First, we investigated whether theta-rhythmical (5 Hz) EC/DG-input triggers hippocampal network dynamics which, except for their rate of occurrence, differ from those elicited by non-theta (0.2 Hz) EC/DG-input (Figure 1C). Repetitive (0.2 Hz) EC/DG-input invariably caused prominent neuronal activity in the DG, only weak neuronal activity in the CA3 region, and no distinct neuronal activity in area CA1 (Figures 1D,E). In contrast, 5 Hz EC/DG-input reliably generated waves of neuronal activity which propagated through the entire trisynaptic circuit of the hippocampus (HTC) (Figures 1D–G). These “HTC-Waves” started to appear in an initially progressive manner approximately one second after the onset of the EC/DG-input and precisely followed the input rhythm (Figures 1D,E; Supplementary Video). Another observation was that neuronal activity in the DG markedly decreased during 5 Hz EC/DG-input (Figure 1E). This decremental activity possibly resulted from a known habituation process that involves GABAergic neurotransmission (Teyler and Alger, 1976). A pharmacological blockade of ionotropic glutamate receptors fully abolished HTC-Waves (Figure 1H).

The CA3 (stratum lucidum) field potential recordings suggest that HTC-Waves involve high-frequency firing of CA3 pyramidal neurons. This is deduced from multiple population spikes which accompanied the typically fast-rising mossy fiber field EPSP (fEPSP) (Toth et al., 2000; Nicoll and Schmitz, 2005) (Figure 1I). Both the rate and the number of these population spikes (173 ± 19 Hz and Figure 1I) correspond to the characteristics of physiological burst firing of CA3 pyramidal neurons (Andersen et al., 2007, p. 157). Population spikes were also interlaced in CA1 stratum radiatum (Schaffer collateral) fEPSPs (Figure 1I), indicating that HTC-Waves elicit action potentials in CA1 pyramidal neurons, which mediate hippocampal output to numerous brain structures, including the entorhinal cortex (Andersen et al., 2007).

We further tested whether non-theta (1 and 20 Hz) EC/DG-input generates activity propagations through the HTC. This was the case, but these activity propagations were considerably less strongly pronounced than those produced by 5 Hz EC/DG-input (Figure 1J). In addition, we found that the aforementioned decline of DG activity more heavily took place during 20 Hz EC/DG-input. The opposite scenario was observed for 1 Hz EC/DG-input (Figure 1K).

If not stated otherwise, all brain slice experiments were carried out in the presence of 0.6 μM bicuculline methiodide (BIM). We used the GABA_A receptor antagonist BIM at this weakly blocking dose (Figure 2A) for the following reasons: first, compared to the relatively short axons of most hippocampal GABAergic interneurons (Andersen et al., 2007), projections from excitatory neurons are cut to a greater extent during preparation of slices. This probably leads to an artificial enhancement of inhibition in large-scale hippocampal circuits in vitro (e.g., Iijima et al., 1996). And second, the voltage-sensitive dye used (Di-4-ANEPPS) possibly slightly potentiated GABA_A receptor function (Mennerick et al., 2010). BIM almost never led to epileptiform activity in the hippocampal subfields under investigation. Nevertheless, we always paid particular attention to such activity and, in the very rare cases in which it was observed (<1% of experiments) (Figure 2B), slices were discarded from subsequent analysis. Although less strongly pronounced, HTC-Waves also reliably occurred in the absence of BIM (Figure 2C).

To ensure proper signal-to-noise ratios and minimize experimental variability in HTC dynamics, we evoked EC/DG-input using relatively high stimulation intensities. In detail, the intensity of voltage stimulation was adjusted in a manner to produce DG-FDSs with peak amplitudes of approximately 80% of the highest attainable value. These FDSs ranged at the end of the linear upturn of the respective input–output curve (Figure 2D). Hence, EC/DG-input probably caused spiking of numerous DG granule cells. Under in vivo conditions, however, these neurons have been found to be sparsely active (Jung and McNaughton, 1993; Leutgeb et al., 2007). We therefore addressed the question whether HTC-Waves likewise can occur if only a small number of trisynaptic interconnections becomes activated. Providing substantial support for this scenario, HTC-Waves were also detectable with considerably lower activity states of the DG and continuously declined, rather than abruptly dropped off in their strength if the stimulation intensity was significantly decreased in a gradual manner (Figure 2E).

