stimulation. A, Confocal images of an acute slice of mouse hippocampus demonstrating expression of ChR2 in the EC and PP fibers (red), and calbindin staining showing cell bodies of GCs (green). Inset, Shows magnified (63×) image of EC fibers expressing ChR2-mCherry and innervating the DG. B, Representative current-clamp traces of a firing dentate granule cell (V_m = −45 mV) during optogenetic stimulation of the PP (5 pulses, 2 ms, 50 Hz) and a single-trial trace; inset zooms over synaptic LLH following activation of PP and disturbing the firing pattern. C, In GCs, optogenetic activation of PP terminals in the molecular layer results in LLH with similar duration (1.19 ± 0.071 s, n = 7) than observed with electrical stimulation (1.67 ± 0.14 s, n = 6; p = 0.0035). D, APFDT analysis of firing GCs shows that optogenetic stimulation of PP (n = 13; single trials appear in light gray and average distribution with SEM in red; black arrows indicate the stimulation time) recreates disruption of the firing pattern. E, Quantitative summary of APFDT analysis; AP frequency is significantly decreased during synaptic LLH (p < 0.0001, n = 13; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or + 0.01; ***: p < or = 0.0001.).

Figure 8.

Nested LLH bursts alter AP frequency distribution during single-input stimulation. A, Simplified schematic representation of the DG showing the stimulation pipette position in the ML during single-input experiments. B, Top left, Representative current-clamp traces of AP firing induced in a GC (subthreshold, V_m = −60 mV) by a single pipette 5 Hz train stimulation, NoLLH (3 s, 200 ms intervals between trains, 3 trains total); black arrows indicate the start of the stimulation and each EPSPs is indicated with a black line. Top right, Representative current-clamp traces (V_m = −60 mV) for LLH/NoLLH stimulation consisted of a 50 Hz bursts (5 pulses) preceding (200 ms) each NoLLH train; red arrow indicates the start of the burst, single pulses indicated by a red line; single trials appear in light gray and red overlay trace represents the average trace of 10 trials. Inset, The enlarged superposition of a NoLLH trace with a LLH/NoLLH trace from the same cell. Note that LLH prevented AP firing in the beginning of the NoLLH train. C, Representative current-clamp traces of a GC firing pattern of over time showing five trials for each NoLLH (black arrows indicates first pulse of stimulation and black lines indicate following EPSPs) alone and when coupled to LLH bursts (LLH/NoLLH; red arrows indicate first pulse and red lines indicate following EPSPs). Note the decrease of APs during LLH/NoLLH stimulation. D, Peristimulus time histogram of AP events in response to PP stimulation over 10 trials (black and red arrows indicate beginning of NoLLH 5 Hz and LLH/NoLLH stimulation, respectively). E, AP frequency distribution (50 ms bins, running average) of GCs during PP NoLLH and LLH/NoLLH stimulation (gray traces represent single cell and red traces represent global average AP frequency distribution, error bars represent SEM; n = 5). F, Top, Schematic representation of APFDT analysis. AP frequency was calculated for each train (3 s) and normalized to the number of trials for each cell during NoLLH and LLH/NoLLH stimulations. Graph below resumes the average AP frequency per train, showing an overall significant decrease when NoLLH is coupled to LLH bursts (average AP frequency per train: NoLLH first train = 0.047 ± 0.01; second train = 0.04 ± 0.007; third train = 0.03 ± 0.005; LLH first train = 0.02 ± 0.009; second train = 0.02 ± 0.009; third train = 0.02 ± 0.008; Mean difference: first train −0.023, p = 0.0004; second train = 0.019, p = 0.0042; third train = 0.02, p = 0.013; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or = 0.01; ***: p < or = 0.0001.).

Figure 9.

Single-input LLH can alter AP frequency distribution of other inputs. A, Simplified schematic representation of the DG showing the of the two stimulation pipettes in the ML during double-input experiments. B, Representative current-clamp traces (V_m = −60 mV) of firing GC during NoLLH input stimulation (5 Hz) and when coupled to the second, LLH input (50 Hz). Inset, Enlarged superposition of NoLLH and LLH input traces. Note the LLH drop induced by the second input bursts (50 Hz) causing loss of NoLLH spikes while it is active. C, Firing pattern of a GC during NoLLH (top) and LLH/NoLLH (bottom) stimulations over time showing that LLH consistently decreases AP fired in response to the initial PP stimulation (NoLLH). D, Peristimulus time histogram of AP events in response of PP stimulation over 10 trials shows decrease number of APs (NoLLH = 4 ± 0.9, LLH/NoLLH= 1.8 ± 0.7 per train, n = 6). E, AP frequency distribution per 5 Hz train (50 ms bins, running average, n = 6) during PP NoLLH and LLH/NoLLH stimulations. F, Quantification of AP frequency average values per train shows a significant decrease during double LLH/NoLLH input compared with NoLLH single-input stimulation (average AP frequency NoLLH: first train = 0.09 ± 0.01, second train = 0.08 ± 0.01, third train = 0.07 ± 0.01; LLH/NoLLH: first train = 0.04 ± 0.006, second train = 0.03 ± 0.005; third train = 0.03 ± 0.004; Mean of differences: first train −0.05405, p < 0.0001; second train −0.04702, p = 0.0022; third train −0.04075, p = 0.0022; non significant (ns): p > 0.05, difference was assumed; *: p < or = 0.05; **: p < or = 0.01; ***: p < or = 0.0001.).