Impaired cerebrospinal fluid (CSF) flow has been proposed as a pathological mechanism for numerous neurodegenerative diseases, including Alzheimer's disease [1,2], Parkinson's disease [3], and Huntington's disease [4]. It has been reported to result from the accumulation of pathological proteins such as tau and amyloid-beta [1,2,5]. In the same context, improving pathological impairments in CSF flow and facilitating the clearance of accumulated toxic waste in the brain to restore intracerebral homeostasis can be proposed as a therapeutic strategy for neurodegenerative diseases. Several pharmacological approaches have been proposed to modulate CSF flow and manage neurodegenerative diseases as ‘glymphatic enhancers’, including dexmedetomidine, an adrenergic alpha-2 agonist, and pan-adrenergic receptor inhibitors [6-8]. However, these pharmacological approaches carry the risk of side effects such as excessive sedation and drowsiness with routine use, limiting their direct application [6].

Vagus nerve stimulation (VNS) is a bioelectrical medicinal intervention providing a non-pharmaceutical treatment option for various diseases [9]. Especially, auricular vagus nerve stimulation (aVNS), stimulating the auricular branch of the vagus nerve innervating the cymba concha, is a non-invasive and accessible neuromodulation technique, and therefore, well-tolerated by patients [10]. It has been demonstrated to be effective on several neurological disorders including dementia and cognitive impairment [11,12], Parkinson’s disease [13] and neuropathic pain [14]. Concerning the mechanism of the therapeutic effects of VNS, modulating CSF flow has been suggested with in vitro and in vivo studies [11,15]. However, despite the potential of VNS to modulate CSF flow, the optimization of stimulation parameters—such as duration, number, intensity, and the frequency of stimulation—has not been thoroughly explored and has largely been approached empirically.

Herein, we aimed to identify the most efficient frequency for modulating CSF flow in relation to the effective parameters of aVNS. To this end, we utilized imaging techniques that allow in vivo observation of CSF and cerebral vasculatures. We investigated the modulation of arterial pulsatility, proposed as a driving factor for CSF flow [16], and the effect of aVNS at different frequencies on CSF influx to determine the effective stimulation frequency.

Animals

Male C57BL/6 mice aged 8 to 10 weeks were used for the studies. The mice were sourced from Koatech. They were kept on a 12-h light/dark cycle with unlimited access to food and water. To minimize diurnal variation of CSF flow, all experiments were conducted at the same time each day. All animal procedures were conducted according to the protocols approved by the Institutional Animal Care and Use Committee of Kyung Hee University (KHUASP-24-240).

aVNS

aVNS was applied based on previous studies [11,14]. Transcutaneous neodymium electrodes were used to stimulate the cymba concha and dorsal auricle innervated by the auricular branch of the vagus nerve [17]. Rectangular bipolar electrical pulses were transduced with the IX-RA-834 (iWorx Systems) and pulse width was determined according to each frequency with 50% duty cycle (2 Hz: 250 ms; 40 Hz: 12.5 ms; 100 Hz: 5 ms). Before stimulating the left auricle, a small amount of electrically conductive gel (SignaGel; Parker Labs) was applied. The total stimulation time was 30 min with a voltage of 5 V and the specified frequency (Supplementary Video 1).

Drugs and tracers

For surgery, stimulation, and imaging, mice were anesthetized with a mixture of Zoletil 50 (Virbac) at a concentration of 25 mg/ml and xylazine (X1126; Sigma-Aldrich) at 5 mg/ml. The anesthesia was induced through an intraperitoneal injection of approximately 0.05 ml of the Zoletil-xylazine (Z/X) mixture, delivering doses of 62.5 mg/kg for Zoletil and 12.5 mg/kg for xylazine. As plasma and CSF tracer, fluorescein isothiocyanate (FITC)–dextran (2,000 kDa; FD2000S, Sigma-Aldrich) was dissolved at a concentration of 0.5% in phosphate-buffered saline (PBS) or artificial CSF (150 mM NaCl, 2.5 mM KCl, 10 mM HEPES, 2 mM CaCl2, 1 mM MgCl2, pH = 7.4).

Cisterna magna cannulation and tracer infusion

The procedure for cisterna magna cannulation was performed based on a previous study [18]. After being mounted on the stereotaxic apparatus, the head was positioned to provide access to the atlanto-occipital membrane and the connective tissue was separated to expose the membrane. A 30G dental needle cannula connected to a polyethylene tube (PE10) and filled with artificial CSF was carefully inserted into the cisterna magna to a depth of 1–2 mm, ensuring the entire bevel was under the dura mater. Mice that exhibited subdural bleeding or significant, continuous CSF leakage were excluded from the study. After drying the membrane surface, the needle was secured in place, and 2–3 drops of 3M Vetbond Tissue Adhesive were applied, followed by the application of dental cement to cover the skull and the exposed membrane. Upon completion of cannulation, a total volume of 10 µl of FITC CSF tracers was infused into the cisterna magna at a rate of 1 µl/min using a syringe pump.

In vivo two photon imaging on vascular pulsation

To compose a cranial window, a 2 mm × 2 mm square-shaped craniotomy was created on the left cerebral cortex, positioned 0.5 mm posterior and 1.5 mm lateral to the bregma keeping the dura mater intact. Mice showing parenchymal damage or significant bleeding were excluded from the study. The exposed cortex was thoroughly washed with artificial CSF, then covered with a cover glass (Matsunami Glass Ind., Ltd.), and sealed with Vetbond. A custom-made metal ring, designed for mounting the mice with a two photon microscopy setup, was fixed on the skull with dental cement.

