Cathodic- and anodic-pulses can alternately activate different sub-populations of neurons during sustained high-frequency stimulation of axons in rat hippocampus

Objective. Charge-balanced biphasic-pulses are commonly used in neural stimulations to prevent possible damages caused by charge accumulations. The lagging anodic-phases of biphasic-pulses may decrease the activation efficiency of stimulations by counteracting the depolarization effect of the leading cathodic-phases. However, a monophasic anodic-pulse alone can itself activate neurons by depolarizing neuronal membrane through a mechanism of virtual cathode. This study aimed to verify the hypothesis that the anodic-phases/pulses in charge-balanced stimulations could play an activation role during sustained high-frequency stimulations (HFSs). Approach. Two types of antidromic HFS (A-HFS) were applied on the alveus of hippocampal CA1 region of anesthetized rats: monophasic-pulse A-HFS of alternate opposite pulses and biphasic-pulse A-HFS with the same frequency of 100 or 200 Hz. The antidromically-evoked population spike was used as a biomarker to evaluate the activation effects of A-HFS pulses. Main results. Despite a significant difference in the initial abilities of anodic- and cathodic-pulses to activate neurons, an anodic-pulse was able to induce similar amount of neuronal firing as a cathodic-pulse during sustained monophasic-pulse A-HFS. Additionally, the amount of neuronal firing induced by the monophasic-pulse A-HFS was similar to that induced by the biphasic-pulse A-HFS consuming a double amount of electrical energy. Furthermore, the alternate cathodic- and anodic-pulses respectively activated different sub-populations of neurons during steady A-HFS. Significance. The anodic-phases/pulses in charge-balanced HFS at axons can play an activation role in addition to a role of charge balance. The study provides important information for designing charge-balanced stimulations and reveals new mechanisms of neural stimulations.


Introduction
Extracellular stimulations of electrical pulse sequences have been utilized in neural therapies such as deep brain stimulation (DBS) for treating certain brain disorders, as well as cochlear implants and intracortical visual stimulations for neural prostheses [1]. To minimize the risks of damages to both stimulated tissues and metal electrodes, charge-balanced biphasic pulses are commonly used [2][3][4]. The leading phase of the biphasic pulse is cathodic and acts as a working phase to activate neurons, while the lagging phase is anodic and acts as a balance phase to deliver opposite charges to neutralize the charges of the leading cathodic-phase, thereby preventing damages caused by possible charge accumulations [5][6][7].
The addition of anodic-phases cannot only double the consumption of electrical energy but also decrease the activation efficiency of stimulations by hyperpolarizing neuronal membranes to counteract the depolarization effect of the leading cathodicphase [6,[8][9][10]. To improve the activation efficiency of biphasic pulses, asymmetric charge-balanced pulses have been utilized, such as a rectangular cathodic-phase followed by a smaller and longer anodic-phase of rectangular or non-rectangular waveform [2,11,12]. In addition, short inter-phasegaps (IPGs) have been inserted between the cathodicand anodic-phases to decrease the counteraction of anodic-phase [9,13].
Nevertheless, a monophasic anodic-pulse itself can activate neurons by depolarizing neuronal membrane. For extracellular stimulations, due to the closed-loop flow of an electrical current, the anodicpulse hyperpolarizing neuronal membrane in the vicinity of electrode must simultaneously depolarize membrane at distant sites on the same neuron [14,15]. Especially for the long and thin axonal structure of neurons, an anodic-pulse can depolarize axonal membrane at flanking regions of the stimulation site and generate propagable action potentials (APs) [14,16]. Although the activation efficiency of an anodic-pulse is much smaller than a cathodicpulse under a normal situation [6,15,17], we proposed here that anodic-pulses/phases could play an important role in activating neurons with certain paradigms of stimulations, such as sustained highfrequency stimulations (HFSs) commonly used in DBS.
HFS sequences of biphasic-pulses around 100-200 Hz have been utilized in clinic DBS [18,19]. In addition, axonal activations by HFS have been shown to play a crucial role in DBS therapy [20,21], because axons occupy a large portion of brain space, and because an axon has a smaller rheobase current than other neuronal elements and is most prone to be activated by narrow pulses of HFS [14,[22][23][24]. However, sustained HFS may generate intermittent axonal blockage to decrease the evoked firing of axons/neurons because of an excessive depolarization of axonal membranes by HFS [25][26][27]. Under this situation, an addition of a reverse effect by anodicpulses could be expected to alleviate the axonal blockage and to increase the activation effect of HFS on axons.
