Modulating the firing of downstream neurons by orthodromic axonal stimulation
To obtain mere unit spikes in extracellular recordings during stimulation, we first determined the stimulation intensity. A single stimulation pulse applied to the Schaffer collaterals of hippocampal CA1 region could activate a bunch of axons to generate action potentials that would spread along the axons and agitate the downstream neurons (see Fig. 2a). If the intensity of the pulse was adequate (e.g., 0.1 mA), a population of downstream neurons would generate action potentials simultaneously. Because of the dense packing of the cell bodies in the hippocampal region [17], these evoked action potentials would superimpose together to form a potential waveform so-called population spike (PS) in the pyramidal layer. The PS would prevent the extraction of unit spikes immediately following stimulation pulses (Fig. 2b). Therefore, smaller intensities were tested until a single pulse only induced unit spikes without PS potentials (Fig. 2c). The weak intensity (e.g., 20 μA) was then used for the stimulation trains with various pulse frequencies (Fig. 1). In the stratum radiatum, the field EPSP (fEPSP) evoked by the first pulse of each stimulation train was distinguishable as long as the inter-pulse-interval was long enough, e.g. 10 ms for 100 Hz frequency (Fig. 2d).
With a lower stimulation frequency (e.g., 20 Hz), unit spikes always followed the pulses (Fig. 3a), indicating a strong control of the stimulation on the neuronal firing. However, the mean firing rate of MUA during the stimulation period (34.2 counts/s) was similar to the value (35.2 counts/s) in the baseline recording before stimulation. When the stimulation frequency was increased to 100 Hz (Fig. 3b), the mean firing rate of MUA during stimulation increased to 115 counts/s. In addition, a silent period (~ 0.3 s) without spikes appeared immediately following the cessation of stimulation (see the curve in the bottom of Fig. 3b).
These results indicated that stimulations of pulse trains could modulate the firing of downstream neurons and increase firing rates of neurons by a higher pulse frequency.
Because the stimulation of afferent axons (i.e., Schaffer collaterals) can simultaneously activate both interneurons and pyramidal cells in the downstream projection region (Fig. 2a), we next compared the firing of the two types of neurons.
Frequency-dependent responses of the two types of neurons to stimulation trains
To investigate the changes of neuronal responses induced by the 0.5-s stimulations, dynamic firing rates of individual interneurons (n = 12) and pyramidal cells (n = 27) were evaluated (Fig. 4a–f left). The mean firing rates in the three periods of before, during and after the stimulations with various pulse frequencies were also evaluated (Fig. 4a–f middle and right). For the period before stimulations, statistical analysis of one-way ANOVA showed no significant differences among the mean baseline firing rates with different frequencies for both interneurons (F5,66 = 0.07, P = 0.99) and pyramidal cells (F5,156 = 0.29, P = 0.91), indicating similar baseline neuronal states. Because the firing of unit spikes stopped for a short period immediately following the end of stimulation train (Fig. 3b), to evaluate the recovery of neuronal firing after stimulation, the post-stimulation firing rate was calculated in the time window of 10–12 s after the cessation of stimulation.
During stimulations with lower pulse frequencies (10 and 20 Hz), the interneurons generated a strong phase-locked firing following each pulse with a mean delay of 6.7 ± 5.9 ms (10 Hz) and 4.3 ± 2.1 ms (20 Hz) between unit spikes and the preceding pulse (Fig. 4a, b). The mean firing rates of interneurons (15.2 ± 7.2 and 21.3 ± 8.5 counts/s for 10 and 20 Hz, respectively) were not significantly different from the mean firing rates before stimulations (paired t test, P > 0.1, n = 12). However, the mean firing rates of pyramidal cells decreased significantly during stimulations (vs. the values before stimulations, paired t test, P < 0.01, n = 27). The mean delay of unit spikes of pyramidal cells was 40.8 ± 13.6 ms (10 Hz) and 19.1 ± 5.5 ms (20 Hz). The long delays with relatively large variances indicated that the firing of pyramidal cells was not phase-locked to the pulses of stimulation.
During stimulations with a higher pulse frequency of 50 Hz, the mean firing rates of interneurons increased significantly, whereas the mean firing rates of pyramidal cells did not change significantly (Fig. 4c).
During stimulations with further higher pulse frequencies (100, 200 and 400 Hz), the mean firing rates of both interneurons and pyramidal cells were significantly greater than the values before stimulation. In addition, during 200 and 400 Hz stimulations, the firing rates of interneurons showed a quick increase to a peak value near the onset of stimulation and then fell slightly to a steady-state (Fig. 4d–f).
Interestingly, the firing rates of interneurons increased significantly with the increase of pulse frequency from 10 to 100 Hz and maintained steady at 100 and 200 Hz (93.2 ± 40.3 and 88.0 ± 54.9 counts/s, respectively) (Fig. 5). These firing rates at 100 and 200 Hz were significantly greater than the values at pulse frequencies of 10 and 20 Hz (15.2 ± 7.2 and 21.3 ± 8.5 counts/s; P < 0.01, post-hoc Bonferroni tests after significant ANOVA P < 0.01, n = 12). The firing rates of interneurons decreased slightly when the pulse frequency increased further to 400 Hz, but the value (65.9 ± 39.5 counts/s) was still significantly greater than the value at low frequency of 10 Hz (post hoc Bonferroni tests, P < 0.01, n = 12). During stimulations with a frequency over 50 Hz, the firing rates of all 12 interneurons increased (Fig. 4c–f middle, blue dots in the scatter plots).