Next, we addressed the question whether HTC-Waves also occur under in vivo conditions. For this purpose, we electrically stimulated the medial PP (Tang and Dani, 2009) in anesthetized mice and conducted field potential recordings in the CA1 stratum pyramidale (Figure 3A). A stepwise increase in the stimulation intensity (0.5 V steps) reliably exposed a threshold (3.4 ± 0.6 V, N = 4 mice) at which CA1 population spikes appeared during 5 Hz PP activation. These population spikes emerged in a nearly identical manner and with virtually equal latencies to onset (~18 ms) as the CA1 neuronal responses in the in vitro experiments. Again, this phenomenon could never be observed if the PP was stimulated at 0.2 Hz (Figure 3B). Only with higher stimulation intensities, the recordings displayed an additional potential deflection (i.e., a fEPSP) which, according to the latency to onset (~10 ms), most likely resulted from direct PP input to area CA3 (Andersen et al., 2007). A possible explanation for this connection is that, due to the distinct dendritic attenuation of direct PP input to CA3 pyramidal neurons (Urban et al., 1998), the concomitant fEPSP merely became detectable under our experimental settings if a relatively high number of CA3-targeting PP fibers was activated.

The delayed and initially progressive appearance of HTC-Waves (Figures 1D,E; Supplementary Video) indicates that the 5 Hz EC/DG-input somehow unlocked a “gate” in the HTC for passages of neuronal activity. We considered frequency facilitation of mossy fiber to CA3 synaptic transmission (Toth et al., 2000; Nicoll and Schmitz, 2005) as the pivotal unlocking mechanism. Corroborating this notion, such frequency facilitation strongly took place during 5 Hz EC/DG-input, while neuronal activity in the DG decreased (Figures 1D,E,I). HTC-Waves and CA3-FDSs disappeared if glutamate release at CA3 mossy fiber terminals was selectively blocked by applying the mGluR2 agonist DCG-IV (Toth et al., 2000; Nicoll and Schmitz, 2005) specifically to area CA3 (Figures 4A–C).

Frequency facilitation at mossy fiber synapses is a short-lasting form of synaptic plasticity (Toth et al., 2000). We therefore postulated that HTC-Waves dissipate within a couple of seconds if 5 Hz EC/DG-input is followed by low-frequency (0.05 Hz) EC/DG-input (Toth et al., 2000). This was the case and reoccurring 5 Hz EC/DG-input again generated HTC-Waves which displayed a delayed and initially progressive emergence (Figure 4D).

HTC-Waves might provide a valuable assay for studying drug effects on polysynaptic activity flow through the hippocampus. For this reason and due to the fact that the hippocampus plays a fundamental role in cognitive processes, we examined the impact of the cognitive enhancer caffeine (Nehlig, 2010) on HTC-Waves. As already suggested by the DCG-IV experiments (Figure 4C), stimulation trains containing 10 stimuli allowed for stable baseline recordings of HTC-Waves (Figures 4B,E). Subsequent bath application of caffeine at a concentration found in the cerebrospinal fluid of humans after intake of 1–2 cups of coffee [5 μM (Fredholm et al., 1999)] boosted HTC-Waves (Figure 4E).

It has been shown that caffeine can mediate a presynaptic form of LTP at CA3-CA1 synapses (Martín and Buno, 2003). On this account, we additionally performed wash-out experiments to test whether such LTP might contribute to the caffeine-induced enhancement of HTC-Waves. Contradicting this possibility, CA1-FDSs reverted to baseline strength if caffeine was removed from the superfusion medium (Figure 4F).

To also investigate potential effects of a natural neuromodulator on HTC-Waves, we administered the stress hormone corticosterone to slices. Corticosterone was chosen since the rodent hippocampus is a target of this glucocorticoid (McEwen, 1999; Andersen et al., 2007), but experimental demonstrations of how it affects neuronal activity flow through the HTC network are lacking to date. As observed for caffeine, corticosterone (100 nM) rapidly amplified HTC-Waves (Figure 4G).