Based on the considerations above, we hypothesized that by separating two phases of biphasic-pulses with an IPG to form a HFS sequence composed of alternate monophasic cathodic-and anodic-pulses in axonal stimulations, the anodic-pulses could activate neurons themselves rather than only balance charges. To test the hypothesis, we utilized the antidromicallyevoked population spike (APS) in rat hippocampus as a biomarker to evaluate the activation effects of different types of pulses in HFS sequences. The results may reveal new effects of anodic-pulses/phases in chargebalanced HFS and provide important information for developing stimulation paradigms to improve activation efficiency and to save electrical energy in neural stimulation therapies.

Animal surgery and electrode implantations
Experiments were performed on adult male Sprague-Dawley rats in a stereotaxic apparatus under anesthesia by urethane (1.25 g kg −1 , i.p.). Surgical procedures and electrode implantations were similar to previous reports [26]. Briefly, a recording electrode (RE) and a stimulation electrode (SE) were inserted into the hippocampal CA1 region of left brain with coordinates in millimeter (AP −3.5; ML 2.7; DV 2.5) and (AP −4.8; ML 2.7; DV 2.3), respectively. The distance between the RE and SE was ∼1.3 mm ( figure 1(A)). The RE was a 16channel array (#Polytrode, A1x16-Poly2-5 mm-50 s-177, Neuro-Nexus Technologies Inc., USA) and was positioned perpendicularly across the layers of CA1 region activated by stimulations from SE. The SE was a bipolar concentric electrode (#CBCSG75, FHC Inc., USA) and was targeted at the axons of CA1 pyramidal neurons, the alveus fiber of CA1 region (figure 1(A)). The diameters of inner and outer poles of the bipolar SE were 75 and 250 µm respectively, and the heights of both poles were 100 µm with a separation distance of 100 µm.
The spontaneous unit spikes as well as the typical waveforms of evoked potentials (including APS at the pyramidal cell layer), serially appearing in the recording channels of RE array, were used to justify the electrode positions instantly during electrode implantations. The RE was first in position. Then the SE was inserted gradually to approach the alveus until the amplitude of APS evoked by a test cathodicpulse (100 µs, 0.3 or 0.4 mA) was not able to increase further, indicating that the inner pole of SE had reached the alveus. After the experiments, the electrode positions were confirmed by stained sections (figure 1(A, right)) of paraformaldehyde-fixed brains by histological methods reported previously [28].

Recording and stimulating
Electrical potentials collected by the RE were amplified 100 times by a 16-channel extracellular amplifier (Model 3600, A-M System Inc., USA) with a band-pass filtering range of 0.3-5000 Hz. Then the amplified signals were sampled by a PowerLab data acquisition system (Model PL3516, ADInstruments Inc., Australia) with a sampling rate of 20 kHz per channel.
Stimulation pulses applied on the alveus were monophasic cathodic-and anodic-pulses, denoted as ⊤ and ⊥ respectively, and symmetric biphasic-pulses with cathodic-phase first and zero IPG, denoted as ┼. The stimuli were all rectangular current pulses with a width of 100 µs per phase and a current intensity of 0.3-0.4 mA per phase, and were generated by a programmable stimulator (Model 3800, A-M System Inc., USA). The current intensity of the pulses was able to evoke APSs with an amplitude approximately 3/4 of the maximal amplitude in the input-output curve of APS amplitudes.
Two types of antidromic HFS (A-HFS) were applied on the alveus: biphasic-pulse A-HFS and monophasic-pulse A-HFS of alternate cathodic-and anodic-pulses. Both types of A-HFS were chargebalanced sequences. The polarity of pulses was defined according to the polarity of current flowing through the inner pole of the bipolar SE. The pulse frequency of A-HFS was 100 or 200 Hz. The frequency of monophasic-pulse A-HFS was defined as the sum number of cathodic-pulses and anodicpulses per second. The duration of A-HFS was 2 min.

Data analysis
The amplitude and latency of evoked APS waveforms were used to evaluate the neuronal responses to different types of pulses. The APS amplitude was measured as the potential drop of the negative peak of APS. The APS latency was the time distance between the stimulation pulse and the negative peak of APS. The evoked APSs were steady in the second minute of the 2 min A-HFS. An average value of APS waveforms calculated by the data in the last 1 s of A-HFS (denoted by a subscript 'end') was termed as a steady value to describe the evoked APSs during the steady period. To clarify illustrations, stimulation artifacts in the A-HFS recordings were removed by a custom-made MAT-LAB program as previous reports [29].
