Advertisement

kHz-frequency electrical stimulation selectively activates small, unmyelinated vagus afferents

Open AccessPublished:October 10, 2022DOI:https://doi.org/10.1016/j.brs.2022.09.015

      Highlights

      • Selective activation of small visceral vagus afferents has not been achieved. .
      • kHz stimulation (kS) activates motor and mostly sensory vagal neurons in brainstem. .
      • kS activates afferent while blocking larger efferent fibers in rats and mice. .
      • Afferent fiber selectivity is frequency- and intensity-dependent. .
      • Selectivity is explained by how fiber size and myelin shape Na-channel responses. .

      Abstract

      Background

      Vagal reflexes regulate homeostasis in visceral organs and systems through afferent and efferent neurons and nerve fibers. Small, unmyelinated, C-type afferents comprise over 80% of fibers in the vagus and form the sensory arc of autonomic reflexes of the gut, lungs, heart and vessels and the immune system. Selective bioelectronic activation of C-afferents could be used to mechanistically study and treat diseases of peripheral organs in which vagal reflexes are involved, but it has not been achieved.

      Methods

      We stimulated the vagus in rats and mice using trains of kHz-frequency stimuli. Stimulation effects were assessed using neuronal c-Fos expression, physiological and nerve fiber responses, optogenetic and computational methods.

      Results

      Intermittent kHz stimulation for 30 min activates specific motor and, preferentially, sensory vagus neurons in the brainstem. At sufficiently high frequencies (>5 kHz) and at intensities within a specific range (7–10 times activation threshold, T, in rats; 15-25 × T in mice), C-afferents are activated, whereas larger, A- and B-fibers, are blocked. This was determined by measuring fiber-specific acute physiological responses to kHz stimulus trains, and by assessing fiber excitability around kHz stimulus trains through compound action potentials evoked by probing pulses. Aspects of selective activation of C-afferents are explained in computational models of nerve fibers by how fiber size and myelin shape the response of sodium channels to kHz-frequency stimuli.

      Conclusion

      kHz stimulation is a neuromodulation strategy to robustly and selectively activate vagal C-afferents implicated in physiological homeostasis and disease, over larger vagal fibers.

      Keywords

      1. Introduction

      Physiological homeostasis in organisms is maintained by the endocrine, immune and nervous system. The autonomic nervous system regulates internal organs and systems through an interconnected network of reflexes. Autonomic reflexes comprise sensory neurons that detect changes in physiological parameters and motor neurons that regulate the function of organs. The vagus nerve is the main conduit of sensory (afferent) and motor (efferent) signals communicated between the brain and body organs. The majority (>80%) of vagus nerve fibers are C-type afferents [
      • Agostoni E.
      • Chinnock J.E.
      • De Daly M.B.
      • Murray J.G.
      Functional and histological studies of the vagus nerve and its branches to the heart, lungs and abdominal viscera in the cat.
      ]. C-afferents are the smallest (0.2–1.5 μm diameter) and slowest (0.2–2 m/s conduction speed) fibers in the vagus and constitute the peripheral branch of axons from sensory neurons in the nodose ganglion. They form the sensory arc of numerous autonomic reflexes involving the gut [
      • Grabauskas G.
      • Owyang C.
      Plasticity of vagal afferent signaling in the gut.
      ,
      • Carabotti M.
      • Scirocco A.
      • Maselli M.A.
      • Severi C.
      The gut-brain axis: interactions between enteric microbiota, central and enteric nervous systems.
      ,
      • Teratani T.
      • Mikami Y.
      • Nakamoto N.
      • Suzuki T.
      • Harada Y.
      • Okabayashi K.
      • et al.
      The liver-brain-gut neural arc maintains the Treg cell niche in the gut.
      ], lungs [
      • Undem B.J.
      • Kollarik M.
      The role of vagal afferent nerves in chronic obstructive pulmonary disease.
      ,
      • Kubin L.
      • Alheid G.F.
      • Zuperku E.J.
      • McCrimmon D.R.
      Central pathways of pulmonary and lower airway vagal afferents.
      ], heart and vessels [
      • Thoren P.N.
      • Donald D.E.
      • Shepherd J.T.
      Role of heart and lung receptors with nonmedullated vagal afferents in circulatory control.
      ] and the immune system [
      • Silverman H.A.
      • Chen A.
      • Kravatz N.L.
      • Chavan S.S.
      • Chang E.H.
      Involvement of neural transient receptor potential channels in peripheral inflammation.
      ,
      • Chavan S.S.
      • Pavlov V.A.
      • Tracey K.J.
      Mechanisms and therapeutic relevance of neuro-immune communication.
      ], and transmit mechanical, chemical, thermal and inflammatory signals from the respective visceral-sensory receptors to the nucleus tractus solitarius of the brainstem [
      • Prescott S.L.
      • Liberles S.D.
      Internal senses of the vagus nerve.
      ].
      Altering activity of vagus C-afferents with bioelectronic devices represents a method to study autonomic reflexes and a therapeutic opportunity for treating cardiovascular [
      • Fallen E.L.
      Vagal afferent stimulation as a cardioprotective strategy? Introducing the concept.
      ,
      • Gronda E.
      • Francis D.
      • Zannad F.
      • Hamm C.
      • Brugada J.
      • Vanoli E.
      Baroreflex activation therapy: a new approach to the management of advanced heart failure with reduced ejection fraction.
      ], gastrointestinal [
      • Payne S.C.
      • Furness J.B.
      • Stebbing M.J.
      Bioelectric neuromodulation for gastrointestinal disorders: effectiveness and mechanisms.
      ,
      • Browning K.N.
      • Verheijden S.
      • Boeckxstaens G.E.
      The vagus nerve in appetite regulation, mood, and intestinal inflammation.
      ] and inflammatory diseases [
      • Silverman H.A.
      • Chen A.
      • Kravatz N.L.
      • Chavan S.S.
      • Chang E.H.
      Involvement of neural transient receptor potential channels in peripheral inflammation.
      ,
      • Komegae E.N.
      • Farmer D.G.S.
      • Brooks V.L.
      • McKinley M.J.
      • McAllen R.M.
      • Martelli D.
      Vagal afferent activation suppresses systemic inflammation via the splanchnic anti-inflammatory pathway.
      ] in which sensory signaling through C-afferents has been implicated. However, it remains largely unexplored, partly because of the lack of a selective electrical stimulus for C-afferent fibers that spares other fibers in the vagus. In 1935, Bugnard and Hill observed that electrical nerve stimulation with trains of stimuli at frequencies above 2 kHz decreased nerve responses [
      • Bugnard L.
      • Hill A.V.
      Electric excitation of the fin nerve of sepia.
      ], and Cattell and Gerard discovered that the effect was due to suppressed nerve conduction [
      • Cattell M.
      • Gerard R.W.
      The "inhibitory" effect of high-frequency stimulation and the excitation state of nerve.
      ]. This stimulation strategy, termed kHz-frequency electrical stimulation, has been extensively used to block conduction in fibers of somatic and autonomic nerves [
      • Ling D.
      • Luo J.
      • Wang M.
      • Cao X.
      • Chen X.
      • Fang K.
      • et al.
      Kilohertz high-frequency alternating current blocks nerve conduction without causing nerve damage in rats.
      ,
      • Pena E.
      • Pelot N.A.
      • Grill W.M.
      Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance.
      ,
      • Patel Y.A.
      • Butera R.J.
      Challenges associated with nerve conduction block using kilohertz electrical stimulation.
      ,
      • Kilgore K.L.
      • Bhadra N.
      Reversible nerve conduction block using kilohertz frequency alternating current.
      ], including selective block of vagus afferents [
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ]. The clinical importance of blocking C-afferents has been established by studies in which kHz stimulation of the abdominal vagus was associated with clinically-significant weight loss in morbidly [
      • Sarr M.G.
      • Billington C.J.
      • Brancatisano R.
      • Brancatisano A.
      • Toouli J.
      • Kow L.
      • et al.
      The EMPOWER study: randomized, prospective, double-blind, multicenter trial of vagal blockade to induce weight loss in morbid obesity.
      ,
      • Apovian C.M.
      • Shah S.N.
      • Wolfe B.M.
      • Ikramuddin S.
      • Miller C.J.
      • Tweden K.S.
      • et al.
      Two-year outcomes of vagal nerve blocking (vBloc) for the treatment of obesity in the ReCharge trial.
      ] and moderately obese individuals [
      • Apovian C.M.
      • Shah S.N.
      • Wolfe B.M.
      • Ikramuddin S.
      • Miller C.J.
      • Tweden K.S.
      • et al.
      Two-year outcomes of vagal nerve blocking (vBloc) for the treatment of obesity in the ReCharge trial.
      ,
      • Morton J.M.
      • Shah S.N.
      • Wolfe B.M.
      • Apovian C.M.
      • Miller C.J.
      • Tweden K.S.
      • et al.
      Effect of vagal nerve blockade on moderate obesity with an obesity-related comorbid condition: the ReCharge study.
      ,
      • Shikora S.A.
      • Wolfe B.M.
      • Apovian C.M.
      • Anvari M.
      • Sarwer D.B.
      • Gibbons R.D.
      • et al.
      Sustained weight loss with vagal nerve blockade but not with sham: 18-month results of the ReCharge trial.
      ].
      Accordingly, we were surprised to observe that kHz stimulation of the rodent vagus nerve activated specific motor and, preferentially, sensory vagus neurons in the brainstem. Using electrical and optogenetic stimulation, IHC imaging, physiological and computational methods, we show here that intermittent kHz stimulation of the vagus preferentially activates C-afferents over larger fibers in the vagus, in a frequency- and intensity-dependent manner, in rats and mice.

      2. Methods

      2.1 Animal preparation, anesthesia, physiological instrumentation

      Forty-two adult male Sprague Dawley rats (age 2–5 months and weight between 300 and 550 gm) and eleven male C57BL/6 mice (2–4 months and weight between 25 and 30 gm) from Charles River and Jackson Laboratory respectively were used in the study under the approval of the Institutional Animal Care and Use Committee at The Feinstein Institutes for Medical Research. Rodents were anesthetized using isoflurane (induction at 4% and maintenance at 1.5–2%) and medical oxygen; anesthesia was maintained throughout the experiment. Body temperature was measured with a rectal probe and maintained between 36.5 and 37.5 °C using a heating pad (78914731, Patterson Scientific) connected to a warm water recirculator (TP-700 T, Stryker). ECG (Fig. 2A–c) was recorded by using 3-lead needle electrodes subcutaneously on the limbs and amplified using a commercial octal bio-amplifier (FE238, ADI). Breathing was monitored by using a temperature probe placed outside of the nostrils along with a bridge amplifier (FE221, ADI); the probe reported changes in air temperature during breathing movements: drop in temperature during inhalation and rise during exhalation (Fig. 2A–b). All physiological signals were first digitized and then acquired at 1-kHz (PowerLab 16/35, ADI) and visualized on LabChart v8 (all from ADInstruments Inc).

      2.2 Surgical preparation and vagus electrode placement

      To expose the cervical vagus nerve (cVN) in the rat model, a midline 3 cm skin incision was given on the neck. Salivary glands were separated, and muscles were retracted to reach the carotid bundle. Under a dissecting microscope, the right cVN was isolated first at the caudal end of nerve and then at rostral end of nerve. The middle portion, between the two isolated sites was left intact within carotid bundle to minimize the extent of surgical manipulation and trauma to the nerve. After isolation of the nerve, a pair of custom-made, tripolar cuff electrodes was placed on the caudal and rostral sites relative to omohyoid muscle (Fig. 2A–a). The cuff electrodes were made using a 2-layer polyimide substrate (total material thickness of 14 μm, total length along the nerve ∼2 mm) and sputter-deposited iridium oxide contacts for low electrode impedances and stable stimulation characteristics [
      • Negi S.
      • Bhandari R.
      • Rieth L.
      • Solzbacher F.
      In vitro comparison of sputtered iridium oxide and platinum-coated neural implantable microelectrode arrays.
      ,
      • Negi S.
      • Bhandari R.
      • Rieth L.
      • Van Wagenen R.
      • Solzbacher F.
      Neural electrode degradation from continuous electrical stimulation: comparison of sputtered and activated iridium oxide.
      ,
      • Levy T.J.
      • Ahmed U.
      • Tsaava T.
      • Chang Y.C.
      • Lorraine P.J.
      • Tomaio J.N.
      • et al.
      An impedance matching algorithm for common-mode interference removal in vagus nerve recordings.
      ]. Electrode contacts had dimensions of 1418 × 167 μm2 with an edge-to-edge spacing of 728 μm and center-to-center spacing of 895 μm. Typical individual electrode impedances in saline ranged from 0.5 to 1.5 kΩ, measured at 1 kHz. The distance between the stimulating electrode (center contact of tripolar cuff) to the most proximal recording electrode on the nerve was measured roughly 5–6 mm. Silicone elastomer (Kwiksil, World Precision Instruments) was placed around the cuff to minimize current leakage during stimulation. In the mouse model, all surgical procedures were identical except the left cVN was targeted, and only one stimulating electrode was placed due to dimensional limit. In addition, for direct laryngeal muscle measurement, the thyroid cartilage was exposed by separating the sternohyoid muscle at the midline using blunt dissection. Using a 29G insulin syringe, a shallow slit was made in the thyroid cartilage just lateral and inferior to the laryngeal prominence. With the needle bevel facing up, the two PTFE-coated platinum-iridium wires were carefully inserted into the underlying laryngeal muscles through the slit guided by the syringe needle.

