Abstract

How and where hippocampal–neocortical interactions required for memory formation take place is a major issue of current research. Using a combined in vivo functional magnetic resonance imaging/electrophysiology approach, we have investigated whether specific frequencies of CA3 neuronal activation, inducing different forms of short-term plasticity at CA1 synapses, contribute to differential activity propagation in brain-wide networks connected to the hippocampus. We report that localized activation of CA3 neurons in dorsal hippocampus produced activity propagation within the hippocampal formation, including the subiculum and entorhinal cortex, which increased monotonically with frequency to a maximum at 20–40 Hz. However, robust extrahippocampal propagation was seen specifically at theta–beta frequencies (10–20 Hz), reaching a network of midline neocortical and mesolimbic structures. Activation in those regions correlated with a frequency-dependent facilitation of spiking activity recorded in CA1. These results provide a mechanistic link between the dynamic properties of short-term plasticity in the efferent synapses of CA3 neurons in CA1 and activity propagation in brain-wide networks, and identify polysynaptic information channels segregated in the frequency domain.

Introduction

A recurrent theme in the study of memory is the need for interactions between the hippocampus and the neocortex when new information is incorporated (McNaughton and Morris 1987; Buzsaki 1989; Squire 1992; McClelland et al. 1995; Morris 2006; Girardeau et al. 2009). However, the spatiotemporal nature and causal consequence of these physiological interactions are not yet fully understood.

A number of studies using different techniques to map brain activity globally, such as detection of 2DG uptake (Bontempi et al. 1999) and immediate early genes (Maviel et al. 2004; Frankland and Bontempi 2005), have been used to identify distributed networks of structures recruited during specific behavioral tasks or stimulation paradigms. These have been valuable in identifying activation in a set of midline structures of the neocortex following hippocampal-dependent learning, but more dynamic accounts of connectivity that causally relates activity in the hippocampus to cortical regions are desirable. An alternative method is the use of functional magnetic resonance imaging (fMRI) with blood–oxygen level-dependent (BOLD) contrast combined with targeted activation of brain nuclei (i.e., by means of electrical microstimulation) and concomitant electrophysiological recordings (Tolias et al. 2005; Angenstein et al. 2007; Canals, Beyerlein, Murayama et al. 2008; Logothetis et al. 2010; Moreno et al. 2013; Alvarez-Salvado et al. 2014; Weitz et al. 2014). We have used this technique to investigate whether short-term plasticity of synapses, induced by distinct frequencies of CA3 neuronal activation and recorded in CA1, conditions activity transfer from the hippocampus to the neocortex. Using this novel experimental approach, we found several functional connectivity patterns for CA3 depending on its frequency of activation. Overall, our findings point to a mechanistic link between the dynamic properties of hippocampal synapses and those of activity propagation in brain-wide neuronal networks.

Materials and Methods

Animals

Data from 13 male Sprague-Dawley rats (250–300 g) are reported from combined electrophysiology and fMRI studies; a further 5 rats were used in pilot work. Animals were purchased from Janvier Labs (France) and maintained under a 12/12-h light/dark cycle (lights on 07:00–19:00 h) at room temperature (22 ± 2°C), with free access to food and water. Rats were housed in groups of 5 and adapted to these conditions for at least 1 week before experimental manipulation. All experiments were approved by the local authorities (IN-CSIC) and were performed in accordance with Spanish (law 32/2007) and European regulations (EU directive 86/609, EU decree 2001-486).

Carbon-Fiber Electrodes

Glass-coated carbon-fiber bipolar electrodes were developed for the present study based on previous reports (Shyu et al. 2004). Individual 7 µm diameter carbon fibers (Goodfellow Cambridge Limited, England) were used. These consisted of bundles of fibers inserted into a theta-shaped glass capillary (World Precision Instruments) previously pulled to form 7 mm long pipettes with ∼200 µm tip diameter and adjusted to produce an electrical impedance of 40–65 kΩ. A regular wire with a pin connector was attached to the pipette, connected to the carbon fibers using silver conductive epoxy resin (RS Components, UK), and isolated with clear epoxy resin. Afterwards, the tip was bent in a flame to form a 90° angle to minimize implant size, and thereby allow close proximity between the magnetic resonance imaging (MRI) array coil and the head of the animal. These electrodes were stereotaxically implanted to target specific locations in the dorsal hippocampus with a final arrangement as shown in Figure 1 and Supplementary Figure 1.

Figure 1.

Experimental setup. (A) Micrograph of the tip of one of the carbon-fiber bipolar microstimulating electrodes used in the study. CF (carbon fiber), S (septum). (B) Coronal section of Paxinos and Watson atlas (Paxinos and Watson 2007) at 3.55 mm posterior to bregma illustrating the position of the microstimulating (asterisk) and recording (diamond) electrodes. (Inset) Schematic representation of the L-shaped electrodes implanted in the rat dorsal hippocampus. The vertical plane (dark gray) depicts the approximate location of the coronal section from the atlas and the MR image in (D). (C) Electrophysiological field potentials recorded along the dorso-ventral extension of CA1 were used to optimize the effective positioning of the carbon-fiber microstimulation in the contralateral CA3 first, and then to quantify CA1 spiking activity during the experiment. Note the typical EPSP (arrow) and PS recorded in the stratum radiatum and pyramidale, respectively. (D) Representative T2-weighted MRI anatomy with overlaid functional maps evoked by CA3 stimulation. Note the very small artifact in the MR image produced by the carbon-fiber electrode (asterisk) and the virtual absence of artifacts induced by the borosilicate recording electrode (diamond). Scale bars: 100 µm (A); 2 ms, 4 mV (C); 5 mm (D).

Figure 1.

Experimental setup. (A) Micrograph of the tip of one of the carbon-fiber bipolar microstimulating electrodes used in the study. CF (carbon fiber), S (septum). (B) Coronal section of Paxinos and Watson atlas (Paxinos and Watson 2007) at 3.55 mm posterior to bregma illustrating the position of the microstimulating (asterisk) and recording (diamond) electrodes. (Inset) Schematic representation of the L-shaped electrodes implanted in the rat dorsal hippocampus. The vertical plane (dark gray) depicts the approximate location of the coronal section from the atlas and the MR image in (D). (C) Electrophysiological field potentials recorded along the dorso-ventral extension of CA1 were used to optimize the effective positioning of the carbon-fiber microstimulation in the contralateral CA3 first, and then to quantify CA1 spiking activity during the experiment. Note the typical EPSP (arrow) and PS recorded in the stratum radiatum and pyramidale, respectively. (D) Representative T2-weighted MRI anatomy with overlaid functional maps evoked by CA3 stimulation. Note the very small artifact in the MR image produced by the carbon-fiber electrode (asterisk) and the virtual absence of artifacts induced by the borosilicate recording electrode (diamond). Scale bars: 100 µm (A); 2 ms, 4 mV (C); 5 mm (D).