Due to the inherent high-frequency discharges of CA3 pyramidal neurons (Figure 1I), we expected HTC-Waves to yield a kind of “theta-burst” activation of CA3-CA1 synapses. Theta-burst stimulation of CA3-CA1 projections in turn is known to effectively induce CA1 LTP (Bliss and Collingridge, 1993). Therefore, we finally tested whether 5 Hz EC/DG-input can elicit CA1 LTP. To enable the imaging of the putative formation of CA1 LTP, we refined our VSDI assay. To accurately gauge the strength of neurotransmission at CA3-CA1 synapses, we repeatedly activated them by directly stimulating the Schaffer collateral-commissural pathway at a low frequency (Bliss and Collingridge, 1993) (Figures 5A–C). Owing to methodological reasons (see “Materials and Methods”), we employed the amplitude of the resultant CA1 (stratum radiatum)-FDSs at imaging frame 13 (28.6 ms) after the stimulation pulse (FDS_F13 amplitude) as measure of changes in the strength of neurotransmission at CA3-CA1 synapses (Figure 5C; “Materials and Methods”). As confirmed by field potential recordings performed in the CA1 stratum radiatum, this analysis showed that 5 Hz EC/DG-input for 6 s was sufficient to evoke CA1 LTP (Figures 5C,D; “Materials and Methods”). The CA1 LTP could be increased in its magnitude by a further train of HTC-Waves (Figure 5E). CA1 LTP critically depended on the number of HTC-Waves (Figure 5F) and the activation of NMDA receptors (Bliss and Collingridge, 1993) as revealed by the administration of the NMDA receptor blocker APV specifically to area CA1 (Figure 5G). 5 Hz EC/DG-input for 6 s also elicited CA1 LTP in the absence of BIM (Figure 5H). In all LTP experiments, we monitored hippocampal activity resulting from EC/DG-input by means of VSDI (Figure 6).

For all experiments, adult (8–12 week-old) male C57BL/6N mice were used. All experimental procedures were approved by the Committee on Animal Health and Care of the local governmental body and performed in strict compliance with the guidelines for the care and use of laboratory animals set by the European Community.

Mice were anesthetized with isoflurane and decapitated. All following steps were done in ice-cold sucrose-based saline (Bischofberger et al., 2006; Refojo et al., 2011; von Wolff et al., 2011) saturated with carbogen gas (95% O₂/5% CO₂). This saline (pH 7.4) consisted of (in mM): 87 NaCl, 2.5 KCl, 25 NaHCO₃, 1.25 NaH₂PO₄, 0.5 CaCl₂, 7 MgCl₂, 25 glucose, and 75 sucrose. After decapitation, the brain was rapidly removed from the cranial cavity and separated into its hemispheres. Next, the hemispheres were prepared for the slicing procedure by a special transversal cut, which is sometimes called “magic cut” (Bischofberger et al., 2006). We cut the hemispheres at angles optimized to conserve the intrahippocampal axonal projections along the DG-CA axis as best as possible (Bischofberger et al., 2006; Refojo et al., 2011; von Wolff et al., 2011). Subsequently, 350 μm-thick horizontal slices containing the hippocampus were cut using a vibratome (HM650V; Thermo Scientific).

After preparation, slices were incubated in carbogenated sucrose-based saline for 30 min at 34°C. Subsequent staining of slices with the voltage-sensitive dye Di-4-ANEPPS (Tominaga et al., 2000; Airan et al., 2007; Refojo et al., 2011; von Wolff et al., 2011) (dissolved in DMSO to a 20.8 mM stock solution) was carried out at room temperature (23–25°C). For staining, slices were kept for 15 min in carbogenated physiological saline containing Di-4-ANEPPS [7.5 μg/ml; <0.1% DMSO (Refojo et al., 2011; von Wolff et al., 2011)]. The physiological saline (pH 7.4) consisted of (in mM): 125 NaCl, 2.5 KCl, 25 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂, 1 MgCl₂, and 25 glucose. Afterwards, slices were stored at room temperature for at least 30 min in Di-4-ANEPPS-free, but BIM (0.6 μM)-containing carbogenated physiological saline.

All brain slice experiments were carried out at room temperature. In the recording chamber, slices were fixed with a platinum frame/nylon string harp and continuously superfused with BIM (0.6 μM)-containing carbogenated physiological saline (4–5 ml/min flow rate). Pilot experiments revealed that HTC-Waves can be reliably evoked in slices that were cut at a distance between 2.3 and 3.0 mm (±0.1 mm) from the ventral base of the brain. For this reason, all in vitro investigations were performed in these particular slices. Accordingly, experiments were conducted in the dorsal hippocampus or in the transition zone from the dorsal to the ventral hippocampus (Maggio and Segal, 2007; Fanselow and Dong, 2010).

We evoked EC/DG-input by electrical stimulation of the PP, which also contains fibers that directly innervate CA3 pyramidal neurons (Andersen et al., 2007). We cut these fibers at the point where they exit the DG. Temporoammonic projections (Andersen et al., 2007) were likewise functionally inactivated. The deafferentations were accomplished by means of a tapered scalpel blade and a custom-made “microknife” (approximately 100 μm blade length) (Figure 7A).

VSDI and data analysis were performed using the MiCAM02 hard- and software package (BrainVision) (Refojo et al., 2011; von Wolff et al., 2011). The tandem-lens fluorescence microscope was equipped with the MiCAM02-HR camera and the 2× and 1× lens at the objective and condensing side, respectively (for further technical details see http://www.scimedia.com). Acquisition settings were as follows: 88 × 60 pixels frame size, 36.4 × 40.0 μm pixel size, and 2.2 ms sampling time.