      2.3 Vagus nerve recording and stimulation

      Neural activity from each contact on the recording electrode was amplified, digitized (30 kS/s, 16bit resolution) and filtered (60-Hz notch), using a 32-channel RHS2000 stim/record headstage and 128ch Stimulation/Recording controller (Intan Technologies); recordings were single-ended, relative to a ground lead placed in the salivary gland. Nerve stimulation was delivered in constant current mode as trains of pulses using an STG4008 stimulus generator (Multichannel Systems). A custom-made buffer amplifier was used to record the induced voltage on the electrode during stimulation. The 2-two stage buffer consistsconsisted of an isolation stage and an attenuation stage with variable gain selectable by the user. The inputs to the buffer were placed in parallel with the stimulation source and the stimulation electrodes. The output of the buffer was connected to the power labs acquisition system and the artifact signal was recorded along with a TTL trigger signal from the stimulator to facilitate syncing during subsequent signal processing.
      To achieve fiber-selective VNS, we used waveform manipulation, by altering the waveform of single pulses in a train, and frequency manipulation, by altering the stimulation frequency of pulses in a train. For experiments related to waveform manipulation, stimulation waveforms were composed of monophasic pulses with varying pulse width, intensity, polarity, and shape. Monophasic pulses were used because they yield lower thresholds and simpler stimulus artifact shapes. Even though monophasic pulses are not charge-balanced, it is unlikely that during acute experiments with relatively low pulsing frequencies, significant charge build-up occurs. Single pulses at 1Hz (30-s on and 10-s off) were used to assess nerve fiber responses. Stimulus trains of 10-s duration, 30Hz frequency were delivered to assess physiological responses. For experiments related to frequency manipulation, focusing on kHz stimulation, all stimuli were delivered as symmetric biphasic square pulse, in 10-s-long trains with varying frequency, pulse width, and intensity, in random order. To assess fiber excitability locally at the site of delivery of kHz frequency stimulation, we used a single stimulating electrode to deliver a 10-s-long kHz train, briefly interrupted once every second by a single probing, monophasic square pulse. Each probing pulse elicits an evoked CAP recorded through a second recording electrode, placed at a known distance from the stimulating electrode. The probing pulse is either 100μs or 600μs long, to elicit A/B- or C-fiber eCAPs. Around each probing pulse, kHz stimulation is discontinued for 30-ms (5 ms before, 25 ms after the pulse's onset), to allow time for compiling evoked CAPs free of electrical artifacts. The addition of the periodic probing pulses does not alter the physiological responses to kHz trains (Fig. S4). The stimulation configuration was tripolar (cathode-center or cathode-corner) as it outperforms bipolar configuration in terms of protection of current leakage for all experiments (Fig. S7). There were at least 15-s long pauses between successive trains to ensure that physiological measurements had reached a steady state before a new train was delivered.
      For optogenetic stimulation, ChAT-IRES-Cre (#006410), Vglut2-IRES-Cre (#016963), and Ai32 ChR2-eYFP (#024109) mice were purchased from the Jackson Laboratory and crossed to produce ChAT-ChR2-eYFP and Vglut2-ChR2-eYFP mice. 8- to 16-week-old mice were anesthetized and the vagus nerve exposed as described earlier. A custom-made optical cuff, consisting of a blue LED light source (XLAMP XQ-E, Cree LED) in a molded silicone enclosure (Nusil MED-4211), was placed on the left cervical vagus nerve and connected to a stimulus generator (Multichannel Systems). Optogenetic stimulation was delivered using 10-sec stimulus trains of 10 ms pulse width and 30 Hz frequency while recording heart rate and breathing rate responses. The breed of mice used in these experiments carry the H134R mutation, which allows following up to 73 Hz. Optical stimulus intensity was varied by gradually increasing the voltage driving the LED, with threshold of visual perception around 2.3V.

      2.4 Definition of threshold

      Threshold is defined as either neural threshold (T) or physiological threshold (PT), depending on the experiment. In all experiments with neural recording and mice with EMG recordings, we initially determined T as the smallest stimulus intensity of a 100-μs square pulse that elicited a discernible raw eCAPs (or EMG in mice experiment) at the recording electrode. In all experiments, we determined PT, as the smallest stimulus intensity (100-μs pulse) which elicited a visible (5–10%) heart rate/respiratory change, whichever was smaller (usually 3 or 4 × T, in agreement with [
      • Ahmed U.
      • Chang Y.C.
      • Lopez M.F.
      • Wong J.
      • Datta-Chaudhuri T.
      • Rieth L.
      • et al.
      Implant- and anesthesia-related factors affecting cardiopulmonary threshold intensities for vagus nerve stimulation.
      ]) In a given experiment using T, subthreshold or suprathreshold intensities refers to intensity values above or below the T; similarly for PT. Detailed information on the types of threshold used in each experiment can be found in Table S1 and S2 [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ].

      2.5 Identification and analysis of neural and EMG signals

      Raw nerve signal traces from both electrodes were filtered using a digital, 2nd order Butterworth high-pass filter with cut-off frequency at 1Hz to remove the DC component. eCAPs elicited from individual pulses or from trains of pulses, were extracted, by averaging individual sweeps of nerve recording traces around the onset of pulses (waveform manipulation experiments) or probing pulse (frequency manipulation experiments). Stimulation artifact was suppressed offline for waveform manipulation experiment by a recently proposed method which subtracts the trace of the stimulation electrode voltage from the eCAPs with proper template matching and an edge effect removal algorithm [
      • Chang Y.
      • Ahmed U.
      • Tomaio J.N.
      • Rieth L.
      • Datta-Chaudhuri T.
      • Zanos S.
      Extraction of evoked compound nerve action potentials from vagus nerve recordings.
      ]. For frequency manipulation, due to the saturation of artifact voltage buffer, same artifact removal algorithm has not been applied.
      Given the rough estimation of distance between the recording and stimulation electrodes (5–6 mm), we fine tune the distance in analysis so that the latency windows can align well with the A-, B- and C-fiber prominent peaks with pre-defined conduction velocity ranges for each fiber type (A: 5–120 m/s; B: 2–8 m/s; C: 0.1–0.8 m/s) [
      • L Parker J.
      • H Shariati N.
      • M Karantonis D.
      Electrically evoked compound action potential recording in peripheral nerves.
      ]. Fig. 2A–d shows representative eCAPs, including activity of different fiber types. To quantify engagement of laryngeal muscles by VNS, we recorded EMG directly from laryngeal muscles in mice [
      • Abbas A.
      • Mughrabi I.T.
      • Zanos S.
      Laryngeal electromyography to estimate A-fiber engagement by vagal stimuli in mice.
      ] or extracted EMG signals from recordings of nerve CAPs, as described in previous reports [
      • Ahmed U.
      • Chang Y.C.
      • Lopez M.F.
      • Wong J.
      • Datta-Chaudhuri T.
      • Rieth L.
      • et al.
      Implant- and anesthesia-related factors affecting cardiopulmonary threshold intensities for vagus nerve stimulation.
      ,
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ,
      • Chang Y.
      • Ahmed U.
      • Tomaio J.N.
      • Rieth L.
      • Datta-Chaudhuri T.
      • Zanos S.
      Extraction of evoked compound nerve action potentials from vagus nerve recordings.
      ]. To extract EMG, signals from 2 contacts in the recording electrode were collected (solid black and dashed red traces in Fig. 2A–d). For the given electrode spacing, A- and B-fiber eCAPs had short latencies (<3 ms, red and green windows), while slower C-fiber eCAPs occurred at longer latencies (>6 ms, yellow window) [
      • Chang Y.
      • Ahmed U.
      • Tomaio J.N.
      • Rieth L.
      • Datta-Chaudhuri T.
      • Zanos S.
      Extraction of evoked compound nerve action potentials from vagus nerve recordings.
      ]. To discriminate C-fiber components from EMG, we reasoned that C-fiber volleys should show a latency difference between the proximal and distal recording contact, spaced apart by a distance of 895 μm, of 1–2 ms, whereas EMG signals should occur simultaneously on both recording contacts [
      • Chang Y.
      • Ahmed U.
      • Tomaio J.N.
      • Rieth L.
      • Datta-Chaudhuri T.
      • Zanos S.
      Extraction of evoked compound nerve action potentials from vagus nerve recordings.
      ] (Fig. 2A–d, grey window), with time window around 2–6 ms, identified with neuromuscular junction blocking agent in our previous study [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ].

      2.6 Analysis of physiological signals

      We computed the magnitude of EMG responses from stimulus-triggered averages of eCAPs (in rats) or of laryngeal EMGs (in mice) averaged as the peak-to-trough amplitude of the (typically biphasic) response within the EMG latency window (Fig. 2A–d, grey window); that amplitude was then normalized by the maximum EMG amplitude in that subject. When probing pulses were used, e.g. in the kHz stimulation experiments, the EMG responses were derived the same way, triggered on the delivery of the but from probing pulses, Using a custom algorithm, ECG peaks corresponding to the R waves were identified, and heart rate (HR) was computed from R-R intervals. We defined stimulus-induced change in HR (ΔHR) as the difference between the mean HR during a 10-s epoch before the onset of the stimulus train and the mean HR during the stimulus train (“VNS”), divided by the mean pre-stimulus HR (Fig. 2A–c). In recordings from the nasal temperature sensor, we identified peaks (end of expiration) and troughs (end of inspiration). We defined the interval between two successive peaks (or two successive troughs) as breathing interval (BI). We defined the stimulus-elicited change in breathing interval (ΔBI) as the difference between the mean pre-stimulus and the mean during-stimulus BI (Fig. 2A–b). For those cases without breath during stimulation period, the breathing interval between last breath of pre-stimulus and first breath post-stimulus was used as mean during-stimulus BI. The measured signals and corresponding derived variables (ECG and ΔHR, and nasal sensor temperature and ΔBI) are shown in Fig. 2A–b and -c respectively. In physiology experiments, the same investigators participated in the experiments and the analysis of experimental data, and were not blinded. All the analyses were performed using MATLAB 2016a software (MathWorks, Natick, MA, USA).

      2.7 Immunohistochemistry

      Rats received VNS intermittently for 30 min (10 s on, 50 s off), and sham group received implant without VNS, were deeply anesthetized with isoflurane and transcardially perfused using 250 ml of 0.9% saline after 1.5 h. The brains were removed immediately and post-fixed for 2 days in 4% paraformaldehyde in 0.1 M PBS. After post fixation, the brainstem was precut by razor blade and sectioned in vibratome with 50 μm thickness (VT1200S, Leica, U.S.A). Sections, acquired from anterior-posterior: 13.4 mm–13.9 mm relative to Bregma, were washed with TBS buffer 3 times for 5 min and subjected to antigen retrieval using 1X SignalStain® Citrate Unmasking Solution (Cell signaling). The citrate buffer was bought to boiling temperature and added to the sections. The well plate was incubated at 85 °C for 10 min. The plate was allowed to cool down and the sections were washed with TBST buffer (1X TBS buffer + 0.025% Tween 20). 5% normal donkey serum and 0.3% Triton-X100 in TBS buffer was used as blocking buffer and the sections were blocked for 1 h at room temp. Sections were first stained with c-Fos and then Choline Acetyltransferase (ChAT) was performed subsequently. c-Fos staining was the done by incubating the sections with the c-Fos antibody (Abcam, ab190289) at 1:1000 dilution in blocking buffer for 3 days placed in shaking incubator at 4 °C. Section were rinsed with the TBST buffer 3 × 5 times and incubated in the donkey anti-rabbit secondary antibody labeled with alexa-fluro 488 (1:500) in blocking buffer for 2 h at room temperature. Sections were washed 3 × 5 min in TBST buffer and incubated with ChAT antibody (Sigma, AB144) at 1:100 dilutions overnight at 4 °C in the blocking buffer. Sections were rinsed the next day with 3 × 5 min in TBS buffer and incubated with anti-goat secondary antibody labeled with alexa fluro 555 (1:200) for 2 h in room temperature. Sections were rinsed in TBST buffer 3 times and incubated with DAPI 1:1000 dilution in TBS buffer for 1 h at room temperature. The section was then rinsed two times with TBS buffer and mounted on to Poly-l-lysine coated glass slides. Cover glass was secured on the top of the sections with VECTASHIELD® PLUS Antifade Mounting Medium (Vector labs, H-1900). Sections from naïve, sham and different VNS treatment groups were processed in parallel.
      DAPI, c-Fos and ChAT expression in the nucleus tractus solitarius (NTS) and dorsal motor nucleus of the vagus nerve (DMV) was captured using all-in-one fluorescence microscope under 20× objective (BZ-×800, Keyence, U.S.A.), with region of interests (medial-lateral: ±2 mm relative to mid sagittal plane, dorsal-ventral: 7.2–8.2 mm relative to brain vertex). After processed by commercial imaging stitching software (BZ-X800 Analyzer), the orientation of the stitched images was first adjusted for cross-animal comparison and consistency. To correct non-uniform illumination and attenuate background, the original image was subtracted from background approximation image estimated from morphological opening method. The ChAT stain was used to identify the DMV region as it is a marker for cholinergic neurons. DAPI is a commonly used stain for cell nuclei; it was used as a positive control for counting c-Fos + cells, but it was not itself quantified in the analysis. Cells expressing c-Fos were then counted bilaterally in 3-4 sections/brain using thresholding, watershed separation, and automatic particle counting tools in ImageJ. In the IHC studies, 3 different investigators performed the physiological experiments, processing of samples and IHC staining, and IHC image analysis; the tissue processing and IHC analysis investigators was blinded with respect to which animals were assigned to what experimental group. All the image post-processing techniques were performed using ImageJ or MATLAB 2016a.

      2.8 CAP, physiological, and neuronal-c-Fos selectivity indices

      To evaluate the fiber selective performance of tested stimulation parameters, for different types of fibers, we defined CAP selectivity indices (CSIs) which aim to maximize target and minimize off-target fiber activity. The CSIs for A-, B-, C-fibers are:
      CSIX=XA+B+CX+ε,forX=A,B,C


      where A, B, C are normalized evoked fiber activity amplitude with respect to their maximums within each animal, and a small constant ε is used to prevent the overflow due to the extremely small fiber activity in the denominator.
      Similar to CSIs, based on an existing relationship/models between fiber activity and physiological response: A-fiber to evoked EMG, B-fiber to HR, and C-fiber to BI (Fig. S2) [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ], we defined physiological selective indices (PSIs) which aim to maximize desired and minimized non-desired physiological effects corresponding to different type of fiber activation. The PSI for A-, B-, C-fibers defined as:
      PSIA=EMG(norm.)ΔHR(norm.)+ΔBI(norm.)+ε


      PSIB=ΔHR(norm.)EMG(norm.)+ΔBI(norm.)+ε


      PSIC=ΔBI(norm.)EMG(norm.)+ΔHR(norm.)+ε


      where EMG(norm.), ΔHR(norm.), and ΔBI(norm.) are normalized physiological responses within each animal, and a small constant ε is used to prevent the overflow, same as CSI. To quantify the performance of PSIs using different stimulation parameters across animals, a 1st order Gaussian model was computed to capture the relationship between computed PSIs and different stimulation intensities.
      Finally, to quantify the immunohistochemistry results in brainstem regions, we further defined neuronal-c-Fos + selectivity index (NcSI) for sensory neurons (mostly related to C-fibers), which is:
      NcSINTS(sensory)=|%ΔcFosNTS%ΔcFosDMV|


      where Δc-Fos in NTS and DMV are normalized with respect to the average number of expressed neurons in corresponding regions in sham group animals.