Electrode Implantation and Electrophysiology

All experiments were performed under urethane anesthesia (1.3 g/kg, i.p.). Stimulating and recording electrodes were implanted using standard surgical and stereotaxic procedures, as described previously (Canals, Beyerlein, Keller et al. 2008; Canals, Beyerlein, Murayama et al. 2008; Canals et al. 2009). A carbon-fiber bipolar stimulating electrode was positioned in the CA3 region of the dorsal hippocampus (from bregma: 3.5 mm anteroposterior and 3.5 mm lateral, initial position 3.6 mm ventral to the dural surface). An electrophysiological multichannel recording electrode (single shank, 100 µm contact spacing, 32 channels; Neuronexus Technologies) was targeted to the contralateral CA1 (3.5 mm caudal and 2.5 mm lateral from bregma). The final position of the bipolar carbon electrode was adjusted to the stratum pyramidale according to electrophysiological potentials recorded in CA1 (Fig. 1C). Stimulation along the implantation tract of the carbon-fiber electrode, from dorsal stratum oriens to the ventral border of CA3 demonstrates 2 peaks of excitatory postsynaptic potentials (EPSPs) recorded in the contralateral CA1, the first evoked by the activation of Schaffer collateral/commissural axons in stratum radiatum and the second by the depolarization of CA3 neurons in the soma cell layer (Supplementary Fig. 2). The second one was selected for the present experiments.

The multichannel electrode was then substituted by an MRI-compatible L-shaped single channel borosilicate electrode to record the population spike (PS) in the contralateral CA1 (Supplementary Fig. 1). Stimulating and recording electrodes were secured to the skull with dental cement (Heraeus Medical, Wehrheim, Germany) and the animal then transferred to the scanner. These arrangements ensured the specificity of the stimulation during fMRI acquisition and simultaneous electrophysiological recording.

After filtering (0.1 Hz–3 kHz) and amplification, the electrophysiological signals were digitized (20 kHz acquisition rate) and stored for offline processing with MatLab (The MathWorks, USA) and Spike2 (Cambridge Electronic Design, Cambridge, USA). The EPSP, reflecting the population synaptic response, was measured as the maximal slope of the negative going field potential recorded in CA1 stratum radiatum, and the PS in the CA1 pyramidal layer, reflecting the summed spiking activity of the recorded population, as the amplitude between the maximal negativity and the preceding positive crest. To correlate the fast electrophysiological responses and the slow BOLD signals for every stimulation protocol used (see below), we calculated the mean PS amplitude in response to all pulse in a stimulation train (equivalent to the area under the curves in Fig. 4D). Those integrated measurements of spiking activity were then correlated to the corresponding BOLD signal amplitudes (see below).

Stimulation Protocols

Charge balanced biphasic 0.1-ms duration pulses were delivered through the carbon-fiber stimulating electrode with a constant current source and a pulse generator (STG2004, Multichannel Systems, Reutlingen, Germany). Importantly, none of the stimulation protocols used induced episodes of spontaneous (epileptiform) activity (Supplementary Fig. 3).

For fMRI data acquisition, multiple stimulation protocols were used (Supplementary Fig. 4). Each protocol consisted of 10 trains of stimulation with trains repeated every 30 s (300 s per protocol) and protocols repeated 3 times per condition. In a first set of conditions, the trains were designed to maintain a constant charge transfer by fixing the total number of pulses to 40 and electric pulse width to 100 µs (biphasic), but varying the stimulation frequencies between 5, 10, 20, and 40 Hz. These protocols yielded stimulation trains of 8, 4, 2, and 1 s, respectively, and were presented in random order. In a second set of conditions, the different stimulation trains were mixed in a single protocol in pseudorandom order (Fig. 4A). In a third set of conditions, the protocols were designed with a constant duration of the stimulus train at 10 and 40 Hz (4 s), thereby delivering a different number of pulses (40 and 160, respectively); or in which the stimulation frequency was increased to 160 Hz (1 s, 160 pulses/train). A final set of conditions consisted of 40 Hz stimulation trains structure into theta- and delta-burst protocols. In these protocols, one stimulation period consisted of short trains of 4 and 8 pulses separated by off periods of 250 and 600 ms and repeated 10 and 5 times, respectively. Stimulation periods were repeated 10 times as before.

MRI Experiments and Data Analysis

For the MRI experiments, the previously prepared urethane-anesthetized animals were placed in a custom-made animal holder with adjustable bite and ear bars, and positioned on the magnet bed. The animals were constantly supplied with 0.8 L/m O2 with a face mask and temperature is kept between 37.0 and 37.5°C through a water heat-pad. The temperature, heart rate, SpO2, and breathing rate were monitored throughout the session (MouseOx, Starr Life Sciences, Oakmont, US). The experiments were carried out in a horizontal 7-T scanner with a 30 cm diameter bore (Biospec 70/30, Bruker Medical, Ettlingen, Germany).

Acquisition was performed in 15 coronal slices using a GE-EPI sequence applying the following parameters: (field of view) FOV, 25 × 25 mm; slice thickness, 1 mm; matrix, 96 × 96; segments, 1; FA, 60°; (time echo) TE, 15 ms; (time repetition) TR, 2000 ms. T2-weighted anatomical images were collected using a rapid acquisition relaxation enhanced sequence (RARE): FOV, 25 × 25 mm; 15 slices; slice thickness, 1 mm; matrix, 192 × 192; TEeff, 56 ms; TR, 2 s; RARE factor, 8. A 1H rat brain receive-only phased-array coil with integrated combiner and preamplifier, and no tune/no match, was employed in combination with the actively detuned transmit-only resonator (Bruker BioSpin MRI GmbH, Germany).

Functional MRI data were analyzed offline using our own software developed in MatLab, which included Statistical Parametric Mapping package (SPM8, www.fil.ion.ucl.ac.uk/spm), Analysis of Functional NeuroImages (AFNI, http://afni.nimh.nih.gov/afni) and FSL Software (FMRIB http://fsl.fmrib.ox.ac.uk/fsl/). After linear detrending, temporal (0.015–0.2 Hz) and spatial filtering (3 × 3 gaussian kernel of 1.5 sigma) of voxel time series, a general linear model (GLM) or cross-correlation analysis was applied with a simple boxcar model shifted forward in time, typically by 2 s, or a boxcar convolved with the hemodynamic response function (HRF) (MatLab). The results were largely comparable with all methods tested (GLM with HRF was used in Fig. 3 and cross-correlation analysis results with HRF are shown in Supplementary Fig. 5). Functional maps were generated from voxels that had a highly significant (P < 0.0001) component for the model and were clustered together in space (cluster size = 14; calculated with Monte Carlo simulation implemented in AFNI).

