Finds documents with both search terms in any word order, permitting "n" words as a maximum distance between them. Best choose between 15 and 30 (e.g. NEAR(recruit, professionals, 20)).
Finds documents with the search term in word versions or composites. The asterisk * marks whether you wish them BEFORE, BEHIND, or BEFORE and BEHIND the search term (e.g. lightweight*, *lightweight, *lightweight*).
This chapter delves into the intricate responses of neuronal axons to high-frequency pulse stimulations, focusing on the induction of axonal block and its implications on neuronal firing patterns. The study highlights the bidirectional propagation of action potentials along axons, the effects of intracellular and extracellular stimulations, and the role of axonal block in neuromodulation techniques. Key findings include the intermittent block of axonal firing during sustained high-frequency stimulation, the extension of the refractory period, and the modulation of post-synaptic neuronal activity. The chapter also explores the distinct recovery phases following stimulation, providing a comprehensive understanding of the dynamic interactions between axonal stimulation and neuronal responses. These insights are crucial for advancing therapeutic approaches in neuromodulation and understanding the underlying mechanisms of neural signal transmission.
AI Generated
This summary of the content was generated with the help of AI.
Abstract
Among neuronal structures, the axonal membrane has the shortest chronaxie, making it most susceptible to narrow pulses commonly used in neuromodulation techniques like deep brain stimulation (DBS). This chapter details our experiments in the rat hippocampal CA1 region, which verify intermittent axonal block caused by high-frequency stimulations (HFS). The experiments include both antidromic HFS (A-HFS) applied to the axons of recorded pyramidal neurons and orthodromic HFS (O-HFS) applied to presynaptic afferent fibers. The results showed that the axonal HFS can extend the refractory period of axons and alter the excitability of somata, causing a silent period without firing immediately after HFS cessation. The experiments also showed non-uniform clustered firing produced by HFS with constant pulse intervals.
5.1 Introduction
The axon is a slender, specialized extension of a neuron. It acts like a cable, conducting action potentials between neurons or from neurons to effectors for information transmission. Under normal physiological conditions, an action potential typically originates at the axon initial segment (AIS) and then propagates reliably along the axon. For simplicity, this book sometimes refers to action potentials from the AIS as “originating from the cell body” to distinguish them from those that start from the middle segment of axon produced by external stimuli. Action potentials travel along the axonal membrane through local currents (Fig. 5.1A). At rest, the axonal membrane maintains a resting potential of approximately − 70 mV, with positive charge outside and negative charge inside. In the excitation area where the action potential is generating, the membrane potential momentarily reverses—becoming negative outside and positive inside. This reversal creates a potential difference between the excited and adjacent resting areas, leading to local currents. As shown by the dashed arrow lines in Fig. 5.1A, current flows from the resting area to the excited area outside the membrane. Inside the membrane, current flows in the opposite direction to complete the current loop. In the resting area, the outward current can cause membrane depolarization. When this depolarization reaches threshold, it can trigger an action potential, turning the resting area into a new excited area. As this process continues, the action potential propagates along the axon. A myelinated axon follows the same conduction mechanism, but its action potential occurs at the nodes of Ranvier during conduction because these nodes have much lower impedance than the myelin-wrapped internodal areas. This results in saltatory conduction—the action potential jumps from node to the next.
Fig. 5.1
Generation and propagation of action potentials along an axon under distinct conditions. A Action potential originating from the axon initial segment (AIS) under normal physiological conditions. B Action potential produced intracellularly in the middle of axon. C Action potential produced by an extracellular pulse in the middle of axon
External stimulation can produce axonal action potentials through intracellular and extracellular methods. Figure 5.1B shows intracellular induction, in which a current is injected into the axon. The induced action potential originates in the middle of the axon and then propagates bilaterally. Similarly, extracellular pulses (typically negative) can also produce bilaterally propagatable action potentials on an axon (Fig. 5.1C). However, the two methods have a key difference: intracellular stimulation causes only depolarization, while extracellular stimulation produces both depolarization and hyperpolarization at different sites along the axon. Immediately at the stimulation site, the transmembrane current from an extracellular negative pulse flows outward, causing depolarization. The current loop can produce inward currents and cause hyperpolarization on both flanking sides. This hyperpolarization can impede action potential conduction, resulting in a lower maximum firing rate than intracellular induction. Nevertheless, extracellular stimulation has the advantage of activating multiple axons simultaneously with a single electrode contact.
Advertisement
Extracellular electrical stimulation is essential for developing clinical neuromodulation techniques, as these interventions require simultaneous activations of massive neurons to achieve therapeutic effects. Many treatments specifically target axons, including spinal cord stimulation (SCS) for pain relief (Lam et al. 2023), vagus nerve stimulation (VNS) for treating refractory epilepsy, depression, and migraines (Gurbani et al. 2016; Rush et al. 2005; Silberstein et al. 2020), and functional electrical stimulation (FES) for restoring limb function. Other successful applications include cochlear stimulation for hearing restoration, sublingual nerve stimulation for sleep apnea treatment, and electrical stimulation to promote axonal regeneration in damaged peripheral nerves (Juckett et al. 2022). These approaches work by directly stimulating nerves—bundles of axons—to either block nerve signal conduction or regulate neuronal and muscular activity. Although deep brain stimulation (DBS) does not specially aim at axons, axonal activity can also play an important role in its effectiveness.
DBS typically targets specific brain regions or neural nuclei. The electrical pulses from DBS electrodes can affect various parts of surrounding neurons, including cell bodies (somata), axons, and dendrites. Axonal membranes are particularly sensitive, having the lowest activation threshold and responding most readily to narrow pulses due to their low chronaxie. Therefore, action potentials can start from the axon—even when the soma is closer to the stimulation electrode (Ranck 1975; Nowak and Bullier 1998; McIntyre and Grill 1999). Studies have shown that high-frequency pulses (around 130 Hz with 60–450 μs width) mainly activate axons rather than somata and dendrites in brain regions such as the ventral nucleus of thalamus and the medial pallidum when treating Parkinson disease or primary tremor (Holsheimer et al. 2000). Additionally, axons occupy more brain space than other neural structures. White matter, which consists of neuronal axons, accounts for about half of the human brain volume (Fields 2008). When DBS pulses are delivered, they first activate nearby axons, whether these axons originate in, terminate in, or simply pass through the stimulation area. The resulting firing travels bilaterally along the axons and can modulate the activity of neural networks. As a result, the axonal responses to high-frequency pulse stimulation play a crucial role in DBS (Lozano et al. 2019; Lee et al. 2019; Chomiak and Hu 2007; Udupa and Chen 2015).
Under normal physiological conditions, the axonal action potentials have four features: bidirectionality, no attenuation, insulation, and resistance to fatigue. Bidirectionality means that an action potential (AP) can propagate in both directions from its origin site on the axon, rather than unidirectionally from soma. The propagation toward axonal terminals is orthodromic—the normal physiological direction, while that toward the soma is antidromic. The “no attenuation” property means that APs can maintain their strength as they travel outward. Insulation comes from the high impedance of myelin sheath which wraps around the axonal membrane and allows each axon in a bundle to conduct signals independently. Although axons typically function independently, they can form connections through specific membrane areas. For example, electrical synapses (also called gap junctions) exist between axonal membranes. These junctions allow a current to flow from highly excited to less excited axons, helping synchronize neuronal populations (Traub et al. 2002; Choi et al. 2021). Although axons are more resistant to fatigue than muscle tissue, they cannot sustain high-frequency firing indefinitely and may experience depolarized axonal block.
In theory, the maximum firing rate (frequency) of an axon is determined by the refractory period of the axonal membrane. This refractory period is typically about 1 ms, allowing axons to conduct action potentials at frequencies up to a kilohertz. Using the HH model introduced in Sect. 1.3, we can simulate continuous action potentials in an axon up to 100 Hz under ideal conditions. However, in real physiological conditions, axons cannot sustain such a high frequency for long periods. They can only maintain such rapid firing briefly. Studies have shown that the axons of brain neurons require about 100 ms for continuous and reliable action potential transmission (Chomiak and Hu 2007; Bucher and Goaillard 2011). This corresponds to a frequency of only 10 Hz.
Advertisement
Note that here the term “frequency” for APs refers to their occurrence rate, similarly to the “frequency” of stimulation pulses. This differs from the frequency components contained in AP and pulse waveforms, as shown by spectrum analysis in Sect. 4.3. For example, a 100 Hz pulse signal means 100 pulses per second. The spectrum of this pulse signal contains a fundamental component at 100 Hz, along with harmonic components at frequencies of 300, 500, 700 Hz and beyond. Following conventions, this book uses Hertz (Hz)—equivalent to s−1—as the unit for pulse frequency. The AP firing frequency of neurons is measured in “Hz” or “count/s”. The electrical stimulation pulses in this book are narrow square waveforms, typically 100 μs in width. Unless specified otherwise, current pulses are used rather than voltage pulses. Their current intensity (amplitude) is measured in milliamperes (mA).
To measure the AP propagations along axons, the most effective way is placing intracellular microelectrodes at multiple sites along an axon to detect membrane potential changes. However, measuring individual axons within the brain is challenging due to their thin, fragile nature. Current technology restricts these measurements to in-vitro experiments. For example, Radivojevic et al. (2017) measured axons of cultured rat cortical neurons using a microelectrode array (MEA) with hundreds of recording contacts to track AP propagation. Their findings showed a decreased reliability in axonal conduction when 100 Hz pulses were applied at the soma for one second. Another study on rat hippocampal slices in-vitro showed that under high potassium conditions or during epileptiform discharges, the axons in Schaffer collaterals failed to conduct APs at frequencies around 50 Hz (Meeks and Mennerick 2004; Meeks et al. 2005). These studies measured extracellular rather than intracellular APs, requiring specific constraints to verify that the measured APs came from the same axon. Therefore, such experiments are usually classified as quasi-single axon studies.
Neurons cultured separately in-vitro or in ex-vivo brain slices often have compromised neural networks that lack the integrity of intact brain networks. This integrity is essential for investigating neuromodulation therapies. Additionally, therapy evaluations require examining population-level neuronal activity, not only single-cell responses. Moreover, the stimulation effects vary across neurons within the stimulation area, as multiple factors influence the distribution of electric field around each neuron: stimulation electrode structure, stimulation model (single-pole, bipolar or multipole), and the distance between neurons and electrodes. As a result, neurons respond differently to stimulation based on their locations. Therefore, studying a single neuron or axon in isolation cannot fully reveal the complex effects of neuromodulation.
In the following chapters, I will present our findings from in-vivo experiments studying the effects of high-frequency stimulation (HFS) pulses on axons in the rat hippocampal region. This brain region has a distinct laminar structure with tightly packed neurons, making it ideal for separately stimulating at axonal fibers and recording at cell bodies (see Chap. 2 for details). Additionally, the hippocampal region is often involved in refractory temporal lobe epilepsy, Alzheimer disease and other disorders. It is also a potential target for DBS (Li and Cook 2018; Laxton et al. 2010). Therefore, studying how hippocampal neurons respond to axonal stimulations can provide insights into the modulation mechanisms of various stimulations and guide the development of new therapeutic approaches.
5.2 Axonal Block Induced by High-Frequency Sustained Stimulation
Sustained extracellular HFS can cause depolarization block at axons, preventing them from generating APs in response to each HFS pulse in in-vitro conditions (Jensen and Durand 2009; Zheng et al. 2011). However, debate has persisted about whether such axonal block can occur in the intact brain during ~ 100 Hz HFS, since axons are known to transmit APs rapidly. Additionally, if HFS-induced axonal block exists, can it simultaneously prevent both AP generation and passage? This section addresses these questions through in-vivo experiments in rat hippocampus, confirming that axonal HFS can produce intermittent block. Additional questions explored include: can axonal HFS affect the somata, and can neurons still fire during HFS-induced axonal block—and if so, how can their firing patterns change?
5.2.1 Antidromic Activation of Pyramidal Neurons by Axonal HFS in the Rat Hippocampal CA1 Region
Figure 5.2A shows the experimental configuration for investigating the effect of axonal HFS on neurons. The antidromic stimulation electrode (ASE) was placed on the alveus—the white matter covering the dorsal surface of the hippocampus. The alveus is composed of the axons of CA1 pyramidal neurons, forming the efferent pathway of hippocampus CA1 region. An array recording electrode (RE) was placed across the CA1 region upstream of the ASE. A single pulse applied to the alveus can produce an APS in the pyramidal cell layer (pcl) as detailed in Chap. 2. Using APS as an index, we can analyze how the alveus axons respond to applied stimuli. The APS can be measured by its amplitude and latency, as shown in the dashed box in Fig. 5.2A. Since the APS results from the antidromic activation of the HFS, the applied HFS is called antidromic HFS (A-HFS).
Fig. 5.2
CA1 neuronal responses to A-HFS trains with different pulse frequencies. A Schematic diagrams show the RE and ASE positions. A typical APS is shown in the dashed box, indicating the definitions of APS amplitude and latency. B Typical pcl recordings during A-HFS trains at 50, 100 and 200 Hz frequencies. The three rows (from top to bottom) show: raw recording (0.3–5000 Hz) of 1-min A-HFS with stimulation artifacts removed, and scatter plots of APS amplitudes and latencies induced by each pulse
As shown in Fig. 5.2B, 1-min A-HFS at three different pulse frequencies initially produced similar APSs. As the A-HFS continued, the APS amplitudes decreased and APS latencies increased, depending on pulse frequencies. Statistical data show that by the end of A-HFS, the APS amplitude dropped to approximately 30–40% (50 Hz), 10–20% (100 Hz), and 5–10% (200 Hz) of its initial value (Feng et al. 2013, 2014). The final APS amplitude was inversely proportional to pulse frequency, with a ratio of 1:2:4 across the three frequencies. During the initial A-HFS period, both the speeds of APS amplitude decrease and latency increase were proportional to pulse frequencies, although the increments in the final APS latencies were similar across all three frequencies (Feng et al. 2014). These results indicated that A-HFS-induced APS changes correlated positively with pulse frequency.
The APS generation involves both the axon and soma (Fig. 5.2A). Changes in either of the structures could result in the APS changes. Our following experiments verified that the A-HFS-induced APS attenuation occurs due to axonal failure rather than somatic failure. The soma can remain its ability to generate APs during sustained A-HFS.
1.
Testing Soma Excitability During A-HFS Using Orthodromic Stimulation
To assess soma excitability during A-HFS, as shown in Fig. 5.3A, orthodromic test stimuli (OTS) were applied every 5 s upstream at the Schaffer collaterals. The first OTS was applied 20 ms after the A-HFS started (Fig. 5.3B), coinciding with the third pulse of A-HFS at 100 Hz. The OPS evoked by this OTS was suppressed due to the refractory period following the large APS evoked by the A-HFS pulse, as OPS had only a slightly longer latency than APS. The fEPSP recorded at the apical dendritic layer (sr) remained similar to baseline recording (Fig. 5.3B, bottom row). As the A-HFS continued, the OTS-evoked OPS grew to baseline level while the APS decreased substantially. The statistical data in Fig. 5.3C showed the changes in APS amplitudes, OPS amplitudes and fEPSP slopes before, during and after A-HFS. After A-HFS ended, the amplitudes of APS evoked by test pulses at 1-min intervals recovered gradually, while the OPS evoked by OTS showed an initial decrease before returning to baseline level. The temporary decrease in OPS suggested a change at the somata, which is further examined in Sect. 5.2.3.
Fig. 5.3
Neuronal responses to orthodromic test stimuli (OTS) during A-HFS. A Schematic diagram showing the positions of the two stimulation electrodes (ASE for A-HFS and OSE for OTS) and the recording array RE. B Typical signals recorded simultaneously at the CA1 soma layer (pcl) and apical dendrite layer (sr) during a 1-min 100 Hz A-HFS. Pulse artifacts were removed. The enlarged insets below show the OTS-evoked OPS and fEPSP waveforms (shaded in green), with red arrows above and green dots below indicating the trimmed artifacts of A-HFS and OTS pulses, respectively. C Statistical data of normalized APS amplitude (upper), OTS-induced OPS amplitude (middle) and fEPSP slope (bottom) pooled from four 100 Hz and five 200 Hz 1-min A-HFS trains with 0.2 Hz OTS (n = 9). Grey shadings with a red bar on top indicate the A-HFS period. Note: To facilitate the comparison between APSs and OPSs during A-HFS, APSs were sampled every 5 s (20 ms before each OTS, with the first APS produced by the first A-HFS pulse). After the A-HFS ended, identical pulses were applied every minute to show APS recovery.
