All experiments were conducted in accordance with standards for the care and use of animals in research outlined in European Directive 2010/63/EU, and were supervised by a consulting veterinarian.

Six male rats weighing about 350 g were housed individually and maintained on a 14 h/10 h light/dark cycle. Animals were placed on water-restricted diet 1 day prior to the beginning of the experiments. To ensure that the animals did not suffer dehydration as a consequence of water restriction, they were allowed to continue the behavioral testing to satiation and were given access to ad lib drinking water for 1 h after the end of each recording session. The animals' body weight and general state of health was monitored throughout the experiments. Out of the six animals trained to perform the behavioral task, four were implanted using microdrives for chronic recordings. Neuronal data suitable for analysis was collected from two animals.

The stimuli were chosen so that they were short sounds, easily perceived by rats. We chose artificial vowels as a more “naturalistic” class of stimuli than pure tones. Artificial vowels are a simplified version of vocalization sounds used by many species of mammals, and have been used in studies of the ascending auditory pathway from the auditory nerve (Cariani and Delgutte, 1996; Holmberg and Hemmert, 2004) to the auditory cortex (Bizley et al., 2009, 2010). The spectra of natural vowels are characterized by “formant” peaks which result from resonances in the vocal tract of the vocalizing animal (Schnupp et al., 2011). Formant peaks therefore carry information about both the size and the configuration of the vocal tract, and human listeners readily categorize vowels according to vowel type (e.g., /a/ vs /o/, Peterson and Barney, 1952) as well as according to speaker type (e.g., male vs. female voice) or speaker identity (Gelfer and Mikos, 2005). Many species of animals, including rats (Eriksson and Villa, 2006), chinchillas (Burdick and Miller, 1975), cats (Dewson, 1964), monkeys (Kuhl, 1991), and many bird species (Kluender et al., 1987; Dooling and Brown, 1990) readily learn to discriminate synthetic vowels.

The stimuli were generated using binary click trains at the fundamental frequency of 330 Hz and 100 ms duration which were bandpass filtered with a bandwidth equal to 1/50th of the formant frequency using Malcolm Slaney's Auditory Toolbox. Formant center frequencies were 6.912 and 15.008 kHz, respectively. The spectra of the stimuli are shown in Figure 1B. The formant frequencies chosen here lie within rat and human sensitivity range (Heffner et al., 1994). As demonstrated in our previous experiments (Itskov et al., 2012) the rats can perceive and discriminate similar artificial vowels. A third, control sound used in passive recordings as a novel stimulus was a naturalistic noise-like sound (see the spectrum in Figure 1B). The stimuli were ramped on and off (supplementary audio files 1–2). They were presented at a sound level of 70 dB SPL, from a speaker located above the experimental apparatus.

Sounds were presented through Visaton FRS 8 speaker, which has a flat frequency response (<±2 dB) between 200 Hz and 10 kHz. The speaker was driven through a standard PC sound card controlled by LabView (National Instruments, Austin, TX, USA).

The arena for the behavioral task consisted of a dim-lit sound-attenuated box (35 × 30 × 40 cm). One of the walls had two round holes (called “nose pokes” throughout the text, see Figure 1A). A water well was placed behind each of the nose pokes. Infrared sensors at the edge of the nose pokes monitored the rat's entry and exit from the holes. Outside the context of the behavioral task, passive exposures to the awake rat were done in an unfamiliar dark plastic bin (33 × 22 × 35 cm) which differed from the training environment by geometry, texture of the floor and the walls, the lighting, and the smell. For recordings during the sleep we used a tall box (26 × 20 × 60 cm) lined with thick cotton tissue. Light sensors, sound, and water delivery were controlled by a custom-written LABview script, operating a National Instruments card (National Instruments Corporation, Austin, TX, USA).

The rats were implanted with chronic recording electrodes. They were anesthetized with a mixture of Zoletil (30 mg/kg) and Xylazine (5 mg/kg) delivered i.p. A craniotomy was made above left dorsal hippocampus, centered 3.0 mm posterior to bregma and 2.5 mm lateral to the midline. A Neuralynx “Harlan 12” microdrive loaded with 12 tetrodes (25 μm diameter Pl/Ir wire, California Fine Wire) was mounted over the craniotomy. Tetrodes were loaded perpendicular to the brain surface and individually advanced in small steps of 40–80 μm, until they reached the CA1 area, indicated by the amplitude and the shape of the sharp wave/ripples. After surgery, animals were given antibiotic (Baytril; 5 mg/kg delivered through the water bottle) and the analgesic caprofen (Rimadyl; 2.5 mg/kg, subcutaneous injection) for postoperative analgesia and prophylaxis against infection, and were allowed to recover for one week to 10 days after surgery before training started.