Regions of interest (ROIs) extracted using a rat atlas registered to the functional images (Schwarz et al. 2006) (Supplementary Fig. 6) were used to compute the amplitude of the evoked BOLD signal responses (as a percentage relative to a prestimulus baseline of 6 s) and volume of brain tissue activated in absolute terms (number of voxels above the statistical threshold) or relative to the ROI (number of active voxels divided by the total number of voxels in the region). The maps shown in Figure 3 represent the group probability for each voxel of being activated by a particular stimulation protocol. A voxel probability of 1 means that 9 out of 9 animals used in that study showed coincident activation of that particular voxels to a certain protocol. The resulting maps are color-coded and thresholded so that probabilities <0.33 (activations found in only 3 out of 9 animals) are not represented. Shown are 7 coronal slices from caudal (left, −8 mm from bregma) to rostral (right, +4 mm from bregma).

Histology

After completion of each experiment, the rats were perfused intracardially with 100 mL of 1% phosphate-buffered saline (PBS) solution and 50 mL of ice-cold 4% paraformaldehyde (PFA). Brains were kept for 24 h on 4% PFA post-fixation at 4° and cut in a fixed material vibratome in 50 µm thick slices. Slices were then stained with 4′,6-diamidino-2-phenylindole (DAPI) for photography under a fluorescence microscope. The position of the electrodes was confirmed by the observable damage of the insertion.

Results

We have combined fMRI and local field potential (LFP) recordings in the rat (Canals, Beyerlein, Murayama et al. 2008; Canals, Beyerlein et al. 2009) to investigate brain-wide patterns of activity propagation as a function of the frequency of neuronal activation in the CA3. A first step was the creation of small bipolar carbon-fiber microstimulating electrodes (Fig. 1A) that preserve the quality of MR images while producing highly localized depolarization (Supplementary Fig. 1). We implanted these and MRI-compatible recording electrodes precisely guided by layer-specific potentials to target the stratum pyramidale of CA3 and CA1, respectively (Fig. 1B, Supplementary Fig. 2). Bipolar depolarizing pulses delivered through the CA3 carbon-fiber electrodes generated localized LFP activation within CA1 (Fig. 1C), but also more widespread fMRI signals that we used as readout of activity propagated from the CA3 network (Fig. 1D).

Activity Propagation in Brain-Wide Hippocampal Networks

We used fMRI stimulation protocols consisting of 10 repetitions of an ON–OFF sequence with a duty-cycle of 30 s, known to produce repetitive and reliably stable BOLD signals (Canals, Beyerlein, Murayama et al. 2008). Stimulation parameters during the ON-period were chosen according to the temporal characteristics of action potential activity of hippocampal neurons observed in the behaving animal (Ranck 1973; Berger et al. 1983). Stimulation in all experiments was set at an intensity evoking a PS in CA1 (Fig. 2A) at 50% of the maximal amplitude identified via an input–output curve (Fig. 2B). Electrophysiological and BOLD signals were recorded during stimulus presentations (Fig. 2CE). We first applied a constant number of depolarizing pulses (40 per duty cycle) to CA3 pyramidal cells, delivered at 4 different frequencies (from 5 to 40 Hz) (Supplementary Fig. 4A–D). This ensures a constant charge transfer at all frequencies. Robust BOLD signal time-courses were routinely obtained in response to every ON-period, both at individual (Fig. 2D,E) and group level (Fig. 2F) analyses.

Figure 2.

Frequency-dependent gating of activity propagation: electrophysiological and BOLD responses. (A) Representative PS recorded in CA1 evoked by single pulse stimulation of contralateral CA3. Data in panels (A)–(E) come from the same animal. (B) Stimulus–response curve for PS amplitude at increasing stimulation intensities. Data represent mean ± SEM of 5 repetitions of each intensity. Final intensity for train stimulation was selected as the one producing 50% of the maximal PS amplitude. (C) PS amplitude measured in the course of a 5 Hz stimulation train (40 pulses in total). Data represent mean ± SEM averaged across 10 train repetitions. (D) Representative BOLD signal time courses in the hippocampus and (E) PFC of the same animal in response to 10 trains of stimulation (indicated by vertical gray bars), delivered at 5 Hz (red), 10 Hz (orange), 20 Hz (green), or 40 Hz (blue). Each panel shows the mean ± SEM of 3 repetitions of the same protocol. (F) Average BOLD signal time courses across animals (n = 9) for the hippocampus (HC) and PFC for each of the frequencies applied. Same color code as in (A–E). Note that BOLD signal responses in the hippocampus are elicited at all stimulation frequencies, contrary to the PFC in which BOLD responses are only evoked at 10–20 Hz stimulation frequencies. Scale bars: 2 mV, 5 ms (A); 20 s (D,E); 3 s, 1% BOLD (F).

Figure 2.

Frequency-dependent gating of activity propagation: electrophysiological and BOLD responses. (A) Representative PS recorded in CA1 evoked by single pulse stimulation of contralateral CA3. Data in panels (A)–(E) come from the same animal. (B) Stimulus–response curve for PS amplitude at increasing stimulation intensities. Data represent mean ± SEM of 5 repetitions of each intensity. Final intensity for train stimulation was selected as the one producing 50% of the maximal PS amplitude. (C) PS amplitude measured in the course of a 5 Hz stimulation train (40 pulses in total). Data represent mean ± SEM averaged across 10 train repetitions. (D) Representative BOLD signal time courses in the hippocampus and (E) PFC of the same animal in response to 10 trains of stimulation (indicated by vertical gray bars), delivered at 5 Hz (red), 10 Hz (orange), 20 Hz (green), or 40 Hz (blue). Each panel shows the mean ± SEM of 3 repetitions of the same protocol. (F) Average BOLD signal time courses across animals (n = 9) for the hippocampus (HC) and PFC for each of the frequencies applied. Same color code as in (A–E). Note that BOLD signal responses in the hippocampus are elicited at all stimulation frequencies, contrary to the PFC in which BOLD responses are only evoked at 10–20 Hz stimulation frequencies. Scale bars: 2 mV, 5 ms (A); 20 s (D,E); 3 s, 1% BOLD (F).