All statistical data were represented as mean ± standard deviation with n representing the number of rats for data collections. Paired t-test, one-way analysis of variance (ANOVA) or two-way ANOVA with post-hoc Bonferroni tests were used to judge the statistical significances of the differences among data groups.

Responses of neuronal populations to monophasic-pulse A-HFS with alternate cathodicand anodic-pulses
A single pulse applied at the alveus can activate the axon fiber and then antidromically evoke APs in the cell bodies of pyramidal neurons in the hippocampal CA1 region. The synchronous firing of the neuronal population forms an APS waveform that can be recorded extracellularly in the pyramidal cell layer to evaluate the stimulation effect [30]. With an identical current intensity, a ⊤ pulse always evoked a larger APS than a ⊥ pulse ( figure 1(A, left)).
At the onset of a 100 Hz monophasic-pulse A-HFS (i.e. 50 Hz for ⊤ and ⊥ pulses respectively), the two types of pulses alternately evoked larger and smaller APSs (figure 1(B, bottom left)), similar to single pulse stimulations at baseline. As the A-HFS continued, the APS amplitudes decreased rapidly, indicating failures of neurons to follow each of the high-frequency pulses. Interestingly, the APSs evoked by ⊥ pulses decreased first to almost disappear and then returned to a steady level, while the APSs evoked by ⊤ pulses decreased monotonically to a steady level. After ∼40 s stimulation, the amplitudes of APSs evoked by both types of pulses became similar (figures 1(B) and (C1)). In the experiments with 100 Hz A-HFS, the amplitudes of initial APS evoked by the first ⊤ and ⊥ pulses (A ⊤ini and A ⊥ini ) were significantly different (figure 1(C2); P < 0.01, paired t-test, n = 15). The mean A ⊥ini was only about half of the mean A ⊤ini . However, the mean amplitudes of steady APSs evoked by the two types of pulses were similar (A ⊤end and A ⊥end in figure 1(C3)). The difference between the amplitudes of the two types of APSs significantly decreased from the initial value ). The A ⊤end and A ⊥end were ∼20% and ∼40% of A ⊤ini and A ⊥ini , respectively (see figures 1(C2) and (C3)), indicating different suppression ratios of APSs evoked by the two types of pulses.
In addition, during the A-HFS, the latencies of all APSs gradually increased with the mean latency following ⊥ pulses shorter than that following ⊤ pulses (figure 1(C4)). The significant differences between the mean latencies of the two types of APSs existed through the entire period of A-HFS (L ⊥ini vs L ⊤ini in figure 1(C5) and L ⊥end vs L ⊤end in figure 1(C6); P < 0.01, paired t-test, n = 15). Furthermore, the latencies of both types of APSs nearly doubled from their initial values to steady values, resulting in the difference of the two latencies significantly increasing from the initial value (∆L ini = L ⊤ini − L ⊥ini = 0.12 ± 0.06 ms) to the steady value (∆L end = L ⊤end − L ⊥end = 0.48 ± 0.08 ms; P < 0.01, paired t-test, n = 15). After the end of A-HFS, the APSs evoked by single test stimulations of the ⊤ and ⊥ pulses all recovered to the baseline level in ∼2 min (figure 1(B, upper right)).
During A-HFS with an increased pulse frequency of 200 Hz, the change of APS amplitudes was similar to the 100 Hz A-HFS (figures 1(D1) and (D2)). The difference between the amplitudes of the two types of APSs significantly decreased from the initial value (∆A ini = 5.7 ± 1.7 mV) to the steady value (∆A end = −0.11 ± 0.43 mV; P < 0.01, paired t-test, n = 13). The APS amplitudes of ⊤ pulses decreased to below 10% of the initial value and that of ⊥ pulses decreased to below 20% (see figures 1(D1) and (D2)), approximately half of the ratios with 100 Hz A-HFS. Again, the mean latency of APS evoked by ⊥ pulses was significantly shorter than that evoked by ⊤ pulses (figures 1(D3) and (D4); P < 0.05 or 0.01, paired t-test, n = 13), with the mean latency difference increasing significantly from the initial value (∆L ini = 0.08 ± 0.14 ms) to the steady value (∆L end = 0.23 ± 0.15 ms, P < 0.05, paired t-test, n = 13).