After recovery from electrode implantation the animals were trained, during which neuronal responses were recorded from the tetrode array using TDT data acquisition equipment (RZ2, Tucker and Davis Technologies, Alachua, FL, USA). We usually recorded from 1 to 4 single units per tetrode. Spikes were presorted automatically using KlustaKwik (Harris et al., 2000), using waveform energy on each of the four channels of the tetrode as coordinates in a 4 dimensional feature space. The result of the automatic clustering was inspected visually after importing the data into MClust (kindly provided by A. D. Redish). Each single unit was recorded only in one session.

The task was designed in a way that allowed us to examine neuronal responses to behaviorally relevant auditory stimuli. We took into account the richness of location-related neuronal signals in hippocampus that are known to influence sensory responses recorded in different spatial locations (Itskov et al., 2012). The animal initiated each trial by poking its snout into a nose poke. During the recording sessions (referred to as “lottery,” see below) the sounds were presented with 50% probability for a given nose poke entry. Trials with and without sound presentation were compared, assuring that the animal was in exactly the same position and motivational state in both types of trials.

To record hippocampal responses to sounds in a context unrelated to the behavioral task we placed the rat in a novel environment 2–3 h after completion of the behavioral session. Three types of sound stimuli (familiar vowel associated with the right well, familiar vowel associated with a left well, and a novel noise-like sound) were presented pseudo-randomly, so that no sound was repeated more than two times in a row. Inter-stimulus interval varied randomly from 25 to 30 s. The sounds were presented irrespectively of the animal's action and location. Each stimulus type was presented 43 times on average. The long inter-stimulus interval was intended to reduce the animal's adaptation to stimulus repetitions (Ulanovsky et al., 2004).

Passive sound presentation during sleep started 5 min after the animal had fallen asleep, as determined by total immobility. The animal was monitored continuously via remote video camera. As was the case during the session in which the rat remained awake in a novel environment, three types of sound stimuli (familiar vowel associated with the right spout, familiar vowel associated with a left spout, and a novel noise-like sound) were presented pseudo-randomly, so that no sound repeated more than two times in a raw. Inter-stimulus interval varied randomly from 25 to 30 s. Each stimulus type was presented 49 times on average.

To ensure the proper timing of the responses, we recorded a copy of the signal sent to the loudspeaker along with the neuronal signals.

Firing rates of hippocampal neurons typically vary over the course of a single trial, since they tend to respond to different aspects of the task. The timing and the latency of the response can vary across individual cells. Taking this into account, to quantify the responses to sounds we used a method intended to (1) take into account multiple time points, (2) provide a single measure of statistical significance of the effect, (3) take into account variability in the timescale of the responses, fast in some cases (on the order of tens of milliseconds, Brankack and Buzsaki, 1986) and slow (hundreds of milliseconds) in others.

To quantify firing rate changes related to animal's location in space we measured firing rate in non-overlapping 50 ms bins during the approach to the nose pokes (see Figure 3A). Firing rates in each time bin were compared using Wilcoxon rank-sum test for two independent samples (Siegel and Castellan, 1988). Benjamini–Hochberg false discovery rate corrections were used to correct for multiple comparisons (FDR, Benjamini and Hochberg, 1995). A neuron was considered to distinguish between the two nose pokes and therefore to be a “location-encoding” neuron if any of the time bins showed a corrected p-value smaller than 0.05.

To quantify a neuron's responses to sound presentation and water reward delivery during the behavioral task, we compared firing rates in the “lucky” trials to firing rates in trials in which no sound was played and no water delivered. To take into account the variability of the responses (e.g., fast transient or sustained responses), we used four a-priori defined bin sizes (50 ms, 500 ms, 1 and 2 s). In case of 50 ms bins, we measured firing rate in six non-overapping bins (0–300 ms). Wilcoxon rank-sum test for two independent samples was used to compare the firing rates, bin by bin, in “lucky” trials versus not reward trials. Benjamini–Hochberg FDR correction were used to correct for testing multiple time points (n = 6 in 50 ms bins) and bin sizes (50 ms, 500 ms, 1 and 2 s, FDR, Benjamini and Hochberg, 1995). A neuron was considered a “sound or reward-encoding” neuron if any of the time bins showed a corrected p-value smaller than 0.05.