The key finding was that while BOLD signal recorded in hippocampus responded to all stimulation frequencies (Fig. 2D,F), it only responded to specific frequencies in extrahippocampal structures, such as the prefrontal cortex (PFC) (Fig. 2E,F). Activity propagation across dorsal and ventral structures of the hippocampal formation including the CA3, CA1, subiculum, entorhinal cortex (EC), and dentate gyrus (DG), as well as the septum, increased with frequency, reaching a plateau at 20–40 Hz (Supplementary Fig 3 and Supplementary Fig. 5, compare with Supplementary Fig. 6). Strikingly, at 10 and 20 Hz, with a peak in the latter, activity spread extensively beyond the hippocampus into neocortical and subcortical regions (Fig 3 and Supplementary Fig. 5). The brain regions receiving this hippocampal activity largely involved structures in midline anterior neocortical networks such as the orbital frontal (OC), medial prefrontal (mPFC), cingulate (Cg), and retrosplenial cortex (Rs), as well as the dorsal (Str) and ventral striatum (Acc). Parietal cortex (PC) was also activated. However, the extrahippocampal spread of activity abruptly declined at 40 Hz (Fig. 3; Supplementary Fig. 5), remaining circumscribed within the hippocampal formation with maximal BOLD amplitude. Therefore, the probability of regional activation within and beyond the hippocampus was highly dependent on the temporal properties of CA3 efferent activation. Detailed analyses of both the magnitude and probability of BOLD demonstrated that this biphasic patter of activity propagation was independent of the statistical threshold applied (P < 10−1 to 10−6; Supplementary Movie 1). The results suggest that hippocampal–neocortical interactions are boosted for activity patterns occurring at mid-frequency ranges.

Figure 3.

Brain-wide and frequency-dependent functional connectivity of dorsal CA3. (A) fMRI BOLD maps (group data, n = 9) overlaid on an anatomical T2-weighted images. Color-code denotes the probability of activating a region by each specific stimulation protocol (denoted on the left column). Representative individual fMRI maps (P < 0.0001, cluster size 14) used to build the probability map shown here are presented in Supplementary Figure 4. Note the extensive extrahippocampal propagation of activity to specific neocortical and subcortical regions at 10–20 Hz. Dorsal and ventral hippocampus (HCd, HCv), entorhinal cortex (Ent), striatum (Str), nucleus accumbens (Acc), retrosplenial (Rs), cingulate (Cg), prelimbic (PrL), infralimbic (IL), and orbitofrontal cortex (OC). (B) Color-coded percentage of brain ROI's activated by a particular stimulation protocol (P < 0.0001, cluster size 14). Two-way ANOVA reveals highly significant effects of frequency (F3, 840 = 189.49, P < 0.0001), ROI (F34, 280 = 21.43, P < 0.0001), and interaction (F102, 840 = 3.91, P < 0.0001). Bonferroni's multiple comparison test shown in Supplementary Table 1. Anatomical locations and abbreviations for all brain ROI's provided in Supplementary Figure 7. Scale bars: 5 mm (A).

Figure 3.

Brain-wide and frequency-dependent functional connectivity of dorsal CA3. (A) fMRI BOLD maps (group data, n = 9) overlaid on an anatomical T2-weighted images. Color-code denotes the probability of activating a region by each specific stimulation protocol (denoted on the left column). Representative individual fMRI maps (P < 0.0001, cluster size 14) used to build the probability map shown here are presented in Supplementary Figure 4. Note the extensive extrahippocampal propagation of activity to specific neocortical and subcortical regions at 10–20 Hz. Dorsal and ventral hippocampus (HCd, HCv), entorhinal cortex (Ent), striatum (Str), nucleus accumbens (Acc), retrosplenial (Rs), cingulate (Cg), prelimbic (PrL), infralimbic (IL), and orbitofrontal cortex (OC). (B) Color-coded percentage of brain ROI's activated by a particular stimulation protocol (P < 0.0001, cluster size 14). Two-way ANOVA reveals highly significant effects of frequency (F3, 840 = 189.49, P < 0.0001), ROI (F34, 280 = 21.43, P < 0.0001), and interaction (F102, 840 = 3.91, P < 0.0001). Bonferroni's multiple comparison test shown in Supplementary Table 1. Anatomical locations and abbreviations for all brain ROI's provided in Supplementary Figure 7. Scale bars: 5 mm (A).

Additional control experiments were performed to test whether repetitive presentation of activity patterns (10 per trial) may condition activity propagation. To this end, all the frequency protocols were scheduled in random order in a single fMRI run. The BOLD signals evoked by such protocol were unchanged, and are shown for the hippocampus and PFC (Fig. 4A), together with the amplitude quantification (Fig. 4B,C). The unchanged nature of the frequency-dependence function indicates that transitions between propagating and nonpropagating activation modes are highly dynamic and independent of stimulation order. A conditional crosstalk between protocols is unlikely.

Figure 4.

Short-term plasticity facilitates CA1 output and boosts system-level interactions. (A) fMRI BOLD signals evoked in the hippocampus (upper panel) and PFC (lower panel) by a pseudorandom presentation of stimulation frequencies. Inset indicates the time of presentation of the different protocols. Colors in all panels represent stimulation frequencies as follows: red (5 Hz), orange (10 Hz), green (20 Hz), and blue (40 Hz). Values represent mean ± SEM. (B) BOLD signal amplitude (% from baseline, n = 9, P < 0.0001, cluster size 14) in dorsal hippocampus and (C) PFC across stimulation frequencies. PFC is composed of frontal, orbital, prelimbic, infralimbic, cingulate, and insular cortices. Note the monotonic increase in hippocampal BOLD signals amplitude (repeated-measures one-way ANOVA F3,24 = 37.05, P < 0.0001) versus the inverted U-shape response for PFC (repeated-measures one-way ANOVA F3,24 = 31.80, P < 0.0001). (D) Representative example in one animal showing the evolution of PS amplitudes in the stimulation train at frequencies ranging from 5 to 40 Hz. Data represents the mean (solid line) and SEM (shaded area) of 10 repetitions for each frequency. (E) Quantification of the output spiking activity at all tested frequencies as the PS amplitude averaged across all delivered pulses, and normalized against 5 Hz. (Inset) Representative PS evoked by the first (black) and last (colored) pulse in the train for each stimulation frequency. Note the PS facilitation at 10 and 20 HZ (repeated-measures one-way ANOVA F3,12 = 8.06, P = 0.0007). (F) Linear correlation between BOLD signals in PFC and spiking activity in CA1 (F1,34 = 10.09, P = 0.0032, r2 = 0.38). Multiple comparisons in all panels with Bonferroni's test: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Asterisks color denotes comparison against 10 (yellow), 20 (green), or 40 (blue) Hz. Scale bars: 5 ms, 2 mV (C).

Figure 4.