From recorded VSDI signals, the fractional change in fluorescence (ΔF/F) was calculated. For all quantifications, Δ F/F values were spatially and temporally smoothed using a 3 × 3 × 3 average filter. VSDI signals presented in images and the video were smoothed with a 5 × 5 × 3 average filter. To attenuate slow signal components produced from bleaching of the voltage-sensitive dye (Carlson and Coulter, 2008) and slight summation of 5 and 20 Hz neuronal responses, we afterwards applied a weak high-pass filter of the MiCAM02 software (τ = 220 ms) to the imaging data. Pixelation of images and the video was reduced by the interpolation function of the MiCAM02 software.

For analysis of neuronal population activity in hippocampal subregions, three standardized ROIs were manually set according to anatomical landmarks. The circular CA3-ROI (r = 4 pixels) was positioned into the CA3 region near the DG, but not overlapping with it. The circular CA1-ROI (r = 4 pixels) was placed into the CA1 subfield with a distance of approximately 200 μm from the visually identified distal end of the stratum lucidum. The CA3-ROI spanned the stratum oriens, stratum pyramidale, and stratum lucidum/radiatum. This was also the case for the CA1-ROI, with the exception of the LTP experiments where it only covered the stratum radiatum (r = 2 pixels). The DG-ROI enclosed the fascia dentata (without the hilus) and was created by the polygon-drawing function of the MiCAM02 software. The average of smoothed Δ F/F values within a particular ROI served as final measure of neuronal population activity.

Field potentials in the CA3 stratum lucidum and CA1 stratum radiatum were recorded using glass microelectrodes (1 MΩ open-tip resistance) that were filled with BIM (0.6 μM)-containing physiological saline. Recording data were low-pass filtered at 1 kHz and digitized at 5 kHz.

All in vivo recordings were performed in the left dorsal hippocampus and carried out under urethane anesthesia (750 mg/kg body weight; i.p.). An additional dose of urethane (250 mg/kg) was added 30–60 min later (Irikura et al., 1995). The temperature of the anesthetized animal was constantly controlled with a rectal probe and maintained at 36°C with a heating pad (homeothermic blanket system; Harvard Apparatus). During preparation of the skull and the recording session, the animal was permanently fixed in a computer-guided stereotactic device (Angle Two™; myNeurolab). The coordinates for electrical stimulation and recording of neuronal activity were as follows: medial PP (stimulation): 0.2 mm posterior and 2.9 mm lateral from lambda, 1.3 mm ventral from brain surface (Tang and Dani, 2009); CA1 stratum pyramidale (recording): 2.2 mm posterior and 1.2 mm lateral from bregma, 1.3 mm ventral from brain surface (Franklin and Paxinos, 1997). For recording, tungsten microelectrodes (TM33A20; World Precision Instruments) were used. A gold wire with a ball-shaped ending served as the indifferent electrode and was placed epidurally through a small hole in the skull (250 μm drilling diameter) at the rostral part of the right os frontale. Recording data were band-pass filtered (2 Hz cutoff frequency for high-pass filter and 1 kHz for low-pass filter) and digitized at 5 kHz. After termination of the experiment, electrical lesions were set at stimulation and recording sites (3 μA for 3 min; negative polarity). Each brain was then cut into 30 μm sections and stained (standard cresyl violet staining) to verify the position of each electrode.

In all experiments, neuronal activity was evoked by square pulse electrical stimuli (200 μs pulse width). For stimulation of the medial PP in vivo, custom-made monopolar tungsten electrodes (Teflon-insulated to the tip of 75 μm diameter) were used.

For electrical stimulation in vitro, we also used custom-made monopolar tungsten electrodes (Teflon-insulated to the tip of 50 μm diameter). These electrodes allowed for a very precise placement into the neuronal tissue and did not interfere with VSDI (e.g., by producing irritating shadows as observed with other electrodes). A highly localized electrical stimulation was achieved by positioning the indifferent electrode far away from the slice in the recording chamber. In LTP experiments, in which two stimulation electrodes were employed, the currently unused electrode was uncoupled from the stimulation device. This was done to prevent an indirect application of voltage to the inactive electrode.