These experiment results show that the CA1 pyramidal neurons can still fire synchronously in respond to low-frequency orthodromic activations while failing to respond to high-frequency pulses at their axons. This indicates that their somata maintain firing ability during sustained A-HFS. The antidromic activation pathway of A-HFS involves only axon and soma, no synaptic transmission (Fig. 5.3A). Therefore, the APS suppression during A-HFS must stem from a failure at the axon rather than the soma. This evidence can strongly support the existence of HFS-induced axonal block.
The APSs shown in Figs. 5.2 and 5.3 were all produced by the stimulation electrode ASE, which delivered A-HFS pulses at a fixed site of the axonal fiber. The APS suppression indicated a decrease in the evoked firing caused by this ASE. This decrease might only result from a trigger failure at the ASE, while the axon could still be able to transmit APs originating from other sources—meaning that the axon may not be truly blocked. To exclude this possibility, we placed a second stimulation electrode at a site sufficiently distant from the A-HFS electrode. This additional electrode delivered a test pulse to verify whether an AP generated away from the A-HFS site could pass through the site. This experiment is described below.
In Fig. 5.4A, two antidromic stimulation electrodes, ASEH and ASET, were placed at the alveus. The ASEH delivered A-HFS, while the ASET delivered testing pulses—ATS (Fig. 5.4B). The APSs evoked by the two electrodes were termed APSH and APST, respectively. Since the ASET was positioned farther from the recording site than the APSH, along a same axon, an AP evoked by the ASET had to cross the affected site of ASEH before reaching the soma at the recording site. The OSE was placed in the Schaffer collaterals. Except for ASET, the positions of the other three electrodes remained the same as in the previous experiments (Fig. 5.3A). The positioning coordinates of the four electrodes, from front to back, were: OSE at 2.2 mm posterior to bregma, 2.0 mm lateral to midline, depth ~ 2.8 mm; RE at 3.5 mm posterior to bregma, 2.7 mm lateral to midline, depth 2.1–2.5 mm; ASEH at 4.8 mm posterior to bregma, 2.7 mm lateral to midline, depth 1.8–2.1 mm; and ASET at 5.8 mm posterior to bregma, 3.0–3.3 mm lateral to midline, depth 2.0–2.2 mm. The distance between ASET and ASEH exceeded 1 mm. Figure 5.4A shows a photograph taken after an experiment, displaying the relative positions of the three stimulation electrodes and one array recording electrode. For details on the imaging method, refer to Sect. 3.2.3.
Fig. 5.4
HFS-induced axonal block preventing activation transmission across the block site. A A photograph showing the electrode positions. B Schematic diagram showing the positions of two stimulation electrodes (ASEH and ASET) on the alveus and the RE in the CA1 region. C Typical recording in the CA1 cell layer (pcl) before, during and after 1-min 100 Hz A-HFS applied by ASEH. The enlarged insets show APSH and APST produced by the pulses applied by ASEH and ASET, denoted by red and green arrows respectively. D and E Changes in the evoked APSH and APST over time after the end of A-HFS, with their baseline waveforms shown at top. The evoked potentials align with the stimulus artifacts. F Comparisons of the amplitude ratios of APSH and APST (P1) at the end of 100 Hz and 200 Hz A-HFS to their baseline values. G Comparisons of the latency increments (ΔL) of APSH and APST (P2) at the end of A-HFS with two different pulse frequencies. The pulse intensity applied by both ASEH and ASET was 0.3 mA, with a pulse width of 100 μs. In the figures F and G, **P < 0.01, paired t-test, n = 12 rats. The signals of APSH and APST are shown in black and green, respectively
During baseline recording before A-HFS (Fig. 5.4C), the APST and the APSH, produced respectively by single pulses from ASET and ASEH, served as baseline controls. The latency (L) of APST was significantly longer than that of APSH, confirming the longer distance between ASET and RE. When a 100 Hz A-HFS was applied by ASEH, each pulse initially induced a large APSH. However, the APSH rapidly decreased to ~ 1/5 of its initial amplitude within seconds and remained stable until the end of 1-min A-HFS.
Immediately after A-HFS ended, the ASET delivered a test pulse 10 ms after the final A-HFS pulse. The evoked APST showed decreased amplitude and altered waveform compared to its baseline control. This APST appeared as a bimodal pattern with troughs P1 and P2 (denoted by two hollow arrows in Fig. 5.4C). While P1 latency was similar to the baseline APST, P2 latency was significantly longer. At 2.5 s post-A-HFS, the bimodal APST became more pronounced. Five seconds after A-HFS, a single pulse from ASEH produced a partially recovered APSH. The amplitude and latency of the APSH then gradually returned to baseline level (Fig. 5.4D). Meanwhile, the APST returned to its baseline unimodal waveform as P2 gradually shifted and merged with P1 (Fig. 5.4E).
The bimodal appearance of APST indicated that the activation area of ASET in the alveus only partially overlapped with that of ASEH. While the ASET-achievable axons within the ASEH activation area were blocked by A-HFS, those activated only by ASET remained unaffected. Therefore, APST appeared bimodally at the end of A-HFS. Its P1 was produced by neurons with axons activated solely by ASET, remaining their original latency unaffected by A-HFS. The P2 was from the axons achievable by both ASET and ASEH, resulting in a prolonged latency caused by A-HFS. Once the A-HFS ended, the prolonged latency and reduced amplitude of P2 gradually recovered, merging with P1, as shown in Fig. 5.4E.
At the end of A-HFS, due to the unaffected part of APST, the ratio of APST amplitude to its baseline amplitude (A1e/A1b) significantly exceeded the value of APSH (Ae/Ab) (Fig. 5.4F). Since ASET was farther from the recording site than ASEH, the latency increment (ΔL) of P2 in APST was also significantly greater than the value in APSH (Fig. 5.4G). Note: The P2 amplitude and latency could not be accurately measured at 10 ms after the end of A-HFS due to the extremely small P2 in the APST waveform. Therefore, the P2 measurements shown in Fig. 5.4G were taken from the APST produced at 2.5 s post-A-HFS (see Fig. 5.4E). Based on the “fast recovery” of APS amplitude described in Sects. 5.2.2 and 5.2.3, this latency measurement can be used to estimate the P2 delay caused by A-HFS.
These results show that the HFS-induced axonal block not only can weaken the activation at the HFS site but also can prevent firing signals from other sites from passing through the block area. Additionally, the increased P2 latency in APST indicates that the axonal block may extend beyond the site directly acted by ASEH. Otherwise, the increment of P2 latency should be similar to that of APSH.
During sustained A-HFS, the increase in APS latency was always accompanied by a decrease in APS amplitude (i.e., fewer firing neurons). This raises a question: does an APS induced by a weak-intensity pulse naturally have a longer latency at baseline? In other words, when fewer neurons are activated, do axons take longer to transmit the activations antidromically to the somata? To determine whether the latency changes were caused by decreased neuron firing or axonal block, we conducted further investigations.
3.
Comparing APS Amplitudes and Latencies between Baseline and A-HFS Periods
Figure 5.5A shows typical scatter plots of the APS amplitude (upper) and latency (lower) produced by each pulse during a 200 Hz A-HFS. Within the first second of A-HFS, the APS amplitude decreased while the APS latency increased rapidly (Fig. 5.5B). The superimposed plot of APS waveforms, extracted at five time points (0, 0.25, 0.5, 0.75 and 1 s of A-HFS), clearly shows these changes (Fig. 5.5C).
Fig. 5.5
Comparisons of APS changes caused by A-HFS and by different intensities in single pulse stimulations. A Typical scatter plots showing APS amplitudes (upper) and latencies (lower) produced by each pulse during 1-min 200 Hz A-HFS with 0.5 mA pulse intensity. B Enlarged plots of the first A-HFS second. C Superimposed APS waveforms extracted at five different time points. D Amplitudes (upper) and latencies (lower) of APSs induced by single pulses with increased intensity in an I-O test. E Superimposed APS waveforms induced by six pulses with intensities ranging from 0.1 to 0.6 mA shown in D. F Comparisons of the amplitude changes (ΔA, upper) and latency changes (ΔL, lower) between APSs induced in I-O tests and in the initial period of A-HFS. The “I-O” group represents the APS changes induced by single-pulse stimulation from 0.1 to 0.6 mA. The three A-HFS groups (50, 100 and 200 Hz) represent the differences between the 1st and 50th APS induced in initial A-HFS periods. *P < 0.02 and **P < 0.001, one-way ANOVA with post hoc Bonferroni tests. “n” represents the rat number.
When single pulse stimulations with varying intensity were applied by the same ASE, the APS amplitude increased with the increase of pulse intensity due to the increased number of firing neurons (Fig. 5.5D, top). This relationship is known as the input–output (I-O) curve (see Sect. 2.3). However, the APS latency showed only slight shortening (Fig. 5.5D, bottom). The superimposed APS waveforms, produced by the pulses of six different intensities (0.1–0.6 mA), showed significant amplitude changes but minimal latency changes (Fig. 5.5E). The mean change in APS amplitudes (denoted as ΔA) in I-O curves was significantly greater than those produced during A-HFS with three different pulse frequencies (Fig. 5.5F, top). Conversely, the mean change in APS latencies (denoted as ΔL) in I-O curves was significantly smaller than those of A-HFS (Fig. 5.5F, bottom).
These results indicate that when produced by a single-pulse at baseline, the firing can take a similar time to antidromically activate the soma regardless of stimulus intensity—whether it is weak to activate only a few axons or strong to activate many. In other words, once an axon is activated, the entire activation process to the soma takes a consistent time, showing similar latency. In contrast, during the initial period of A-HFS, the APS latency increased much more than in I-O testing, even though the APS amplitude during A-HFS showed smaller changes. This substantial latency change suggests that A-HFS applied to the alveus could affect both axons and somata.
5.2.2 Extension of Refractory Period by Axonal HFS
The HFS-induced axonal block is not complete. Each A-HFS pulse can still induce a small APS. This raises a question: are these small APSs produced by a same group of unblocked axons firing persistently following each pulse, or by different groups of axons taking turns to fire? If the latter is true, this means that A-HFS could extend the refractory period of axons. Our following experiment confirms this hypothesis.
1.
Two Distinct Recovery Phases After A-HFS
As shown in Fig. 5.6A, 1-min A-HFS trains with identical pulses but three different frequencies (50, 100 and 200Hz) were applied to the alveus. At the onset of A-HFS, each pulse produced a large APS, which then gradually decreased in size. The APS suppression occurred more rapidly and intensely at higher pulse frequencies (a and b in Fig. 5.6B). During the recovery period after the A-HFS ended, ATSs with identical pulse parameters were applied as single-pulses. Although the APS was suppressed more at the end of 200 Hz A-HFS than at 50 and 100 Hz, it unexpectedly recovered within 3 s to an amplitude comparable to those seen with 50 and 100 Hz stimulations. Furthermore, the partially recovered APSs were all similar at both 3 s and 13 s after A-HFS (c and d in Fig. 5.6B). This similarity contrasted with the frequency-dependent APS amplitudes observed at the end of A-HFS.
Fig. 5.6
Changes in APS amplitudes during and after A-HFS at three different pulse frequencies. A Typical recordings of 1-min A-HFS at 50, 100, and 200 Hz pulse frequencies, with pulse intensity of 0.3 mA. The scatter plots show APS amplitudes induced by each A-HFS pulse. B Enlarged APS signals induced at the onset and end of A-HFS (a and b), as well as APSs induced by single ATS pulses at 3 and 13 s after the end of A-HFS (c and d).
To determine the full recovery course of A-HFS-induced block at different pulse frequencies, we continued to apply ATS pulses for 5 min following the A-HFS (Fig. 5.7A). The ATS pulses were applied at 10-s intervals for the first 2 min (starting 3 s after A-HFS), then at 30-s intervals for the remaining 3 min, with the end of A-HFS served as time 0. The APS recovery occurred in two distinct phases—fast and slow. During the fast phase, the APS amplitude recovered quickly within seconds following A-HFS. In the subsequent slow phase, the complete recovery of APS took several minutes (Fig. 5.7A). We distinguished the two phases based on the lowest point of the small trough in the APS amplitude after 3 s (Fig. 5.7A1). The mean trough time across all 70 trains of 50, 100 and 200 Hz A-HFS was (15.9 ± 5.5)s, with no significant difference among the three frequency groups. The time required for the APS to recover to 90% of baseline amplitude (T90%) was also not significantly different among these groups (Fig. 5.7B). The mean T90% for all A-HFS was 1.20 ± 0.37 min (n = 70).
Fig. 5.7
Two distinct phases in APS recovery after the end of A-HFS. A Typical changes of the APS amplitudes normalized by baseline value during and following 1-min 50, 100 and 200 Hz A-HFS trains. Each data point (except the first one) during the A-HFS period represents mean APS amplitude per second. Each point before and following A-HFS represents an APS evoked by an ATS pulse. B No significant difference in the recovery time T90% for 50, 100 and 200 Hz HFS trains. C The mean recovery of fast phase (ΔA1) significantly increased with frequency. D No significant difference in the recovery of slow phase (ΔA2) among the three frequency groups. E Total APS suppression (i.e., ΔA = ΔA1 + ΔA2) significantly increased with frequency. F Mean APS amplitude at the last 1s of 1-min A-HFS (Aend) significantly decreased with frequency. G Cumulative APS area within the last 1s of A-HFS normalized to the area of the 1st APS of A-HFS. In B–G: **P < 0.001, one-way ANOVA with post hoc Bonferroni test.
The fast-phase recovery of normalized APS amplitude (ΔA1) correlated significantly with A-HFS frequency—higher frequencies resulted in greater ΔA1 (Fig. 5.7C). However, the slow-phase recovery (ΔA2) was independent of A-HFS frequency (Fig. 5.7D). As a result, the total recovery across both phases (ΔA1 + ΔA2)—i.e., the suppression of APS amplitude by A-HFS—correlated significantly with A-HFS frequency (Fig. 5.7E).
The frequency independence of slow-phase recovery indicates a commonality in neuronal responses across different A-HFS frequencies. During the late steady period of A-HFS, while the amplitude of each APS correlated negatively with pulse frequency, the number of produced APSs per second was proportional to it. This suggests that similar amounts of neuronal firing occurred during identical time periods. APS is a waveform formed by integrating the synchronous firing of a neuronal population. Both its amplitude and area can reflect the integrated AP number—and thus the number of firing neurons (Theoret et al. 1984). To compare neuronal firing during the steady period of A-HFS across different frequencies, we calculated the sum of APS areas in the final second of A-HFS. Although the mean APS amplitudes at 100 and 200 Hz were significantly smaller than at 50 Hz (Aend, Fig. 5.7F), these frequencies produced 2 and 4 times the APS counts, respectively. Therefore, the normalized cumulative APS areas showed no significant difference among the three A-HFS frequencies (Fig. 5.7G). This similarity in neuronal firing amounts may explain why the slow-phase recovery was frequency independent.
2.
Fast Recovery Following A-HFS
In contrast to the slow-phase recovery, the fast-phase APS recovery was significantly frequency dependent (Fig. 5.7C). To investigate the fast phase with a higher time resolution, ATS was applied at 20 or 100 ms after A-HFS. To prevent interactions between these close ATS pulses, the 20 and 100 ms tests were conducted in separated A-HFS trials. For 50 Hz A-HFS, the pulse interval (IPI) was 20 ms, allowing the final pulse of A-HFS to serve as a 20 ms ATS. Similarly, the final pulses of 100 and 200 Hz A-HFS served as 10 and 5 ms ATS, respectively (Fig. 5.8A). To improve measurement accuracy, we used the mean amplitude of evoked APS in the final second of A-HFS (Aend) to represent the amplitude of APS evoked by these “ATS"s. We compared the Aend with the other APSs evoked within 3 s post-A-HFS for the three frequencies of 50, 100 and 200 Hz (Fig. 5.8B–D). The APS had already recovered to a level similar to the 3 s level across all three frequencies by 100 ms post-A-HFS. Notably, while the Aend of 200 Hz A-HFS was significantly smaller than those of 50 and 100 Hz A-HFS (Fig. 5.7F), at 20 ms post-A-HFS, the APS jumped to a level exceeding both 50 and 100 Hz A-HFS (Fig. 5.8E).