      2.9 Computational model of vagus fibers

      To explore potential mechanisms for part of the in vivo differential intensity-dependency of kHz stimulation between myelinated and unmyelinated axons, a computational model of vagus fibers was reconstructed using COMSOL Multiphysics v. 5.4 (COMSOL Inc., Burlington, MA) based on our previous modelling studies [
      • Dokos S.
      Modelling organs, tissues, cells and devices using MATLAB and COMSOL Multiphysics.
      ,
      • Lin Q.
      • Shivdasani M.N.
      • Tsai D.
      • Chang Y.C.
      • Jayaprakash N.
      • Zanos S.
      • et al.
      A computational model of functionally-distinct cervical vagus nerve fibers.
      ]. Model parameters defining myelin structure and ion channels were adapted from the McIntyre, Richardson and Grill (MRG), and the Schwarz, Reid and Bostock (SRB) models [
      • Schwarz J.R.
      • Reid G.
      • Bostock H.
      Action-potentials and membrane currents in the human node of ranvier.
      ]. Two nerve fiber subtypes were simulated: myelinated fiber (myelin diameter 5 μm, axon diameter 2.6 μm) and unmyelinated fibers (axon diameter 1.3 μm). All model parameters are listed in Table S3. A step-by-step model reconstruction method of both myelinated and unmyelinated fibers can be found in our supplementary materials.
      As shown in Fig. S14A, the extracellular environment was modeled by a 1000-μm long, 40-μm diameter cylinder (approximated by a homogeneous conductivity of 1 S/m) surrounding the 1D nerve fiber [
      • Dokos S.
      Modelling organs, tissues, cells and devices using MATLAB and COMSOL Multiphysics.
      ,
      • Lin Q.
      • Shivdasani M.N.
      • Tsai D.
      • Chang Y.C.
      • Jayaprakash N.
      • Zanos S.
      • et al.
      A computational model of functionally-distinct cervical vagus nerve fibers.
      ]. Two 50-μm electrodes (edge-edge distance 50 μm) were placed on the surface of the cylinder with the electrode edges forming a 60° angle with the nerve fiber. The first electrode was the cathode and the second was designated as ground. The electrode-fiber distance was set to be 20 μm based on our histology data of electrode-nerve distance (Fig. S15). Different from the in vivo setting, we used a bipolar electrode configuration in the model, given the similar in vivo performance of bipolar and tripolar configurations with regard to activation of functionally distinct nerve fiber activation [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ].
      The stimulus waveform included a wide range of frequencies ranging from 0.1-kHz to 12-kHz sinusoid kHz stimulation, with a duration of 50 ms. We only simulated kHz-frequency stimulation in the modelling studies, i.e. no test pulses were simulated. A no-flux (i.e insulating) boundary condition was implemented for Vi and Ve at the ends of the fiber. The mesh for the myelinated fibers was set to a total of 20 elements for each myelin segment and 3 elements for each node. The mesh for unmyelinated axon was set to a total of 20 elements for each fiber segment. The length of the nodes was set to 1 μm in a myelinated fiber [
      • McIntyre C.C.
      • Richardson A.G.
      • Grill W.M.
      Modeling the excitability of mammalian nerve fibers: influence of afterpotentials on the recovery cycle.
      ]. The mesh resolution was confirmed by the fact that of further increasing the mesh number could not change the simulation results. The node and myelin diameters used in the model were estimated based on histological data from rat cervical nerves [
      • Lin Q.
      • Shivdasani M.N.
      • Tsai D.
      • Chang Y.C.
      • Jayaprakash N.
      • Zanos S.
      • et al.
      A computational model of functionally-distinct cervical vagus nerve fibers.
      ]. The model's predictive ability was validated by in vivo compound nerve action potential recordings from rats [
      • Lin Q.
      • Shivdasani M.N.
      • Tsai D.
      • Chang Y.C.
      • Jayaprakash N.
      • Zanos S.
      • et al.
      A computational model of functionally-distinct cervical vagus nerve fibers.
      ], and experimentally observed spiking activities and conduction velocities from rats, pigs and humans [
      • McIntyre C.C.
      • Richardson A.G.
      • Grill W.M.
      Modeling the excitability of mammalian nerve fibers: influence of afterpotentials on the recovery cycle.
      ,
      • Jayaprakash N.
      • Toth V.
      • Song W.
      • Vardhan A.
      • Levy T.
      • Tomaio J.
      • et al.
      Organ- and function-specific anatomical organization and bioelectronic modulation of the vagus nerve.
      ,
      • Parker J.L.
      • Shariati N.H.
      • Karantonis D.M.
      Electrically evoked compound action potential recording in peripheral nerves.
      ,
      • Metcalfe B.W.
      • Nielsen T.N.
      • Donaldson N.N.
      • Hunter A.J.
      • Taylor J.T.
      First demonstration of velocity selective recording from the pig vagus using a nerve cuff shows respiration afferents.
      ,
      • Baker M.
      • Bostock H.
      • Grafe P.
      • Martius P.
      Function and distribution of three types of rectifying channel in rat spinal root myelinated axons.
      ].
      Node and myelin structures in the model fibers were characterized by partial differential equations (PDEs). Myelin was approximated by a distributed resistance in parallel with a capacitance (Fig. S14B). We approximated the MRG double cable structure by a single-cable model of the vagus nerve to reduce the computational complexity. The membrane dynamics at the node follows SRB formulations. Models for all fiber types shared ion channel parameters (see Tables S3 and S4) but had fiber-specific physical parameters.
      The extracellular potential distribution Ve was calculated using:
      (1ρe(Ve))=0


      where ρe is the extracellular resistivity. The intracellular potential Vi was calculated separately for the myelin and node compartments:
      (rnρn(Vi))+2CnVit=2(iionCnVet)


      (rmyρmy(Vi))+2CmyVit=2CmyVet


      where rn and rmy are the nodal and myelin radius respectively. Membrane potential Vm was determined from the difference between the intracellular and extracellular potentials.

      2.10 Statistical analysis

      Analysis of Covariance (ANCOVA) was used to compare the neural responses (A-, B-, C-), physiological responses (EMG, HR, BI), and proposed CSIs and PSIs for different stimulus manipulations (categorical independent variable) and intensity (continuous independent variable). Linear regression was used to compare the same stimulus parameter with different intensity. One-way analysis of variance (ANOVA) and Tukey's post-hoc tests were used to compare the histological results in brainstem, and two sample t-test was used for corresponding NcSI. Comparisons were deemed statistically significant for p < 0.05 for all analyses. All statistical analyses were performed on MATLAB (Mathworks).

      3. Results

      3.1 Intermittent kHz stimulation activates motor and, preferentially, sensory vagus neurons in the brainstem

      kHz stimulation has been shown to block nerve fibers of different types [
      • Kilgore K.L.
      • Bhadra N.
      Reversible nerve conduction block using kilohertz frequency alternating current.
      ], and several studied investigated its effect on neurons and potential mechanisms of action [
      • Kilgore K.L.
      • Bhadra N.
      Reversible nerve conduction block using kilohertz frequency alternating current.
      ,
      • Kilgore K.L.
      • Bhadra N.
      Nerve conduction block utilising high-frequency alternating current.
      ,
      • Kameneva T.
      • Maturana M.I.
      • Hadjinicolaou A.E.
      • Cloherty S.L.
      • Ibbotson M.R.
      • Grayden D.B.
      • et al.
      Retinal ganglion cells: mechanisms underlying depolarization block and differential responses to high frequency electrical stimulation of ON and OFF cells.
      ]. To determine the effect of kHz stimulation delivered to cervical vagus on the activation level of neurons associated with afferent and efferent vagus fibers, we quantified a marker of neuronal activation, c-Fos expression, in sensory and motor vagus neuronal populations in the brainstem. In anesthetized rats, we delivered for 30 min sham or intermittent VNS (10 s on, 50 s off), comprising either kHz trains (8-kHz frequency, 40μs pulse width, 2 mA intensity) or “standard” frequency trains (30 Hz, 40μs, 2 mA) (Fig. 1A ). We then counted c-Fos+ neurons in the nucleus tractus solitarius (NTS), a sensory region receiving projections from A- and C-afferents, with the latter likely outnumbering the former [
      • Berthoud H.R.
      • Neuhuber W.L.
      Functional and chemical anatomy of the afferent vagal system.
      ,
      • Beaumont E.
      • Campbell R.P.
      • Andresen M.C.
      • Scofield S.
      • Singh K.
      • Libbus I.
      • et al.
      Cervical vagus nerve stimulation augments spontaneous discharge in second- and higher-order sensory neurons in the rat nucleus of the solitary tract.
      ], and in the dorsal motor nucleus of the vagus (DMV), a motor region with cholinergic (ChAT+) cells providing efferent fibers, both myelinated [
      • Neuhuber W.L.
      • Berthoud H.R.
      Functional anatomy of the vagus system - emphasis on the somato-visceral interface.
      ] and unmyelinated [
      • Pelot N.A.
      • Grill W.M.
      Effects of vagal neuromodulation on feeding behavior.
      ], to the vagus (Fig. 1B). Sham VNS was associated with minimal c-Fos expression in NTS and DMV (Fig. 1B–a). To our surprise, after 30 min of intermittent kHz VNS, we observed increased, compared to sham, c-Fos expression, stronger in the ipsilateral sensory, NTS region, and weaker in the motor, DMV region (Fig. 1B–b). As expected, 30-Hz VNS induced strong c-Fos expression in both ipsilateral (to VNS) sensory, NTS, and motor, DMV, regions (Fig. 1B–c). Overall, kHz VNS induces a 1.7-fold increase, compared to sham, in c-Fos expression in NTS, with standard VNS producing a comparable 2.2-fold increase (Fig. 1C); at the same time, kHz stimulation induces a non-significant (0.9-fold) increase in c-Fos in DMV, much smaller than that induced by standard VNS (2.4-fold) (Fig. 1C–D). This preferential activation of NTS over DMV by kHz stimulation is demonstrated by a sensory neuronal-c-Fos + selectivity index (NcSI), defined as c-Fos + cell count ratio of NTS to DMV (Fig. 1E) (Suppl. Figure S1D). Counts of c-Fos + neurons in the sham stimulation condition are moderately greater than naïve animals in NTS, and no different in DMV (Fig. 1C and D, and Suppl. Figure S1). Interestingly, 30Hz VNS causes a moderate increase of c-Fos + cells in contralateral NTS and DMV (0.85- and 1.1-fold, respectively), whereas kHz VNS did not significantly affect contralateral c-Fos expression (Suppl. Figure S1B-D), therefore eliciting a more “lateralized” neuronal response. These results indicate that kHz VNS activates motor and, primarily, sensory brainstem neurons that are associated with efferent and afferent vagus fibers, respectively.
      Fig. 1
      Fig. 1Intermittent kHz-frequency stimulation activates motor and, preferentially, sensory vagus neurons in the brainstem, in the rat model. (A) Time course of experiments to quantify c-Fos-expressing neurons in the brainstem after VNS. (B) Representative immunohistochemistry images of sections across ipsilateral, to VNS, sensory and motor brainstem regions (yellow contours): nucleus tractus solitaries (NTS) and dorsal motor nucleus of the vagus (DMV), identified by DAPI (blue) and ChAT(red) (1st row), each stained for c-Fos (green) (2nd row). The insets (3rd row) show ipsilateral DMV at higher magnification; arrows point to cells positive for c-Fos. (C) c-Fos+ cell numbers (mean ± SE) in ipsilateral NTS, in different groups of animals: naïve (white), sham stimulation (blue), kHz VNS (8-kHz, dark red), 30 Hz VNS (light green). Statistical comparisons between groups use one-way ANOVA and Tukey's post-hoc tests (∗p < 0.05, ∗∗p < 0.005, ∗∗p < 0.0005). (D) Same as (C) but in ipsilateral DMV. (E) NTS neuronal-c-Fos+ selectivity index (NcSI) (mean ± SE), calculated as the fold change of c-Fos+ expression, with respect to sham average, in the NTS region over the one in DMV region, for the kHz stimulation group and the 30 Hz group. Statistical comparison between groups uses 2-sample t-test (p < 0.05).
      Fig. 2
      Fig. 2kHz stimulation differentially controls asynchronous activity and excitability of vagus fibers in the rat model. (A) Schematic of a typical physiology experiment. Nerve electrodes are placed on the cervical vagus nerve for stimulation and recording of nerve activity; a nasal sensor and ECG leads record air flow and ECG, respectively (panel a). During electrical stimulation of the vagus (VNS), changes in breathing interval (BI), typically slowing of breathing or apnea (panel b), and in heart rate (HR), usually bradycardia (panel c), are observed. Single stimuli evoke compound action potentials (eCAPs), extracted from electroneurogram (ENG) (panel d), with early, intermediate and late components representing evoked volleys in A-, B- and C-fibers, respectively [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ]. (B) Optogenetic VNS delivered to ChAT-ChR transgenic mice causes bradycardia (panel a), whereas when delivered to VGluT-ChR mice causes slowing of breathing (panel b), in a dose-dependent manner. (C) Mean (±SE, N = 4 animals) of normalized magnitude of physiological responses (EMG, red; ΔHR, green; ΔBI, yellow), elicited by kHz VNS of different intensities. (Linear Regression, p < 0.05 for intensity, across all physiological responses (EMG, HR, BI)). (D) Representative physiological responses elicited with kHz VNS (8-kHz, 970 ms ON, 30 ms OFF, 10s). (Panel a) As intensity increases, from bottom to top, suppressed EMG, minimal ΔHR and robust ΔBI responses are observed. (Panel b) eCAPs encompassing A- and B-fiber responses were compiled by delivering, intermittently with kHz trains, single “probing” pulses (PP) of 100 us PW during the OFF window (every 1 s, for 10 s). A- or B-fiber evoked activity is mostly blocked at kHz stimulus intensities above ∼5 × T. C-fiber responses are not evoked with short probing pulses. (Panel c) Same as (b), but this time showing eCAPs triggered on 600 μs-long probing pulses, to evoke C-fiber activity. In contrast to A- and B-activity, evoked C-fiber activity is maintained at kHz stimulus intensities 5-10 × T and progressively disappears at higher intensities. The C-fiber activity is shown in monopolar recording from thesingle proximal contact (dash traces), and also in differential recording across 2results from both recording contacts (red traces) to ensure that evoked EMG activity has eliminate minor contribution to it of EMG. (E) Mean (±SE, N = 4 animals) of normalized amplitude of A-, B-, C-fiber evoked activity (red, green and yellow bars, respectively), for kHz stimulation of different intensities; 100 μs probing pulses used for A- and B-fibers, 600 μs probing pulses used for C-fibers (ANOVA, p < 0.05 for intensity, across A-, B- and C-fiber responses).