These results indicated that despite the significant difference in their initial activations on neurons, the ⊥ pulses were able to activate neurons as efficiently as ⊤ pulses during sustained A-HFS of alternate pulses when the evoked APSs were suppressed. Previous studies have also shown suppressions of APS during biphasic-pulse A-HFS [26,31]. We next compared the APSs evoked by the two types of A-HFS (monophasic vs biphasic) with a same pulse frequency and a same intensity, but with a different consumption of electrical energy.

Comparisons of APSs evoked by monophasic-pulse A-HFS and by biphasic-pulse A-HFS
At the initial period of a 100 Hz biphasic-pulse A-HFS, large APS (∼10 mV) followed each pulse and then decreased rapidly in seconds (figure 2(A)). After tens of seconds of continuous stimulation, the APS amplitudes decreased to a steady level that was similar to the steady level of monophasic-pulse A-HFS (figure 2(B, left)). Meanwhile, the latencies of APSs evoked by biphasic-pulses increased to a steady level between the latencies of APS evoked by the two opposite pulses of monophasic-pulse A-HFS ( figure 2(B, right)).
Statistical data showed that significant differences existed among the amplitudes of initial APSs of biphasic-pulses (A ┼ ini ) and of monophasic-pulses (A ⊤ini and A ⊥ini ). Both A ┼ ini and A ⊤ini were significantly greater than A ⊥ini (figure 2(C1), * * P < 0.01, two-way ANOVA with post-hoc Bonferroni tests, n = 8). During the steady period, the significant differences disappeared (figure 2(C2)). In addition, significant differences existed among the latencies of APSs evoked by the three types of pulses in both initial and steady periods of A-HFS. The latencies of APSs evoked by biphasic-pulses (L ┼ ini and L ┼ end ) were significantly longer than those evoked by anodic-pulses (L ⊥ini and L ⊥end ) and were significantly shorter than those evoked by cathodic-pulses (L ⊤ini and L ⊤end ) (figures 2(C3) and (C4), * * P < 0.01 or * P < 0.05, twoway ANOVA with post-hoc Bonferroni tests, n = 8). Similar results were obtained by comparing the amplitudes and latencies of APSs of the three types of pulses during A-HFS with a higher frequency of 200 Hz, except that the latency L ┼ end was significantly shorter than both latencies L ⊤end and L ⊥end ( figure 2(D)).
To compare the mean amounts of neuronal firing evoked by monophasic-pulse A-HFS and biphasicpulse A-HFS, the amplitudes of APSs evoked by cathodic-and anodic-pulses in monophasic A-HFS were averaged (the rightmost bar with two colors in figures 2(C1), (C2), (D1) and (D2)). For 100 Hz A-HFS, at the initial period, the average amplitude (A ⊤ini + A ⊥ini )/2 of monophasic-pulses was significantly smaller than the amplitude A ┼ini of biphasic-pulses (figure 2(C1), P < 0.05, paired t-test, n = 8). At the steady period, the (A ⊤end + A ⊥end )/2 became similar to A ┼end (figure 2(C2)), indicating similar neuronal firing evoked by A-HFS of alternate monophasic-and biphasic-pulses. Similar results of APS amplitudes were obtained during A-HFS with a higher frequency of 200 Hz (figures 2(D1) and (D2)).
In addition, to compare the recoveries of neuronal activity after the two types of A-HFS, single test pulses with the identical parameters as A-HFS pulses were repeatedly applied with an interval of ∼30 s after the end of A-HFS. For the both types of A-HFS, over 90% changes in the APS amplitudes and latencies recovered in ∼2 min following the end of A-HFS, indicating no obvious neuronal damages caused by these A -HFS (figure 3).
These results showed that although significant differences existed in the initial APSs, the alternate monophasic-pulse A-HFS and the biphasic-pulse A-HFS suppressed APSs to a similar level during the steady period of A-HFS. The suppressed APSs indicated that each of the pulses only activated a small fraction of the neuronal population that was covered by the pulses in baseline situation or at the onset of A-HFS. Because the activation sites of the ⊤ and ⊥ pulses along an axon should be different [7,14,32], we hypothesized that the suppressed APSs evoked by the ⊤ and ⊥ pulses of monophasic-pulse A-HFS could be formed by the firing from different subpopulations of neurons, which was different from biphasic-pulse A-HFS. The hypothesis was verified next.