Fig. 5.8
Fast-phase of APS recovery following A-HFS at three different frequencies. A Schematic diagram showing the timing of ATS pulses at 20ms, 100 ms and 3 s following A-HFS. The red arrows and black vertical lines represent stimulus pulses. B–D Comparisons of the mean APS amplitudes during the final second of A-HFS (Aend) with APS amplitudes at 20ms, 100ms and 3 s following A-HFS at 50, 100 and 200 Hz, respectively. All APS amplitudes were normalized to baseline values. E Summary data from plots (B–D). Error bars are omitted for clarity since they are shown in plots (B–D). *F = 5.27, P = 0.017, 50 versus 200 Hz, one-way ANOVA with post hoc Bonferroni tests.
The results show that a substantial portion of APS suppression produced by higher frequencies (100 and 200 Hz) can recover even within 20 ms. This rapid recovery suggests that APS suppression could be related to changes in the refractory period. The A-HFS might extend the refractory period beyond the inter-pulse-interval (IPI) of 200 and 100 Hz A-HFS pulses (5 or 10 ms), causing the axons to fire action potentials intermittently in response to only a portion of A-HFS pulses. To verify this hypothesis, we examined changes in axonal refractory period during A-HFS.
3.
Extension of Refractory Period
The refractory period (RP) is a phase when a neuron cannot generate a second action potential right after firing. Changes in RP can be measured using the evoked APSs from paired-pulses. To evaluate axonal RP during A-HFS, we inserted a prolonged IPI (a gap) to produce a sufficiently large APS based on the fast recovery described above.
As shown in Fig. 5.9A, at the onset of a 200 Hz A-HFS with an IPI of 5 ms, each pulse induced a large APS, indicating an RP of less than 5 ms. We denoted the amplitudes of the initial three APSs as A1, A2 and A3, respectively. During the 1-min A-HFS, prolonged IPIs of 100 ms were inserted every 2 s to create 29 gaps. The amplitude of APS immediately following the gap (denoted as A1 and indicated by a red dot in Fig. 5.9A) was significantly greater than the amplitude of the preceding APS that followed normal 5 ms IPI (denoted as A0) (Fig. 5.9B). However, the second APS (A2) after A1 was significantly smaller than A1 and nearly disappeared in the late A-HFS period. Statistical data (Fig. 5.9C) showed that during the initial period of A-HFS, the amplitude ratio of the first paired-APS (i.e., A2/A1 for the 1th and 2nd APS of A-HFS) was 90 ± 11% (n = 7), indicating thatA2 was only slightly suppressed. However, during the final 30 s A-HFS period, the A2/A1 fell below 7.0%, indicating that A2, induced after a 5 ms delay, was mostly suppressed by the gradually extended RP of A1. For the third APS (A3) after A1, generated ~ 10 ms after A1, the A3/A1 was around 60% in the final 30 s of A-HFS. Disregarding the ignorable A2, this indicates that the axonal RP had exceeded 10 ms.
Fig. 5.9
Assessing the changes in neuronal refractory period (RP) during 200 Hz A-HFS by using the amplitude ratios of paired APS. ATop: typical neuronal responses to 1-min 200Hz A-HFS with prolonged 100 ms IPI gaps inserted every 2s. Red dots below the signal denote the evoked APSs following the gaps. Middle: Enlarged plots showing the evoked APSs at A-HFS onset and around three of the total 29 gaps. The APSs (A1) following the gaps are shown in red. Bottom: scatter plot of APS amplitudes evoked by each A-HFS pulse, with A1 data points highlighted in red. B The amplitudes of APSs before (A0) and after (A1) the gaps, normalized to the amplitude of the very first APS at A-HFS onset. C Ratios of A2/A1 and A3/A1 during A-HFS. The definitions of A0 toA3 are illustrated in (A).
To further confirm the RP extension, we shortened the gaps from 100 to 20ms and inserted them in both 100 and 200 Hz A-HFS. For 100 Hz A-HFS, statistical data showed that the mean normalized amplitudes of A1 remained above 50% (n = 8, Fig. 5.10A), indicating sufficient APS recovery following the 20 ms gaps. After the initial 4 s A-HFS, the A2 was suppressed significantly with A2/A1 < 10% (n = 8, Fig. 5.10B), indicating that the RP of A1 exceeded 10 ms. For 200 Hz A-HFS with 20 ms gaps, the data showed that A2/A1 < 6.0% andA3/A1 < 20% for most A-HFS period with sufficiently large A1 > 60% (n = 6, Fig. 5.10C, D), also indicating an extended RP beyond 10 ms.
Fig. 5.10
Assessment of RP changes during 100 and 200 Hz A-HFS with 20 ms gaps. A Normalized amplitudes of the preceding APS (A0) and subsequent APS (A1) relative to the 20 ms gaps during 100 Hz A-HFS. B Ratios of A2/A1 during 100Hz A-HFS. C Same measurements as in (A), but for 200 Hz A-HFS. D Ratios of A2/A1 and A3/A1 during 200 Hz A-HFS. See Fig. 5.9A for the definitions of A0 to A3.
The mechanism of RP extension produced by sustained A-HFS is not yet fully understood. Based on a mechanism of potassium ions (K+) accumulation in peri-axonal spaces during sustained axonal membrane activation (Bellinger et al. 2008), we simulated the reactions of axonal membrane by computational modeling (Guo et al. 2018; Zheng et al. 2020). The results showed that when A-HFS pulses continuously depolarize the axonal membrane, the K+ channels open and release K+ into the narrow peri-axonal spaces, increasing the extracellular K+ concentration ([K+]o). Since K+ clearance cannot keep pace with its accumulation, [K+]o remains high within IPIs. This elevated [K+]o keeps the axonal membrane in a sustained depolarization state, leading to partial inactivation of membrane sodium (Na+) channels (refer to Sect. 1.3). The inactivated Na+ channels need multiple IPIs to recover sufficiently for the axon to fire again, resulting in an intermittent axonal block.
Our simulation results showed that small APSs induced during the steady period of 100 and 200 Hz A-HFS (with IPI of 10 and 5 ms) are generated by different neurons firing in turns, rather than the same neuronal group firing following each pulse. When A-HFS is applied to axons, it causes them to enter an intermittent block state where they can fire APs in response to only some pulses, not every pulse. Nevertheless, this intermittent firing is not necessarily evenly distributed over time. For details on the firing patterns of individual pyramidal neurons during A-HFS, please refer to Sect. 5.2.5. In addition, the above experimental results cannot rule out the possibility that sustained A-HFS can cause some neurons to become “silent” and cease firing completely. Each neuron's response to stimulation depends on its distance from the stimulation electrode. While some neurons fire intermittently due to extended refractory periods, others may become completely unresponsive, remaining silent until they are able to recover after A-HFS ends. Furthermore, A-HFS with gaps represents one of the stimulation patterns with varying frequencies (i.e., varying IPIs), which are detailed in Chap. 6.
The RP extension can explain the fast-phase recovery following A-HFS but not the slow-phase recovery that lasts several minutes, suggesting that other mechanisms could be involved in APS suppression during A-HFS. Although A-HFS was directly applied to the alveus (the axons of CA1 pyramidal neurons), the sustained stimulation could affect the somata. To investigate this issue, we used current source density analysis to examine changes in AP generation and conduction near the somata during antidromic propagation of A-HFS excitation.
5.2.3 Effect of Axonal A-HFS on the Somata of Pyramidal Neurons
To reveal possible somatic changes produced by sustained axonal A-HFS, we compared soma reactions between the initial and late A-HFS periods. As shown in Fig. 5.11A, to obtain large APSs during late A-HFS comparable to the initial period, we deleted one pulse every 20 pulses in the final 20 s of 1-min 100Hz A-HFS. Each pulse deletion created a 20 ms gap. Due to fast recovery, the APS (A1) induced by the first post-gap pulse was more than twice the size of the pre-gap APS (A0) (Figs. 5.10A and 5.11A). During the initial 2 s period of A-HFS, as the APS decreased gradually, we picked an APS with an amplitude equivalent to A1, which we termed “A1_equivalence”. The lower left corner of Fig. 5.11A shows superimposed waveforms of five APSs to illustrate their differences: the APS induced by the first A-HFS pulse, the A1_equivalence, the A0, the A1, and the recovered APS at 2 min after A-HFS ended. Statistical data in Fig. 5.11B show significant differences in the amplitudes and latencies of these APSs. While the amplitudes ofA1 and A1_equivalence were similar (indicating a similar number of firing neurons), the meanA1 latency was significantly longer than that of A1_equivalence. However, no significant difference existed between the latencies of A1 and A0. We hypothesized that the extended A1 latency might be attributed to a change in excitation conduction speed around the soma. Therefore, we utilized current source density (CSD, see Sect. 4.4) to measure the conduction speed of the APS current sink across the CA1 cell body layer (pcl).
Fig. 5.11
Changes in APS current sink conduction during different periods of A-HFS. A Typical recording at the pcl during 1-min 100 Hz A-HFS, with 20 ms gaps inserted every 0.2 s in the final 20 s of A-HFS. The waveforms of the 1st APS, A1_equivalence, A0 (pre-gap) and A1 (post-gap) during A-HFS, as well as the APS evoked 2 min after the end of A-HFS, are shown below. A superimposed plot of these APS waveforms is shown in the lower left corner. B Comparisons of the amplitudes (B1) and the latencies (B2) among the four APSs. C Examples of 9-channel evoked potentials across the pcl at the four periods. D CSD curves derived from the evoked potentials in (C), with grey shadings indicating CSD current sinks and arrows showing their travel directions. E Comparisons of the travel speeds of CSD sinks among the four APSs. In (B) and (E), n.s. not significant; *P < 0.05, **P < 0.01, one-way ANOVA with post-hoc Bonferroni tests; n = 10 rats.
Figure 5.11C shows evoked potentials from nine recording contacts on a 16-channel electrode array. These contacts were vertically arranged across a 400 μm span containing the pcl. This figure shows four episodes of 9-channel evoked potentials during different periods of A-HFS: the 1st APS, A1_equivalence, A0 (pre-gap) and A1 (post-gap). Figure 5.11D shows the CSD curves calculated from the potential signals, with grey shadings indicating CSD current sinks. The sink travel trajectory—which can reflect action potential conduction—moved consistently across all four periods, starting above the pcl before moving sequentially downward to the soma and apical dendrites. This pattern aligns with previous findings (Kloosterman et al. 2001). The travel speed of A1 sink was significantly slower than that of A1_equivalence but similar to A0 (Fig. 5.11E). This indicates that sustained A-HFS affected the somatic region by decreasing the excitation conduction speed. Furthermore, the recovery of conduction speed was slower than the restoration of activated neuron numbers.
A-HFS-induced decrease in the conduction speed of antidromic activations across somata indicated certain change. This change around somata could also affect the conduction of orthodromic activations from presynaptic neurons. To test this hypothesis, we applied orthodromic test stimulations (OTS) every 10s to the Schaffer collaterals as shown in Fig. 5.12A (refer to Fig. 5.3A for electrode positions). During A-HFS, while the amplitude of OPS induced by each OTS pulse remained its baseline level (Fig. 5.12B1), the OPS latency increased significantly (Fig. 5.12B2) and returned to baseline level about 2 min after A-HFS ended.
Fig. 5.12
Changes in OPS current sink conduction during different periods of A-HFS. A Typical recording at the pcl during 1-min 100 Hz A-HFS with OTS pulses applied every 10 s. The enlarged insets below show the OPSs evoked by OTS pulses respectively at baseline before A-HFS, at 10 s and 40 s of A-HFS and at 2 min following A-HFS. The OPS waveforms from these four periods are overlapped in the lower right corner. B Changes in OPS amplitudes (B1) and latencies (B2) before, during and following A-HFS. Pink shadings denote the A-HFS period. C Examples of 9-channel OTS-evoked potentials across the pcl at the four periods. D Corresponding CSD curves derived from the evoked potentials in (C), with grey shadings indicating current sinks and arrows showing their travel directions. E Changes in the travel speeds of CSD sinks before, during and following A-HFS. In (B) and (E): *P < 0.05, **P < 0.01, repeated-measure ANOVA with post-hoc Bonferroni tests, n = 10.
Based on the evoked OPSs recorded by the nine channels across the pcl (Fig. 5.12C), we calculated their CSD curves (Fig. 5.12D). The CSD current sinks of the OPSs originated from the apical dendrites and travelled sequentially toward the soma and then basal dendrites. During A-HFS, the travel speed of these current sinks decreased gradually and significantly, but returned to baseline level within 2 min after the A-HFS ended (Fig. 5.12E). While axonal A-HFS did not significantly affect the somatic ability to respond to OTS pulses, it did significantly slow the conduction of orthodromic activation across the somata.
The decrease of activation conduction speed suggests reduced excitability of the pyramidal neurons. However, during A-HFS, the amplitude of OTS-evoked OPS did not decrease but even slightly increased (see Figs. 5.3C and 5.12B). This could be attributed to the continuous replenishment of antidromic activations from A-HFS, which compensated for the reduced excitability of pyramidal neurons themselves. Supporting this hypothesis, we observed a temporary decrease in OPS amplitude immediately after A-HFS ended, as shown in Fig. 5.3C. To further verify this hypothesis, we investigated the firing changes of individual neurons before and after A-HFS, as shown below.
5.2.4 Effect of Axonal HFS on Neuronal Excitability
Figure 5.13A shows the firing of individual neurons following A-HFS. The wideband signal from the soma layer of CA1 region (pcl) was filtered to produce MUA signal above 500 Hz. The MUA contained clear unit spikes from pyramidal neurons (Pyr) and interneurons (IN) before and after A-HFS. Although the stimulus artifacts of A-HFS were removed before filtering (detailed in 4.5.2), the evoked APSs during A-HFS had much larger amplitudes than unit spikes, making the spikes indistinguishable during A-HFS period. Though pyramidal neurons are more numerous in the pcl than interneurons, interneurons fire at a much higher rate. Therefore, when interneuron firing is captured in the MUA signal, their spikes typically outnumber those of pyramidal neurons, as shown in Fig. 5.13A.
Fig. 5.13
Temporary silence of neuronal firing immediately following A-HFS. ATop: typical recording in the pcl of CA1 region before, during, and after 1-min 100 Hz A-HFS. Middle: Enlarged signals showing the initial large APSs and the final small APSs evoked by A-HFS pulses. Red arrows with dashed lines denote removed pulse artifacts. Bottom: MUA signal produced by filtering the artifact-free wide-band signal. B and C Mean normalized firing rates of pyramidal cells (Pyr) and interneurons (IN) before and after A-HFS (at 5 s time bins). Typical spike waveforms and firing patterns for both neuron types are illustrated above the plots. Red dashed horizontal lines denote the mean baseline firing rates. Due to variations in silent period durations across individual neurons (some shorter than 5 s), the plots with 5 s time bins cannot capture clear silent period.
The signals in Fig. 5.13A revealed a notable phenomenon. Before A-HFS, spontaneous spike firing occurred in the baseline recording. During A-HFS, each pulse triggered an APS, indicating that neurons maintained continuous firing throughout the stimulation, although the APS amplitude decreased significantly in the late A-HFS period. However, immediately after A-HFS ended, a silent period without spikes appeared in the MUA signal, lasting several seconds before the firing gradually returned to baseline level (Fig. 5.13A, bottom).
To investigate the firing of individual neurons, four-channel MUA signals recorded around the pcl were used for spike detection and sorting (refer to Sect. 4.1 for detailed method). Single unit spikes (SUA) of pyramidal neurons and interneurons were identified based on their firing patterns and spike waveforms (Barthó et al. 2004). As shown in Fig. 5.13B, C, pyramidal neurons usually fire in bursts with wider rising-phase in their spike waveforms, while interneurons fire more regularly with narrower rising-phase waveforms. We analyzed unit spikes from 104 pyramidal neurons and 66 interneurons in 40 rat experiments with 1-min 100 Hz A-HFS. During the 1-min baseline recordings before A-HFS, the pyramidal neurons showed a mean firing rate of 4.8 ± 4.6 spikes/s, while interneurons fired at 10.3 ± 9.4 spikes/s. Based on the average baseline firing rate of each neuron, we created histogram plots of normalized firing rates for both neuron types using 5 s bins (Fig. 5.13B, C). After A-HFS, pyramidal neurons showed significantly longer silent and recovery periods than interneurons (P < 0.05 for silent period, 21.9 ± 22.9 s vs. 11.2 ± 8.9 s; P < 0.01 for recovery period, 2.9 ± 1.5 min vs. 0.76 ± 0.60 min, unpaired t-test). Note that these mean durations were the statistical data from individual neurons, differing from the aggregated histogram data shown in Fig. 5.13B, C.