      3.2 kHz stimulation differentially affects asynchronous activity and excitability of vagus fiber types

      Our finding that kHz VNS increases activity of motor and, preferentially, sensory vagus neurons in the brainstem agrees with prior reports that kHz stimulation is capable of not only achieving nerve conduction block [
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ,
      • Patel Y.A.
      • Butera R.J.
      Challenges associated with nerve conduction block using kilohertz electrical stimulation.
      ] but also excitation [
      • Bowman B.R.
      • McNeal D.R.
      Response of single alpha motoneurons to high-frequency pulse trains. Firing behavior and conduction block phenomenon.
      ] of different fiber types, depending on intensity. To investigate whether fiber engagement by kHz stimulation can explain this finding, we assessed changes in ongoing, asynchronous activity of different fiber types during nerve stimulation. We use the term 'asynchronous' to denote fiber activity that is elicited by trains of stimuli, especially kHz-frequency trains, but is not time-locked to individual stimuli; in contrast, we use term ‘evoked” to denote fiber activity that is triggered by single stimuli and is captured by stimulus-evoked compound action potential responses (eCAPs). Short (10 s-long) trains of kHz stimulation (8-kHz, at a range of intensities) were delivered in the vagus of anesthetized rats while recording several physiological parameters and nerve potentials (Fig. 2A). Changes in laryngeal EMG, HR and BI during VNS are real-time physiological responses driven by and correlating with asynchronous activity in A-, B- and C- fibers, respectively (Suppl. Fig. S2), due to temporal summation of post-synaptic potentials [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ]. Likewise, optogenetic stimulation of cholinergic B-fibers in ChAT-ChR mice causes bradycardia; in VGluT-ChR mice, stimulation of glutamatergic sensory afferents, of which C-fibers constitute the vast majority, slows down breathing, in a dose-dependent manner (Fig. 2B). We used those responses to assess changes in asynchronous fiber activity elicited by VNS [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ], since electrical artifacts during kHz stimulus trains preclude direct recording of fiber potentials from the nerve. At low intensities, kHz stimulation elicits A-fiber-associated EMG (Fig. 2C; 2D-a; Suppl. Video 1D) and B-fiber associated HR responses; with increasing intensity, those responses are progressively suppressed and almost completely abolished at intensities >6–7 times threshold intensity ( × T) (Fig. 2C; Fig. 2D–a; Suppl. Video 1E). In contrast, C-afferent-associated BI responses appear at intensities >6-7 × T, continue increasing up to 15 × T and are eventually blocked above 30 × T (Fig. 2C; Fig. 2D–a; Suppl. Video 1F). The results were replicated in mice, with an intensity window of 10–30 × T (Suppl. Fig. S3). These findings indicate that during kHz stimulation of relatively high intensity, asynchronous activity of C-afferents remains robust, whereas that of A- and B-fibers is suppressed.
      The following are the supplementary data related to this article:
      Our study provides the first in vivo experimental evidence on simultaneous block of larger efferent fibers and activation of smaller afferent fibers by kHz stimuli in the vagus, and confirm previous hypotheses on the mode of action of kHz nerve stimulation, e.g. [
      • Bowman B.R.
      • McNeal D.R.
      Response of single alpha motoneurons to high-frequency pulse trains. Firing behavior and conduction block phenomenon.
      ]. We next assessed changes in fiber excitability during VNS, by delivering single “probing” pulses, every 1 s, throughout the 10-s kHz trains and measured fiber-specific, synchronous eCAPs [
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ] (Fig. 2D–b and c). Addition of probing pulses does not alter the physiological responses to kHz stimulation (Suppl. Fig. S4). Compared to pre-stimulation levels, eCAP amplitude progressively decreases as kHz stimulus intensity increases (Fig. 2D–b and c). This suppression of excitability occurs almost immediately upon onset of kHz stimulation (Suppl. Fig. S5, 2nd eCAP) and affects all fiber types. However, whereas A- and B- fiber excitability is almost completely abolished at intensities >7-8 × T, C-fiber excitability is suppressed at a much slower rate and C-fibers are still significantly excitable at intensities 7-20 × T (Fig. 2E). Importantly, A-, B- and C-fiber excitability returns to pre-stimulation levels after the end of each kHz stimulus train (Suppl. Fig. S5), suggesting that kHz stimulation within this intensity range (<20 × T for rats, <30 × T for mice) produces reversible effects. These findings indicate that kHz stimulation of relatively high intensity minimally suppresses excitability of C-fibers, while effectively blocking excitability of larger fibers.

      3.3 kHz stimulation selectivity for C-afferents arises from an interaction between stimulus frequency and intensity

      Stimulation at relatively low kHz-frequencies (e.g. 1-kHz) elicits similar HR and BI responses to those elicited by 30-Hz trains with matched duration, intensity and PW (Suppl. Fig. S13B), indicating limited selectivity for C-afferents. At higher frequencies (>5-kHz), high intensity stimulation results in similar BI responses as 30-Hz trains but with minimal HR responses, indicating selective activation of C-afferents (Fig. 3A and Suppl. Fig. S13A). Overall, the higher the frequency of kHz stimulation, the smaller the HR effect is for a similar, to 30-Hz VNS, BI response (Fig. 3B). In rats, selectivity for C-afferents is maximized at frequencies of 5-kHz or above, and intensities of 8–10 × T (Fig. 3C). In experiments in mice, kHz stimulus intensities 15-25 × T elicit similar BI responses as 30 Hz VNS, but with a much smaller HR response (Suppl. Fig. S3).
      Fig. 3
      Fig. 3Stimulus frequency interacts with stimulus intensity to convey C-fiber selectivity in the rat model. (A) Representative heart rate (HR) and breathing responses (airflow) elicited by stimuli of 12.5-kHz (40 μs PW, panel c and d), next to their 30 HZ, PW- and intensity-matched controls from the same animal (panel a and b). The HR responses are highly suppressed by 12.5-kHZ stimuli even with high VNS intensity, compared with 30Hz, whereas the BI responses remain similar. (B) Mean (±SE, N = 5 animals) normalized ΔHR and ΔBI responses for 1 kHz (PW = 500μs, panel a), 5 kHz (PW = 100μs, panel b), 12.5-kHz (PW = 40μs, panel c) frequency stimuli (red curves) along with responses to their corresponding 30 Hz controls (black curves), as a function of stimulus intensity respectively. (ANCOVA, ΔHR: p < 0.05 for all pair-wise stimuli and intensity, and their interaction; ΔBI: p > 0.05 for all pair-wise stimuli and p < 0.05 intensity, p < 0.05 for stimuli/intensity interaction). (C) panel a: Mean (±SE, N = 5 animals) normalized ΔHR responses to trains of stimuli of varying frequencies (intensity ranging from 3 × PT to 10 × PT, shown in different color curves) and of identical PW (40 μs), as a function of frequency. (ANCOVA, p < 0.05 for frequency and intensity, and p < 0.05 for interaction). panel b: Same as (C) but for normalized ΔBI responses. (ANCOVA, p < 0.05 for frequency and intensity, and p < 0.05 for interaction). (D) Mean (±SE, N = 4 animals) of CSI and normalized PSI values for A- (panel a), B- (panel b) and C-fibers (panel c) as function of 8-kHz stimulation intensity (ANCOVA, p < 0.05 for all types of CSI and intensity, and their interaction; p < 0.05 for all types of PSI, and PSI/intensity interaction). kHz stimulation of high intensity produces highest selectivity for C-fibers (yellow window in panel c). (panel d) C-fiber PSI values at different kHz stimulus intensities and Gaussian fits in individual animals. Average RMSE for fits: 0.169.
      Given the non-monotonic dependency of fiber selectivity on stimulus frequency and intensity, and the variability between animals (Fig. 3D), optimal parameters for selectively activating C-fibers will have to be determined on an individual animal basis, ideally in real-time. We sought to optimize stimulus parameters on the basis of fiber-specific selectivity indices. Physiological and eCAP selectivity indices (PSI and CSI, respectively) are compiled from real-time physiological and eCAP responses to stimuli (Suppl. Fig. S6C; detailed equations given in Supplement), allowing direct comparisons of the selectivity profile of different parameters within the same subject, and across subjects [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ,
      • Chang Y.
      • Ahmed U.
      • Tomaio J.N.
      • Rieth L.
      • Datta-Chaudhuri T.
      • Zanos S.
      Extraction of evoked compound nerve action potentials from vagus nerve recordings.
      ]. To validate this approach, we calculated PSI and CSI resulting from stimuli with known fiber selectivity profile. PSI and CSI for A-fibers were highest when we delivered 30-Hz VNS trains of short-square pulses (100 μs) at low intensity (1-3 × T) (Suppl. Figs. S6–S10 for rats, and Suppl. Figs. S11–12 for mice), which is expected given the low activation threshold of A-fibers [
      • Ahmed U.
      • Chang Y.C.
      • Cracchiolo M.
      • Lopez M.F.
      • Tomaio J.N.
      • Datta-Chaudhuri T.
      • et al.
      Anodal block permits directional vagus nerve stimulation.
      ]. Similarly, PSI and CSI for B-fibers were highest when we delivered 30-Hz VNS with long-square (>500 μs) or quasi-trapezoidal pulses at intermediate intensities (Suppl. Figs. S6–S10 for rats, and Suppl. Figs. S11–S12 for mice), a finding that is potentially due to anodal block of A-fibers [
      • Ahmed U.
      • Chang Y.C.
      • Cracchiolo M.
      • Lopez M.F.
      • Tomaio J.N.
      • Datta-Chaudhuri T.
      • et al.
      Anodal block permits directional vagus nerve stimulation.
      ,
      • Tosato M.
      • Yoshida K.
      • Toft E.
      • Struijk J.J.
      Quasi-trapezoidal pulses to selectively block the activation of intrinsic laryngeal muscles during vagal nerve stimulation.
      ,
      • Vuckovic A.
      • Tosato M.
      • Struijk J.J.
      A comparative study of three techniques for diameter selective fiber activation in the vagal nerve: anodal block, depolarizing prepulses and slowly rising pulses.
      ]. When applied to kHz stimulus trains, both PSI and CSI report similar changes for each of the 3 fiber types (Fig. 3D-, a-c). For C-fibers in particular, selectivity indices are maximal at relatively high kHz stimulation intensities (Fig. 3D–c). Even though stimulus intensity for optimal selectivity in engaging C-fibers is different between animals, it always follows a similar trend (Fig. 3D–d and Suppl. Fig. S3D). Together, these results indicate that kHz stimulation of relatively high frequency and intensity activates small, unmyelinated C-afferents, while inhibiting activation of larger, myelinated fibers.

      3.4 A possible cell membrane mechanism of kHz stimulation selectivity for C-afferents

      Neuro-electric computational models of nerve fibers can reveal how single channel biophysics and anatomical properties of axons can account for the effects of electrical stimuli on fiber activity and excitability [
      • Pelot N.A.
      • Catherall D.C.
      • Thio B.J.
      • Titus N.D.
      • Liang E.D.
      • Henriquez C.S.
      • et al.
      Excitation properties of computational models of unmyelinated peripheral axons.
      ]. To gain mechanistic insight into fiber selectivity at the level of the cell membrane, we simulated voltage responses to kHz trains in larger, myelinated fibers (2.6 μm axonal diameter, 5 μm myelin diameter) and in smaller, unmyelinated fibers (1.3 μm axonal diameter). At low frequencies (<1-kHz) neither fiber type is blocked, but A-fibers are activated at relatively low intensities (Fig. 4A, a-c). At higher frequencies (>2-kHz), large myelinated fibers are blocked at low intensities, whereas unmyelinated fibers are progressively activated at increasingly higher intensities (Fig. 4A, d-g) (Fig. 4B), in agreement with the stimulus intensity window we observe experimentally. The fiber selectivity of high-frequency stimulation can be explained by the different activation and inactivation dynamics of sodium currents, which are present in fibers of both diameters. At low intensities, action potentials (APs) are elicited in large fibers (Fig. 4B-a1) as activation (m) and inactivation (h) gates are fully functional in a physiological range, while small fibers are unresponsive (Fig. 4B-b1). At intermediate intensities, large fibers are quickly blocked, as the activation and inactivation gates start to passively follow the extracellular voltage changes (Fig. 4B-a2), while small fibers remain unresponsive (Fig. 4B-b2). At high intensities, APs are now elicited in small fibers as sodium channel gates (m and h) become activated and inactivated in a physiological range (Fig. 4B-b3). Small unmyelinated fibers are engaged at higher intensities compared to fibers of larger size and also compared to myelinated fibers of the same size (Fig. 4C), suggesting that both fiber size and presence of myelin shape fiber responses to kHz stimulation.
      Fig. 4
      Fig. 4kHz stimulation effects on large and small fibers can be explained in computer simulations by how sodium channel responses to stimuli are shaped by axonal size and myelination. Physiologically realistic neuro-electric models were implemented to simulate the responses of single fibers to kHz electrical stimuli. (A) Simulated normalized spike rate elicited in the myelinated (A-type) and un-myelinated (C-type) fiber models using 0.1-kHz, 0.5-kHz, 1-kHz, 2-kHz, 4-kHz, 8-kHz and 12-kHz sinusoidal stimulation, across multiple stimulus intensities. In panel (f), arrowheads point to the 3 stimulus intensities used to compile traces in (B). (B) Examples of membrane voltage (Vm) trajectories (black trace) and stimulus artifacts (superimposed grey trace), along with steady-state activation (m3) and inactivation (h) gating variables of sodium current in neurites below the electrode, during 8-kHz stimulus trains. Traces are shown for A-fibers (top “a” panels) and C-fibers (bottom “b” panels), at 3 stimulus intensities (3.3 μA, left column panels a1 and b1; 8 μA, middle column panels a2 and b2; 13 μA, right column panels a3 and b3). In panels a1 and b3, short snippets of Vm traces, and the corresponding gating variable time-courses, are shown magnified. (C) Normalized spike rate elicited by 8-kHz stimulation in the myelinated A-fiber (black), un-myelinated C-fiber (red), and un-myelinated fiber with the same diameter as an A-fiber (blue).