Short-term plasticity facilitates CA1 output and boosts system-level interactions. (A) fMRI BOLD signals evoked in the hippocampus (upper panel) and PFC (lower panel) by a pseudorandom presentation of stimulation frequencies. Inset indicates the time of presentation of the different protocols. Colors in all panels represent stimulation frequencies as follows: red (5 Hz), orange (10 Hz), green (20 Hz), and blue (40 Hz). Values represent mean ± SEM. (B) BOLD signal amplitude (% from baseline, n = 9, P < 0.0001, cluster size 14) in dorsal hippocampus and (C) PFC across stimulation frequencies. PFC is composed of frontal, orbital, prelimbic, infralimbic, cingulate, and insular cortices. Note the monotonic increase in hippocampal BOLD signals amplitude (repeated-measures one-way ANOVA F3,24 = 37.05, P < 0.0001) versus the inverted U-shape response for PFC (repeated-measures one-way ANOVA F3,24 = 31.80, P < 0.0001). (D) Representative example in one animal showing the evolution of PS amplitudes in the stimulation train at frequencies ranging from 5 to 40 Hz. Data represents the mean (solid line) and SEM (shaded area) of 10 repetitions for each frequency. (E) Quantification of the output spiking activity at all tested frequencies as the PS amplitude averaged across all delivered pulses, and normalized against 5 Hz. (Inset) Representative PS evoked by the first (black) and last (colored) pulse in the train for each stimulation frequency. Note the PS facilitation at 10 and 20 HZ (repeated-measures one-way ANOVA F3,12 = 8.06, P = 0.0007). (F) Linear correlation between BOLD signals in PFC and spiking activity in CA1 (F1,34 = 10.09, P = 0.0032, r2 = 0.38). Multiple comparisons in all panels with Bonferroni's test: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Asterisks color denotes comparison against 10 (yellow), 20 (green), or 40 (blue) Hz. Scale bars: 5 ms, 2 mV (C).

Short-Term Plasticity at CA3→CA1 Synapses and Activity Propagation

Frequency-dependent and short-term changes in synaptic transmission in the Schaffer collateral and spiking activity in CA1 have been previously described from low-to-high stimulation frequencies (Alger and Nicoll 1982; Herreras et al. 1987; Dutar and Nicoll 1988; Davies and Collingridge 1993; Davies and Collingridge 1996). Accordingly, we measured CA1 PS amplitude corresponding to the 4 stimulation protocols (5, 10, 20, and 40 Hz) recorded in our combined fMRI-electrophysiology experiments. We observed that spiking activity was largely facilitated in the range of propagating frequencies (10–20 Hz). However, comparable facilitation did not occur at 5 Hz and fell away at 40 Hz (Fig. 4D). We computed the averaged PS amplitude across all delivered pulses as an integral measure of the output spiking activity, for each animal at each tested frequency. The results of this analysis unveiled an inverted U-shape curve with maxima at 10–20 Hz and minima at 5 and 40 Hz (Fig. 4E), reminiscent of those found for the fMRI signals in extrahippocampal areas (Fig. 4C). Accordingly, we plotted fMRI magnitude as a function of PS amplitude and observed a liner correlation. This finding suggests a mechanistic link between local PS facilitation as a consequence of short-term plasticity and activity propagation to extrahippocampal targets (Fig. 4F).

Frequency Determines Hippocampal–neocortical Interactions

Further control studies were necessary as it should be noted that, in addition to alteration in frequency, the choice of 40 pulses per duty cycle meant that the duration of stimulation covaried. We therefore dissected the potential contributions of stimulus frequency versus duration. Train duration was kept constant to 4 s, with the stimulation frequency alternating between 10 and 40 Hz, thereby delivering 40 and 160 pulses, respectively (Supplementary Fig. 4, compare B,E). Extensive and indistinguishable activations in the hippocampal formation were found with both protocols (Fig. 5A), with neocortical propagation occurring at 10 Hz but not at 40 Hz (Fig. 5B), even though 4 times more pulses were delivered to CA3 in the second protocol. In a next set of experiments, even higher stimulation frequencies were tested by delivering 160 pulses in 1 s (160 Hz; Supplementary Fig. 4F), leading to the identical result as 40 Hz stimulation (Fig. 5). Given observations suggesting that spike amplitude would first climb and then decline at 40 Hz (i.e., Fig. 4D), we also examined high frequency (40 Hz) stimulation bursts of 4–8 pulses separated by 250 or 600 ms, respectively (theta-to-delta burst stimulation). Once again, we observed that this produced extensive hippocampal activation in the absence of extrahippocampal propagation, (Fig. 5, data for both burst protocols pooled in the same analysis for simplicity). Overall, these findings robustly indicate that the frequency of hippocampal activation is the critical factor determining the extent of propagation to cortical and subcortical structures.

Figure 5.

Frequency is the key parameter determining extrahippocampal activity propagation. (A) BOLD signal responses in the hippocampus and (B) PFC to different stimulation protocols varying the frequencies, number of pulses (p) and structure (bursts). Burst protocols consisted of 4 or 8 pulses separated by 250 or 600 ms, respectively. Both burst protocols produced the same result and has been pooled together for simplicity. Data points represent mean ± SEM (n = 3). Note the absence of extrahippocampal propagation at 40 Hz even if long (4 s) stimulation trains equivalent to those used at 10 Hz with 4 times the number of pulses are used. Note also the absence of propagation at higher frequencies (160 Hz) or when 40 Hz activation is patterned into theta- or delta-burst protocols. Repeated-measures One-way ANOVA for hippocampal (F4,8 = 1.9, P = 0.2) and PFC activations (F4,8 = 127.9, P < 0.0001). Multiple comparisons in all cases with Bonferroni's test: ****P < 0.0001; NS (nonsignificant differences, P > 0.05).

Figure 5.

Frequency is the key parameter determining extrahippocampal activity propagation. (A) BOLD signal responses in the hippocampus and (B) PFC to different stimulation protocols varying the frequencies, number of pulses (p) and structure (bursts). Burst protocols consisted of 4 or 8 pulses separated by 250 or 600 ms, respectively. Both burst protocols produced the same result and has been pooled together for simplicity. Data points represent mean ± SEM (n = 3). Note the absence of extrahippocampal propagation at 40 Hz even if long (4 s) stimulation trains equivalent to those used at 10 Hz with 4 times the number of pulses are used. Note also the absence of propagation at higher frequencies (160 Hz) or when 40 Hz activation is patterned into theta- or delta-burst protocols. Repeated-measures One-way ANOVA for hippocampal (F4,8 = 1.9, P = 0.2) and PFC activations (F4,8 = 127.9, P < 0.0001). Multiple comparisons in all cases with Bonferroni's test: ****P < 0.0001; NS (nonsignificant differences, P > 0.05).

Discussion

The main findings of this study are that (1) localized activation of a discrete area of CA3 neurons propagates across hippocampal subfields, including the Schaffer collateral/commissural output of CA3 to CA1, and may then invade spatially diverse neocortical and subcortical territories; (2) propagation inside the hippocampal formation increases monotonically with frequencies from 5 to 40 Hz, reaching a plateau at 20–40 Hz; (3) robust propagation beyond the hippocampal formation occurs specifically at frequencies between 10 and 20 Hz, independently of train duration; (4) extrahippocampal activity propagation in fMRI measurements correlates with the magnitude of spiking activity recorded in CA1, which is dynamically determined by short-term plasticity; (5) stimulation at 10 and 20 Hz induces frequency facilitation of CA1 PS output; and (6) the afferent neurons receiving propagated hippocampal activity configure a network of frontoparietal neocortical structures and mesolimbic dopaminergic regions.