These results showed firing suppression in both pyramidal neurons and interneurons in the post-stimulation period, including a complete silent phase. It can explain the decreased amplitude of OTS-induced OPS following A-HFS, as shown in Fig. 5.3C. The temporary pause and reduction in interneuron firing indicated weakened inhibitory synaptic effects on pyramidal neurons (see Sect. 2.3 for details of local inhibitory circuits in the CA1 region). This should have increased, rather than decreased, the excitability of pyramidal neurons. Therefore, the post-stimulation firing suppression likely stemmed from decreased excitability of pyramidal neuron somata, caused by sustained A-HFS acting on their axons. However, this decreased excitability was not observed during A-HFS (Fig. 5.3C middle line and Fig. 5.12B1), possibly due to an offsetting effect created by the continuous A-HFS activation travelling to the somata. Once this activation ceased, decreased somatic excitability emerged immediately. In fact, during A-HFS, while OPS amplitude remained largely unchanged, OPS latency increased significantly (Fig. 5.12B2) and the travel speed of activations around the soma decreased (Fig. 5.12E). These results indicate decreased neuronal excitability. Moreover, the occurrence and recovery of this somatic change was temporally consistent with the slow phase of APS recovery after the end of A-HFS (Xu et al. 2023, and Fig. 5.7A), indicating that somatic changes may constitute another mechanism for APS decrease during A-HFS. Therefore, the effect of sustained A-HFS on axons can extend beyond the stimulation site to other neuronal structures, like somata.
5.2.5 Clustered Firing in Pyramidal Neurons During A-HFS with Fixed-IPI
Pyramidal neurons in the hippocampal CA1 region typically generate action potentials in a burst pattern (Fig. 5.13B). The above studies have shown that during sustained A-HFS at the axons of pyramidal neurons, each pulse can induce a small APS at the pcl. Since an APS represents integration of action potentials from a neuronal population, a regular APS sequence does not necessarily mean uniform firing of individual neurons. We next investigated the firing pattern of individual pyramidal neurons during A-HFS, to explore whether their firing maintained a non-uniform, burst-like characteristic.
Obtaining spikes from individual pyramidal neurons during their axons directly under A-HFS is challenging in intact brain, as these spikes become integrated into APS waveforms. In our in-vivo rat experiments, we typically used a bipolar stimulation electrode (CBCSG75, as detailed in Sect. 3.4.2) to deliver biphasic pulses with 100 μs width per phase and approximate 0.3 mA intensity. In normal baseline conditions, a single such pulse induced a large APS with an amplitude usually reaching ~ 10 mV. Although the APSs decreased to 1–2 mV during sustained 100 Hz A-HFS, they remained significantly larger than typical unit spikes from individual neurons, making it difficult to distinguish the spikes contained in APS waveforms. Therefore, to detect pyramidal neuron spikes during A-HFS, the APS must be minimized. While reducing pulse intensity can decrease the APS by activating fewer neurons, this approach creates a new challenge: a weaker stimulation can only produce a smaller activation scope, making it harder to reach the stimulated neurons for recording. In addition, to ensure reliable spike identification, we need to record spikes with sufficiently large amplitudes. This requires positioning the recording contact very close to a stimulated neuron's soma, demanding precise placements of both recording and stimulation electrodes.
To meet the requirement, we initially used single test pulses with a common intensity of 0.3 mA to position the stimulation electrode and the recording array to produce a satisfactory APS. We then gradually lowered the pulse intensity to around 0.05 mA and carefully adjusted the positions of both electrodes to obtain large spikes from stimulated pyramidal neurons during A-HFS. Notably, recording of large spikes in baseline spontaneous firing does not necessarily mean a successful experiment, because the neurons producing these spikes could lie outside the activation scope of the low-intensity stimulation and thus remained unmodulated by A-HFS. Such neurons would continue fire spontaneously during A-HFS. Additionally, under normal physiological conditions, not every pyramidal neuron is in a firing state—many remain silent (Andersen et al. 2007). These silent neurons could be activated and fire during A-HFS if they fall within the activation scope of A-HFS. Therefore, some firing can be unmodulated by A-HFS, while some can emerge only during A-HFS but absence in baseline recording.
Through our trials, we obtained the required spike data from individual pyramidal neurons modulated by low-intensity A-HFS. As shown in Fig. 5.14A, the spontaneous firing of pyramidal neurons in baseline recording appeared as bursts, which resulted in a clear peak at 3–8 ms in the inter-spike-intervals (ISI) histogram (Fig. 5.14B). We calculated the firing rates per second of single pyramidal neurons during 1-min baseline recordings. Figure 5.14C shows an example. The non-uniform burst firing caused substantial changes in these firing rates, characterized by a large coefficient of variation (CV). The CV—the ratio of the standard deviation to the mean—can evaluate the uniformity of neuronal firing (Di Miceli et al. 2020). A larger CV indicates more uneven firing. During 1-min baseline recordings in 18 rat experiments, we collected spontaneous firing from 27 pyramidal neurons, with a mean firing rate of 3.1 ± 1.7 spikes/s and the mean CV of the per-second firing rates of 1.2 ± 0.5.
Fig. 5.14
Unit spikes of pyramidal neurons during baseline recording and during low-intensity A-HFS. A A typical spontaneous burst of a pyramidal neuron extracted from a baseline MUA signal with a bandwidth of 0.5–5 kHz. B Typical ISI histogram of baseline firing from a pyramidal neuron. C Scatter plot showing firing rates per second from a pyramidal neuron during 1-min baseline recording. D Recording example of pyramidal neuron firing during 0.04 mA low-intensity A-HFS at 100 Hz for 2 min. The first row shows the wideband raw recording at 0.3–5000 Hz. The second row shows the corresponding MUA signal (0.5–5 kHz) after high-pass filtering. The third and fourth rows show enlarged episodes of the initial and steady A-HFS periods. The red arrows with dashed lines indicate removed stimulus artifacts. The red dots indicate unit spikes. E Scatter plot of the spike latencies during the A-HFS shown in (D)
Figure 5.14D shows the spikes of a pyramidal neuron with an amplitude of approximately 0.7 mV recorded during a 2-min train of low-intensity 100 Hz A-HFS at 0.04 mA. In the initial period of A-HFS, each evoked spike overlapped with an APS. Due to its large amplitude, these spikes were visible at the tip of APSs in the wideband raw recording signal. As the A-HFS continued and the APSs decreased, the spikes became clearer. The spikes were more distinguishable in the MUA signal after high-pass filtering (denoted by red dots in the enlarged insets in Fig. 5.14D), which were strongly phase-locked with the A-HFS pulses. The spike latencies gradually increased from ~ 2 ms at the A-HFS onset to around 3 ms in a dozen seconds, and then remained stable during the late period of A-HFS (Fig. 5.14E). This pattern matched the change in APS latencies as shown in Figs. 5.2B and 5.5A, which verified that the spikes of this pyramidal neuron were produced by direct A-HFS activation on its axon.
During 28 trains of low-intensity 100 Hz A-HFS from 18 rats, we recorded 42 stimulated pyramidal neurons with spike amplitudes exceeding 300 μV to ensure accurate spike detection with minimal APS interference. We identified spikes of CA1 pyramidal neurons based on their rising phase width of > 0.7 ms (Barthó et al. 2004) and their phase-locked short latencies to the A-HFS pulses as shown in Fig. 5.14E. During A-HFS, the mean spike rate increased significantly to 15.5 ± 9.4 spikes/s (n = 42) compared to the baseline value of 3.1 ± 1.7 spikes/s (n = 27, t-test, P < 0.01), indicating enhanced neuronal firing by the stimulation. Meanwhile, the mean CV of the per-second firing rates during 2-min A-HFS decreased to 0.6 ± 0.3 (n = 42) compared to the baseline value of 1.2 ± 0.5 (n = 27, t-test, P < 0.01), indicating that the firing became uniform. Although the firing rate increased significantly during A-HFS, it was still well below the 100 Hz pulse frequency, meaning that the neurons did not fire following each pulse. Moreover, the pyramidal neurons exhibited two distinct firing patterns: non-clustered (regular) firing in 12 neurons (29%) and clustered (non-uniform) firing in the remaining 30 neurons (71%).
An example of non-clustered firing is shown in Fig. 5.15. During the 2-min A-HFS, although the spike ISIs varied as the pyramidal neuron firing at different IPIs (Fig. 5.15A), the ISI histogram was smooth (Fig. 5.15B), and the per-second firing rates were stable with a CV of only 0.6 (Fig. 5.15C). Figure 5.16 shows an example of clustered firing. During the 2-min A-HFS, the pyramidal neuron fired in alternating periods of intensive spikes and silence (Fig. 5.15A). Initially, its firing rate was higher, with clusters appearing as spikes following several successive pulses. Later, the spikes in clusters no longer followed each pulse successively, instead following every two, three, or more pulses. Two distinct peaks appeared in the ISI histogram of the clustered firing (Fig. 5.15B): one at 10–50 ms (corresponding to intra-cluster ISI) and the other at 100–200 ms (corresponding to inter-cluster ISI). The per-second firing rates changed smoothly, starting higher in the initial A-HFS period before gradually decreasing to a stable level, with a CV of only 0.7 (Fig. 5.15C). The firing rates for both non-clustered and clustered patterns followed a similar trend to the changes of APS amplitudes during A-HFS at a commonly used intensity (e.g., 0.3 mA in Fig. 5.2B). Furthermore, the ISI histogram was modulated by the10ms pulse interval and its integer multiples, indicating that the firing of pyramidal neurons was directly controlled by the stimulation pulses.
Fig. 5.15
Example of non-clustered firing of a CA1 pyramidal neuron produced by A-HFS. A MUA signal containing the spikes from a pyramidal neuron which was modulated by A-HFS pulses. The enlarged views at two time scales show the details of initial and steady A-HFS periods. Red dots below the signals denote the spikes, with ISI values shown in blue. B ISI histogram of the non-clustered spike sequence shown in (A). C Scatter plot of the per-second firing rates with a CV of 0.6. From Yuan (2023)
Example of clustered firing of a CA1 pyramidal neuron produced by A-HFS. A MUA signal containing the spikes from a pyramidal neuron which was modulated by A-HFS pulses. B ISI histogram of the clustered spike sequence from (A), showing two distinct peaks. C Scatter plot of the per-second firing rates with a CV of 0.7. From Yuan (2023)
We measured the spike rates from individual pyramidal neurons during low-intensity A-HFS at frequencies of 50, 100, 133 and 200 Hz, respectively. Their mean spike rates showed no significant differences across these pulse frequencies. Additionally, each neuron kept its firing pattern (clustered or non-clustered) during A-HFS trains, regardless of the pulse frequency (Yuan 2023).
It is an interesting finding that A-HFS with uniform pulses can produce non-uniform clustered firing in individual pyramidal neurons, distinct from the regular APS of neuronal populations. Although this clustered firing somewhat resembled burst firing, the two patterns differed in two points. First, the amplitudes of intra-cluster spikes remained stable—the mean amplitudes of first and last spikes in clusters were 374 ± 178 μV and 336 ± 175 μV respectively (n = 30 pyramidal neurons). In contrast, the spike amplitudes in spontaneous bursts decreased substantially. In baseline recordings from 18 rat experiments, bursts of 3–5 spikes showed mean amplitudes of the first and last spikes of 512 ± 293 μV and 342 ± 168 μV respectively (n = 27 pyramidal neurons). Second, the intra-cluster ISIs were much longer than the intra-burst ISIs (Figs. 5.14A, B and 5.16A, B). These distinctions indicate different underlying mechanisms for the two firing patterns.
Intrinsic bursts of CA1 pyramidal neurons are generated in the soma through local regenerative self-excitation, triggered by a large after-depolarization (ADP). This ADP occurs through persistent Na+ flows at or near the soma (Azouz et al. 1996; Yue et al. 2005). Consecutive spikes within a burst progressively decrease in amplitude due to increased inactivation of fast Na+ channels (Jensen et al. 1996). Unlike these intrinsic bursts, each spike in A-HFS-induced clusters was triggered by stimulation pulse with a phase-locked latency, resulting in longer ISIs than those in bursts. The increased spike rate also indicated the excitatory modulation of A-HFS on the neuronal firing.
The results of our computational modeling revealed the following possible mechanism underlying the clustered firing (Yuan et al. 2025). During the initial period of A-HFS, intensive firing produces an intermittent axonal block caused by increased [K+]o in the peri-axonal space and increased Na+ channel inactivation (Guo et al. 2018; Zheng et al. 2020; Brazhe et al. 2011). This block decreases the amplitudes of evoked action potentials (AP) at the initial and adjacent axon nodes around stimulation site. The weakened AP eventually loses its ability to propagate and fails to reach the soma, resulting in pause of soma firing. This pause allows the initial AP to gradually recover and grow strong enough to trigger soma AP again. Subsequently, a cluster of APs continues due to the feedback effects of after-hyperpolarization (AHP) from the AIS and soma until the initial AP is weakened again. Additionally, our simulation showed that neurons with axons closer to the stimulation electrode tend to generate non-uniform clustered firing, while those farther away tend to generate regular firing. Nevertheless, further investigations are needed to verify the mechanism and uncover additional ones.
The absence of burst firing in the pyramidal neurons during A-HFS may result from inhibitory effects of local feedback circuits. When a pulse of A-HFS antidromically triggers somata to fire, the firing can immediately activate the interneurons in feedback inhibitory circuits to suppress subsequent burst firing (Knowles and Schwartzkroin 1981; Davies et al. 1990; Kepecs and Fishell 2014). Additionally, antidromic activation travelling along axons during A-HFS can activate interneurons through axonal branches thereby inhibiting the somata of pyramidal neurons (Ye et al. 2022; Dudok et al. 2021). As shown in Chap. 2, the experiments of single-pulse and paired-pulse stimulations have revealed that the behavior of CA1 pyramidal neurons was modulated by local inhibitory circuits, resulting in single spike in the evoked APS. When GABA antagonist (e.g., PTX) blocked the inhibitory synapses, the evoked APS appeared as burst-like firing with multiple peaks (Fig. 2.9). Therefore, the activations of interneurons in inhibitory circuits could have prevented burst firing during A-HFS.
The non-uniform clustered firing induced by a uniform incentive in the rat brain in-vivo provides new evidence for nonlinear dynamics of neuronal firing in intact brains. The alternating periods of clustered firing and silence resemble bistable behavior—a typical property of nonlinear bifurcation dynamics observed previously in neural computational models with high [K+]o (Wu and Shuai 2012; Contreras et al. 2021; Hahn and Durand 2001). However, this behavior has limited support from real neurons in the brain. Additionally, the non-uniform firing originated in directly stimulated neurons can propagate to post-synaptic neurons through neuronal networks, resulting in widespread effects, as shown in our another report (Wang et al. 2024). The following sections will describe how high-frequency axonal stimulations affect post-synaptic neurons downstream of the stimulation site.
5.3 Effect of High-Frequency Axonal Stimulation on Post-synaptic Neurons
The neuronal responses to the axonal A-HFS described above were produced by antidromic propagation of axonal activation to the neuronal somata, which involving only the axon and soma of same neurons, without involving synaptic transmissions. However, the activations initiating in the middle site of axons can travel bilaterally—both antidromically toward the soma and orthodromically toward axonal terminals. The orthodromic propagation can activate downstream post-synaptic neurons through synaptic transmissions. This orthodromic pathway is more complex than the antidromic one, involving axonal conduction, synaptic transmission, integration of post-synaptic potentials, and activation of the post-synaptic neuronal soma finally. To investigate the responses of post-synaptic neurons to orthodromic high-frequency stimulation (O-HFS), we still took the hippocampal CA1 neurons as the recording target. That is, remain the same recording site as in A-HFS, but move the stimulation site to the Schaffer collaterals—the afferent axon fiber of the CA1 region thereby positioning the stimulation site upstream of the recording site. Although the axons of the Schaffer collaterals somewhat differ from those of the alveus (Andersen et al. 2007), the HFS effects on the axons remained similar. The following O-HFS experiments further demonstrate that HFS can generate axonal block.