      4. Discussion

      Vagal C-afferents mediate numerous functions, including sensing of nutrients [
      • Grabauskas G.
      • Owyang C.
      Plasticity of vagal afferent signaling in the gut.
      ], regulation of appetite and glucose metabolism [
      • Bodenlos J.S.
      • Schneider K.L.
      • Oleski J.
      • Gordon K.
      • Rothschild A.J.
      • Pagoto S.L.
      Vagus nerve stimulation and food intake: effect of body mass index.
      ], effects of gut microbiome on brain function and cognition [
      • Carabotti M.
      • Scirocco A.
      • Maselli M.A.
      • Severi C.
      The gut-brain axis: interactions between enteric microbiota, central and enteric nervous systems.
      ], neural regulation of breathing [
      • Undem B.J.
      • Kollarik M.
      The role of vagal afferent nerves in chronic obstructive pulmonary disease.
      ], modulation of immune responses to lung infections, and shaping of emotional responses by feedback from the body [
      • Papasavas P.
      • El Chaar M.
      • Kothari S.N.
      • American Society for M.
      • Bariatric Surgery Clinical Issues C.
      American Society for Metabolic and Bariatric Surgery position statement on vagal blocking therapy for obesity.
      ]. Vagal C-afferents also constitute the afferent arm of cardiovascular reflexes [
      • Katona P.G.
      • Poitras J.W.
      • Barnett G.O.
      • Terry B.S.
      Cardiac vagal efferent activity and heart period in the carotid sinus reflex.
      ], neuroimmune circuits in the gut [
      • Teratani T.
      • Mikami Y.
      • Nakamoto N.
      • Suzuki T.
      • Harada Y.
      • Okabayashi K.
      • et al.
      The liver-brain-gut neural arc maintains the Treg cell niche in the gut.
      ], and certain inflammatory reflexes [
      • Silverman H.A.
      • Chen A.
      • Kravatz N.L.
      • Chavan S.S.
      • Chang E.H.
      Involvement of neural transient receptor potential channels in peripheral inflammation.
      ], neuroimmune circuits that maintain immunological homeostasis throughout the body [
      • Chavan S.S.
      • Pavlov V.A.
      • Tracey K.J.
      Mechanisms and therapeutic relevance of neuro-immune communication.
      ].
      Our finding that kHz nerve stimulation leads to sustained activation of C-fiber-associated vagal neurons is novel, in light of several past studies that have reported kHz-elicited block of vagus fibers in general [
      • Kilgore K.L.
      • Bhadra N.
      Reversible nerve conduction block using kilohertz frequency alternating current.
      ,
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ,
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ] and of C-fibers in particular [
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ]. Both transient and sustained activation of axons from kHz stimuli has been described in the dorsal columns of the spinal cord as part of a “kHz stimulus-onset response” [
      • Crosby N.D.
      • Janik J.J.
      • Grill W.M.
      Modulation of activity and conduction in single dorsal column axons by kilohertz-frequency spinal cord stimulation.
      ]; the same mechanism could also mediate the effects of kHz stimulation on vagal C-afferents.
      kHz stimulation of the sub-diaphragmatic vagus is used clinically in the treatment of obesity [
      • Apovian C.M.
      • Shah S.N.
      • Wolfe B.M.
      • Ikramuddin S.
      • Miller C.J.
      • Tweden K.S.
      • et al.
      Two-year outcomes of vagal nerve blocking (vBloc) for the treatment of obesity in the ReCharge trial.
      ], putatively acting by blocking conduction in gut-innervating vagus sensory neurons that signal satiety and/or in vagus efferent fibers involved in the control of gastrointestinal fluid release and motility [
      • de Lartigue G.
      Role of the vagus nerve in the development and treatment of diet-induced obesity.
      ]. However, our finding indicates that part of the effect of kHz stimulation in obesity may be mediated by activation of fibers rather than by nerve block [
      • Johannessen H.
      • Revesz D.
      • Kodama Y.
      • Cassie N.
      • Skibicka K.P.
      • Barrett P.
      • et al.
      Vagal blocking for obesity control: a possible mechanism-of-action.
      ], as has been recently suggested [
      • Pelot N.A.
      • Behrend C.E.
      • Grill W.M.
      Modeling the response of small myelinated axons in a compound nerve to kilohertz frequency signals.
      ].
      In our study, we documented neuronal activation by quantifying c-Fos immunoreactivity in neurons in the NTS, a brainstem region that directly receives input from C-afferent-bearing vagal sensory neurons, and the DMV, from which much of motor vagus signaling is communicated to the periphery via cholinergic vagal efferents [
      • Baker E.
      • Lui F.
      Neuroanatomy, vagal nerve nuclei.
      ]. In the naïve and sham groups, we observed more c-Fos-expressing cell in NTS than DMV, likely reflecting the larger number of sensory fibers in the vagus nerve. We found that 30 min of intermittent kHz stimulation results in increased expression of c-Fos in both those regions, at lower, but still comparable, levels to those produced by “standard” 30 Hz VNS (Fig. 1B–D). Collection of tissue in our study occurred 1.5 h after the end of stimulation (Fig. 1A), a time point within the window for strong c-Fos protein expression following neuronal activation, suggesting that these c-Fos expression levels likely represent peak neuronal responses to acute nerve stimulation. Compared to 30 Hz VNS, kHz VNS elicited a relatively stronger response in NTS than in DMV neurons (Fig. 1C–E, Supple. Fig. S1D), indicating preferential activation of afferent vagal fibers, of which the vast majority are C-type, over efferent fibers [
      • Yuan H.
      • Silberstein S.D.
      Vagus nerve and vagus nerve stimulation, a comprehensive review: Part I.
      ]. VNS at 30 Hz causes robust efferent physiological effects, including changes in heart rate and blood pressure [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ], which are sensed by vagal sensory neurons bilaterally, eliciting bilateral, reflexive increases in c-Fos expression; in contrast, kHz stimulation preferentially activates sensory neurons ipsilateral to the stimulated nerve (Suppl. Figure S1B-D). Mechanical stimulation of the nerve during acute placement of vagus electrode had a moderate effect on c-Fos expression in NTS, and no effect in DMV (Fig. 1B–D).
      Classic c-Fos reporters have limited temporal resolution: their dependence on transcriptional processes quantified through an immunohistochemical assay preclude them from resolving the detailed time course of neuronal activation [
      • Hudson A.E.
      Genetic reporters of neuronal activity: c-Fos and G-CaMP6.
      ]. In our report, we theorized that c-Fos expression in NTS neurons reflects primarily their activation due to stimulation of C-afferent vagus fibers. Neurons in the NTS receive direct inputs from the central axons of sensory vagal neurons in the nodose ganglion. Given the abundance of C-fibers in the vagus, most nodose sensory neurons have peripheral axons composed of C-fibers; it is therefore likely that most NTS neurons that are activated by VNS are also activated by stimuli that preferentially stimulate C-fibers. We also theorized that c-Fos expression in DMV neurons reflects primarily their direct activation by antidromic stimulation of their efferent axons. Subsets of NTS neurons project synaptically to DMV neurons, and stimuli that engage afferent C-fibers may also, indirectly, activate DMV neurons. However, it is likely that direct, antidromic activation of DMV neurons by efferent stimulation is more pronounced than indirect, synaptically-mediated activation by inputs from sensory neurons: in the first case, the c-Fos read-out is 1 cell away from the stimulus with no intervening synapses, in the second it is 3 cells and 2 synapses away from the stimulus. Other studies have used the c-Fos expression in these 2 regions to estimate activation of afferent and efferent vagal pathways [
      • Cunningham J.T.
      • Mifflin S.W.
      • Gould G.G.
      • Frazer A.
      Induction of c-Fos and DeltaFosB immunoreactivity in rat brain by Vagal nerve stimulation.
      ,
      • Huffman W.J.
      • Subramaniyan S.
      • Rodriguiz R.M.
      • Wetsel W.C.
      • Grill W.M.
      • Terrando N.
      Modulation of neuroinflammation and memory dysfunction using percutaneous vagus nerve stimulation in mice.
      ]. Another limitation of the c-Fos approach is that it cannot resolve cell inhibition by stimuli, therefore the degree to which kHz or 30 Hz stimuli result in inhibition of neurons in the brainstem is unknown. Finally, measurement of c-Fos expression in our study was limited to 2 vagal brainstem nuclei, one sensory and one motor. We did not assess c-Fos expression in other regions of the central nervous system with extra-vagal motor neurons that provide efferent fibers to the vagus, e.g., unmyelinated post-ganglionic fibers “hitchhiking” into the vagus from the sympathetic chain [
      • Cunningham J.T.
      • Mifflin S.W.
      • Gould G.G.
      • Frazer A.
      Induction of c-Fos and DeltaFosB immunoreactivity in rat brain by Vagal nerve stimulation.
      ,
      • Huffman W.J.
      • Subramaniyan S.
      • Rodriguiz R.M.
      • Wetsel W.C.
      • Grill W.M.
      • Terrando N.
      Modulation of neuroinflammation and memory dysfunction using percutaneous vagus nerve stimulation in mice.
      ,
      • Patel Y.A.
      • Kim B.S.
      • Rountree W.S.
      • Butera R.J.
      Kilohertz electrical stimulation nerve conduction block: effects of electrode surface area.
      ], and therefore the effect of kHz stimulation on those neurons is unknown.
      Direct recordings of ongoing fiber activity during stimulation were not available to us, as intracellular recordings from single axons are not feasible and extracellular recordings are obscured by the kHz-frequency electrical artifact. Therefore, to demonstrate that kHz stimulation, together with several other factors, including nerve morphology, electrode geometry and stimulus waveform, preferentially activates C-afferents over other fiber types in the vagus in our setup, we relied on physiological responses to fiber engagement from some of the end-organs they innervate, as proxies for stimulus-elicited asynchronous fiber activity, e.g. Refs. [
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ,
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ,
      • Patel Y.A.
      • Kim B.S.
      • Rountree W.S.
      • Butera R.J.
      Kilohertz electrical stimulation nerve conduction block: effects of electrode surface area.
      ,
      • Buckley U.
      • Chui R.W.
      • Rajendran P.S.
      • Vrabec T.
      • Shivkumar K.
      • Ardell J.L.
      Bioelectronic neuromodulation of the paravertebral cardiac efferent sympathetic outflow and its effect on ventricular electrical indices.
      ,
      • Bhadra N.
      • Kilgore K.L.
      Direct current electrical conduction block of peripheral nerve.
      ]. We use asynchronous activity to estimate the level of fiber activation by trains of stimuli, and stimulus-evoked activity to estimate the level of fiber excitability (or “responsiveness” to stimuli) at the moment of delivery of single, probing pulses. The correspondence between physiological responses to trains of stimuli and asynchronous fiber activity relies on correlation studies described previously [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ]; however, we have not provided a direct demonstration between physiological markers and ongoing activity of single fibers, and that is a limitation of our approach.
      Stimulus-elicited changes in breathing interval (BI) were used to estimate C-afferent activity, since engagement of C-fibers, either optogenetically (Fig. 2B–b) or electrically [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ] is associated with alterations in the breathing pattern. We found that kHz stimulation elicits changes in BI and simultaneously minimizes effects on heart rate, an index of B-fiber activation, and on laryngeal EMG, an index of A-fiber activation (Fig. 2C–E; Suppl. Video 1D-F; Suppl. Figs. S3–S5) [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ]. When HR changes occur, especially at high stimulus intensities (Fig. 2D–a, Figs. S3A–b), those could be due to C-afferent-mediated vago-vagal reflexes, with slow-onset and less pronounced cardio-inhibition (e.g. Fig. 2B–b), as opposed to the fast-onset, intense HR drop seen with direct B-fiber activation [
      • McAllen R.M.
      • Shafton A.D.
      • Bratton B.O.
      • Trevaks D.
      • Furness J.B.
      Calibration of thresholds for functional engagement of vagal A, B and C fiber groups in vivo.
      ]. Even though measurement of physiological responses to neurostimulation is necessary to establish nerve activation or conduction block, it is not sufficient. For example, nerve block is not necessarily the cause of a given physiological effect, as that could be achieved through other mechanisms, such as facilitation [
      • Pelot N.A.
      • Behrend C.E.
      • Grill W.M.
      Modeling the response of small myelinated axons in a compound nerve to kilohertz frequency signals.
      ]. Accordingly, physiological responses to kHz stimuli indicating fiber activation need to be corroborated by direct measurement of changes in fiber excitability, by recording nerve compound action potentials in response to single “probing” pulses (eCAPs) [
      • Patel Y.A.
      • Saxena T.
      • Bellamkonda R.V.
      • Butera R.J.
      Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
      ,
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ,
      • Patel Y.A.
      • Butera R.J.
      Differential fiber-specific block of nerve conduction in mammalian peripheral nerves using kilohertz electrical stimulation.
      ]. eCAPs in response to probing stimuli before, during and after kHz trains are consistent with sequential conduction block as stimulus intensity increases (Fig. 2D and Suppl. Fig. S5), which starts with A-fibers, then encompasses B-fibers and finally affects C-fibers (Fig. 2D–E), as described before [
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ,
      • Yi G.
      • Grill W.M.
      Kilohertz waveforms optimized to produce closed-state Na+ channel inactivation eliminate onset response in nerve conduction block.
      ]. This non-monotonic relationship establishes an “intensity window”, within which excitability of A- and B-fibers is almost completely blocked and C-fiber excitability is only partially suppressed, and provides a physiological basis for the C-afferent selectivity of kHz stimulation (Fig. 2E). Notably, asynchronous activity and excitability are 2 different measures of the state of nerve fibers: during kHz stimulation, A- and B-fibers have both minimal activity and excitability, whereas C-afferent have higher activity and at the same time have partially suppressed excitability, compared to baseline. Indeed, C-afferent-associated BI responses reach a maximum at intensities producing only half-maximal block of C-fibers in eCAPs (Fig. 2C–E), indicating that action potentials in C-afferents are still elicited by kHz trains, even if excitability of those fibers is partially suppressed. This mechanism is consistent with our finding that C-afferent-bearing NTS neurons are preferentially activated against Aα/B-fiber-associated DMV neurons, but at relatively lower levels compared to NTS neuronal activation by non-selective 30 Hz VNS (Fig. 1). These changes in fiber excitability occur and cease rapidly after the onset and offset of kHz trains, indicating that 10-s long trains have a tightly controlled, temporary effect with no damage to nerve fibers [
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ] (Suppl. Fig. S5). Slowing in conduction velocity, more prominent in slow fibers, is observed in our experiments (Fig. 2D and Suppl. Fig. S5), consistent with previous reports.
      By simultaneously varying the frequency and intensity of kHz stimuli, we found an interaction between the 2 variables, similar to what has been documented previously in peripheral nerves and retinal neurons: high stimulus intensities (7-10 × T in rats, 15-25 × T in mice) and frequencies >5 kHz convey the highest C-afferent selectivity (Fig. 3B), with some variability in the optimal stimulus intensity in terms of thresholds (T) between subjects (Fig. 3D). We compiled selectivity indices (SIs) to normalize and compare responses and determine optimal stimulus parameters across subjects (Figs. 1E and 3D). These indices account for the fact that the effects of electrical stimulation depend on the nerve-electrode interface and the underlying nerve anatomy [
      • Losanno E.
      • Badi M.
      • Wurth S.
      • Borgognon S.
      • Courtine G.
      • Capogrosso M.
      • et al.
      Bayesian optimization of peripheral intraneural stimulation protocols to evoke distal limb movements.
      ], both of which are highly variable, and that practically any set of nerve stimulation parameters engages multiple fiber populations, in different degrees, e.g. Ref. [
      • Ahmed U.
      • Chang Y.C.
      • Cracchiolo M.
      • Lopez M.F.
      • Tomaio J.N.
      • Datta-Chaudhuri T.
      • et al.
      Anodal block permits directional vagus nerve stimulation.
      ]. The nerve potential measurements required for CAP SIs are surgically challenging and often noisy, and the IHC procedures required for neuronal c-Fos SIs are not feasible in real-time. In contrast, compiling SIs from noninvasive physiological responses that represent fiber engagement can be implemented in real-time, and is practical and feasible in experimental animals and human subjects [
      • Chang Y.C.
      • Cracchiolo M.
      • Ahmed U.
      • Mughrabi I.
      • Gabalski A.
      • Daytz A.
      • et al.
      Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
      ]. In our study, physiological SIs compiled from A-, B- or C-fiber selective stimuli are strongly correlated with CAP SIs compiled from the same stimuli (Fig. 3, Suppl. Fig. S6), suggesting that they are good indicators of selective fiber engagement. Such indices could be programmed into research or bedside systems for automatic calibration and optimization of VNS therapies targeting different fiber populations in individual patients. For example, B-fiber indices may facilitate calibration of VNS in treating heart failure [
      • Gold M.R.
      • Van Veldhuisen D.J.
      • Hauptman P.J.
      • Borggrefe M.
      • Kubo S.H.
      • Lieberman R.A.
      • et al.
      Vagus nerve stimulation for the treatment of heart failure: the INOVATE-HF trial.
      ] or cardiac arrhythmias [
      • Nasi-Er B.G.
      • Wenhui Z.
      • HuaXin S.
      • Xianhui Z.
      • Yaodong L.
      • Yanmei L.
      • et al.
      Vagus nerve stimulation reduces ventricular arrhythmias and increases ventricular electrical stability.
      ], and C-fiber indices in application of VNS in inflammatory [
      • Sundman E.
      • Olofsson P.S.
      Neural control of the immune system.
      ,
      • Tanaka S.
      • Abe C.
      • Abbott S.B.G.
      • Zheng S.
      • Yamaoka Y.
      • Lipsey J.E.
      • et al.
      Vagus nerve stimulation activates two distinct neuroimmune circuits converging in the spleen to protect mice from kidney injury.
      ] or metabolic disorders [
      • Pardo J.V.
      • Sheikh S.A.
      • Kuskowski M.A.
      • Surerus-Johnson C.
      • Hagen M.C.
      • Lee J.T.
      • et al.
      Weight loss during chronic, cervical vagus nerve stimulation in depressed patients with obesity: an observation.
      ]. In our study, the duration of kHz trains, as well the intervals between trains, are likely to have contributed to the C-afferent selective effect. kHz trains produce transient fiber excitation followed by longer-lasting inhibition of nerve fibers [
      • Pelot N.A.
      • Grill W.M.
      In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
      ]. The intermittent, rather than continuous, time course of stimulus trains in our study may be responsible for the sustained activation. By controlling the duty cycle, the ON/OFF epochs of kHz stimulation, the balance between the brief onset excitation and longer-lasting inhibition can be further optimized to increase the level of C-afferent activation and/or selectivity. Optimizing the duty cycle is important for another reason: a small duty cycle (short ON epochs) might lead to limited selectivity, as a minimum duration is required to block large fibers, and a large duty cycle might end up blocking both large and small fibers.
