Experiments were performed on 9 adults Sprague-Dawley rats (male, 274 $\pm$ 56 g). The animals were anesthetized with urethane (1.25 g/kg, i.p.) and were placed in a rat stereotaxic apparatus (Stoelting Co.). The temperature of the rat body was maintained at ~37 ${}^{\circ }$C. A part of the skull above the left-brain hemisphere was removed to implant the recording and stimulating electrodes. The recording electrode (RE), a 16-channel array (Model A1x16-Poly2-5mm-50s-177, NeuroNexus Technologies), was implanted into the left hippocampal CA1 region (AP 3.5 mm and ML 2.7 mm from bregma, DV ~2.5 mm from the surface of the brain). The stimulation electrode (SE), a concentric bipolar electrode of stainless-steel with a distance of 100 $\mu$m between the inner and outer poles (Model CBCSG75, FHC), was implanted into the afferent axons (i.e., the Schaffer collaterals) of the CA1 region (AP 2.2 mm and ML 2.0 mm from bregma, DV ~2.8 mm from the surface of the brain) for orthodromically stimulating the CA1 neurons. The final positions of the two electrodes were judged according to the patterns of the evoked potentials and the unit spike signals that appeared serially in the recording array (Kloosterman et al., 2001). Two separate screw anchors were fixed in the nose bone and served as the ground electrode and reference electrode.

The electrical signals collected by the recording electrode were first amplified 100-fold by a 16-channel microelectrode amplifier (3600, A-M system) with a band-pass filtering range of 0.3-5000 Hz. The amplified signals were then sampled with a rate of 20 kHz by a PowerLab data-acquisition system (PL 3516, ADInstruments).

The sinusoidal waveforms of stimulation were originated from a signal generator (DG1032Z, REGOL Technologies). The potential waveforms were then input to an analog stimulus isolator (2200, A-M Systems) to generate the continuous current sinusoids with symmetrical waveforms between their positive and negative phases that were delivered into the stimulation electrode. The sinusoidal frequency was 50 Hz. The applied stimulation of the sinusoidal waveform was verified by a measurement in a cup of saline to mimic brain tissue with the same electrodes as used in the rat experiments. The peak-to-peak current intensity of the sinusoidal waveforms was adjusted to 40-70 $\mu$A (57.2 $\pm$ 10.7 $\mu$A, n = 9) to modulate the unit firing neurons without evoking population spikes. The duration of the sinusoidal stimulation was 1 min.

Also, the signal of sinusoidal stimulation, i.e., the outputs of the DG1032Z signal generator, were simultaneously recorded with the neuronal signals. The polarity of the sinusoidal recording was the same as the inner pole of the concentric bipolar stimulation electrode. Because the surface area of the inner pole was much smaller than that of the outer pole, the inner pole contributed significant stimulation effects with a much higher current density surrounding the pole.

On the recording electrode array, the signals of four adjacent channels located in the pyramidal layer of the CA1 region were used to extract the single unit spikes of pyramidal neurons and interneurons. The unit spikes of the two types of neurons were detected and sorted by the previously described methods (Feng et al., 2017; Wang et al., 2018). Briefly, the raw recording signals were filtered ($>$ 500 Hz) by a built-in high-pass digital filter of PowerLab Chart v7.0 software (ADInstruments) to remove the stimulation artifact of sinusoids and the local field potential (LFP) in the low-frequency bands to generate multiple unit activity (MUA) signals. Then, unit spikes in the MUA were detected by an amplitude threshold at 5-fold standard deviations of the MUA signals. Finally, the principal components and the amplitudes of the spike waveforms obtained from the 4-channel unit signals by a MATLAB program were used to sort the spikes of individual neurons by an open-source clustering software (SpkieSort 3D, www.Neuralynx.com). The unit spikes of pyramidal neurons and interneurons were distinguished according to the widths of spike waveforms. The unit spikes with a width of rising phase $>$ 0.7 ms and 0.4 ms were defined as the spikes of putative pyramidal neurons and putative interneurons, respectively (Csicsvari et al., 1998, 1999). The distributions of inter-spike intervals (ISI) of the sorted units were calculated to verify the clustering procedure. Spikes with an ISI histogram without clear refractory (~1 ms) were not taken as single unit spikes (Barthó et al., 2004).

The peri-stimulus time histograms (PSTH, bin = 1 ms) of the unit spikes per neuron and minute was calculated. The distribution relationship between unit spikes and sinusoidal phases was evaluated by a statistical method of circular distribution, i.e., a circular PSTH. For control, a mimic circular PSTH was obtained by relating the unit spikes in the 1-min pre-stimulation period to a 1-min mimic 50 Hz sinusoidal signal with continuous 20 ms cycles.

The firing rhythms of both types of neurons were analyzed by using a spectrum of auto-correlograms (ACs) of the binary sequences of unit spikes as following (Leblois et al., 2010; McConnell et al., 2012). The values of ACs were first calculated and normalized to eliminate the influence of firing rates. The mean components of ACs were removed. The length of ACs was 2 seconds (-1 to 1 s). Then, the power spectrum density (PSD) of ACs of the individual neurons was calculated by Welch’s modified periodogram by using a Hanning window with an overlap of 50%. The cumulative power in the frequency band of theta rhythm (2-5 Hz) of the urethane anesthetized rats (Buzsáki, 2002) and around the sinusoidal frequency (49-51 Hz) were compared between the periods of 1-min sinusoidal stimulation and the 1-min pre-stimulation of baseline recording.

All values were given as mean $\pm$ standard deviation. Statistical significance in differences between two data groups was evaluated using the paired t-test or t-test performed by SPSS software.