5.3.1 Responses of Downstream Post-Synaptic Neurons to Sustained O-HFS Trains at Different Frequencies
As shown in Fig. 5.17, when 1-min O-HFS trains at 50, 100, or 200 Hz were applied to the Schaffer collaterals, the initial responses of CA1 neurons were similar, regardless of the pulse frequencies. The first O-HFS pulses induced similar orthodrminc population spikes (OPS), denoted as 1st OPS, with similar latencies. Then, in the subsequent tens of milliseconds, OPSs disappeared due to the effect of local inhibitory circuits (see Sect. 2.3 for details) before reappearing temporarily. With the lower frequency of 50 Hz, the OPSs continued for tens of seconds, sometimes persisting until the end of the 1-min O-HFS (Fig. 5.17A). The OPSs in the late O-HFS period resembled the epileptiform activity induced by epileptic agents (refer to Chap. 8), showing no stable latencies with the O-HFS pulses. At 100 and 200 Hz frequencies, no obvious OPS appeared in the late O-HFS periods (Fig. 5.17B, C). With similar pulse intensity (around 0.3 mA) and 1st OPS amplitude (around 8 mV), the mean duration with OPSs during 50 Hz O-HFS was 40.7 ± 20.8 s (n = 8), significantly longer than those during 100 Hz O-HFS of 9.3 ± 12.8 s (n = 8) and 200 Hz O-HFS of 0.7 ± 1.1 s (n = 9) (Feng et al. 2013). Although O-HFS at 100 and 200 Hz delivered 2 and 4 times the stimulation pulses of that at 50 Hz, they generated much less the synchronous firing—OPSs. This suggests that higher pulse frequencies can weaken synchronous responses in downstream neurons.
Fig. 5.17
Neuronal responses recorded in the pcl of rat hippocampal CA1 region during O-HFS at three different pulse frequencies. A–C Examples of 1-min O-HFS trains at 50 (A), 100 (B) and 200 Hz (C), with enlarged insets showing evoked OPSs
Similar to the A-HFS periods described in the previous sections, the axonal block caused by O-HFS at the lower frequency of 50 Hz may be weaker. This can lead to stronger activations of Schaffer collateral axons, causing increased activations of downstream neurons and resulting in a longer period of synchronous firing with OPSs. In contrast, O-HFS at higher frequencies (100 and 200 Hz) may produce deeper axonal blocks, resulting in weaker activation of downstream neurons that cannot sustain the synchronous firing.
Like the A-HFS experiment shown in Fig. 5.3, we conducted an O-HFS experiment to investigate whether post-synaptic pyramidal neurons retained their ability to fire action potentials during the late period of O-HFS when no obvious OPS occurred. In this experiment (Fig. 5.18A), while 1-min O-HFS was applied to the Schaffer collaterals, an antidromic test stimulus (ATS) was applied to the alveus every 5s to antidromically activate CA1 pyramidal neurons at the recording site. We hypothesized that if the failure of O-HFS pulses to induce OPS was due to presynaptic and/or synaptic problems, the somata of CA1 pyramidal neurons should still be able to respond to ATS activations by firing action potentials. Our results confirmed this hypothesis (Fig. 5.18B, C). When O-HFS pulses at 200 Hz no longer induced OPSs, ATS pulses persisted to induce large APSs. This indicated that the absence of OPS during late O-HFS was not caused by failures in the post-synaptic neurons themselves. The O-HFS activation pathway involves afferent axons (Schaffer collaterals), pre- and post-synaptic structures, and the integration of post-synaptic potentials. While this experiment cannot pinpoint the exact failure locations, multiple mechanisms may be involved and act jointly: axons under O-HFS can experience depolarization block, and presynaptic structure at axonal terminals may fail to transmit activation reliably due to neurotransmitter depletion from intensive synaptic activations by O-HFS (Anderson et al. 2006; Iremonger et al. 2006).
Fig. 5.18
Neuronal responses to antidromic test stimuli (ATS) during O-HFS. A Schematic diagram showing the electrode positions. B Typical recording at pcl with 1-min 200 Hz O-HFS (stimulus artifacts removed). The enlarged insets below show the ATS-evoked APSs (shaded in grey). Red arrows above the waveforms denote the trimmed artifacts of 200 Hz O-HFS pulses, while the green dots below denote the ATS artifacts. C Changes in mean normalized amplitudes of OPS (upper) and APS (bottom) pooled from four 100 Hz and six 200 Hz 1-min O-HFS trains with 0.2 Hz ATS (n = 10). APSs were induced every 5 s during O-HFS and every minute after O-HFS. For the comparison between OPSs and APSs, during O-HFS, OPS samples were taken every 5 s (20 ms before each ATS, with the 1st OPS induced by the first O-HFS pulse). After O-HFS ended, identical orthodromic stimuli were applied per minute to show OPS recovery. Modified from Feng et al. (2013)
Since A-HFS on the alveus axons of CA1 pyramidal neurons can induce axonal block (Sect. 5.2), the O-HFS on the Schaffer collaterals (the axons of CA3 pyramidal neurons) may also induce axonal block. The two axon types have some differences: the alveus axons have thin myelin sheaths, while the Schaffer collateral axons are unmyelinated (Debanne et al. 2011; Andersen et al. 2007). Nevertheless, studies have shown that unmyelinated axons can also generate and even be more prone to axonal block (Chomiak and Hu 2007; Kim et al. 2012; Burke et al. 2004). Therefore, we tentatively assume that O-HFS on the Schaffer collaterals can induce more severe axonal block, resulting in no synchronous firing in the downstream neurons during the late steady period of O-HFS at pulse frequencies over 100 Hz. However, this does not necessarily mean that O-HFS has no activation effect on these neurons, it may drive the neurons to fire asynchronously. To verify the hypothesis, we investigated the behavior of individual neurons during O-HFS, as shown below.
5.3.2 Asynchronous Firing in Downstream Post-Synaptic Neurons Responding to O-HFS
As shown in Fig. 5.19A top, when we applied 1 min 100 Hz O-HFS, no obvious OPS events occurred in the late 40 s period except for the initial O-HFS period. We termed this late period the “steady period”. After stimulation artifact removal, the wideband recording signal was filtered to produce MUA signal. During the steady period, MUA spikes continued until the end of O-HFS with a firing rate exceeding baseline level (Fig. 5.19A bottom). However, upon the end of O-HFS, a silent period without spikes occurred, after which spikes gradually recovered (Fig. 5.19B). Statistical data showed that the mean MUA rate during the late 40 s O-HFS period was 77 ± 44 spikes/s, significantly higher than the baseline rate 26 ± 21 spikes/s (n = 29, Fig. 5.19C). The silent period after O-HFS lasted 15.6 ± 7.7 s. Within 2 min after O-HFS ended, the mean MUA rate returned to baseline level. Note: The silent period here likely stemmed from CA3 pyramidal neurons’ silence following A-HFS of their axons (as the Schaffer collateral stimulation was antidromic when viewed from the somata of these CA3 neurons), similar to the firing silence of CA1 pyramidal neurons following alveus A-HFS shown in Sect. 5.2.4. Without the inputs from upstream CA3 pyramidal neurons, the downstream CA1 neurons thus ceased firing.
Fig. 5.19
Increased firing in the downstream CA1 neurons by 100 Hz O-HFS at the Schaffer collaterals. A Example recording of 1-min 100 Hz O-HFS with stimulus artifacts removed (top), its enlarged episodes (middle), and its MUA signal (bottom). Red dots denote the unit spikes. B Peri-stimulus histogram of MUA firing rates of the signal shown in (A) (bin size = 1 s). The initial 20 s O-HFS period was skipped due to the interference of OPS events. C Comparison of mean MUA firing rates among the three periods: baseline before O-HFS, during O-HFS, and after O-HFS (measured from the second minute after the end of O-HFS to skip the silent period). **P < 0.001, paired t-test, n = 29.
The silent period after O-HFS indicated that the MUA appearance during the O-HFS was produced by the stimulation pulses. Moreover, during the late O-HFS period, when OPS events were absent due to axonal block, the MUA increased rather than decreased compared to baseline (Fig. 5.19C), because the axonal block was intermittent, not complete. Since the activated axonal terminals of the Schaffer collaterals can excite both pyramidal neurons and interneurons in the post-synaptic region (refer to Fig. 2.6A), the MUA signal should have contained the firing from both neuron types. We next classified the mixed unit spikes to investigate their firing alterations respectively.
1.
Increased Firing of Both Pyramidal Neurons and Interneurons by O-HFS
By analyzing MUA signals during 1 min 100 Hz O-HFS from 29 rat experiments, we identified SUA spikes from 214 neurons, including 154 (72%) pyramidal neurons and 60 (28%) interneurons. During the steady period (the late 40 s O-HFS period), the mean firing rate of pyramidal neurons was 5.7 ± 6.5 spikes/s, significantly higher than their baseline rate of 1.5 ± 2.0 spikes/s (Fig. 5.20A1). The mean firing rate of interneurons during the steady period was 24.5 ± 28.3 spikes/s, also significantly higher than their baseline rate of 8.8 ± 11.1 spikes/s (Fig. 5.20B1).
Fig. 5.20
Changes in the firing rates of pyramidal neurons and interneurons induced by 1 min 100 Hz O-HFS. A Firing changes in pyramidal neurons. A1: Comparison of mean firing rates between baseline (pre-O-HFS) and during steady O-HFS period (late 40 s). A2: Three response types in the 154 pyramidal neurons (excited, inhibited, and no response) defined by thresholds of ± 20% differences from baseline rate. B Firing changes in interneurons. B1 and B2: Corresponding analysis for 60 interneurons, matching (A1) and (A2) respectively. In (A1) and (B1), **P < 0.01, Wilcoxon signed rank test.
Based on firing rates during O-HFS, we defined three neuronal response types: “excited” neurons (with a rate exceeding 120% of baseline), “inhibited” neurons (with a rate below 80% of baseline), and “no response” neurons (with a rate between 80 and 120% of baseline). Their distributions among the 154 pyramidal neurons and 60 interneurons are shown respectively in Fig. 5.20A2, B2. Even for the excited neurons, the mean firing rates of pyramidal neurons (6.1 ± 5.6 spikes/s) and interneurons (41.7 ± 28.2 spikes/s) were far below the O-HFS pulse frequency of 100 Hz. Similar results were observed during the steady period of 1 min O-HFS at a higher frequency of 200 Hz (refer to Feng et al. 2017).
2.
Firing Patterns during Sustained O-HFS
Under normal physiological conditions, pyramidal neurons commonly fire action potentials in a burst pattern, while interneurons fire in a regular pattern (Barthó et al. 2004; Csicsvari et al. 1998). As shown in Fig. 5.21A1, in the baseline recording before O-HFS, the 130 “excited” pyramidal neurons exhibited burst firing, resulting in a peak around 5 ms in their average ISI histogram, with a cumulative ISI probability of 45 ± 25% in the range of 0–8 ms. However, as shown in Fig. 5.21A2, during the steady period of 100 Hz O-HFS, the pyramidal neurons no longer fired in bursts. The ISI probability in the range of 0–8 ms dropped to only 4.6 ± 3.4% without obvious peak. Instead, small peaks appeared in the ISI histogram at integer multiples of the 10 ms pulse interval of O-HFS, indicating that stimulation pulses modulated the firing of pyramidal neurons.
Fig. 5.21
Changes in the firing patterns of pyramidal neurons and interneurons by O-HFS. A and B Comparisons of the average ISI histograms of the 130 “excited” pyramidal neurons (A) and 43 “excited” interneurons (B), showing baseline vs. steady period of 100 Hz O-HFS. Each plot includes a typical episode of unit spikes, an average ISI histogram (with grey shading indicating the range of one standard deviation), and an enlarged view of the 0–25 ms ISI range. The cumulative probabilities in the two ISI ranges (0–8 ms and 9–11 ms) are denoted.
Unlike the pyramidal neurons, in baseline recording, the average ISI histogram of 43 “excited” interneurons was relatively flat, with a cumulative ISI probability of only 5.6 ± 8.4% in the 0–8 ms range (Fig. 5.21B1). However, during the steady period of 100 Hz O-HFS, peaks appeared around integer multiples of the 10 ms pulse interval in the ISI histogram (Fig. 5.21B2). In the 9–11 ms ISI range (containing the first peak), the cumulative ISI probability of 21 ± 17% was significantly greater than the 3.3 ± 3.3% at baseline (P < 0.01, Wilcoxon signed rank test, n = 43), indicating the strong modulation effect of O-HFS pulses on the interneuron firing. Since the firing rate of interneurons (~ 40 spikes/s) was much higher than that of pyramidal neurons (~ 6 spikes/s), the interneuron spikes represented by the 21% probability in the 9–11 ms ISI range was much more than the corresponding 3.7% of pyramidal neuron spikes. This firing difference between the two neuron types may be caused by the lower action potential threshold of interneurons (Csicsvari et al. 1998).
These results indicate that O-HFS can modulate and increase neuronal firing. Unlike single pulse stimulation, the absence of evoked-population spikes (PS) during steady O-HFS period suggests that sustained axonal HFS can weaken the synchronization in pulse-evoked neuronal firing. We next examined the changes in neuronal firing synchrony during O-HFS.
3.
Asynchronous Firing during Sustained O-HFS
As described in Sect. 4.1.4, we used recordings from four adjacent channels on the electrode array to classify and distinguish spikes from individual neurons. To study firing synchrony between neurons, we analyzed spikes from pairs of interneurons simultaneously recorded on these four channels. The two interneurons in each pair were denoted as “Neuron-1” and “Neuron-2” (Fig. 5.22A). During baseline recording, the two neurons fired asynchronously (Fig. 5.22B1). We denoted the interval between adjacent spikes of the paired neurons as Δt. The Δt histogram was flat without obvious peak (Fig. 5.22B2). The synchronous firing ratio of the paired neurons was only 8%. Note: we defined synchronous firing as spike pairs with Δt < 2 ms, and the synchronous firing ratio between Neuron-1 and Neuron-2 as F/(F1 · F2)1/2—where F was the firing rate of spike pairs with Δt < 2 ms, F1 and F2 were the firing rates of Neuron-1 and Neuron-2 respectively (Quian Quiroga et al. 2002).
Fig. 5.22
O-HFS-induced asynchronous firing of paired interneurons. A Unit spikes from two interneurons (Neuron-1 and Neuron-2) recorded simultaneously by four adjacent channels in the electrode array. The spike waveforms from both neurons in four channels (Ch1-4) were superimposed, together with their averaged waveforms shown in red. The neurons showed largest spike amplitudes in Ch2 and Ch3 respectively, so these channels were used to analyze the firing relationship between the paired neurons. B Baseline firing. B1: Asynchronous firing of the paired neurons. B2: Histogram of firing probability versus interval (Δt) between adjacent spikes from the paired neurons. C Synchronous firing of the paired neurons induced by a single pulse at a weak 50 μA intensity. C1: Example of evoked spikes from both neurons. C2: Schematic diagram showing simultaneous excitation of the neurons by the orthodromic stimulation of Schaffer collaterals. C3: Histogram of Δt from 40 single-pulse trials. D Example of asynchronous firing of paired neurons during steady period of 100 Hz O-HFS (D1) and corresponding Δt histogram (D2). E Comparison of synchronous firing ratios of paired neurons among baseline firing, single-pulse evoked firing, and the firing during steady O-HFS period. **P < 0.001, paired t-test, n = 5 pairs.
As shown in Fig. 5.22C1, an orthodromic test stimulus (OTS) of 50μA (weak intensity) induced paired neurons to fire synchronously with a latency difference of Δt ≈1ms. This synchronous firing indicated that the stimulation at the Schaffer collaterals activated both neurons simultaneously (Fig. 5.22C2). The latencies of induced spikes were less than 5 ms, indicating only monosynaptic transmission involved in the activation pathway (Buzsáki 1984). In 40 repetitions of the weak OTS (single-pulse stimulation), the synchronous firing ratio was ~ 85%, resulting in a peak near 0 ms in the Δt histogram (Fig. 5.22C3). Note: to prevent from generating population spikes, we applied the OTS at a very low intensity. At such intensity level, both neurons did not always fire simultaneously—in the remaining ~ 15% of trials, at least one neuron failed to respond to the OTS.