Synaptic Physiology Level

Our results suggest that the output of CA3 implements a frequency-dependent gating mechanism that promotes long-range interactions at 10–20 Hz. The primary aim of our study was to delineate the characteristics of any hippocampal gating of propagation rather than identify mechanisms, but several possibilities can be envisioned based on the previous work. For example, we show that the spiking activity of CA1 pyramidal neurons is facilitated in a train at propagating frequencies, a finding in good agreement with extensive previous literature in hippocampal slices demonstrating frequency-facilitation of neural transmission in CA1 at theta–beta ranges (Alger and Nicoll 1982a, b; Dutar and Nicoll 1988; Davies and Collingridge 1993; Davies and Collingridge 1996). This facilitation has been shown to be the combined result of presynaptic GABAB-mediated decrease in synaptic inhibition and the concomitant enhancement of glutamate release. Conversely, at higher frequencies (starting at 40–50 Hz), a GABAA-mediated inhibitory current supported by recurrent interneuronal networks builds up progressively in the train strongly curtailing CA1 pyramidal cell firing (Pouille and Scanziani 2004). The dynamic nature of these opposing short-term plasticity processes is clearly illustrated in the behavior of the PS amplitude at 40 Hz (i.e., Fig. 4D, blue tracing) in which an initial facilitation is then efficiently curtailed after several pulses.

It is therefore likely that the range of frequency-dependent facilitation of activity propagation beyond hippocampus found in our study is determined by interplay of the biophysical properties of pre- and postsynaptic elements in the CA3–CA1 circuits, including principal cells and interneurons, and their short-term plasticity. While the failure of fMRI signals to propagate at high frequencies occurs simultaneously with maximal amplitude signals locally in the hippocampus, the failure at lower frequencies coincides with the lowest activation inside the hippocampus (Fig. 4). We hypothesize (see above) that the lack of spiking facilitation at low frequencies precludes the boosting of fMRI signals propagation, however, a contribution of limited sensitivity to detect BOLD signals at low frequencies cannot be fully discarded (Canals, Beyerlein, Murayama et al. 2008). Future electrophysiological recordings simultaneously in the hippocampus and those cortical regions targeted by activity propagation will try to clarify this extent focusing on ongoing activity propagation (Fernández-Ruiz et al. 2012). Also, a potential contribution of the septum, which receives monosynaptic inputs from CA3 neurons and is activated in our fMRI experiments, cannot be ruled out with the present experiments. In addition to the identified functional roles of the dynamic CA3–CA1 interaction such as the modulation of synaptic plasticity (Davies et al. 1991; Mott and Lewis 1991; Manabe et al. 1993; Perkel and Nicoll 1993) and temporal fidelity of pyramidal cell firing (Pouille and Scanziani 2001; Schaefer et al. 2006), our results suggest a new role of this local hippocampal circuit in routing activity propagation according to input frequency.

Synapse to Network Transformations

Activation of CA3 neurons at 10–20 Hz is routed towards a distributed but highly reproducible network of brain structures known to be critically related to memory formation, including the PFC. Interestingly, in previous fMRI studies we showed that long-term potentiation (LTP) of synaptic strength in the DG—the accepted experimental analog of learning at the synaptic level—induces a functional reorganization of the network such that test stimulation delivered at 10 Hz which produces activations initially circumscribed to the hippocampal formation, started to propagate to the PFC and ventral striatum (Canals, Beyerlein et al. 2009). Similarly, electrophysiological evidence from other groups further has shown that LTP induction in the same pathway facilitates the disynaptic transfer of activity at 10 Hz from the DG to CA3 (Yeckel and Berger 1998). In this context, the previously described functional reorganization of hippocampal outputs after LTP induction in the DG (Canals, Beyerlein et al. 2009) could be the result of 2 processes coordinated in the frequency domain (theta–beta range), (1) the enhanced transfer of activity from DG to CA3 (Yeckel and Berger 1998) and (2) the specific channeling to neocortical and striatal regions supported by the CA3 network as shown here. In the context of memory formation, it is tempting to hypothesize that changes of synaptic strength at particular network inputs (i.e., EC→DG) and the pacing of information coding in frequency bands favoring long-range hippocampal interactions (i.e., 10–20 Hz) may represent one possible routing instruction for hippocampal–neocortical interactions.

The present fMRI study draws attention, by virtue of observation of widespread neuronal activations, to important network implications of short-term synaptic plasticity. Whether long-term modulations of the same synapses have further functional consequences on the described frequency-dependent gating of activity propagation is an interesting question that requires further investigation.

Systems Level

Work over the past years points to cortical oscillations as a framework for organizing interactions among functionally specialized neurons in distributed brain networks (Singer 1993; Buzsaki 2006; Colgin and Moser 2010; Engel and Fries 2010; Lisman and Jensen 2013). For example, beta oscillations have recently been identified as a potential mechanism for synchronizing the evolution of neural representations in dispersed neural circuits after hippocampal-dependent learning (Igarashi et al. 2014). Similarly, theta rhythms have been proposed to coordinate hippocampal–prefrontal interactions in a spatial memory task (Jones and Wilson 2005; Siapas et al. 2005). Here, using a combination of fMRI and electrophysiological techniques, we demonstrate activity propagation to a distributed network of structures connected with the hippocampus at 10–20 Hz. Furthermore, long-range propagation at these frequencies to midline neocortical, prefrontal and sensory-motor networks together with ventromedial striatum (i.e., see Fig. 3B) is consistent with widespread beta oscillations recorded in those same regions commonly found both in sensory, attentional, and learning processes (Fries et al. 2001; Ravel et al. 2003; Martin et al. 2007; Berke et al. 2008; Colgin et al. 2009; Engel and Fries 2010; Howe et al. 2011; Igarashi et al. 2014). We are aware of the fundamental differences between recorded LFP signals oscillating at different frequencies and reflecting the orchestrated organization of multiple synaptic inputs (not necessarily matching the firing frequency of individual cells), and the activation frequencies experimentally imposed in our study by direct depolarizing pulses. We believe, however, based on the strong correspondence between the stimulation parameters used here and the temporal characteristics of action potential activity of hippocampal neurons observed in the behaving animal (Ranck 1973; Berger et al. 1983), that our results help to understand long-range interactions across neuronal populations. Although the specific range of relevant frequencies may vary for naturally occurring neuronal activations, the concept of frequency-dependent gating of activity propagation shown in our study and, most importantly, the unveiled polysynaptic routes of communication, would still be valid. Activity propagation segregated in the frequency domain most likely illustrates direct information transfer across brain regions, as discussed above, but may also represent a mechanism by which the CA3 output in the dorsal hippocampus coordinates the oscillatory framework that synchronizes activity across larger and functionally related territories. While a high level of coherence is a likely signature of long-range communication (von Stein and Sarnthein 2000), how such coherence is achieved mechanistically remains unclear. Overall, our results suggest that direct activity transfer at theta–beta frequencies between the hippocampus and connected structures may drive transient windows of strong functional coupling, organizing oscillations in dispersed networks and enhancing coherence as a consequence.