      Our finding that kHz stimulation selectively activates C-afferents over larger efferent fibers has important implications for physiological and translational research. Controlled, selective engagement of distinct nerve fiber types, separately from other fiber populations in the same nerve, is required for the study of their physiological and translational roles [
      Bioelectronics SPARC at NIH
      ]. Selective activation of C-afferents in the vagus is possible using optogenetic nerve stimulation [
      • Rajendran P.S.
      • Challis R.C.
      • Fowlkes C.C.
      • Hanna P.
      • Tompkins J.D.
      • Jordan M.C.
      • et al.
      Identification of peripheral neural circuits that regulate heart rate using optogenetic and viral vector strategies.
      ] but that is currently only practical with mice in acute experiments, with limited value in preclinical models of chronic disease and unclear clinical applicability. On the other hand, electrical stimulation of the vagus can be delivered acutely or chronically, in various animal models, including mice [
      • Mughrabi I.T.
      • Hickman J.
      • Jayaprakash N.
      • Papadoyannis E.S.
      • Abbas A.
      • Chang Y.-C.
      • et al.
      An implant for long-term cervical vagus nerve stimulation in mice.
      ], and in humans. However, there is no standard electrical stimulus that selectively activates vagal C-fibers. That is mainly because of the natural recruitment order of nerve fibers: larger fibers (A- and B-type) are activated well before smaller fibers can be engaged, and at high stimulus intensities, required to activate C-fibers, larger fibers are also maximally activated [
      • McAllen R.M.
      • Shafton A.D.
      • Bratton B.O.
      • Trevaks D.
      • Furness J.B.
      Calibration of thresholds for functional engagement of vagal A, B and C fiber groups in vivo.
      ,
      • Yoo P.B.
      • Lubock N.B.
      • Hincapie J.G.
      • Ruble S.B.
      • Hamann J.J.
      • Grill W.M.
      High-resolution measurement of electrically-evoked vagus nerve activity in the anesthetized dog.
      ]. The lack of a C-fiber-selective electrical stimulus hinders the study of the many interoceptive functions and autonomic reflexes in which C-afferents are involved, the translational testing of VNS in animal models of disease and, ultimately, the therapeutic potential of VNS. It is important to note here that the same kHz stimuli that activate C-afferents are also likely to activate unmyelinated efferents, as the 2 fiber types share similar morphological characteristics. Selective activation of C-afferents over C-efferents has not been demonstrated in this study, and will require additional optimization steps, e.g., by leveraging the spatial arrangement of different fiber types within the cervical vagus [
      • Ahmed U.
      • Chang Y.C.
      • Zafeiropoulos S.
      • Nassrallah Z.
      • Miller L.
      • Zanos S.
      Strategies for precision vagus neuromodulation.
      ].
      We used an biophysically plausible computational model of axons to study the effects of kHz stimuli in different regimes, from subthreshold facilitation to suprathreshold block [
      • Joseph L.
      • Butera R.J.
      High-frequency stimulation selectively blocks different types of fibers in frog sciatic nerve.
      ,
      • Neudorfer C.
      • Chow C.T.
      • Boutet A.
      • Loh A.
      • Germann J.
      • Elias G.J.
      • et al.
      Kilohertz-frequency stimulation of the nervous system: a review of underlying mechanisms.
      ], and to gain insight into possible membrane-level mechanisms that could contribute to the differential fiber responses to these stimuli. Our simulations suggest that large, myelinated and small, unmyelinated fibers undergo both activation and block, and that selectivity for one or the other can be attained by modulating stimulus frequency and intensity (Fig. 4A), in agreement with our experimental findings. In the model, parameters related to voltage-gated channel kinetics and ion channel distributions were kept the same for large and small fibers; those fibers differed only with regard to their physical properties, i.e. axonal size and presence of myelin [
      • Vetter P.
      • Roth A.
      • Hausser M.
      Propagation of action potentials in dendrites depends on dendritic morphology.
      ,
      • Spruston N.
      Pyramidal neurons: dendritic structure and synaptic integration.
      ,
      • Wang J.
      • Jacoby R.
      • Wu S.M.
      Physiological and morphological characterization of ganglion cells in the salamander retina.
      ]. In large, myelinated fibers, the activation function increases with longer internodal distance and larger axonal radius [
      • Rattay F.
      Analysis of models for external stimulation of axons.
      ]. The decreased axial resistance of larger fibers also leads to greater electrical connection between adjacent axonal segments, preventing the intracellular potential from ‘floating’ with changes in extracellular potential and rendering the axon more responsive to gradients in the extracellular potential. Larger fibers are more robustly affected by kHz-induced depolarization and therefore exhibit lower activation and blocking thresholds. The effect of myelin does not qualitatively alter but it does magnify this effect by significantly increasing the internodal distance, thereby increasing the voltage gradients at each node [
      • Rattay F.
      Analysis of models for external stimulation of axons.
      ] (Fig. 4C). Therefore, morphological differences between fibers with otherwise identical ionic channels can lead to different activation and blocking thresholds, potentially explaining part of the intensity-dependency of kHz stimulation. The distribution of different ion channels in functionally distinct vagus fibers likely affects selective fiber activation. For example, in larger myelinated fibers, ion channels are mainly concentrated in the nodes of Ranvier, while in smaller unmyelinated fibers ion channel are distributed in a more widespread manner along the axon. In addition, different ion channel subtypes could play distinct roles during selective stimulation. Sodium channels play a dominant role during kHz stimulation [
      • Yi G.
      • Grill W.M.
      Kilohertz waveforms optimized to produce closed-state Na+ channel inactivation eliminate onset response in nerve conduction block.
      ,
      • Guo T.
      • Tsai D.
      • Yang C.Y.
      • Al Abed A.
      • Twyford P.
      • Fried S.I.
      • et al.
      Mediating retinal ganglion cell spike rates using high-frequency electrical stimulation.
      ]. Other ion channels may also be involved, including delayed rectifier potassium and T-type low-voltage-activated calcium channels, all of which play significant roles in the responses of retinal ganglion neurons to kHz stimuli [
      • Kameneva T.
      • Maturana M.I.
      • Hadjinicolaou A.E.
      • Cloherty S.L.
      • Ibbotson M.R.
      • Grayden D.B.
      • et al.
      Retinal ganglion cells: mechanisms underlying depolarization block and differential responses to high frequency electrical stimulation of ON and OFF cells.
      ], with the caveat that retinal neurons have different biophysical properties than all other neurons. Because of the limited data on the intrinsic ion channel diversity of functionally distinct vagal fibers [
      • Pelot N.A.
      • Catherall D.C.
      • Thio B.J.
      • Titus N.D.
      • Liang E.D.
      • Henriquez C.S.
      • et al.
      Excitation properties of computational models of unmyelinated peripheral axons.
      ,
      • Schild J.H.
      • Clark J.W.
      • Hay M.
      • Mendelowitz D.
      • Andresen M.C.
      • Kunze D.L.
      A- and C-type rat nodose sensory neurons: model interpretations of dynamic discharge characteristics.
      ], our study modeled nerve fibers equipped with a minimal set of ion channels, the universally present sodium channels, to gain insight into how morphological differences between fibers could explain part of the experimental effect. In addition, our model only simulated the nerve fibers with a homogeneous surrounding environment as saline solution. However, surrounding tissue environment such as endoneurium, perineurium, and epineurium will influence the nerve response to electrical stimulation [
      • Pelot N.A.
      • Grill W.M.
      Effects of vagal neuromodulation on feeding behavior.
      ], so may qualitatively alter the relative selectivity between myelinated and unmyelinated axons. Future modeling efforts will include anatomically realistic information to better simulate the potential nerve response under electrical stimulation.
      The frequency-dependency and fiber-selectivity of activation and block from kHz stimuli can be explained by the fiber-specific refractory period [
      • Tackmann W.
      • Lehmann H.J.
      refractory period in human sensory nerve fibres.
      ,
      • Kimura J.
      • Yamada T.
      • Rodnitzky R.L.
      Refractory period of human motor nerve fibres.
      ] and passive time constant [
      • Neudorfer C.
      • Chow C.T.
      • Boutet A.
      • Loh A.
      • Germann J.
      • Elias G.J.
      • et al.
      Kilohertz-frequency stimulation of the nervous system: a review of underlying mechanisms.
      ] in response to individual stimuli. At subthreshold intensities, for a given fiber type, membrane voltage modulations occurring faster than the membrane time constant, result in charge accumulation, which can further facilitate spiking [
      • Neudorfer C.
      • Chow C.T.
      • Boutet A.
      • Loh A.
      • Germann J.
      • Elias G.J.
      • et al.
      Kilohertz-frequency stimulation of the nervous system: a review of underlying mechanisms.
      ,
      • Heffer L.F.
      • Sly D.J.
      • Fallon J.B.
      • White M.W.
      • Shepherd R.K.
      • O'Leary S.J.
      Examining the auditory nerve fiber response to high rate cochlear implant stimulation: chronic sensorineural hearing loss and facilitation.
      ] (Fig. 4B-a1 and b3). On the other hand, at suprathreshold intensities, sodium channel inactivation induced by tonic membrane depolarization to consecutive stimuli within refractory period is likely the major mechanism underlying action potential block (Fig. 4B-a2 and a3) [
      • Kilgore K.L.
      • Bhadra N.
      Nerve conduction block utilising high-frequency alternating current.
      ,
      • Bhadra N.
      • Lahowetz E.A.
      • Foldes S.T.
      • Kilgore K.L.
      Simulation of high-frequency sinusoidal electrical block of mammalian myelinated axons.
      ]. Intensities that are supra-threshold for large fibers promote conduction block (Fig. 4B-a2 and a3), while at the same still being sub-threshold for small C-fibers, facilitating asynchronous C-afferent activity (Fig. 4B-b3). By combining the frequency- and intensity-dependencies of kHz stimulation, the resulting block and activation windows for different fiber types can be leveraged for fiber-selective stimulation (Fig. 4A–e through g). The nerve-specific properties may also explain our finding that, during kHz stimulation, C-fiber eCAPs start to decline at lower intensities than those at which the respective physiological response, i.e. breathing changes, reaches its maximum (yellow bars in Fig. 2C and E). As stimulus intensity increases, more asynchronous action potentials are elicited from C-fibers, thereby leading to greater changes in breathing. However, that might also transiently bring more C-fibers into their absolute refractory period, even for a very short time, so fewer C-fibers will be able to generate action potentials in response to the probing stimulus, thus leading to smaller C-fiber eCAPs. It has recently been reported that neural block threshold increases monotonically with increasing frequency between 10 and 300 kHz, when kHz trains are symmetrical [
      • Pena E.
      • Pelot N.A.
      • Grill W.M.
      Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance.
      ]. Even though we used symmetrical kHz stimuli, in both experimental (Fig. 3C) and modeling studies (Fig. 4), we found decreased excitation and block thresholds with increasing frequency for a myelinated fiber. A possible reason of this discrepancy could be the different criteria of neural block used in other studies: for example, block of spike propagation along the axon, adopted in Ref. [
      • Kilgore K.L.
      • Bhadra N.
      Nerve conduction block utilising high-frequency alternating current.
      ], vs. block of the node under the stimulation electrode, adopted in our study. In addition, our model suggests increased excitation thresholds for a small unmyelinated fiber, in contrast to recent experimental studies that looked at kHz stimulation for unmyelinated nerve stimulation [
      • Guo T.
      • Tsai D.
      • Yang C.Y.
      • Al Abed A.
      • Twyford P.
      • Fried S.I.
      • et al.
      Mediating retinal ganglion cell spike rates using high-frequency electrical stimulation.
      ,
      • Guo T.
      • Yang C.Y.
      • Tsai D.
      • Muralidharan M.
      • Suaning G.J.
      • Morley J.W.
      • et al.
      Closed-loop efficient searching of optimal electrical stimulation parameters for preferential excitation of retinal ganglion cells.
      ,
      • Muralidharan M.
      • Guo T.
      • Shivdasani M.N.
      • Tsai D.
      • Fried S.
      • Li L.
      • et al.
      Neural activity of functionally different retinal ganglion cells can be robustly modulated by high-rate electrical pulse trains.
      ]. This discrepancy might be due to the different total charge delivered by the different stimulation waveforms used in those studies vs. our model. The sinusoid stimulation used in our model delivered the identical total charge across all stimulation frequencies, while the total charge from rectangular waveforms used in other studies was dependent on frequency.
      Our study has several limitations. Anesthesia suppresses excitability of sensory vagal neurons, resulting in increased VNS threshold for eliciting changes in breathing, as well as of motor vagal neurons, resulting in increased heart rate threshold compared to awake state [
      • Ahmed U.
      • Chang Y.C.
      • Lopez M.F.
      • Wong J.
      • Datta-Chaudhuri T.
      • Rieth L.
      • et al.
      Implant- and anesthesia-related factors affecting cardiopulmonary threshold intensities for vagus nerve stimulation.
      ]. Overall, it is likely that all physiological changes in response to the same parameters used in our study would be “brisker” in the awake state. However, the balance of afferent and efferent responses is unlikely to be significantly affected by anesthesia [
      • Ahmed U.
      • Chang Y.C.
      • Lopez M.F.
      • Wong J.
      • Datta-Chaudhuri T.
      • Rieth L.
      • et al.
      Implant- and anesthesia-related factors affecting cardiopulmonary threshold intensities for vagus nerve stimulation.
      ], leaving the effect of stimulation parameters as the main driver for the differences documented in the study.
      Second, in addition to selective nerve fiber block, kHz stimulation transiently evokes an onset response, in both afferent and efferent directions [
      • Kilgore K.L.
      • Bhadra N.
      Reversible nerve conduction block using kilohertz frequency alternating current.
      ,
      • Patel Y.A.
      • Butera R.J.
      Challenges associated with nerve conduction block using kilohertz electrical stimulation.
      ,
      • Crosby N.D.
      • Janik J.J.
      • Grill W.M.
      Modulation of activity and conduction in single dorsal column axons by kilohertz-frequency spinal cord stimulation.
      ]. In our study, at intensities above those that block A- and B-fibers, brief laryngeal muscle contractions are observed at the onset of kHz-frequency trains, likely caused by transient activation of large efferents, quickly followed by EMG block (Fig. S5A; Supple Video 1E and 1F). Likewise, transient spiking in larger fibers is also seen upon the onset of kHz stimulation in the computational model (Fig. 4B-a2-a3, black traces <1 ms after onset). This transient onset response of larger fibers may contribute to c-Fos expression in brainstem neurons (Fig. 1 and Suppl. Fig. S1). Real-time measures of neuronal activity, including single unit recordings or optical imaging, will be needed to determine the detailed time course of responses of different neuronal populations to kHz-frequency stimulation.
      Third, kHz stimulation may cause nerve damage, as it deposits significantly more power in tissue compared to stimuli of lower frequencies. For example, in spinal cord, high frequency stimulation may induce significant temperature increases, something that might alter the effects of kHz stimulation over time. In a recent study of kHz stimulation of the vagus nerve, it was shown that eCAP responses to probing stimuli during kHz stimulation were blocked, and eventually recovered up to certain intensity level, beyond which block was irreversible; the higher the intensity, the longer the recovery time. In our study, Khz stimulus train duration was shorter than that study (10 s versus 25 s) and the interstimulus intervals were longer (5–15 min versus 100 s), which likely allowed full recovery of eCAPs and physiological responses. In our study, all responses elicited below 10 × T intensity (8-kHzKHz, 40-μs pulse width, 2–2.5 mA) were fully recoverable (Fig. 2C and S5). However, at very high intensities, at which C-fiber eCAPs and breathing responses were abolished (30 × T, 6–7.5 mA), the effect was irreversible (data not shown). These findings indicates that complete block of C-fibers in rodents using high intensity kHz stimulation, while maintaining the integrity of the interface and the nerve, might be challenging and will certainly require subject-personalized dosing.
      In our study we used mice and rats, as recruitment of vagal fibers is well-documented in rodents and the classification of fiber types based on conduction velocity and anatomical characteristics was originally done in rodents [
      • Woodbury D.M.
      • Woodbury J.W.
      Effects of vagal stimulation on experimentally induced seizures in rats.
      ]. In addition, rodent models are well suited to study fiber-specific stimulation protocols, as the fascicular structure of the vagus in rodents is simple, compared to that of large animals, while the characteristics of each fiber type remain similar. Compared with clinical VNS, we used much shorter pulse width (100-μs) as our standard stimulus for determining threshold intensities. However, to translate our findings to large animals or human, the stimulation strategies described here need to be extensively validated. In large vagus nerves, the stimulation electrode and the nerve fibers of interest lie at a greater distance, and are separated by several anatomical layers of connective, vascular and neural tissue [
      • Nicolai E.N.
      • Settell M.L.
      • Knudsen B.E.
      • McConico A.L.
      • Gosink B.A.
      • Trevathan J.K.
      • et al.
      Sources of off-target effects of vagus nerve stimulation using the helical clinical lead in domestic pigs.
      ]. Special electrode designs with multiple contacts will likely be needed to allow single contacts to lie closer to specific fascicles and fiber populations [
      • Aristovich K.
      • Donega M.
      • Fjordbakk C.
      • Tarotin I.
      • Chapman C.A.R.
      • Viscasillas J.
      • et al.
      Model-based geometrical optimisation and in vivo validation of a spatially selective multielectrode cuff array for vagus nerve neuromodulation.
      ]. An additional limitation is the need for relatively high stimulus intensities to activate small C-fibers, even in the small rodent nerve; in larger nerves, these intensities will likely be even higher, creating challenges with regard to the voltage compliance to be handed by stimulus generators, to tissue safety limits, to the risk of accidental activation of nearby neural pathways or even muscle activation through current leakage. Even if reduced, the laryngeal response to kHz stimulation may be associated with significant pain. Finally, since C-fiber activation is associated with, potentially unwanted, breathing changes, kHz stimulation paradigms will need to be optimized to reduce the impact of such effects on patients.