Unlike the weak OTS, during the steady period of 1 min 100 Hz O-HFS, even with a strong intensity of 0.3 mA, the pulses rarely induced paired neurons to fire synchronously (Fig. 5.22D1). Small peaks with intervals of about 10 ms appeared in the Δt histogram, indicating the modulation effect of 100 Hz O-HFS on neuronal firing (Fig. 5.22D2), but with a synchronous firing ratio of only 9%. Among the 5 pairs of interneurons obtained from 5 rat experiments, the mean synchronous firing ratio during O-HFS was 10.4% ± 12.4%—significantly lower than the 86.0% ± 3.3% induced by weak OTS, but close to the baseline ratio of 6.8 ± 7.1% (Fig. 5.22E). This result indicates that sustained O-HFS can generate asynchronous firing in post-synaptic interneurons.
Stimulating the Schaffer collaterals can activate both pyramidal neurons and interneurons through monosynaptic transmissions. Interneurons have a lower action potential threshold than pyramidal neurons. Additionally, interneurons are fewer in number, comprising only about 10% of CA1 neurons, and are sparsely distributed across CA1 layers (Andersen et al. 2007; Csicsvari et al. 1998). Due to these characteristics, pairs of innervated interneurons can be reliably activated by single OTS pulses with weak intensity in baseline trials without inducing PS events. The spikes of paired interneurons can be clearly identified following the weak pulses, without contamination from PS or from spikes of pyramidal neurons (Fig. 5.22C). However, it is difficult to activate paired pyramidal neurons simultaneously without inducing PS events, because of their densely packed population and higher action potential threshold. Therefore, we could not repeat the baseline OTS experiment for paired pyramidal neurons by using the extracellular recording approach.
Nevertheless, we calculated a synchronous firing ratio for each pair of simultaneously recorded neurons (Feng et al. 2017). For example, four pyramidal neurons obtained in the same experiment formed 6 pairs, with each neuron participating in three pairs. From the baseline recordings of 29 rat experiments containing 3–11 pyramidal neurons, we obtained a total of 684 pairs. Similarly, for the “excited” neurons (including both pyramidal neurons and interneurons) collected during O-HFS at 100 and 200 Hz, the same pairing analysis was performed. The mean synchronous firing ratios during the steady O-HFS period were small for both neuron types and both pulse frequencies (Fig. 5.23)—less than 3% for pyramidal neurons and less than 7% for interneurons. They all showed no significant differences from the baseline values.
Fig. 5.23
Comparisons of synchronous firing ratios for paired pyramidal neurons (A) and paired interneurons (B) among baseline firing and during the steady period of 100 and 200 Hz 1 min O-HFS trains.
These results show that single-pulses applied to afferent axons can induce synchronous firing in downstream post-synaptic neurons, whereas sustained O-HFS can produce asynchronous firing. The asynchronous firing of post-synaptic neurons can reflect asynchronous excitatory inputs from presynaptic neurons due to HFS-induced intermittent axonal block in the afferent pathway.
5.3.3 Higher-Frequency Axonal HFS Can Increase Firing Randomness Rather Than Firing Rate
Clinical application of deep brain stimulation (DBS) for disorders, such as Parkinson's disease, has shown that HFS with pulse frequencies above 90 Hz usually has better therapeutic effects, while frequencies below 60 Hz are usually ineffective or even detrimental (McConnell et al. 2012; Birdno and Grill 2008; Kuncel et al. 2006). Previous studies have also shown that HFS at around 130Hz is more effective in controlling epilepsy than lower frequencies (Vonck et al. 2013; Jobst 2010). Although many studies on DBS mechanisms have focused mainly on whether target neurons are excited or inhibited based on changes in firing rates (Vitek 2002; Florence et al. 2016), the desynchronization effect of HFS has also gained significant attentions (Medeiros and Moraes 2014; Popovych and Tass 2014). Many neurological disorders, such as motor disorders and epilepsy, are associated with increased synchronous and rhythmic oscillations in the firing of neuronal populations (Hammond et al. 2007; Gatev et al. 2006; Rampp and Stefan 2006; Jiruska et al. 2013). Studies have shown that HFS can reduce the pathological oscillations and synchronous firing of target neurons (Eusebio et al. 2011; Wingeier et al. 2006; Deniau et al. 2010; Medeiros and Moraes 2014). Therefore, the therapeutic effect of DBS may stem from a desynchronization effect of HFS, rather than just changing neuronal firing rates (Hashimoto et al. 2003; McCairn and Turner 2009). In our O-HFS experiments in rat hippocampal CA1 region, we investigated the changes in asynchronous effects of HFS at different pulse frequencies by analyzing the distributions of evoked spikes.
As shown in Fig. 5.24, during O-HFS with pulse frequencies at 50, 100 and 200 Hz respectively, we recorded both the evoked OPS in the pyramidal cell layer (pcl) and the fEPSP in the apical dendritic layer (sr). The first O-HFS pulses always induced large OPS and fEPSP, indicating a highly synchronized response of downstream neurons (enlarged view ① in Fig. 5.24A–C). The O-HFS period with OPS events was longer at 50 Hz than at higher frequencies of 100 and 200Hz. After OPS disappeared in the late O-HFS period, unit spikes continued in the pcl layer. Meanwhile, small oscillations exhibited in the sr layer, pacing with the O-HFS pulses and particularly evident at 50 and 100 Hz frequencies (enlarged view ②). These oscillations reflected fluctuations in post-synaptic potentials produced by the O-HFS activations from the Schaffer collaterals. As pulse frequency increased, the oscillation amplitudes decreased substantially. Let A1 represent the amplitude of fEPSP induced by the first O-HFS pulse and A2 represent the mean amplitude of the oscillations in the final second of O-HFS. At 50, 100 and 200 Hz, A2 was only about 15%, 2%, and 0.2% of A1, respectively (Fig. 5.24D). This indicated that higher pulse frequencies resulted in smaller fluctuations in post-synaptic potentials by each pulse. Note: A2 was not the magnitude of post-synaptic potential (fEPSP). The fEPSPs induced by O-HFS pulses fused together as shown in the recordings at the CA1 sr layer in the enlarged view denoted by ①. Our AC coupling recording could not capture the DC offset component in the fused waveforms.
Fig. 5.24
Changes of evoked potentials during O-HFS with three different frequencies. A–C Typical recordings in the CA1 pyramidal cell layer (pcl) and stratum radiatum (sr) collected by two contacts separated by 0.2 mm during 1 min O-HFS at pulse frequencies of 50 (A), 100 (B) and 200 Hz (C), respectively. The neuronal responses at the O-HFS initial and late periods (denoted by “①” and “②”) are enlarged. D Comparison of the amplitudes of sr oscillations (A2) in the final O-HFS second among the three pulse frequencies. With the similar amplitudes at the O-HFS onset (A1, listed below), the mean A2 was suppressed more at higher frequencies. The A2 values were calculated by superposing and averaging the inter-pulse signals in the final second of O-HFS, as shown in the right insets of figures A-C. Green waveforms are superposed signals, while black ones are average waveforms. Two repeated inter-pulse signals are shown in each inset to highlight oscillations. Red arrows denote removed stimulus artifacts. E Comparisons of MUA rates between baseline and late 30 s periods of 1 min O-HFS at the three pulse frequencies respectively.
Comparing the firing rates of MUA recorded at the pcl layer during the late 30 s of 1 min O-HFS showed that the neuronal firing at the three stimulation frequencies significantly exceeded baseline levels. However, the mean MUA rates showed no significant differences across the three frequencies (Fig. 5.24E).
These results showed that varying pulse frequencies from 50 to 200 Hz had no significant effect on the MUA firing rates in downstream neurons, despite the four-fold difference in the electrical energy delivered by the O-HFS trains. Nevertheless, the fluctuations in post-synaptic potentials were frequency dependent (Fig. 5.24D), suggesting possible differences in firing patterns. This led us to hypothesize that the extra electrical energy from higher frequency pulses could be used to modulate firing timing rather than firing rates. To investigate this, we analyzed the post-stimulation time histogram (PSTH) of MUA during O-HFS at three different pulse frequencies.
As a control, we created a mimic PSTH for baseline recording by adding virtual 10 ms intervals (100 Hz) to the baseline MUA signal. The PSTH showed an even distribution, indicating random spontaneous firing (Fig. 5.25A). During the steady period of 50 Hz O-HFS, the PSTH of MUA was highly non-uniform with a sharp peak around 10 ms in the 20 ms inter-pulse-intervals (IPI). Most spikes occurred between 8 and 13 ms (Fig. 5.25B1). The PSTH distributions became progressively more uniform as pulse frequency increased (Fig. 5.25B2, B3). In these PSTHs, spike counts were measured across different time spans—20, 10 and 5 ms (4:2:1) for 50, 100 and 200 Hz respectively. To compare them with a same time span, we divided the PSTHs of 100 and 200 Hz into two and four equal segments respectively, then connected them to form PSTHs with a 20 ms time span, matching the 50 Hz O-HFS. We also created a mimic PSTH of baseline MUA with 20 ms intervals to serve as a control (Fig. 5.25C). These PSTHs with identical time spans clearly showed that increasing pulse frequency flattened the MUA distribution in the intervals, indicating weakened phase-locking between neuronal firing and O-HFS pulses.
Fig. 5.25
Changes in PSTHs of MUA signals induced by O-HFS at various pulse frequencies. A Creation of mimic PSTH for 30 s baseline MUA signal to serve as a control. Left: typical baseline MUA signal (30 s) divided into virtual 10 ms intervals. Right: superimposed signals of all 10 ms intervals from the 30 s MUA (top) and the corresponding mimic PSTH (bottom). The blue horizontal line denotes the average PSTH value. B Typical plots of MUA PSTH for the final 30 s of 1 min O-HFS at 50, 100, and 200 Hz pulse frequencies. Top: superimposed signals of all inter-pulse intervals. Bottom: corresponding PSTH plots. In these PSTHs, red and blue horizontal lines denote their own averages (Cave) and the baseline averages, respectively. ∆C represents the difference between peak and average values. Pink bins indicate PSTH values exceeding 1.2 times the baseline average (termed as excitatory bins). C Comparisons of PSTH distributions under an identical 20 ms interval by dividing the PSTHs of 100 and 200 Hz into two and four equal segments, respectively. D and E Comparisons of peak coefficient (∆C/Cave) and duty ratio (DR) of excitatory phases among O-HFS with three pulse frequencies.
We used the index of peak coefficient (ΔC/Cave) to measure the non-uniformity of PSTH distribution (Fig. 5.25B). Here, Cave represented the mean of PSTH and ∆C was the difference between the peak and mean PSTH values. A higher PSTH peak resulted in a greater peak coefficient, while a flat PSTH yielded a peak coefficient close to 0. In addition, we used the duty ratio (DR) of excitatory phase—the percentage of excitatory bins in PSTH—to measure the concentration of spikes in the inter-pulse-intervals. An excitatory bin was defined as a PSTH bin (0.5 ms) with a spike count above 120% of the mean value of baseline mimic PSTH (Wang et al. 2018). Our statistical analysis showed that as pulse frequency increased, the mean peak coefficient decreased significantly (Fig. 5.25D), while the DR of excitatory phases increased significantly (Fig. 5.25E). These results indicated that with similar MUA rates produced by O-HFS (Fig. 5.24E), a lower frequency of 50 Hz generated phase-locked firing, whereas higher frequencies of 100 or 200 Hz produced more random firing.
An MUA signal contains action potentials from multiple neurons. Its distribution may not represent the firing patterns of individual neurons. A uniform MUA pattern could result from either random firing of individual neurons or varying firing timings among neurons that maintain different phase-locked firing with O-HFS pulses. To clarify this, we examined the PSTHs of single unit activity (SUA) from individual neurons after spike sorting. As O-HFS frequency increased, the PSTHs of SUA signals from both pyramidal neurons and interneurons were similar to those of MUA (Wang et al. 2018), indicating that the randomization observed in MUA distributions at higher O-HFS frequencies stemmed from the random firing of individual neurons. The results suggest that axonal HFS with a sufficiently high frequency can generate random rather than phase-locked activation in downstream neurons, leading to asynchronous firing.
The above investigations show that O-HFS can induce a frequency-dependent axonal block in the Schaffer collaterals at pulse frequencies ranging in 50–200 Hz. During this type of axonal block, O-HFS can still activate downstream post-synaptic neurons and increase their firing rates, indicating partial or intermittent axonal blocks. As pulse frequencies increase, neuronal firing becomes more random. Although the mean firing rates during O-HFS showed no significant differences across this frequency range, the rate peaked at 100 Hz and declined slightly at 200 Hz (Fig. 5.24E). Previous studies have shown that HFSs in peripheral nerves with pulse frequencies up to thousands of hertz can completely block axonal membranes from firing (McGee et al. 2015). This suggests that pulse frequencies above 200 Hz may lead to further decreases in neuronal firing rates.
5.4 Activation of Interneurons in Local Inhibitory Circuits by HFS at Efferent Fibers
Section 5.3 describes how O-HFS modulates downstream post-synaptic neurons, including pyramidal neurons and interneurons. The O-HFS was applied at the Schaffer collaterals—one of the major afferent pathways in the hippocampal CA1 region. Besides the excitatory projection pathways, neurons also form connections through local circuits, such as the inhibitory circuits in the hippocampal CA1 region described in Sect. 2.3. As shown in Fig. 5.26A, we hypothesized that antidromic activation from the alveus (the CA1 efferent fibers) could activate interneurons in local inhibitory circuits in addition to activating the pyramidal neurons’ somata. To test this hypothesis, we analyzed the SUA from interneurons (IN) during A-HFS at the alveus.
Fig. 5.26
Responses of interneurons (IN) to antidromic stimulation at the CA1 alveus. A Schematic diagram of the local inhibitory circuits and electrode positions. Pyramidal neurons (Pyr) receive inhibitory inputs from both feedforward and feedback inhibitory circuits. INs in these inhibitory circuits are activated by excitatory synapses and then act on Pyr somata or dendrites through inhibitory synapses. B IN spikes and an APS evoked by a single-pulse applied at the alveus. The enlarged spike waveform and the ISI histogram of the IN firing from baseline recording (shown in the lower left) were used to identify the IN firing.
The excitability of pyramidal neurons (CA1 principal neurons) is regulated by both local feedforward and feedback inhibitory circuits composed of interneurons (Pelkey et al. 2017). As shown in Fig. 5.26A, interneurons in the feedforward circuits receive activation from the Schaffer collaterals alongside pyramidal neurons. The activated interneurons then inhibit the pyramidal neurons through inhibitory synapses. In the feedback circuits, interneurons are activated by adjacent pyramidal neurons and subsequently inhibit pyramidal neurons in return (see Sect. 2.3 for details). In the A-HFS experiment, the stimulating electrode (ASE) was placed about 1.3 mm away from the recording electrode (refer to Sect. 5.2.1 and Fig. 5.4A for electrode positions). Therefore, the activation produced by A-HFS on the alveus may antidromically propagate along the axons to the vicinity of the recording electrode (RE) to then activate nearby interneurons recorded by the RE.
As shown in Fig. 5.26B, a single 0.05 mA pulse applied to the alveus antidromically activated a population of pyramidal neurons near the RE, generating an APS with a latency of ~ 1.3 ms. Immediately following this APS, an IN spike appeared with a latency of ~ 2.4 ms. This latency of IN spike was longer than the APS latency and aligned with previous reports of monosynaptic transmission delays (Csicsvari et al. 1998), indicating that the IN was in a feedback inhibition circuit. Although the IN spike had a much smaller amplitude than the APS, their latency difference made the IN spike clearly distinguishable after the APS (Fig. 5.26B).
However, when an IN spike occurred too close to an APS, it could be covered by the APS. To ensure accurate IN spike detection, only INs with sufficiently large spikes (> 100 μV) were used. Due to the similar spectral ranges of unit spikes and population spikes, we could not extract IN spikes by using a filter to remove LFP and APS (refer to Sect. 4.1.4). Instead, we used a window detection method to extract spikes. As described in Sect. 4.1.5, spikes were directly detected in the raw wideband (0.3–5000 Hz) recordings from four adjacent channels and then used for spike sorting to isolate unit spikes from individual neurons. Finally, we identified IN spikes based on two features: a rising phase less than 0.4 ms and a relatively flat ISI histogram in baseline firing (Fig. 5.26B, lower left).