Cathodic-and anodic-pulses activate different sub-populations of neurons during steady period of monophasic-pulse A-HFS
We utilized the theory of refractory period following neuronal firing to test whether or not two consecutive pulses activate a same population of neurons [33,34].
When pairs of biphasic-pulses with identical parameters and with a short inter-pulse interval (IPI) were applied in baseline situation, the second pulse (termed as test pulse) only induced a small APS with an IPI of 1.5 ms following the first pulse (termed as control pulse), and induced no APS with an IPI of 0.8 ms ( figure 4(A)). The result indicated a refractory period of ∼1 ms under the baseline situation. However, during sustained 100 Hz A-HFS of the biphasicpulses with the identical parameters, an additional pulse, even inserted 5 ms following a pulse of A-HFS, induced no APS ( figure 4(B)), indicating a refractory period longer than 5 ms at this time. The result was consistent with previous report that the refractory period of neuronal firing may be extended by A-HFS [31].
As a comparison, pairs of monophasic-pulses were applied with a leading ⊤ pulse followed by a ⊥ pulse in baseline situation. The ⊥ pulse only induced  a small APS with an IPI of 1.5 ms following the ⊤ pulse, and induced no APS with an IPI of 0.8 ms due to refractory period ( figure 4(C)). This result suggested that the population of neurons activated by a ⊥ pulse was included completely in the population of neurons activated by a ⊤ pulse in baseline situation.
During sustained 100 Hz A-HFS of alternate ⊤ and ⊥ pulses, an additional pulse of ⊤ or ⊥ inserted 5 ms following the two types of pulse generated different APSs ( figure 4(D)). When the polarity of inserted pulse was same as the preceding A-HFS pulse, the inserted pulse induced no APS (figures 4(D1) and (D2)), similar to the results in inserting biphasicpulse into A-HFS of biphasic-pulses ( figure 4(E)). However, when the polarity of inserted pulse was opposite to the preceding pulse, it induced a substantial APS (figures 4(D3) and (D4)) with a mean amplitude of ∼60%-70% of the APS induced by an immediately previous pulse with the same polarity ( figure 4(E)). In addition, the inserted pulse resulted in no APS following the subsequent pulse with the same polarity (denoted by the hollow triangles in figures 4(D3) and (D4)). This again indicated the effect of an extended refractory period following the APS induced by the inserted pulse.
These results suggested that the biphasic-pulses controlled a same population of neurons either in baseline or during sustained A-HFS of biphasicpulses. However, the monophasic-pulses of opposite polarities controlled different sub-populations of neurons during sustained A-HFS of alternate monophasic-pulses, although in baseline a ⊥ pulse only controlled a fraction of neurons that was covered by a ⊤ pulse.

Discussion
The novel findings of this study include: (a) an anodic-pulse can induce similar amount of neuronal firing as a cathodic-pulse during sustained A-HFS of alternate monophasic-pulses, despite a significant difference in their initial abilities of neuronal activations. (b) The cathodic-and anodic-pulses can activate different sub-populations of neurons, respectively. (c) The monophasic-pulse A-HFS can induce similar amount of neuronal firing as the biphasicpulse A-HFS that consumes a double amount of electrical energy. Possible mechanisms and implications underlying these results are discussed below.

Possible underlying mechanisms of axonal HFS with alternate cathodic-and anodic-pulses
It is interesting that an anodic-pulse can have an activation ability equivalent to that of a cathodicpulse during sustained A-HFS. Generally, the depolarization effect of an anodic-pulse is much weaker than that of a cathodic-pulse with a same stimulation intensity, because an anodic-pulse depolarizes axonal membrane at the flanking regions of stimulation site through a mechanism of virtual cathode (figure 5(A)) [7,14,16,17]. Therefore, in the baseline situation (or at the initial of monophasic-pulse A-HFS), an anodicpulse can only activate the axons in a smaller zone close to the SE with a smaller magnitude of depolarization at the flanking regions (indicated by zone1 in figures 5(A1) and (A2)), while a cathodic-pulse can activate both the close and distant axons with a greater magnitude of depolarization at the center region immediately under the stimulation site (indicated by zone1 and zone2 in figures 5(A1) and (A3)). This was confirmed by the significant smaller APS evoked by an anodic-pulse than that by a cathodic-pulse in the initial of monophasic-pulse A-HFS (figures 1(C2) and (D1)). In addition, a cathodic-pulse can generate hyperpolarization in the flanking regions that may hinder the propagation of AP initiated at the center region [32], thereby causing a longer latency of APS than that of APS induced by an anodic-pulse (figures 1(C5) and (D3)).