To quantify sound-evoked responses during passive presentation, we compared post-stimulus firing rate with the baseline. To take into account the variability of responses we again used four a-priori defined bin sizes (50 ms, 500 ms, 1 and 2 s). For each analysis, baseline firing rates were defined as the firing rate in the time immediately preceding the presentation of the sounds. We used baseline bin size equivalent to the time bin size used to measure evoked activity (50 ms, 500 ms, 1 or 2 s).

In the case of the smallest bin size, firing rate was calculated in 50 ms non-overlapping bins in first 300 ms after sound onset, yielding six individual time points. Firing rate values in each of them were compared to firing rate values in 50 ms baseline (sign test for dependent measurements, Siegel and Castellan, 1988). This yielded six tests for each neuron, which were corrected for using FDR. Latency of the responses was calculated as time of the first bin after sound onset with activity significantly different from baseline.

In the case of 500 ms bins, we took a mean firing rate in 500 ms baseline and compared it with mean firing rate in the 500 ms post-stimulus interval. In the case of 1 s bins we took a mean firing rate in 1 s baseline and compared it with mean firing rate in 1 s post-stimulus interval. In the case of 2 s bins we took a mean firing rate in 2 s baseline and compared it with mean firing rate in 2 s post-stimulus. This yielded four measurements for each neuron: p-value for 50 ms bins (already corrected for comparing multiple bins), and p-values for 500 ms, 1 and 2 s bins. These p-values were corrected for multiple testing using FDR. A neuron was considered sound-responsive if any of the time bins showed a corrected p-value smaller than 0.05.

For the purposes of illustration, the firing rate was calculated in a 50 ms wide window sliding in steps of 25 ms along the whole duration of the trial (see Figures 3–4).

After conclusion of the recording experiments, the animals were overdosed with intraperitoneal injection of the anesthetic urethane and transcardially perfused with 4% paraformaldehyde solution. Brains were sectioned in the coronal plane and stained with cresyl-violet. Electrode tracks were localized on the serial sections in CA1 field.

This study set out to test how neurons in rat hippocampus represent auditory stimuli in the context of a behavioral task, during passive exposure and during sleep. Familiar sounds that during the training preceded water reward were played in the novel context (new enclosure) or during sleep.

Rats were given four pre-training sessions, during which they learned to alternate between two water spouts to get water reward (see Figure 2A and Materials and Methods). In the main experimental session (“lottery”, Figure 2B) the water reward contingencies were changed: now water was given only in 50% of the entries to the nosepokes. On these “lucky” trials one of the sounds was played immediately upon the nosepoke entry. Distinct sounds were used for each of the two nose pokes (see Figure 1B and Materials and Methods).

In awake rats, hippocampal neurons exhibit strong spatial selectivity: they encode the animal's position, head direction, and other spatial variables (O'Keefe and Dostrovsky, 1971, reviewed in Moser et al., 2008). Consistent with the existing literature, the neurons recorded in this task exhibited significant sensitivity to animal's location in space. Forty-one percent of tested neurons (24 out of 58) discriminated between right and left nosepokes (see Figure 3A; p < 0.05, corrected, rank sum test on firing rates 400 ms before the nose pokes, see Materials and Methods).

On the “lucky” trials (50% of trials, selected randomly), upon entry to the nosepoke the rats heard a sound and received water reward after a 3 s delay. Two different sounds were used in two different locations (nose pokes). In “unlucky” trials (50% of trials) the rat heard no sound and received no water reward (see Figure 2B and Materials and Methods). The rats stayed still with their snouts in the nosepoke waiting for 3 s between the nosepoke entry/sound onset and the water reward delivery. The average time of waiting in no-sound-no-reward trials was 5.09 ± 0.19 s (SEM). Waiting time decreased over the course of the session (p < 0.001, paired-sample t-test). To examine responses to sound and reward, we aligned neuronal activity on the entry to the nosepoke and compared firing rate observed in the two types of trials, separately for the two water wells. The responses are illustrated on Figures 3B–D; the line at 3 s denotes reward delivery. Out of 58 cells recorded in the task, 11 cells responded in the first 300 ms upon sound presentation (19%, p < 0.05, rank sum tests done in 50 ms windows, corrected for testing six time points and also four different bin sizes, see Materials and Methods). Fourteen (24%) responded upon the reward delivery (Figure 3B). There were no cells with responses to both sound and reward.