We obtained unit spikes from one IN in each of the 19 rat experiments. Their firing triggered by the alveus stimulation had a mean latency of 2.7 ± 0.45 ms, indicating monosynaptic activation. The mean spike amplitude of these INs reached 287 ± 110 μV in the maximum amplitude recording channel. However, even these “super large” spikes would be submerged in the large APSs generated during the initial A-HFS period. Therefore, to investigate IN firing through the entire A-HFS period, we used weak-intensity A-HFS of 0.06 ± 0.02 mA in 9 rats. In the remaining 10 rats, we applied strong-intensity A-HFS of 0.33 ± 0.08 mA to examine IN firing during the steady A-HFS period when APS declined.
Figure 5.27 shows an example of low-intensity A-HFS. In the 2 min baseline spontaneous MUA signal, IN spikes were clearly distinguishable (Fig. 5.27A) with a mean firing rate of 9.4 Hz. During the subsequent 2min 100 Hz A-HFS with a weak-intensity of 0.05 mA, each pulse induced an APS with an amplitude of approximately 1.9 mV at the initial period, followed most by an IN spike (Fig. 5.27B). The coupling ratio between the IN firing and the A-HFS pulses was ~ 80%. As the A-HFS continued, the induced APS declined (Fig. 5.27C), accompanied by decreases in both the IN firing rate and its coupling ratio (Fig. 5.27B, bottom right).
Fig. 5.27
Modulation of weak A-HFS on IN firing. A Typical wideband baseline recording and its high-pass filtered MUA signal with clear IN spikes. B Evoked APSs and IN spikes during 2 min A-HFS with a weak-intensity of 0.05 mA. Blue dots denote IN spikes. C Scatter plots showing APS amplitudes (black) and latencies (brown) during A-HFS. D Two-dimensional raster plot of IN spikes during the A-HFS shown in (B), together with the PSTH plots of the first second and late 60 s period (61–120 s) of A-HFS shown on the left and right, respectively. E Mean IN firing rates per second before, during and after weak A-HFS (n = 9 INs). The error bars represent one standard deviation. F–H Comparisons of the mean IN firing rates (F), the mean IN spike latencies (G) and the mean interquartile range (H, IQRPST) of IN PSTHs. #P < 0.05, ##P < 0.01, repeated measurements one-way ANOVA with post hoc Bonferroni tests, 9 rats. **P < 0.01, paired t-test, 9 rats.
The two-dimensional raster plot showed that the IN initially fired in phase-lock with the pulses during the first second of A-HFS (Fig. 5.27D, left). Subsequently, the firing gradually dispersed and decreased in rate (Fig. 5.27D, middle and right). The raster plot showed the timing of each IN spike within the 10ms A-HFS inter-pulse-intervals, known as the post-simulation time (PST). The PST data were used to create the PSTH (see Sect. 4.2.4 for details). The PSTHs (bin = 0.5 ms) for the IN spikes during the first second and late 60 s of A-HFS periods are shown in the left and right plots of Fig. 5.27D. To quantitatively assess the phase-locking between IN spikes and A-HFS pulses, we calculated the interquartile range (IQR) of the PSTH data, denoted as IQRPST, representing the range containing the middle 50% of spikes in the PSTH. A smaller IQRPST value indicated stronger phase-locking between spikes and pulses.
As shown in the left of Fig. 5.27D, during the first second of A-HFS, the example IN had a firing rate (FR) of 47 Hz, with spikes concentrating within 3.0–5.5 ms of the 10 ms IPI. Its IQRPST was only 0.83 ms, indicating strong phase-locking between the IN firing and A-HFS pulses. However, during the final 60 s of 2 min A-HFS, the mean firing rate decreased to 13.8 Hz—though still above the baseline rate of 9.4 Hz. The IQRPST increased to 1.63 ms, with some spikes no longer phase-locked to the pulses. Additionally, as the APS latency gradually increased (Fig. 5.27C), the latency of IN spikes also increased from 3.00 ms at the onset of A-HFS to a mean value of 6.75 ms in the final 60 s of A-HFS, as shown by the peak time of the PSTH plotted in the right of Fig. 5.27D.
Statistical data in Fig. 5.27E, F show that the mean IN firing rate during the first A-HFS second was significantly higher than both baseline and the late 60 s A-HFS period. The mean rate in the late 60 s was also significantly higher than baseline. Additionally, the mean IN spike latency in the late 60 s A-HFS was significantly longer than the value in the first second (Fig. 5.27G), while the IQRPST also showed a significant increase (Fig. 5.27H). These results indicate that axonal block produced by sustained A-HFS reduced the stimulation activation on the INs.
To enhance the A-HFS effect on INs, we increased the pulse intensity to around 0.3 mA, which was strong enough to induce an APS with about 3/4 maximum amplitude in single-pulse stimulation. While IN spikes were distinct in baseline recording (Fig. 5.28A), the large evoked APSs obscured the IN spikes during the initial period of 100 Hz A-HFS (Fig. 5.28B). After about ten seconds of A-HFS, as the APS declined (Fig. 5.28C) and spike latency increased, the evoked IN spikes became visible following APSs (denoted by blue dots in the enlarged insets of Fig. 5.28B). The IN firing rate was 92 Hz at the 11th second of A-HFS with an IQRPST of 0.60 ms in the corresponding PSTH (Fig. 5.28D). This high rate largely persisted until the end of A-HFS, resulting in a mean rate of 78.0 Hz and a still small IQRPST of 0.75 ms in the late 60 s A-HFS. From the 11th second to the late 60 s of A-HFS, the mean APS latency increased slightly from 3.10 to 3.43 ms (Fig. 5.28C), while the mean latency of IN spikes increased from 5.25 to 6.25 ms.
Fig. 5.28
Modulation of strong A-HFS on IN firing. A Typical wideband baseline recording and its high-pass filtered MUA signal with distinct spikes from an IN. B Evoked APSs and IN spikes during 2 min A-HFS with a strong-intensity of 0.3 mA. C Scatter plots showing the APS amplitudes (black) and the latencies (brown) during A-HFS. D Two-dimensional raster plot of the IN spikes during the A-HFS shown in (B), together with per-second firing rates (orange curve) and the PSTH plots for the 11th second (left) and late 60 s period (61–120 s) (right).
Figure 5.29 shows comparisons between the weak- and strong-intensity A-HFS groups. The mean APS amplitude induced by the first pulse was significantly greater in the strong-intensity group than the weak-intensity group (Fig. 5.29A). During the steady A-HFS period (late 60 s), despite significant decreases, the APS amplitude difference between the two groups persisted (Fig. 5.29B), indicating the substantial difference in the numbers of activated pyramidal neurons. Although A-HFS doubled the APS latency, both weak- and strong-intensity groups showed similar APS latencies since they shared the same activation pathway (Fig. 5.29C, D).
Fig. 5.29
Comparisons between the two A-HFS groups with weak- and strong-intensity pulses. A–D Comparisons of mean initial APS amplitudes (A), mean APS amplitudes in late 60 s of A-HFS (B), mean initial APS latencies (C), and mean APS latencies in late 60 s of A-HFS (D). E–G Comparisons of IN spike amplitudes (E), baseline IN firing rates (F), and IN firing rates in late 60 s of A-HFS (G). H Comparison of IQRPST of IN firing in late 60 s of A-HFS. I Comparison of the IN spike latencies in late 60 s of A-HFS. J Comparison of latency increments between IN spikes and APSs evoked during weak A-HFS. For IN spikes, the latency increment was the peak time of PSTH in late 60 s minus the latency following the 1st pulse; for APS, it was the mean latency in late 60 s minus the latency following the 1st pulse. *P < 0.05, **P < 0.01, in (A-I) unpaired t-test, in (J) paired t-test, n = 9 rats for weak-intensity group, n = 10 rats for strong-intensity group.
Neither IN spike amplitudes nor baseline IN firing rates showed significant differences between the two groups (Fig. 5.29E, F). However, during the late 60 s of A-HFS, the strong-intensity group showed a significantly higher mean IN firing rate (Fig. 5.29G) and stronger firing phase-locking, evidenced by its significantly smaller IQRPST value (Fig. 5.29H). Since both groups shared the same activation pathway, the latencies of IN spikes showed no significant differences between the groups (Fig. 5.29I). However, IN and APS had different activation pathways, leading to a significant difference in their latency increments. Throughout the weak-intensity A-HFS period with distinguishable IN spikes, the mean latency increment of IN spikes (3.41 ± 0.61 ms) significantly exceeded the APS latency increment (1.71 ± 0.45 ms) (n = 9, Fig. 5.29J). This shows that the IN activation pathway—involving synaptic transmission—produced longer delays than the antidromic activation pathway of pyramidal neurons.
Unlike the downstream IN activation by O-HFS at the Schaffer collaterals described in Sect. 5.3, A-HFS at the alveus (the CA1 efferent fiber) activates the INs located upstream of the stimulation site. Based on the CA1 local circuits shown in Fig. 5.26A, this A-HFS activation on IN may involve the following process: First, the stimulation activates the alveus axons to generate excitation that propagates antidromically along the axon. This excitation then turns to axonal branches and propagates in an orthodromic direction to the synapses at the axonal terminals. Through synaptic transmissions, the IN in the feedback inhibition circuit is activated (as denoted by the solid orange curve in the figure). Due to the intermittent depolarization block on the axons caused by sustained A-HFS, the excitation propagation to the IN reduces, resulting in a significant decrease in the IN firing rate during the steady A-HFS period compared to the initial period (Fig. 5.27). Nevertheless, IN can still maintain a firing rate of ~ 40 Hz during the steady period of strong-intensity A-HFS (Fig. 5.29G), achieving a coupling ratio of ~ 40% between the IN firing and the 100 Hz pulses. However, the APS amplitude in the steady A-HFS period, which reflects the number of firing pyramidal neurons, drops to only ~ 15% of the initial value (Fig. 5.29A, B).
During the steady period, despite its weakened effect, A-HFS can still activate INs sufficiently to fire at a high rate. This may stem from IN characteristics. While INs comprise only about 10% of total neurons in the hippocampal CA1 region (Andersen et al. 2007), each IN receives excitatory inputs from many pyramidal neurons, and it possesses a low activation threshold (Csicsvari et al. 1998). As a result, an IN requires only a few simultaneous excitatory inputs to fire. Even weak-intensity A-HFS can produce IN firing rates up to ~ 60 spikes/s in the first second of stimulation (Fig. 5.27F), while strong-intensity A-HFS can sustain high IN firing rates throughout its steady period.
The HFS-induced partial axonal block can extend the latency of neuronal responses (both APS and IN spike). While this latency extension remains consistent along the same activation pathway, IN responses show longer extensions because its activation pathway involves additional parts such as axonal branches and synaptic transmissions.
5.5 Summary
Among neuronal structures, the axonal membrane has the shortest chronaxie, making it most sensitive to the narrow pulses commonly used in neuromodulation techniques like DBS. Axons can conduct action potentials quickly and reliably. However, under sustained HFS, axons can experience depolarization block, preventing them from generating action potentials in response to every stimulation pulse. This chapter details our experiments in the rat hippocampal CA1 region to verify intermittent axonal block produced by HFS at frequencies ranging from tens to hundreds of hertz. We found that when A-HFS on the alveus—the axons of CA1 pyramidal neurons—fail to induce synchronized action potentials in their somata through antidromic activation, single-pulse stimulations on their afferent fibers can still orthodromically activate these somata to fire through dendrite excitatory inputs. This finding can clearly demonstrate axonal block produced by sustained HFS. Furthermore, we revealed that this axonal block can not only weaken the excitation from the HFS electrode but also block excitation from other sites passing through the blocked region.
Our experiments using A-HFS with inserted gaps showed that axonal HFS can extend the refractory period of axons, resulting in intermittent axonal block. Additionally, sustained HFS at axons can alter the somata excitability, slowing activation conduction around the soma and causing a silent period of firing immediately after A-HFS cessation. Moreover, with the intermittent axonal block, the stimulated pyramidal neurons can exhibit non-uniform clustered firing even under the stimulations of periodic pulses at constant intervals. This non-uniform response may stem from the nonlinear dynamic properties of neurons. Nevertheless, the HFS-induced clustered firing differed from the spontaneous burst firing commonly observed in pyramidal neurons.
Experiments of O-HFS at the Schaffer collaterals showed that HFS-induced axonal block can attenuate and smooth the excitatory stimulation effect on the downstream postsynaptic neurons, resulting in asynchronous firing after initial synchronous firing of OPS events. Although weakened, sustained O-HFS can still elevate the firing rates of postsynaptic neurons (both pyramidal neurons and interneurons) in the O-HFS frequency range from 50 to 200 Hz, while neuronal firing can become more random and less synchronized at higher frequencies. The silent period of neuronal firing following O-HFS can verify that the increased firing during O-HFS is stimulation-induced.
During steady HFS periods with different pulse frequencies (50, 100 and 200 Hz) and their corresponding electrical energies (1×, 2×, and 4×), neither A-HFS nor O-HFS showed significant differences in the amounts of evoked neuronal firing. However, higher frequencies did produce greater firing dispersity. This finding aligns with observations that neuronal desynchronization needs a sufficiently high frequency when treating Parkinson's disease by DBS (Brown et al. 2004). Thus, in HFS with a higher-frequency, some of the electrical energy may serve to randomize firing rather than increase it. The mechanism by which sustained HFS achieves this function can be the intermittent depolarization block of neuronal membranes, with a large amount of electrical energy from the stimulation being used to maintain this block state.
Many brain diseases involve synchronized, rhythmic neuronal firing. For example, in Parkinson’s disease, movement disorders are associated with increased synchronous bursts and low-frequency oscillatory discharges in basal and thalamic neurons (Birdno and Grill 2008; Gale et al. 2008). Epilepsy is typically characterized by excessive synchronous discharges in neuronal populations (Le Van Quyen et al. 2003; Lopes da Silva et al. 2003). Studies have suggested that desynchronizing neuronal discharges may be a key mechanism for DBS therapy (Medeiros and Moraes 2014; McConnell et al. 2012; Wilson et al. 2011). Effective DBS treatment for movement disorders has been considered using HFS to create new firing patterns that replace the pathological oscillations and synchronous activity (Llinás et al. 1999; Birdno and Grill 2008). Similarly, certain stimulation therapies have controlled seizures by desynchronizing epileptic activity in neuronal networks (Cota et al. 2009, Medeiros and Moraes 2014). Our experimental results in this chapter confirm the asynchronous effect of HFS and support desynchronization as a mechanism of DBS action.