However, during sustained 100 and 200 Hz A-HFS, similar small APSs were evoked by the both types of pulses (figure 1). The suppression of APS is due to depolarization blockage induced in axons as reported previously with A-HFS [25,26,35,36], which may be caused by an increase of extracellular potassium ([K + ] o ) in the peri-axonal space by axonal HFS [27,37,38]. Nevertheless, an intriguing thing is that the suppression ratio of APS by anodicpulses was smaller than that by cathodic-pulses ( figure 3(B)), thereby resulting in the similar small APSs evoked by the both types of pulses. This could be due to the fact that the anodic-pulses generated weaker depolarizations and less accumulations of [K + ] o thereby resulting in a weaker depolarization blockage and a smaller suppression ratio of APS.
Presumably, during steady period of A-HFS, the cathodic-pulses may no longer activate the axons in the zone closer to the electrode because of a stronger depolarization blockage generated by the pulses, but the anodic-pulses can. Thus, the anodicpulses and the cathodic-pulses may activate the close zone and distant zone, respectively (indicated by zone1 and zone2 in figure 5(B1)). The experiment data of refractory period tests support this inference ( figure 4). The two zones could overlap without a definite boundary line indicated by the black dot line in the figure 5(B1). Furthermore, each of the pulses, either anodic-or cathodic-pulses, may only activate a fraction of the axons in zone1 or zone2 (indicated by the radial dot lines in the zones in figure 5(B1)), thereby resulting in suppressed APSs. That is, the axons could only intermittently follow some of the pulses, not every pulse, to fire APs (figures 5(B2) and (B3)) due to a possible mechanism of [K + ] o accumulation [27,38]. In addition, the difference in APS latencies evoked by the two types of pulses persisted through the entire A-HFS due to the differences in initial sites of APs and in the obstruction of hyperpolarization by cathodic-pulses. And, the latencies were all prolonged by the depolarization caused by [K + ] o accumulation during sustained A-HFS.
In the initial seconds of A-HFS, with the activation efficiency of pulses decreasing, the anodic-and cathodic-pulses may compete to activate axons. Due to its stronger ability of depolarization, the cathodicpulses may first obtain a dominant position to prevent the anodic-pulses to activate axons and result in a sharp decrease of anodic-evoked APS even to a transient disappearance ( figure 1(C1)). Then, after the depolarization blockages cause cathodic-pulses fail to activate the axons adjacent to the stimulation site, the anodic-pulses may obtain the opportunity to activate some axons through 'virtual cathode' at the flanks of stimulation site, and the anodic-evoked APSs reappear.
Furthermore, the distance between the SE and the target axons is one of the crucial factors determining the activation ability of delivered pulses [39][40][41]. In this study, the bipolar SE was positioned with its inner pole at the alveus ( figure 1(A)). With a pulse intensity of 0.3 or 0.4 mA, the APS evoked by a separate cathodic-pulse was about twice as large as the APS evoked by a separate anodic-pulse in baseline, then both types of APSs became similar during the steady period of monophasic-pulse A-HFS (figure 1). If the electrode was a little far away from the alveus, an anodic-pulse would hardly activate the axons while a cathodic-pulse would be able to activate a small number of the axons to generate a small APS. In this case, during sustained A-HFS, the anodic-pulses would still not activate the axons to generate APS and only cathodic-pulses would induce suppressed APSs (data not shown).
It is impossible that the differences in the APSs evoked by anodic-and cathodic-pulses were caused by an activation of different elements of neurons. The stimulation was impossible to directly activate the somata around the RE that was ∼1.3 mm from the stimulation site ( figure 1(A)). Also, the evoked potentials (APS) were impossible to be induced by synaptic inputs on dendrites since the APS latencies were all smaller than 2 ms at the initial period of A-HFS (figures 1(C5), (D3) and 2(C3), (D3)). In addition, the typical waveforms of evoked potentials serially appearing in the 16-channel RE ensured that the APSs were induced antidromically by activations of neuronal axons [42].