The electrode employed for evocation of EC/DG-input was placed on the visually identified PP near its entry zone to the DG. Direct stimulation of the Schaffer collateral-commissural pathway in LTP experiments was performed near the CA3 averted boundary of the CA1-ROI, which was used to monitor putative alterations in the strength of neurotransmission at CA3-CA1 synapses. This was done to antidromically (Bliss and Collingridge, 1993) activate a large fraction of the CA3-CA1 synapses that were also activated by the later induced HTC-Waves. Thereby, commissural fibers and Schaffer collaterals which were cut somewhere in the CA3 or CA1 region during preparation of slices became also recruited. To minimize the probability that neuronal activity within the CA1-ROI overly resulted from activation of such “irrelevant” fibers, we used a stimulation intensity which produced a strong fEPSP and, thus, a concomitant population spike (Figures 7B,C).

In contrast to field potential recordings, VSDI displays action potentials and EPSPs always as signals exhibiting the same deflection course (Carlson and Coulter, 2008). Therefore, neither the peak amplitude nor the slope of CA1-FDSs could be used as proper measure of changes in the strength of neurotransmission at CA3-CA1 synapses. To overcome this problem and thereby become also able to conduct the LTP experiments via VSDI, we employed the amplitude of a later CA1-FDS component as such a measure. Several sets of pilot experiments identified the amplitude of CA1-FDSs at imaging frame 13 (28.6 ms) after the stimulation pulse (FDS_F13 amplitude) to be suited. As shown by simultaneous field potential recordings, the FDS_F13 amplitude reflects the amplitude of the glutamatergic fEPSP at a certain time point of its decaying phase. At this time point, the fEPSPs did not anymore mingle with action potentials fired by CA1 pyramidal neurons (Figure 7B). The intensity of voltage stimulation was adjusted in a manner to produce a FDS_F13 amplitude with a value in the range between 60–80% of the peak amplitude of the CA1-FDS (Figures 5C, 7B). If performed in this way, the FDS_F13 amplitude could be increased by enhancing the stimulation intensity. An increase in the FDS_F13 amplitude was also seen after adding the AMPA receptor potentiator cyclothiazide at a concentration of 100 μM (Eder et al., 2003) to the superfusion medium. The same effects could be observed in the field potential recordings (Figure 7B). Additionally vindicating the use of the FDS_F13 amplitude in the LTP experiments, 5 Hz EC/DG-input for 6 s, which evoked CA1 LTP as measured with VSDI, caused a long-lasting increase in the slope and the amplitude of fEPSPs 28.6 ms after the stimulation pulse (Figure 7C).

For a non-invasive pressure application of a given drug to a defined hippocampal subfield, we positioned a glass microelectrode (0.2–0.3 MΩ open-tip resistance) within a certain region close to the surface of the slice under investigation. The electrode was filled with physiological saline containing BIM (0.6 μM), Fluo (10 μM), and the drug to be applied. Fluo, whose fluorescence was excited by means of light emitted from a UV diode, served to visually monitor the spread of the pressure-ejected fluid via a binocular. A spatially restricted spread was achieved by capitalizing on the stream of the superfusion medium through the recording chamber, by appropriately adjusting the strength of the pressure applied to the electrode, and by a specific orientation of the slice within the recording chamber (Figure 7D). Since the drugs (DCG-IV and APV) had to rapidly penetrate the neuronal tissue by diffusion, we used somewhat higher concentrations of them than typically bath-applied to reach a maximal pharmacological effect. In all sets of experiments, control applications of Fluo alone were carried out. To verify the spatial specificity of the Fluo-guided drug administration technique, we bath-applied DCG-IV at a concentration of 1 μM (Kew et al., 2002; Nicoll and Schmitz, 2005) to slices. Consistent with its inhibitory effect on glutamatergic neurotransmission at PP synapses (Kew et al., 2002), DCG-IV decreased neuronal activity in the DG (Figure 7E). This effect was not observed if DCG-IV was applied specifically to area CA3 (Figure 4C). We furthermore bath-applied APV at a concentration of 50 μM to slices. This was done to determine the concentration of pressure-applied APV that leads to a complete block of NMDA receptors. Bath-applied APV reduced the FDS_F13 amplitude of CA1-FDSs almost to the same extent as 200 μM APV that was pressure-applied to the CA1 subfield for 1 min (Figure 5G).

BIM, caffeine, corticosterone, Di-4-ANEPPS, Fluo, Ponceau S, urethane, and all salts for the saline solutions were purchased from Sigma-Aldrich. APV and NBQX were from Ascent Scientific, cyclothiazide and DCG-IV from Tocris, and isoflurane from Abbott. Corticosterone was dissolved in ethanol. During experimentation (including baseline recording), ethanol was present at a concentration of 0.009%.

Two-tailed t-tests were used for statistical analyses and run in SigmaStat (Systat Software), with significance declared at p < 0.05. Data are given as mean ± s.e.m.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.