Concluding Remarks

Our results highlight the important role of short-term plasticity at CA3 synapses to route activity propagation and determine hippocampal–neocortical interactions. We show that activity paced at 10–20 Hz facilitates CA1 pyramidal cell output and promotes long-range interactions with a set of neocortical and striatal regions previously shown to be critically involved in memory formation.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org

Funding

This work was supported by grants from the Spanish MINECO (BFU2012-39958) and CONSOLIDER (CSD2007-00023) to S.C., a Royal Society International Exchange Grant and by an Advanced Investigator Grant from the European Research Council to R.M. A.M. holds a studentship from the University of Edinburgh. The Instituto de Neurociencias is “Centre of Excellence Severo Ochoa”.

Notes

We thank Begoña Fernández, Jesús Pacheco-Torres and Vicente Pallarés for excellent technical assistance. Conflict of Interest: None declared.

References

Alger
BE
,
Nicoll
RA
.
1982
.
Feed-forward dendritic inhibition in rat hippocampal pyramidal cells studied in vitro
.
J Physiol
 .
328
:
105
123
.
Alger
BE
,
Nicoll
RA
.
1982
.
Pharmacological evidence for two kinds of GABA receptor on rat hippocampal pyramidal cells studied in vitro
.
J Physiol
 .
328
:
125
141
.
Alvarez-Salvado
E
,
Pallares
V
,
Moreno
A
,
Canals
S
.
2014
.
Functional MRI of long-term potentiation: imaging network plasticity
.
Philos Trans R Soc Lond B Biol Sci
 .
369
:
20130152
.
Angenstein
F
,
Kammerer
E
,
Niessen
HG
,
Frey
JU
,
Scheich
H
,
Frey
S
.
2007
.
Frequency-dependent activation pattern in the rat hippocampus, a simultaneous electrophysiological and fMRI study
.
Neuroimage
 .
38
:
150
163
.
Berger
TW
,
Rinaldi
PC
,
Weisz
DJ
,
Thompson
RF
.
1983
.
Single-unit analysis of different hippocampal cell types during classical conditioning of rabbit nictitating membrane response
.
J Neurophysiol
 .
50
:
1197
1219
.
Berke
JD
,
Hetrick
V
,
Breck
J
,
Greene
RW
.
2008
.
Transient 23–30 Hz oscillations in mouse hippocampus during exploration of novel environments
.
Hippocampus
 .
18
:
519
529
.
Bontempi
B
,
Laurent-Demir
C
,
Destrade
C
,
Jaffard
R
.
1999
.
Time-dependent reorganization of brain circuitry underlying long-term memory storage
.
Nature
 .
400
:
671
675
.
Buzsaki
G
.
2006
.
Rhythms of the brain
 .
New York: Oxford University Press
.
Buzsaki
G
.
1989
.
Two-stage model of memory trace formation: a role for “noisy” brain states
.
Neuroscience
 .
31
:
551
570
.
Canals
S
,
Beyerlein
M
,
Keller
AL
,
Murayama
Y
,
Logothetis
NK
.
2008
.
Magnetic resonance imaging of cortical connectivity in vivo
.
Neuroimage
 .
40
:
458
472
.
Canals
S
,
Beyerlein
M
,
Merkle
H
,
Logothetis
NK
.
2009
.
Functional MRI evidence for LTP-induced neural network reorganization
.
Curr Biol
 .
19
:
398
403
.
Canals
S
,
Beyerlein
M
,
Murayama
Y
,
Logothetis
NK
.
2008
.
Electric stimulation fMRI of the perforant pathway to the rat hippocampus
.
Magn Reson Imaging
 .
26
:
978
986
.
Colgin
LL
,
Denninger
T
,
Fyhn
M
,
Hafting
T
,
Bonnevie
T
,
Jensen
O
,
Moser
MB
,
Moser
EI
.
2009
.
Frequency of gamma oscillations routes flow of information in the hippocampus
.
Nature
 .
462
:
353
357
.
Colgin
LL
,
Moser
EI
.
2010
.
Gamma oscillations in the hippocampus
.
Physiology (Bethesda)
 .
25
:
319
329
.
Davies
CH
,
Collingridge
GL
.
1993
.
The physiological regulation of synaptic inhibition by GABAB autoreceptors in rat hippocampus
.
J Physiol
 .
472
:
245
265
.
Davies
CH
,
Collingridge
GL
.
1996
.
Regulation of EPSPs by the synaptic activation of GABAB autoreceptors in rat hippocampus
.
J Physiol
 .
496
(Pt 2)
:
451
470
.
Davies
CH
,
Starkey
SJ
,
Pozza
MF
,
Collingridge
GL
.
1991
.
GABA autoreceptors regulate the induction of LTP
.
Nature
 .
349
:
609
611
.
Dutar
P
,
Nicoll
RA
.
1988
.
A physiological role for GABAB receptors in the central nervous system
.
Nature
 .
332
:
156
158
.
Engel
AK
,
Fries
P
.
2010
.
Beta-band oscillations—signalling the status quo?
Curr Opin Neurobiol
 .
20
:
156
165
.
Fernández-Ruiz
A
,
Makarov
VA
,
Benito
N
,
Herreras
O
.
2012
.
Schaffer-specific local field potentials reflect discrete excitatory events at gamma frequency that may fire postsynaptic hippocampal CA1 units
.
J Neurosci
 .
32
:
5165
5176
.
Frankland
PW
,
Bontempi
B
.
2005
.
The organization of recent and remote memories
.
Nat Rev Neurosci
 .
6
:
119
130
.
Fries
P
,
Reynolds
JH
,
Rorie
AE
,
Desimone
R
.
2001
.
Modulation of oscillatory neuronal synchronization by selective visual attention
.
Science
 .
291
:
1560
1563
.
Girardeau
G
,
Benchenane
K
,
Wiener
SI
,
Buzsaki
G
,
Zugaro
MB
.
2009
.
Selective suppression of hippocampal ripples impairs spatial memory
.
Nat Neurosci
 .
12
:
1222
1223
.
Herreras
O
,
Solis
JM
,
Martin del Rio
R
,
Lerma
J
.
1987
.
Characteristics of CA1 activation through the hippocampal trisynaptic pathway in the unanaesthetized rat
.
Brain Res
 .
413
:
75
86
.
Howe
MW
,
Atallah
HE
,
McCool
A
,
Gibson
DJ
,
Graybiel
AM
.
2011
.
Habit learning is associated with major shifts in frequencies of oscillatory activity and synchronized spike firing in striatum