      5. Conclusion

      Despite the physiological and translational significance of small, unmyelinated vagus C-afferents, their selective activation using electrical vagus stimulation has not been achieved to date. In this study, we show that intermittent kHz-frequency electrical stimulation, widely used to block nerve conduction, can preferentially activate C-afferents while blocking larger nerve fibers, in a stimulus intensity- and frequency-dependent manner. This is demonstrated in both mice and rats, suggesting a mechanism that is not species-specific.

      Credit author statement

      YCC: conceived and designed experiments, performed experiments, analyzed and interpreted experimental results, and wrote the paper, UA: conceived and performed rat experiments, NJ: designed and performed anatomy experiments, YCW: designed and performed anatomy experiments, IM: designed and performed experiments in mice, MG: designed and performed experiments in mice, AA: designed and performed experiments in mice, QL: designed and performed the computational stimulation, and wrote the paper, TG: designed and performed the computational stimulation, and wrote the paper, SD: designed and performed the computational stimulation, and wrote the paper, AG: performed rat experiments, critically reviewed the paper, AD: performed the histological analysis, JA: performed the histological analysis, SD: critically reviewed the paper, TDC: critically reviewed the paper, YAA: critically reviewed the paper, SZ: conceived and designed experiments, secured funding, managed the research team, analyzed and interpreted experimental results and wrote the paper.

      Data and material availability

      All data that support the findings of this study are available from the corresponding author upon reasonable request.