Furthermore, our experiments showed that both O-HFS at afferent fibers and A-HFS at efferent fibers can increase the firing of interneurons. These interneurons can regulate pyramidal neurons in a wide area through feedforward and feedback inhibitory circuits (refer to Sect. 2.3). Particularly, the A-HFS effect on interneurons suggests a potential DBS target along the hippocampal efferent pathway to control disorders like epilepsy. The hippocampus is frequently involved in epilepsy, with its CA3 region being particularly prone to epileptiform activity due to abundant excitatory connections among CA3 pyramidal neurons (Andersen et al. 2007). The abnormal activity can spread to the CA1 region and subsequently to other brain regions through the alveus efferent fibers. By activating interneurons, electrical stimulation at the efferent fibers may inhibit CA1 pyramidal neurons, potentially interrupting and eliminating the spread of epileptiform activity. Further research is needed to validate this hypothesis.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits any noncommercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if you modified the licensed material. You do not have permission under this license to share adapted material derived from this chapter or parts of it.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Andersen P, Morris R, Amaral D et al (2007) The hippocampus book. Oxford University Press
Anderson TR, Hu B, Iremonger K et al (2006) Selective attenuation of afferent synaptic transmission as a mechanism of thalamic deep brain stimulation-induced tremor arrest. J Neurosci 26(3):841–850CrossRef
Azouz R, Jensen MS, Yaari Y (1996) Ionic basis of spike after-depolarization and burst generation in adult rat hippocampal CA1 pyramidal cells. J Physiol 492(Pt 1):211–223CrossRef
Barthó P, Hirase H, Monconduit L et al (2004) Characterization of neocortical principal cells and interneurons by network interactions and extracellular features. J Neurophysiol 92(1):600–608CrossRef
Bellinger SC, Miyazawa G, Steinmetz PN (2008) Submyelin potassium accumulation may functionally block subsets of local axons during deep brain stimulation: a modeling study. J Neural Eng 5(3):263–274CrossRef
Birdno MJ, Grill WM (2008) Mechanisms of deep brain stimulation in movement disorders as revealed by changes in stimulus frequency. Neurotherapeutics 5(1):14–25CrossRef
Brazhe AR, Maksimov GV, Mosekilde E et al (2011) Excitation block in a nerve fibre model owing to potassium-dependent changes in myelin resistance. Interface Focus 1(1):86–100CrossRef
Brown P, Mazzone P, Oliviero A et al (2004) Effects of stimulation of the subthalamic area on oscillatory pallidal activity in Parkinson’s disease. Exp Neurol 188(2):480–490CrossRef
Bucher D, Goaillard JM (2011) Beyond faithful conduction: Short-term dynamics, neuromodulation, and long-term regulation of spike propagation in the axon. Prog Neurobiol 94(4):307–346CrossRef
Burke D, Cappelen-Smith C, Kuwabara S (2004) Conduction block in demyelinated axons precipitated by normally innocuous physiological processes. Suppl Clin Neurophysiol 57:191–194CrossRef
Buzsáki G (1984) Feedforward inhibition in the hippocampal-formation. Prog Neurobiol 22(2):131–153
Choi U, Wang H, Hu M et al (2021) Presynaptic coupling by electrical synapses coordinates a rhythmic behavior by synchronizing the activities of a neuron pair. Proc Natl Acad Sci USA 118(20):e2022599118CrossRef
Chomiak T, Hu B (2007) Axonal and somatic filtering of antidromically evoked cortical excitation by simulated deep brain stimulation in rat brain. J Physiol 579(Pt 2):403–412CrossRef
Contreras SA, Schleimer JH, Gulledge AT et al (2021) Activity-mediated accumulation of potassium induces a switch in firing pattern and neuronal excitability type. PLoS Comput Biol 17(5):e1008510CrossRef
Cota VR, Medeiros Dde C, Vilela MR et al (2009) Distinct patterns of electrical stimulation of the basolateral amygdala influence pentylenetetrazole seizure outcome. Epilepsy Behav 14(Suppl 1):26–31CrossRef
Csicsvari J, Hirase H, Czurko A et al (1998) Reliability and state dependence of pyramidal cell-interneuron synapses in the hippocampus: An ensemble approach in the behaving rat. Neuron 21(1):179–189CrossRef
Davies CH, Davies SN, Collingridge GL (1990) Paired-pulse depression of monosynaptic GABA-mediated inhibitory postsynaptic responses in rat hippocampus. J Physiol 424:513–531CrossRef
Debanne D, Campanac E, Bialowas A et al (2011) Axon physiology. Physiol Rev 91(2):555–602CrossRef
Deniau JM, Degos B, Bosch C et al (2010) Deep brain stimulation mechanisms: beyond the concept of local functional inhibition. Eur J Neurosci 32(7):1080–1091CrossRef
Di Miceli M, Husson Z, Ruel P et al (2020) In silico hierarchical clustering of neuronal populations in the rat ventral tegmental area based on extracellular electrophysiological properties. Front Neural Circuits 14:51CrossRef
Dudok B, Szoboszlay M, Paul A et al (2021) Recruitment and inhibitory action of hippocampal axo-axonic cells during behavior. Neuron 109(23):3838-3850.e8CrossRef
Eusebio A, Thevathasan W, Gaynor LD et al (2011) Deep brain stimulation can suppress pathological synchronisation in Parkinsonian patients. J Neurol Neurosurg Psychiatry 82(5):569–573CrossRef
Feng Z, Wang Z, Guo Z et al (2017) High frequency stimulation of afferent fibers generates asynchronous firing in the downstream neurons in hippocampus through partial block of axonal conduction. Brain Res 1661:67–78CrossRef
Feng Z, Yu Y, Guo Z et al (2014) High frequency stimulation extends the refractory period and generates axonal block in the rat hippocampus. Brain Stimul 7(5):680–689CrossRef
Feng Z, Zheng X, Yu Y et al (2013) Functional disconnection of axonal fibers generated by high frequency stimulation in the hippocampal CA1 region in-vivo. Brain Res 1509:32–42CrossRef
Fields RD (2008) White matter matters. Sci Am 298(3):42–49
Florence G, Sameshima K, Fonoff ET et al (2016) Deep brain stimulation: more complex than the inhibition of cells and excitation of fibers. Neuroscientist 22(4):332–345CrossRef
Gale JT, Amirnovin R, Williains ZM et al (2008) From symphony to cacophony: pathophysiology of the human basal ganglia in Parkinson disease. Neurosci Biobehav Rev 32(3):378–387CrossRef
Gatev P, Darbin O, Wichmann T (2006) Oscillations in the basal ganglia under normal conditions and in movement disorders. Mov Disord 21(10):1566–1577CrossRef
Guo Z, Feng Z, Wang Y et al (2018) Simulation study of intermittent axonal block and desynchronization effect induced by high-frequency stimulation of electrical pulses. Front Neurosci 12:858CrossRef
Gurbani S, Chayasirisobhon S, Cahan L et al (2016) Neuromodulation therapy with vagus nerve stimulation for intractable epilepsy: a 2-year efficacy analysis study in patients under 12 years of age. Epilepsy Res Treat 2016:9709056
Hahn PJ, Durand DM (2001) Bistability dynamics in simulations of neural activity in high-extracellular-potassium conditions. J Comput Neurosci 11(1):5–18CrossRef
Hammond C, Bergman H, Brown P (2007) Pathological synchronization in Parkinson’s disease: networks, models and treatments. Trends Neurosci 30(7):357–364CrossRef
Hashimoto T, Elder CM, Okun MS et al (2003) Stimulation of the subthalamic nucleus changes the firing pattern of pallidal neurons. J Neurosci 23(5):1916–1923CrossRef
Holsheimer J, Demeulemeester H, Nuttin B et al (2000) Identification of the target neuronal elements in electrical deep brain stimulation. Eur J Neurosci 12(12):4573–4577CrossRef
Iremonger KJ, Anderson TR, Hu B et al (2006) Cellular mechanisms preventing sustained activation of cortex during subcortical high-frequency stimulation. J Neurophysiol 96(2):613–621CrossRef
Jensen AL, Durand DM (2009) High frequency stimulation can block axonal conduction. Exp Neurol 220(1):57–70CrossRef
Jensen MS, Azouz R, Yaari Y (1996) Spike after-depolarization and burst generation in adult rat hippocampal CA1 pyramidal cells. J Physiol 492(Pt 1):199–210CrossRef
Jiruska P, de Curtis M, Jefferys JGR et al (2013) Synchronization and desynchronization in epilepsy: controversies and hypotheses. J Physiol 591(4):787–797CrossRef
Jobst B (2010) Brain stimulation for surgical epilepsy. Epilepsy Res 89(1):154–161CrossRef
Juckett L, Saffari TM, Ormseth B et al (2022) The effect of electrical stimulation on nerve regeneration following peripheral nerve injury. Biomolecules 12(12):1856CrossRef
Kepecs A, Fishell G (2014) Interneuron cell types are fit to function. Nature 505(7483):318–326CrossRef
Kim E, Owen B, Holmes WR et al (2012) Decreased afferent excitability contributes to synaptic depression during high-frequency stimulation in hippocampal area CA1. J Neurophysiol 108(7):1965–1976CrossRef
Kloosterman F, Peloquin P, Leung LS (2001) Apical and basal orthodromic population spikes in hippocampal CA1 in vivo show different origins and patterns of propagation. J Neurophysiol 86(5):2435–2444
Knowles WD, Schwartzkroin PA (1981) Local circuit synaptic interactions in hippocampal brain slices. J Neurosci 1(3):318–322CrossRef
Kuncel AM, Cooper SE, Wolgamuth BR et al (2006) Clinical response to varying the stimulus parameters in deep brain stimulation for essential tremor. Mov Disord 21(11):1920–1928CrossRef
Lam CM, Latif U, Sack A et al (2023) Advances in spinal cord stimulation. Bioengineering 10(2):185CrossRef
Laxton AW, Tang-Wai DF, McAndrews MP et al (2010) A phase I trial of deep brain stimulation of memory circuits in Alzheimer’s disease. Ann Neurol 68(4):521–534CrossRef
Le Van QM, Navarro V, Martinerie J et al (2003) Toward a neurodynamical understanding of ictogenesis. Epilepsia 44(Suppl 12):30–43
Lee DJ, Lozano CS, Dallapiazza RF et al (2019) Current and future directions of deep brain stimulation for neurological and psychiatric disorders. J Neurosurg 131(2):333–342CrossRef
Li MCH, Cook MJ (2018) Deep brain stimulation for drug-resistant epilepsy. Epilepsia 59(2):273–290CrossRef
Llinás RR, Ribary U, Jeanmonod D et al (1999) Thalamocortical dysrhythmia: a neurological and neuropsychiatric syndrome characterized by magnetoencephalography. Proc Natl Acad Sci USA 96(26):15222–15227CrossRef
Lopes da Silva F, Blanes W, Kalitzin SN et al (2003) Epilepsies as dynamical diseases of brain systems: basic models of the transition between normal and epileptic activity. Epilepsia 44(Suppl 12):72–83CrossRef
Lozano AM, Lipsman N, Bergman H et al (2019) Deep brain stimulation: Current challenges and future directions. Nat Rev Neurol 15(3):148–160CrossRef
McCairn KW, Turner RS (2009) Deep brain stimulation of the globus pallidus internus in the Parkinsonian primate: local entrainment and suppression of low-frequency oscillations. J Neurophysiol 101(4):1941–1960CrossRef
McConnell GC, So RQ, Hilliard JD et al (2012) Effective deep brain stimulation suppresses low-frequency network oscillations in the basal ganglia by regularizing neural firing patterns. J Neurosci 32(45):15657–15668CrossRef
McGee MJ, Amundsen CL, Grill WM (2015) Electrical stimulation for the treatment of lower urinary tract dysfunction after spinal cord injury. J Spinal Cord Med 38(2):135–146CrossRef
McIntyre CC, Grill WM (1999) Excitation of central nervous system neurons by nonuniform electric fields. Biophys J 76(2):878–888CrossRef
Medeiros Dde C, Moraes MF (2014) Focus on desynchronization rather than excitability: a new strategy for intraencephalic electrical stimulation. Epilepsy Behav 38:32–36CrossRef
Meeks JP, Jiang X, Mennerick S (2005) Action potential fidelity during normal and epileptiform activity in paired soma–axon recordings from rat hippocampus. J Physiol 566(2):425–441CrossRef
Meeks JP, Mennerick S (2004) Selective effects of potassium elevations on glutamate signaling and action potential conduction in hippocampus. J Neurosci 24(1):197–206CrossRef
Nowak LG, Bullier J (1998) Axons, but not cell bodies, are activated by electrical stimulation in cortical gray matter. Exp Brain Res 118(4):489–500CrossRef
Pelkey KA, Chittajallu R, Craig MT et al (2017) Hippocampal GABAergic inhibitory interneurons. Physiol Rev 97(4):1619–1747CrossRef
Popovych O, Tass PA (2014) Control of abnormal synchronization in neurological disorders. Front Neurol 5:268CrossRef
Quian Quiroga R, Kreuz T, Grassberger P (2002) Event synchronization: a simple and fast method to measure synchronicity and time delay patterns. Phys Rev E Stat Nonlin Soft Matter Phys 66(4 Pt 1):041904MathSciNetCrossRef
Radivojevic M, Franke F, Altermatt M et al (2017) Tracking individual action potentials throughout mammalian axonal arbors. eLife 6:e30198
Rampp S, Stefan H (2006) Fast activity as a surrogate marker of epileptic network function? Clin Neurophysiol 117(10):2111–2117CrossRef
Ranck JB (1975) Which elements are excited in electrical stimulation of mammalian central nervous system: a review. Brain Res 98(3):417–440CrossRef
Rush AJ, Marangell LB, Sackeim HA et al (2005) Vagus nerve stimulation for treatment-resistant depression: a randomized, controlled acute phase trial. Biol Psychiatry 58(5):347–354CrossRef
Silberstein SD, Yuan H, Najib U et al (2020) Non-invasive vagus nerve stimulation for primary headache: a clinical update. Cephalalgia 40(12):1370–1384CrossRef
Theoret Y, Brown A, Fleming SP et al (1984) Hippocampal field potential: a microcomputer aided comparison of amplitude and integral. Brain Res Bull 12(5):589–595CrossRef
Traub RD, Draguhn A, Whittington MA et al (2002) Axonal gap junctions between principal neurons: a novel source of network oscillations, and perhaps epileptogenesis. Rev Neurosci 13(1):1–30CrossRef
Udupa K, Chen R (2015) The mechanisms of action of deep brain stimulation and ideas for the future development. Prog Neurobiol 133:27–49CrossRef
Vitek JL (2002) Mechanisms of deep brain stimulation: Excitation or inhibition. Mov Disord 17(S3):S69–S72CrossRef
Vonck K, Sprengers M, Carrette E et al (2013) A decade of experience with deep brain stimulation for patients with refractory medial temporal lobe epilepsy. Int J Neural Syst 23(1):1250034CrossRef
Wang Z, Feng Z, Wei X (2018) Axonal stimulations with a higher frequency generate more randomness in neuronal firing rather than increase firing rates in rat hippocampus. Front Neurosci 12:783CrossRef
Wang Z, Feng Z, Yuan Y et al (2024) Dynamics of neuronal firing modulated by high-frequency electrical pulse stimulations at axons in rat hippocampus. J Neural Eng 21(2):026025CrossRef
Wilson CJ, Beverlin B, Netoff T (2011) Chaotic desynchronization as the therapeutic mechanism of deep brain stimulation. Front Syst Neurosci 5:50CrossRef
Wingeier B, Tcheng T, Koop MM et al (2006) Intra-operative STN DBS attenuates the prominent beta rhythm in the STN in Parkinson’s disease. Exp Neurol 197(1):244–251CrossRef
Wu X, Shuai J (2012) Multistability in a neuron model with extracellular potassium dynamics. Phys Rev E Stat Nonlin Soft Matter Phys 85(6 Pt 1):061911CrossRef
Xu Y, Feng Z, Yuan Y et al (2023) Effects on somata by high-frequency electrical stimulation at the axons of hippocampal pyramidal neurons. Prog Biochem Biophys 50(3):561–572 (in Chinese)
Ye X, Feng Z, Wang Z et al (2022) Activating interneurons in local inhibitory circuits by high-frequency stimulations at the efferent fibers of pyramidal neurons in rat hippocampal CA1 region. Brain Sci 12(10):1350CrossRef
Yuan Y (2023) Investigation of firing patterns and post-stimulation effects of pyramidal neurons under axonal high-frequency stimulation in rat hippocampus. PhD dissertation, Zhejiang University, Hangzhou (in Chinese)
Yuan Y, Feng Z, Wang Z (2025) Cluster neuronal firing induced by uniform pulses of high-frequency stimulation on axons in rat hippocampus. IEEE Trans Biomed Eng 72(3):1108–1120CrossRef
Yuan Y, Feng Z, Yang G et al (2022) Suppression of neuronal firing following antidromic high-frequency stimulations on the neuronal axons in rat hippocampal CA1 region. Front Neurosci 16:881426CrossRef
Yue C, Remy S, Su H et al (2005) Proximal persistent Na+ channels drive spike afterdepolarizations and associated bursting in adult CA1 pyramidal cells. J Neurosci 25(42):9704–9720CrossRef
Zheng F, Lammert K, Nixdorf-Bergweiler BE et al (2011) Axonal failure during high frequency stimulation of rat subthalamic nucleus. J Physiol 589(Pt 11):2781–2793CrossRef
Zheng L, Feng Z, Hu H et al (2020) The appearance order of varying intervals introduces extra modulation effects on neuronal firing through non-linear dynamics of sodium channels during high-frequency stimulations. Front Neurosci 14:397CrossRef