The aim of this experiment was to characterize the representation of sound identity during passive listening. We played three different sounds to the animals outside the context of the behavioral task, in a novel box. Two of the sounds were artificial vowel stimuli associated with water in the two different nosepokes and familiar to the animals and one of the sounds was a dissimilar novel band pass filtered noise stimulus. Our intention was to compare the responses to familiar and relevant sounds with the responses to novel and behaviorally irrelevant ones. The animals were placed in a novel context 24 h after they performed the “lottery” behavioral task. Sounds were triggered with no relation to the animal's posture or movement, in a random sequence lasting approximately 50 min with a random interstimulus interval ranging from 25 to 30 s. The rats did not respond to the sounds with any overt behavior.

Three out of 25 neurons recorded during the awake passive conditions were classified as sound-sensitive (see Materials and Methods, p-values were 0.0029, 0.0000016, and 0.01, corrected). The responses are illustrated in Figure 4A. The latencies of response to passive sound presentation were less than 150 ms in two cases and less than 50 ms in the third case (see Figure 4A). Two of the neurons showed a significantly increased firing rate in response to the sound, and one a decreased the firing rate. It would be interesting to know whether the neurons that responded to sounds were also responsive to events occurring within the task, such as sounds, reward, or locations. Although we were able to keep the same neurons across the recording sessions, as evidenced by the neuronal waveforms, autocorrelations and firing rates, we feel that the number of recorded neurons does not allow us to make any conclusive statements regarding a single neuron's properties across time and context. None of the three sound-responding neurons discriminated between different sounds i.e., their responses to different sounds were identical.

During sleep sessions, the rats were placed in a small familiar box in which they soon curled up and fell asleep. Animals' activity was monitored remotely via webcam, placed above the box. Five minutes after the animals curled up and fell asleep, the sound stimuli were turned on. The animal's sleep status was constantly monitored online by the experimenter and sound presentation did not cause the animals to wake up. The sounds were presented the entire time when the animals were asleep (117–180 times, 49 presentations per sound type on average) and were turned off when they awoke. Only the trials where sounds did not cause any visible behavioral response from the sleeping animal were analyzed further.

The responses to sounds during sleep are illustrated in Figures 4B–D and Figure 5. Twenty-four percent of recorded neurons (8 out of 33) exhibited significant sound-evoked responses. Across neurons, latency ranged from 50 to 150 ms. In five cases firing rate decreased after sound presentation, in three cases the firing rate increased.

Figure 5 illustrates neurons with significant differences between baseline and post-stimulus activity lasing more than 1 s, even when considered in short 50 ms long bins (e.g., black line in Figure 5C). In some cases an initial time-locked response was followed by a long sustained activity response (Figures 5A,B,C,D).

In five cases we observed significant differences between mean firing rate in 2 s post-stimulus and the baseline (p < 0.05, corrected). A neuron in Figure 5 (panels F-G-H-I-J) illustrates the case. A short-latency onset response was followed by a sustained, long lasting response (Figure 5F). Panel I plots average firing rate in 2 s pre-and post-stimulus windows. p-value is indicated at the outset (p = 0.0035, sign test for dependent measures). Both neurons depicted on this figure had narrow spikes (Figures 5E and J), 0.2 ms when measured on 25% of spike amplitude. Both of these neurons are likely to be putative interneurons because of the narrow spike width and the absence of the 3–4 ms peak in the autocorrelation plot (data not shown). Unfortunately, limited size of our dataset does not allow us to make any further conclusions on this issue.

Long lasting responses to sounds were unique to responses recorded during sleep (Figure 6). The responses to sounds during sleep were on average almost twice as long as the response to sounds presented during passive wakefulness (Figure 6F). Responses to sounds during sleep also appeared slightly earlier than the responses during passive wakefulness (Figure 6E).

The majority of cells did not discriminate between different sounds. We found only one neuron with significant sound identity-dependent responses (p = 0.0101, Kruskal–Wallis test) comparing responses to three sounds, two artificial vowels and the novel noise-like sound.

Taken together, these results indicate that hippocampal neurons showed sound-evoked responses during awake passive listening. One quarter of recorded neurons showed reliable sound-evoked responses during sleep. Short latencies of the responses (less than 50 ms) were observed alongside with long-lasting sustained activity.

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.