Another interesting finding of the present study is that with 100 and 200 Hz A-HFS, the amount of neuronal firing induced by the monophasic-pulse A-HFS was similar to that induced by the biphasic-pulse A-HFS (figure 2). Presumably, during sustained A-HFS with uniform biphasic-pulses, without the competition of separate anodic-pulses, the neurons in the close zone (zone1 in figure 5(B1)) can be either activated intermittently by the biphasic-pulses or totally blocked. In addition, the neurons in the distant zone (zone2 in figure 5(B1)) can be also activated intermittently by the biphasic-pulses but with a smaller firing amount per pulse than that induced by cathodicpulses during monophasic-pulse A-HFS, because the double frequency of cathodic-phases in the biphasicpulse A-HFS may generate a stronger depolarization blockage than monophasic-pulse A-HFS. Thus, the sum firing in the entire zone (zone1 plus zone2) evoked by the biphasic-pulses is similar to that evoked by monophasic-pulses, despite the fact that the electrical energy consumed by biphasic-pulses is twice as monophasic-pulses.
Taken together, with a proper position of SE at target axons, due to the putative mechanism of intermittent depolarization blockage of axons and the different initial sites of APs induced by the two types of pulses, the anodic-pulses can play an activation role similar to the cathodic-pulses in sustained monophasic-pulse A-HFS of alternate polarities. Nevertheless, the putative mechanisms need to be verified by more direct evidence in future studies with development of experimental technology.

Implications of HFS with alternate phases
Our present study firstly showed that sustained HFS can change the relationship of activation efficiencies between cathodic-pulses and anodic-pulses, which provides new information for designing chargebalanced stimulations for neural stimulations. The monophasic-pulse A-HFS used here (100 or 200 Hz) is a type of charge-balanced stimulation, equivalent to a sequence of biphasic-pulses (50 or 100 Hz) with an IPG (10 or 5 ms). Previous studies have shown that an IPG of a fraction of millisecond to several milliseconds can reduce the reverse effect of the anodic-phase on the leading cathodic-phase in charge-balanced stimulations [9,10,43], so did the IPG of 5 or 10 ms used in this study. In a further step, here we showed that the anodic-phase/pulse itself can also play an activation role to induce neuronal firing similar as the cathodic-phase/pulse. It means that a waveform design of the anodic-phase other than rectangular pulse may not be necessary. Utilizing the activation effect of anodic-pulses in continuous stimulations can save the electrical energy of implanted stimulators. Moreover, the quick recovery of neuronal responses following the A-HFS indicated that the monophasic-pulse A-HFS at 100-200 Hz would not damage brain tissues ( figure 3). This type of HFS was as safe as biphasic-pulse HFS.
Furthermore, the activation effect of monophasic-pulse HFS is different from that of biphasic-pulse HFS since our data showed that the opposite pulses of monophasic-pulse A-HFS can activate different sub-populations of neurons to fire alternately ( figure 4). It provides a new stimulation pattern with certain selective activations of neurons.
The alternate activations of sub-populations of axons may activate the post-synaptic neurons in a distinct way that is worthy of further investigations.
In addition, the present study provides new evidence to support the point that axonal HFS of 100-200 Hz can generate failures (such as depolarization blockage) at axons to prevent them to respond every pulse of HFS successfully [25,26,35,36]. Otherwise, if the axons had been able to follow every pulse of the axonal HFS reliably to generate and conduct APs, the suppression of APS would have been caused by failures at somata. Under this situation, the somata of the neurons would have received similar antidromic inputs of the APs initiated by whatever cathodic-and anodic-pulses at axons, losing a marker of the pulse types. Therefore, the fact that different sub-populations were activated separately by the two types of pulses can only be due to failures at the axons immediately under HFS.
Nevertheless, the results of HFS of alternate monophasic-pulses observed in the present study were only from the rat hippocampal CA1 region. Further studies are needed to verify their universality in other brain regions as well as in spinal nerve, auditory nerve and so on. Moreover, the applicability of the type of HFS with alternate monophasic-pulses in neural stimulations for treating neural disorders needs further investigations of animal studies and clinical verifications.

Conclusions
The novel finding of the study is that sustained high-frequency axonal stimulations can enable an anodic-pulse/phase to activate neurons parallel to a cathodic-pulse/phase, rather than to only act as a balance phase. The finding not only provides new information for developing efficient and power-saving stimulations but also reveals new mechanisms of neural stimulations.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.