Fig. 2.

Sinusoidal stimulation induced the phase-locked firing in downstream neurons. (A) Examples of neuronal firing at pre-stimulation of baseline recording and during stimulation. The dashed gray and solid red sinusoidal waveforms denote the mimic stimuli at baseline and the real stimuli during stimulation. (B) The mean circle PSTHs of the firing of both types of neurons at baseline and during stimulation. For comparison, the baseline PSTHs (in blue) are repeated in the PSTHs during stimulation. Black arrows denote the mean phases. Note that different scales are used in the PSTHs for clear illustrations.

During the period of sinusoidal stimulations at the Schaffer collaterals of the hippocampal CA1 region (Fig. 1A), the neuronal firing in the downstream area increased (Fig. 1B). The unit spikes of neurons rode on the noticeable fluctuations of sinusoidal artifacts in the raw recordings. The extracted MUA signal clearly showed that most of the unit spikes appeared around the trough of the applied sinusoidal waveforms (Fig. 1B, expended plots on the lower middle, red sinusoidal curves denote the stimulations applied at the Schaffer collaterals). After the end of the stimulation, the unit firing of neurons fell immediately. It returned to the baseline level and returned theta rhythm in the LFP within tens of seconds. In the 9 experiment preparations, we obtained single-unit spikes of 40 neurons, including 23 putative pyramidal neurons and 17 putative interneurons. During the 1-min 50 Hz sinusoidal stimulation, the mean firing rate of unit spikes of the pyramidal neurons and the interneurons significantly increased from 2.8 $\pm$ 1.2 spikes/s and 14.1 $\pm$ 5.0 spikes/s in pre-stimulation to 7.8 $\pm$ 4.1 spikes/s and 50.6 $\pm$ 14.7 spikes/s during stimulation, respectively (Fig. 1C). Also, the increments of pyramidal neurons (5.0 $\pm$ 3.8 spikes/s) were significantly smaller than that of interneurons (36.2 $\pm$ 15.9 spikes/s, P 0.01, t-test). The results indicated an excitatory effect of the sinusoidal stimulations on the downstream neurons. Next, we investigated the phase-locked relationship between the firing of the two types of neurons and the sinusoidal stimulation.

In the baseline recordings of pre-stimulation, the unit spikes of both pyramidal neurons and interneurons appeared uniformly in the phases of mimic sinusoidal cycles without an apparent phase-locked relationship (Fig. 2A upper). The mean PSTHs of the neuronal firing were distributed evenly in the mimic sinusoidal cycles (Fig. 2B upper). However, during the real 50-Hz sinusoidal stimulation, the firing of the two types of neurons was phase-locked with the sinusoidal waveforms (Fig. 2A Lower). The mean PSTHs showed that the most firing of pyramidal neurons appeared in the phase range of 180-320${}^{\circ }$ (i.e., 10 - 18 ms) of the 20-ms sinusoidal cycles with a mean firing phase of 260 $\pm$ 61${}^{\circ }$ (n = 23; Fig. 2B lower left), while the firing of interneurons appeared in the phase range of 80-120${}^{\circ }$ (i.e., 4.5 - 6.7 ms) with a mean firing phase of 102 $\pm$ 16${}^{\circ }$ (n = 17; Fig. 2B lower right). The results indicated strong phase-locked relationships between the firing of the two types of neurons and the sinusoidal waveforms of stimulations. Furthermore, the distribution peak in the interneuronal PSTH was narrower than that of the pyramidal neuron PSTH.

Taken together, the results showed that the sinusoidal stimulations of afferent axons entrained the firing of both pyramidal neurons and interneurons in the downstream region and increased their firing. Next, we investigated whether or not the entrainment effects of sinusoidal stimulation could alter the rhythms of neuronal firing.

Fig. 3.

Sinusoidal stimulation changed the firing rhythms of downstream neurons. (A) Typical examples of the neuronal firing and auto-correlation profiles of the two types of neurons at 1-min periods of pre-stimulation and during stimulation. (B) The unit-power spectrum densities (unit-PSD) of pyramidal neurons and interneurons showed in (A). (C) and (D) Comparisons of cumulative powers in the frequency bands of theta rhythm (2-5 Hz) and around the sinusoidal stimulation (49-51 Hz) between the periods of pre-stimulation and during stimulation for pyramidal neurons and interneurons, respectively. ${}^{**}$P 0.01, paired t-test.

During the baseline recordings, the spontaneous firing of the two types of neurons had an unmistakable theta rhythm (2-5 Hz) in the auto-correlation function of the spike sequences (Fig. 3A upper). The rhythms generated a peak at 2-5 Hz in the spectrum of the unit spike sequences of neuronal firing (Fig. 3B blue curves). However, during the period of sinusoidal stimulations, the theta rhythm in the neuronal firing was replaced by a new rhythm at the sinusoidal frequency (i.e., 50 Hz; Fig. 3A lower), indicating by the peaks at 50 Hz in the spectrums (Fig. 3B red curves).

To confirm the changes of firing rhythms, we calculated the cumulative powers in the frequency bands of theta rhythm (2-5 Hz) and around the sinusoidal frequency (49-51 Hz) in the spectrums of the neuronal firing of all the pyramidal neurons and interneurons respectively in the pre-stimulation and during stimulation periods. The stimulations significantly decreased the mean power of theta rhythms (Fig. 3C) and significantly increased the mean power of 49-51 Hz (Fig. 3D) in the firing of the two types of neurons.

The results indicated that the sinusoidal stimulations altered the rhythms in neuronal firing from a theta rhythm of pre-stimulation to a new rhythm at stimulation frequency.