.
Proc Natl Acad Sci USA
 .
108
:
16801
16806
.
Igarashi
KM
,
Lu
L
,
Colgin
LL
,
Moser
MB
,
Moser
EI
.
2014
.
Coordination of entorhinal-hippocampal ensemble activity during associative learning
.
Nature
 .
510
:
143
147
.
Jones
MW
,
Wilson
MA
.
2005
.
Theta rhythms coordinate hippocampal-prefrontal interactions in a spatial memory task
.
PLoS Biol
 .
3
:
e402
.
Lisman
JE
,
Jensen
O
.
2013
.
The theta-gamma neural code
.
Neuron
 .
77
:
1002
1016
.
Logothetis
NK
,
Augath
M
,
Murayama
Y
,
Rauch
A
,
Sultan
F
,
Goense
J
,
Oeltermann
A
,
Merkle
H
.
2010
.
The effects of electrical microstimulation on cortical signal propagation
.
Nat Neurosci
 .
13
:
1283
1291
.
Manabe
T
,
Wyllie
DJ
,
Perkel
DJ
,
Nicoll
RA
.
1993
.
Modulation of synaptic transmission and long-term potentiation: effects on paired pulse facilitation and EPSC variance in the CA1 region of the hippocampus
.
J Neurophysiol
 .
70
:
1451
1459
.
Martin
C
,
Beshel
J
,
Kay
LM
.
2007
.
An olfacto-hippocampal network is dynamically involved in odor-discrimination learning
.
J Neurophysiol
 .
98
:
2196
2205
.
Maviel
T
,
Durkin
TP
,
Menzaghi
F
,
Bontempi
B
.
2004
.
Sites of neocortical reorganization critical for remote spatial memory
.
Science
 .
305
:
96
99
.
McClelland
JL
,
McNaughton
BL
,
O'Reilly
RC
.
1995
.
Why there are complementary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory
.
Psychol Rev
 .
102
:
419
457
.
McNaughton
BL
,
Morris
RGM
.
1987
.
Hippocampal synaptic enhancement and information-storage within a distributed memory system
.
Trends Neurosci
 .
10
:
408
415
.
Moreno
A
,
Jego
P
,
de la Cruz
F
,
Canals
S
.
2013
.
Neurophysiological, metabolic and cellular compartments that drive neurovascular coupling and neuroimaging signals
.
Front Neuroenergetics
 .
5
:
3
.
Morris
RGM
.
2006
.
Elements of a neurobiological theory of hippocampal function: the role of synaptic plasticity, synaptic tagging and schemas
.
Eur J Neurosci
 .
23
:
2829
2846
.
Mott
DD
,
Lewis
DV
.
1991
.
Facilitation of the induction of long-term potentiation by GABAB receptors
.
Science
 .
252
:
1718
1720
.
Paxinos
G
,
Watson
C
.
2007
.
The rat brain in stereotaxic coordinates
 .
New York
:
Academic Press, Elsevier
.
Perkel
DJ
,
Nicoll
RA
.
1993
.
Evidence for all-or-none regulation of neurotransmitter release: implications for long-term potentiation
.
J Physiol
 .
471
:
481
500
.
Pouille
F
,
Scanziani
M
.
2001
.
Enforcement of temporal fidelity in pyramidal cells by somatic feed-forward inhibition
.
Science
 .
293
:
1159
1163
.
Pouille
F
,
Scanziani
M
.
2004
.
Routing of spike series by dynamic circuits in the hippocampus
.
Nature
 .
429
:
717
723
.
Ranck
JB
Jr
.
1973
.
Studies on single neurons in dorsal hippocampal formation and septum in unrestrained rats. I. Behavioral correlates and firing repertoires
.
Exp Neurol
 .
41
:
461
531
.
Ravel
N
,
Chabaud
P
,
Martin
C
,
Gaveau
V
,
Hugues
E
,
Tallon-Baudry
C
,
Bertrand
O
,
Gervais
R
.
2003
.
Olfactory learning modifies the expression of odour-induced oscillatory responses in the gamma (60–90 Hz) and beta (15–40 Hz) bands in the rat olfactory bulb
.
Eur J Neurosci
 .
17
:
350
358
.
Schaefer
AT
,
Angelo
K
,
Spors
H
,
Margrie
TW
.
2006
.
Neuronal oscillations enhance stimulus discrimination by ensuring action potential precision
.
PLoS Biol
 .
4
:
e163
.
Schwarz
AJ
,
Danckaert
A
,
Reese
T
,
Gozzi
A
,
Paxinos
G
,
Watson
C
,
Merlo-Pich
EV
,
Bifone
A
.
2006
.
A stereotaxic MRI template set for the rat brain with tissue class distribution maps and co-registered anatomical atlas: application to pharmacological MRI
.
Neuroimage
 .
32
:
538
550
.
Shyu
BC
,
Lin
CY
,
Sun
JJ
,
Sylantyev
S
,
Chang
C
.
2004
.
A method for direct thalamic stimulation in fMRI studies using a glass-coated carbon fiber electrode
.
J Neurosci Methods
 .
137
:
123
131
.
Siapas
AG
,
Lubenov
EV
,
Wilson
MA
.
2005
.
Prefrontal phase locking to hippocampal theta oscillations
.
Neuron
 .
46
:
141
151
.
Singer
W
.
1993
.
Synchronization of cortical activity and its putative role in information processing and learning
.
Annu Rev Physiol
 .
55
:
349
374
.
Squire
LR
.
1992
.
Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans
.
Psychol Rev
 .
99
:
195
.
Tolias
AS
,
Sultan
F
,
Augath
M
,
Oeltermann
A
,
Tehovnik
EJ
,
Schiller
PH
,
Logothetis
NK
.
2005
.
Mapping cortical activity elicited with electrical microstimulation using FMRI in the macaque
.
Neuron
 .
48
:
901
911
.
von Stein
A
,
Sarnthein
J
.
2000
.
Different frequencies for different scales of cortical integration: from local gamma to long range alpha/theta synchronization
.
Int J Psychophysiol
 .
38
:
301
313
.
Weitz
AJ
,
Fang
Z
,
Lee
HJ
,
Fisher
RS
,
Smith
WC
,
Choy
M
,
Liu
J
,
Lin
P
,
Rosenberg
M
,
Lee
JH
.
2014
.
Optogenetic fMRI reveals distinct, frequency-dependent networks recruited by dorsal and intermediate hippocampus stimulations
.
Neuroimage
 .
107C
:
229
241
.
Yeckel
MF
,
Berger
TW
.
1998
.
Spatial distribution of potentiated synapses in hippocampus: dependence on cellular mechanisms and network properties
.
J Neurosci
 .
18
:
438
450
.