The relative changes in firing rates of pyramidal cells with the increase of pulse frequency were similar to the changes of interneurons. During stimulations with a frequency below 20 Hz, more than 80% of the pyramidal cells decreased firing. During stimulations with a frequency of 100–400 Hz, 63–89% of the 27 pyramidal cells increased firing (Fig. 4a–f middle, red dots in the scatter plots). The firing rates of both types of neurons returned to baseline levels after stimulations (Fig. 4a–f right).
In addition, during stimulations with an identical pulse frequency, the mean firing rate of interneurons was always significantly greater than the value of pyramidal cells (4.3 to 8.8 multiples; Fig. 5; t test, P < 0.01). During baseline recordings before stimulation, the firing rates of interneurons were also significantly greater than the values of pyramidal cells (3.2 to 4.4 multiples; see Fig. 4a–f right; t test, P < 0.01). Taking the 100 Hz stimulation for example (Fig. 4d right), the firing rates of interneurons during baseline (17.2 ± 9.4 counts/s) and during stimulation (93.2 ± 40.3 counts/s) were all significantly greater than the corresponding values of pyramidal cells (6.1 ± 5.3 counts/s during baseline and 13.8 ± 16.9 counts/s during stimulation). Nevertheless, stimulations increased the difference between the firing rates of the two types of neurons from 3.2 multiples to 6.7 multiples.
To examine whether or not the successive short-stimulation trains with a higher frequency generated a long-term potentiation (LTP) in synaptic transmission, we took the first pulse of a train as a test pulse (Fig. 2d) to compare the slopes of fEPSP evoked by the first pulses of the first train and the last train (tenth). In the stimulation session of 100 Hz, the typical frequency to induce LTP, after nine preceding trains, the mean slope of fEPSP in the beginning of the tenth train (0.32 ± 0.31 mV/ms) was similar to that of the first train (0.31 ± 0.27 mV/ms; n = 7 rats, P > 0.1 paired t test). Similarly, in the stimulation session of 50 Hz, the mean slopes of fEPSP induced by the first pulses of trains were also similar (last train 0.30 ± 0.28 mV/ms vs. first train 0.32 ± 0.29 mV/ms; n = 7 rats, P > 0.1 paired t test). This indicated that the stimulation trains with a weak current intensity used in the present study did not induce obvious LTP.
These results indicated that during stimulations, the firing rates of interneurons and pyramidal cells increased in proportion to the pulse frequency till 100 Hz and then saturated even with the stimulation frequency reaching up to 400 Hz. In addition, the increases in firing rates of interneurons were greater than those of pyramidal cells. Compared to baseline values, the firing of pyramidal cells was suppressed by stimulations with lower frequencies but was enhanced by stimulations with higher frequencies. The firing of interneurons did not change significantly by stimulations with lower frequencies but was enhanced by stimulations with higher frequencies. The increase of neuronal firing by stimulation at a higher frequency was not caused by long-term changes in synaptic transmission.
Neuronal activity following the cessation of stimulation
Immediately following the cessation of stimulation, neurons stopped firing for a short period and then recovered gradually (Figs. 3b, 6a). Therefore, we next evaluated the post-effects of stimulation on the two types of neurons by the length of silent periods.
For the interneurons (Fig. 6b), the mean length of silent periods increased as the stimulation frequency increased from 10 to 100 Hz and then did not change significantly up to the frequency of 400 Hz. The mean lengths of silent periods with stimulations over 100 Hz (100 Hz: 1.1 ± 0.6 s; 200 Hz: 1.1 ± 0.4 s; 400 Hz: 1.0 ± 0.4 s) were all significantly longer than the values at lower frequencies (10 Hz: 0.3 ± 0.2 s; 20 Hz: 0.4 ± 0.2 s) (ANOVA P < 0.01; post-hoc Bonferroni tests, P < 0.01, n = 12).
For the pyramidal cells (Fig. 6b), the mean length of silent periods increased as the stimulation frequency increased from 10 to 400 Hz. However, only the silent period at 400 Hz (4.2 ± 3.5 s) was significantly longer than the silent periods at lower frequencies (10 Hz: 1.9 ± 1.8 s; 20 Hz: 2.3 ± 2.4 s) (ANOVA, P < 0.05; post-hoc Bonferroni tests, P < 0.05, n = 27). The lack of significant differences among other groups might be caused by the great variances of silent periods of pyramidal cells. In addition, the mean duration of silent periods of pyramidal cells (> 2 s) was always significantly longer than the mean value of interneurons (≤ 1.1 s, Fig. 6b).
These results showed that the silent period of interneurons increased significantly with stimulation frequencies over 100 Hz, which was parallel to the changes of neuronal firing rates during stimulations (Fig. 5).