      Declaration of competing interest

      The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: SZ and YCC have a provisional patent application that includes aspects of the research presented in this paper. The other authors declare no conflict of interest.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Agostoni E.
        • Chinnock J.E.
        • De Daly M.B.
        • Murray J.G.
        Functional and histological studies of the vagus nerve and its branches to the heart, lungs and abdominal viscera in the cat.
        J Physiol. 1957; 135: 182-205
        • Grabauskas G.
        • Owyang C.
        Plasticity of vagal afferent signaling in the gut.
        Medicina. 2017; 53: 73-84
        • Carabotti M.
        • Scirocco A.
        • Maselli M.A.
        • Severi C.
        The gut-brain axis: interactions between enteric microbiota, central and enteric nervous systems.
        Ann Gastroenterol. 2015; 28: 203-209
        • Teratani T.
        • Mikami Y.
        • Nakamoto N.
        • Suzuki T.
        • Harada Y.
        • Okabayashi K.
        • et al.
        The liver-brain-gut neural arc maintains the Treg cell niche in the gut.
        Nature. 2020; 585: 591-596
        • Undem B.J.
        • Kollarik M.
        The role of vagal afferent nerves in chronic obstructive pulmonary disease.
        Proc Am Thorac Soc. 2005; 2 (discussion 71-2): 355-360
        • Kubin L.
        • Alheid G.F.
        • Zuperku E.J.
        • McCrimmon D.R.
        Central pathways of pulmonary and lower airway vagal afferents.
        J Appl Physiol. 2006; 101: 618-627
        • Thoren P.N.
        • Donald D.E.
        • Shepherd J.T.
        Role of heart and lung receptors with nonmedullated vagal afferents in circulatory control.
        Circ Res. 1976; 38: 2-9
        • Silverman H.A.
        • Chen A.
        • Kravatz N.L.
        • Chavan S.S.
        • Chang E.H.
        Involvement of neural transient receptor potential channels in peripheral inflammation.
        Front Immunol. 2020; 11590261
        • Chavan S.S.
        • Pavlov V.A.
        • Tracey K.J.
        Mechanisms and therapeutic relevance of neuro-immune communication.
        Immunity. 2017; 46: 927-942
        • Prescott S.L.
        • Liberles S.D.
        Internal senses of the vagus nerve.
        Neuron. 2022; S0896–6273 (01037-0)
        • Fallen E.L.
        Vagal afferent stimulation as a cardioprotective strategy? Introducing the concept.
        Ann Noninvasive Electrocardiol : the official journal of the International Society for Holter and Noninvasive Electrocardiology, Inc. 2005; 10: 441-446
        • Gronda E.
        • Francis D.
        • Zannad F.
        • Hamm C.
        • Brugada J.
        • Vanoli E.
        Baroreflex activation therapy: a new approach to the management of advanced heart failure with reduced ejection fraction.
        J Cardiovasc Med. 2017; 18: 641-649
        • Payne S.C.
        • Furness J.B.
        • Stebbing M.J.
        Bioelectric neuromodulation for gastrointestinal disorders: effectiveness and mechanisms.
        Nat Rev Gastroenterol Hepatol. 2019; 16: 89-105
        • Browning K.N.
        • Verheijden S.
        • Boeckxstaens G.E.
        The vagus nerve in appetite regulation, mood, and intestinal inflammation.
        Gastroenterology. 2017; 152: 730-744
        • Komegae E.N.
        • Farmer D.G.S.
        • Brooks V.L.
        • McKinley M.J.
        • McAllen R.M.
        • Martelli D.
        Vagal afferent activation suppresses systemic inflammation via the splanchnic anti-inflammatory pathway.
        Brain Behav Immun. 2018; 73: 441-449
        • Bugnard L.
        • Hill A.V.
        Electric excitation of the fin nerve of sepia.
        J Physiol. 1935; 83: 425-438
        • Cattell M.
        • Gerard R.W.
        The "inhibitory" effect of high-frequency stimulation and the excitation state of nerve.
        J Physiol. 1935; 83: 407-415
        • Ling D.
        • Luo J.
        • Wang M.
        • Cao X.
        • Chen X.
        • Fang K.
        • et al.
        Kilohertz high-frequency alternating current blocks nerve conduction without causing nerve damage in rats.
        Ann Transl Med. 2019; 7: 661
        • Pena E.
        • Pelot N.A.
        • Grill W.M.
        Non-monotonic kilohertz frequency neural block thresholds arise from amplitude- and frequency-dependent charge imbalance.
        Sci Rep. 2021; 11: 5077
        • Patel Y.A.
        • Butera R.J.
        Challenges associated with nerve conduction block using kilohertz electrical stimulation.
        J Neural Eng. 2018; 15031002
        • Kilgore K.L.
        • Bhadra N.
        Reversible nerve conduction block using kilohertz frequency alternating current.
        Neuromodulation : journal of the International Neuromodulation Society. 2014; 17: 242-255
        • Patel Y.A.
        • Saxena T.
        • Bellamkonda R.V.
        • Butera R.J.
        Kilohertz frequency nerve block enhances anti-inflammatory effects of vagus nerve stimulation.
        Sci Rep. 2017; 739810
        • Sarr M.G.
        • Billington C.J.
        • Brancatisano R.
        • Brancatisano A.
        • Toouli J.
        • Kow L.
        • et al.
        The EMPOWER study: randomized, prospective, double-blind, multicenter trial of vagal blockade to induce weight loss in morbid obesity.
        Obes Surg. 2012; 22: 1771-1782
        • Apovian C.M.
        • Shah S.N.
        • Wolfe B.M.
        • Ikramuddin S.
        • Miller C.J.
        • Tweden K.S.
        • et al.
        Two-year outcomes of vagal nerve blocking (vBloc) for the treatment of obesity in the ReCharge trial.
        Obes Surg. 2017; 27: 169-176
        • Morton J.M.
        • Shah S.N.
        • Wolfe B.M.
        • Apovian C.M.
        • Miller C.J.
        • Tweden K.S.
        • et al.
        Effect of vagal nerve blockade on moderate obesity with an obesity-related comorbid condition: the ReCharge study.
        Obes Surg. 2016; 26: 983-989
        • Shikora S.A.
        • Wolfe B.M.
        • Apovian C.M.
        • Anvari M.
        • Sarwer D.B.
        • Gibbons R.D.
        • et al.
        Sustained weight loss with vagal nerve blockade but not with sham: 18-month results of the ReCharge trial.
        J Obesity. 2015; (2015)365604
        • Negi S.
        • Bhandari R.
        • Rieth L.
        • Solzbacher F.
        In vitro comparison of sputtered iridium oxide and platinum-coated neural implantable microelectrode arrays.
        Biomed Mater. 2010; 515007
        • Negi S.
        • Bhandari R.
        • Rieth L.
        • Van Wagenen R.
        • Solzbacher F.
        Neural electrode degradation from continuous electrical stimulation: comparison of sputtered and activated iridium oxide.
        J Neurosci Methods. 2010; 186: 8-17
        • Levy T.J.
        • Ahmed U.
        • Tsaava T.
        • Chang Y.C.
        • Lorraine P.J.
        • Tomaio J.N.
        • et al.
        An impedance matching algorithm for common-mode interference removal in vagus nerve recordings.
        J Neurosci Methods. 2019; 330108467
        • Ahmed U.
        • Chang Y.C.
        • Lopez M.F.
        • Wong J.
        • Datta-Chaudhuri T.
        • Rieth L.
        • et al.
        Implant- and anesthesia-related factors affecting cardiopulmonary threshold intensities for vagus nerve stimulation.
        J Neural Eng. 2021; 18
        • Chang Y.C.
        • Cracchiolo M.
        • Ahmed U.
        • Mughrabi I.
        • Gabalski A.
        • Daytz A.
        • et al.
        Quantitative estimation of nerve fiber engagement by vagus nerve stimulation using physiological markers.
        Brain Stimul. 2020; 13: 1617-1630
        • Chang Y.
        • Ahmed U.
        • Tomaio J.N.
        • Rieth L.
        • Datta-Chaudhuri T.
        • Zanos S.
        Extraction of evoked compound nerve action potentials from vagus nerve recordings.
        in: 41st annual international conference of the IEEE engineering in medicine and biology society. EMBC, 2019: 6278-6281 (2019)
        • L Parker J.
        • H Shariati N.
        • M Karantonis D.
        Electrically evoked compound action potential recording in peripheral nerves.
        Bioelectron Med. 2018; 1: 71-83
        • Abbas A.
        • Mughrabi I.T.
        • Zanos S.
        Laryngeal electromyography to estimate A-fiber engagement by vagal stimuli in mice.
        in: 2021 10th international IEEE/EMBS conference on neural engineering (NER). 2021: 1121-1124
        • Dokos S.
        Modelling organs, tissues, cells and devices using MATLAB and COMSOL Multiphysics.
        Springer, Berlin, Germany2017
        • Lin Q.
        • Shivdasani M.N.
        • Tsai D.
        • Chang Y.C.
        • Jayaprakash N.
        • Zanos S.
        • et al.
        A computational model of functionally-distinct cervical vagus nerve fibers.
        in: Annual international conference of the IEEE engineering in medicine and biology society IEEE engineering in medicine and biology society annual international conference. 2020: 2475-2478 (2020)
        • Schwarz J.R.
        • Reid G.
        • Bostock H.
        Action-potentials and membrane currents in the human node of ranvier.
        Eur J Phys. 1995; 430: 283-292
        • McIntyre C.C.
        • Richardson A.G.
        • Grill W.M.
        Modeling the excitability of mammalian nerve fibers: influence of afterpotentials on the recovery cycle.
        J Nurophysiol. 2002; 87: 995-1006
        • Jayaprakash N.
        • Toth V.
        • Song W.
        • Vardhan A.
        • Levy T.
        • Tomaio J.
        • et al.
        Organ- and function-specific anatomical organization and bioelectronic modulation of the vagus nerve.
        bioRxiv. 2022; (2022.03.07)483266
        • Parker J.L.
        • Shariati N.H.
        • Karantonis D.M.
        Electrically evoked compound action potential recording in peripheral nerves.
        Bioelectron Med. 2017; 1: 71-83
        • Metcalfe B.W.
        • Nielsen T.N.
        • Donaldson N.N.
        • Hunter A.J.
        • Taylor J.T.
        First demonstration of velocity selective recording from the pig vagus using a nerve cuff shows respiration afferents.
        Biomed Eng Lett. 2018; 8: 127-136
        • Baker M.
        • Bostock H.
        • Grafe P.
        • Martius P.
        Function and distribution of three types of rectifying channel in rat spinal root myelinated axons.
        J Physiol. 1987; 383: 45-67
        • Kilgore K.L.
        • Bhadra N.
        Nerve conduction block utilising high-frequency alternating current.
        Med Biol Eng Comput. 2004; 42: 394-406
        • Kameneva T.
        • Maturana M.I.
        • Hadjinicolaou A.E.
        • Cloherty S.L.
        • Ibbotson M.R.
        • Grayden D.B.
        • et al.
        Retinal ganglion cells: mechanisms underlying depolarization block and differential responses to high frequency electrical stimulation of ON and OFF cells.
        J Neural Eng. 2016; 13016017
        • Berthoud H.R.
        • Neuhuber W.L.
        Functional and chemical anatomy of the afferent vagal system.
        Auton Neurosci : basic & clinical. 2000; 85: 1-17
        • Beaumont E.
        • Campbell R.P.
        • Andresen M.C.
        • Scofield S.
        • Singh K.
        • Libbus I.
        • et al.
        Cervical vagus nerve stimulation augments spontaneous discharge in second- and higher-order sensory neurons in the rat nucleus of the solitary tract.
        Am J Physiol Heart Circ Physiol. 2017; 313: H354-H367
        • Neuhuber W.L.
        • Berthoud H.R.
        Functional anatomy of the vagus system - emphasis on the somato-visceral interface.
        Auton Neurosci : basic & clinical. 2021; 236102887
        • Pelot N.A.
        • Grill W.M.
        Effects of vagal neuromodulation on feeding behavior.
        Brain Res. 2018; 1693: 180-187
        • Patel Y.A.
        • Butera R.J.
        Challenges associated with nerve conduction block using kilohertz electrical stimulation.
        J Neural Eng. 2018; 15031002
        • Bowman B.R.
        • McNeal D.R.
        Response of single alpha motoneurons to high-frequency pulse trains. Firing behavior and conduction block phenomenon.
        Appl Neurophysiol. 1986; 49: 121-138
        • Pelot N.A.
        • Grill W.M.
        In vivo quantification of excitation and kilohertz frequency block of the rat vagus nerve.
        J Neural Eng. 2020; 17026005
        • Ahmed U.
        • Chang Y.C.
        • Cracchiolo M.
        • Lopez M.F.
        • Tomaio J.N.
        • Datta-Chaudhuri T.
        • et al.
        Anodal block permits directional vagus nerve stimulation.
        Sci Rep. 2020; 10: 9221
        • Tosato M.
        • Yoshida K.
        • Toft E.
        • Struijk J.J.
        Quasi-trapezoidal pulses to selectively block the activation of intrinsic laryngeal muscles during vagal nerve stimulation.
        J Neural Eng. 2007; 4: 205-212
        • Vuckovic A.
        • Tosato M.
        • Struijk J.J.
        A comparative study of three techniques for diameter selective fiber activation in the vagal nerve: anodal block, depolarizing prepulses and slowly rising pulses.
        J Neural Eng. 2008; 5: 275-286
        • Pelot N.A.
        • Catherall D.C.
        • Thio B.J.
        • Titus N.D.
        • Liang E.D.
        • Henriquez C.S.
        • et al.
        Excitation properties of computational models of unmyelinated peripheral axons.
        J Nurophysiol. 2021; 125: 86-104
        • Bodenlos J.S.
        • Schneider K.L.
        • Oleski J.
        • Gordon K.
        • Rothschild A.J.
        • Pagoto S.L.
        Vagus nerve stimulation and food intake: effect of body mass index.
        J Diabetes Sci Technol. 2014; 8: 590-595
        • Papasavas P.
        • El Chaar M.
        • Kothari S.N.
        • American Society for M.
        • Bariatric Surgery Clinical Issues C.
        American Society for Metabolic and Bariatric Surgery position statement on vagal blocking therapy for obesity.
        Surg Obesity Related Dis: official journal of the American Society for Bariatric Surgery. 2016; 12: 460-461
        • Katona P.G.
        • Poitras J.W.
        • Barnett G.O.
        • Terry B.S.
        Cardiac vagal efferent activity and heart period in the carotid sinus reflex.
        Am J Physiol. 1970; 218: 1030-1037
        • Crosby N.D.
        • Janik J.J.
        • Grill W.M.
        Modulation of activity and conduction in single dorsal column axons by kilohertz-frequency spinal cord stimulation.
        J Nurophysiol. 2017; 117: 136-147
        • de Lartigue G.
        Role of the vagus nerve in the development and treatment of diet-induced obesity.
        J Physiol. 2016; 594: 5791-5815
        • Johannessen H.
        • Revesz D.
        • Kodama Y.
        • Cassie N.
        • Skibicka K.P.
        • Barrett P.
        • et al.
        Vagal blocking for obesity control: a possible mechanism-of-action.
        Obes Surg. 2017; 27: 177-185
        • Pelot N.A.
        • Behrend C.E.
        • Grill W.M.
        Modeling the response of small myelinated axons in a compound nerve to kilohertz frequency signals.
        J Neural Eng. 2017; 14046022
        • Baker E.
        • Lui F.
        Neuroanatomy, vagal nerve nuclei.
        StatPearls, Treasure Island (FL)2021
        • Yuan H.
        • Silberstein S.D.
        Vagus nerve and vagus nerve stimulation, a comprehensive review: Part I.
        Headache. 2016; 56: 71-78
        • Hudson A.E.
        Genetic reporters of neuronal activity: c-Fos and G-CaMP6.
        Methods Enzymol. 2018; 603: 197-220
        • Cunningham J.T.
        • Mifflin S.W.
        • Gould G.G.
        • Frazer A.
        Induction of c-Fos and DeltaFosB immunoreactivity in rat brain by Vagal nerve stimulation.
        Neuropsychopharmacology. 2008; 33 (official publication of the American College of Neuropsychopharmacology): 1884-1895
        • Huffman W.J.
        • Subramaniyan S.
        • Rodriguiz R.M.
        • Wetsel W.C.
        • Grill W.M.
        • Terrando N.
        Modulation of neuroinflammation and memory dysfunction using percutaneous vagus nerve stimulation in mice.
        Brain Stimul. 2019; 12: 19-29
        • Patel Y.A.
        • Kim B.S.
        • Rountree W.S.
        • Butera R.J.
        Kilohertz electrical stimulation nerve conduction block: effects of electrode surface area.
        IEEE Trans Neural Syst Rehabil Eng : a publication of the IEEE Engineering in Medicine and Biology Society. 2017; 25: 1906-1916
        • Buckley U.
        • Chui R.W.
        • Rajendran P.S.
        • Vrabec T.
        • Shivkumar K.
        • Ardell J.L.
        Bioelectronic neuromodulation of the paravertebral cardiac efferent sympathetic outflow and its effect on ventricular electrical indices.
        Heart Rhythm. 2017; 14: 1063-1070
        • Bhadra N.
        • Kilgore K.L.
        Direct current electrical conduction block of peripheral nerve.
        IEEE Trans Neural Syst Rehabil Eng : a publication of the IEEE Engineering in Medicine and Biology Society. 2004; 12: 313-324
        • McAllen R.M.
        • Shafton A.D.
        • Bratton B.O.
        • Trevaks D.
        • Furness J.B.
        Calibration of thresholds for functional engagement of vagal A, B and C fiber groups in vivo.
        Bioelectron Med. 2018; 1: 21-27
        • Patel Y.A.
        • Butera R.J.
        Differential fiber-specific block of nerve conduction in mammalian peripheral nerves using kilohertz electrical stimulation.
        J Nurophysiol. 2015; 113: 3923-3929
        • Yi G.
        • Grill W.M.
        Kilohertz waveforms optimized to produce closed-state Na+ channel inactivation eliminate onset response in nerve conduction block.
        PLoS Comput Biol. 2020; 16e1007766
        • Losanno E.
        • Badi M.
        • Wurth S.
        • Borgognon S.
        • Courtine G.
        • Capogrosso M.
        • et al.
        Bayesian optimization of peripheral intraneural stimulation protocols to evoke distal limb movements.
        J Neural Eng. 2021; 18
        • Gold M.R.
        • Van Veldhuisen D.J.
        • Hauptman P.J.
        • Borggrefe M.
        • Kubo S.H.
        • Lieberman R.A.
        • et al.
        Vagus nerve stimulation for the treatment of heart failure: the INOVATE-HF trial.
        J Am Coll Cardiol. 2016; 68: 149-158
        • Nasi-Er B.G.
        • Wenhui Z.
        • HuaXin S.
        • Xianhui Z.
        • Yaodong L.
        • Yanmei L.
        • et al.
        Vagus nerve stimulation reduces ventricular arrhythmias and increases ventricular electrical stability.
        Pacing Clin Electrophysiol : PACE. 2019; 42: 247-256
        • Sundman E.
        • Olofsson P.S.
        Neural control of the immune system.
        Adv Physiol Educ. 2014; 38: 135-139
        • Tanaka S.
        • Abe C.
        • Abbott S.B.G.
        • Zheng S.
        • Yamaoka Y.
        • Lipsey J.E.
        • et al.
        Vagus nerve stimulation activates two distinct neuroimmune circuits converging in the spleen to protect mice from kidney injury.
        Proc Natl Acad Sci U S A. 2021; 118
        • Pardo J.V.
        • Sheikh S.A.
        • Kuskowski M.A.
        • Surerus-Johnson C.
        • Hagen M.C.
        • Lee J.T.
        • et al.
        Weight loss during chronic, cervical vagus nerve stimulation in depressed patients with obesity: an observation.
        Int J Obes. 2007; 31: 1756-1759
        • Bioelectronics SPARC at NIH
        Nat Biotechnol. 2014; 32: 855
        • Rajendran P.S.
        • Challis R.C.
        • Fowlkes C.C.
        • Hanna P.
        • Tompkins J.D.
        • Jordan M.C.
        • et al.
        Identification of peripheral neural circuits that regulate heart rate using optogenetic and viral vector strategies.
        Nat Commun. 2019; 10: 1944
        • Mughrabi I.T.
        • Hickman J.
        • Jayaprakash N.
        • Papadoyannis E.S.
        • Abbas A.
        • Chang Y.-C.
        • et al.
        An implant for long-term cervical vagus nerve stimulation in mice.
        bioRxiv. 2020; (2020.06.20)160473
        • Yoo P.B.
        • Lubock N.B.
        • Hincapie J.G.
        • Ruble S.B.
        • Hamann J.J.
        • Grill W.M.
        High-resolution measurement of electrically-evoked vagus nerve activity in the anesthetized dog.
        J Neural Eng. 2013; 10026003
        • Ahmed U.
        • Chang Y.C.
        • Zafeiropoulos S.
        • Nassrallah Z.
        • Miller L.
        • Zanos S.
        Strategies for precision vagus neuromodulation.
        Bioelectron Med. 2022; 8: 9
        • Joseph L.
        • Butera R.J.
        High-frequency stimulation selectively blocks different types of fibers in frog sciatic nerve.
        IEEE Trans Neural Syst Rehabil Eng : a publication of the IEEE Engineering in Medicine and Biology Society. 2011; 19: 550-557
        • Neudorfer C.
        • Chow C.T.
        • Boutet A.
        • Loh A.
        • Germann J.
        • Elias G.J.
        • et al.
        Kilohertz-frequency stimulation of the nervous system: a review of underlying mechanisms.
        Brain Stimul. 2021; 14: 513-530
        • Vetter P.
        • Roth A.
        • Hausser M.
        Propagation of action potentials in dendrites depends on dendritic morphology.
        J Neurophysiol. 2001; 85: 926-937
        • Spruston N.
        Pyramidal neurons: dendritic structure and synaptic integration.
        Nat Rev Neurosci. 2008; 9: 206-221
        • Wang J.
        • Jacoby R.
        • Wu S.M.
        Physiological and morphological characterization of ganglion cells in the salamander retina.
        Vision Res. 2016; 119: 60-72
        • Rattay F.
        Analysis of models for external stimulation of axons.
        IEEE (Inst Electr Electron Eng) Trans Biomed Eng. 1986; 33 (BME): 974-977
        • Guo T.
        • Tsai D.
        • Yang C.Y.
        • Al Abed A.
        • Twyford P.
        • Fried S.I.
        • et al.
        Mediating retinal ganglion cell spike rates using high-frequency electrical stimulation.
        Front Neurosci. 2019; 13
        • Schild J.H.
        • Clark J.W.
        • Hay M.
        • Mendelowitz D.
        • Andresen M.C.
        • Kunze D.L.
        A- and C-type rat nodose sensory neurons: model interpretations of dynamic discharge characteristics.
        J Nurophysiol. 1994; 71: 2338-2358
        • Tackmann W.
        • Lehmann H.J.
        refractory period in human sensory nerve fibres.
        Eur Neurol. 1974; 12: 277-292
        • Kimura J.
        • Yamada T.
        • Rodnitzky R.L.
        Refractory period of human motor nerve fibres.
        J Neurol Neurosurg Psychiatr. 1978; 41: 784-790
        • Heffer L.F.
        • Sly D.J.
        • Fallon J.B.
        • White M.W.
        • Shepherd R.K.
        • O'Leary S.J.
        Examining the auditory nerve fiber response to high rate cochlear implant stimulation: chronic sensorineural hearing loss and facilitation.
        J Nurophysiol. 2010; 104: 3124-3135
        • Bhadra N.
        • Lahowetz E.A.
        • Foldes S.T.
        • Kilgore K.L.
        Simulation of high-frequency sinusoidal electrical block of mammalian myelinated axons.
        J Comput Neurosci. 2007; 22: 313-326
        • Guo T.
        • Yang C.Y.
        • Tsai D.
        • Muralidharan M.
        • Suaning G.J.
        • Morley J.W.
        • et al.
        Closed-loop efficient searching of optimal electrical stimulation parameters for preferential excitation of retinal ganglion cells.
        Front Neurosci. 2018; 12: 168
        • Muralidharan M.
        • Guo T.
        • Shivdasani M.N.
        • Tsai D.
        • Fried S.
        • Li L.
        • et al.
        Neural activity of functionally different retinal ganglion cells can be robustly modulated by high-rate electrical pulse trains.
        J Neural Eng. 2020; 17045013
        • Woodbury D.M.
        • Woodbury J.W.
        Effects of vagal stimulation on experimentally induced seizures in rats.
        Epilepsia. 1990; 31: S7-S19
        • Nicolai E.N.
        • Settell M.L.
        • Knudsen B.E.
        • McConico A.L.
        • Gosink B.A.
        • Trevathan J.K.
        • et al.
        Sources of off-target effects of vagus nerve stimulation using the helical clinical lead in domestic pigs.
        J Neural Eng. 2020; 17046017
        • Aristovich K.
        • Donega M.
        • Fjordbakk C.
        • Tarotin I.
        • Chapman C.A.R.
        • Viscasillas J.
        • et al.
        Model-based geometrical optimisation and in vivo validation of a spatially selective multielectrode cuff array for vagus nerve neuromodulation.
        J Neurosci Methods. 2021; 352109079