In vivo imaging was conducted using a Ti:sapphire laser system (Chameleon; Coherent) equipped with a two photon microscope (FVMPE-RS; Olympus) after a week of recovery. Followed by an aVNS session of 30 min, 0.05 ml of FITC dissolved in PBS was injected retro-orbitally to visualize blood vessel. The imaging of arterial vasomotion was performed with a water immersion objective lens (LUMFLN60XW, 1.1 NA; Olympus). An excitation laser wavelength of 920 nm was used to visualize the FITC tracer. The imaging plane was set at a depth of 30–50 µm to ensure visualization of pial artery confirmed by pulsation and flow. We imaged independent arteries that did not overlap in flow by checking the flow in all imaging fields, and 1–3 vessels were identified in each mouse. Images were captured at a resolution of 512 × 512 pixels (0.414 µm per pixel). Time-series images of 1,000 frames were acquired for 65.714 sec.

In vivo transcranial macroscopic imaging for CSF

Transcranial macroscopic imaging was conducted according to a previous study [19]. Under Z/X anesthesia, the entire calvarium was exposed followed by cisterna magna cannulation. After injecting the CSF tracers, the mice were positioned in a fluorescence microscopy with a digital camera (DP23M, Olympus) and an objective lens (PLAPON1.25X, 0.04NA; Olympus). A custom-made metal ring covering the entire calvarium was used for stable fixation. The entire skull was illuminated using a light source (U-HGLGPS; Olympus). Once the influx of the CSF tracer around the middle cerebral artery (MCA) observed, imaging session commenced using cellSens software (Olympus). The images, with a resolution of 1,544 × 1,038, were taken at 12-sec intervals for 60 min, resulting in a total of 301 slices. All imaging sessions were conducted with the same acquisition parameters including exposure time and gain.

Image analysis

All images were processed and analyzed using ImageJ (1.54f; National Institutes of Health) and MATLAB (R2024a; Mathworks Inc.), with the entire process being standardized. For vasomotion analysis, the acquired framescan images were exported as TIFF files and motion corrected using the NoRMCorre algorithm for non-rigid motion correction [20]. From the motion-corrected vascular images, vessel diameter was extracted using VasoMetrics, an ImageJ plugin for spatiotemporal analysis of microvascular diameter [21]. Vasomotion event detection was performed using Python code from a previous study [22]. Raw data were preprocessed with a Savitzky-Golay filter (window size 13, polynomial order 3), and peaks were identified using the find_peaks function (height 2, width 5). For CSF influx analysis, macroscopic CSF images were imported into ImageJ and analyzed using 200-pixel diameter round regions of interest on both MCA areas.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 10 (GraphPad Software Inc.) and Microsoft Excel. Data are presented as mean ± SEM in the graphical representations. Detailed descriptions on the statistical methods and the exact data points are described in the figure legends.

Low-frequency aVNS increased vasomotion events

Arterial pulsation has been proposed as the main factor that drives CSF flow in the perivascular space (PVS) [23]. Furthermore, a potential strategy to modulate CSF flow involving the boost of arterial pulsation has been reported [22]. It has also been demonstrated that VNS affects hemodynamic factors such as blood pressure, heart rate, and blood flow volume [24-27]. Taken together, we investigated whether aVNS modulates arterial vasomotion, resulting in CSF flow enhancement. To evaluate the arterial vasomotion by aVNS, we utilized in vivo two photon imaging to acquire cortical vascular dynamics (Fig. 1A). At 15 Hz vascular imaging, arteries exhibited more pronounced pulsating motion compared to veins. Raw traces from the artery showed greater diameter changes than those from the vein (Fig. 1B). Vasomotion events were determined from the filtered trace with high and wide amplitude changes (Fig. 1C). After 30 min of aVNS, characteristic movements of arterial vessels were observed depending on the frequency (Fig. 1D and Supplementary Video 2). Quantification of vasomotion revealed that low-frequency aVNS at 2 Hz significantly increased vasomotion compared to the sham group (Fig. 1E; p = 0.0098, vs. Sham), however, stimulation at 40 Hz and 100 Hz did not show statistical significance compared to the sham group (p > 0.999, Sham vs. 40 Hz; p = 0.4254, Sham vs. 100 Hz). These results suggest that low-frequency aVNS promotes arterial vasomotion, demonstrating its potential to enhance CSF flow.

Figure 1. Low-frequency (2 Hz) auricular vagus nerve stimulation (aVNS) promotes the vasomotion events number, rather than 40 Hz and 100 Hz. (A) In vivo two photon imaging setup and representative image showing artery and vein. Scale bar = 50 µm. (B) Representative traces from vessels show variable changes in diameter, with arteries exhibiting more pronounced diameter changes compared to veins. (C) Example traces showing raw trace (gray), filtered trace (blue) and detected vasomotion (green ×-shaped mark). A Savitzky-Golay filter was applied to the raw trace, and the peaks were identified. Scale bar = 2 µm in diameter and 10 sec in time. (D) Sample time series of arterial diameter changes following aVNS. (E) Quantified vasomotion events after aVNS (n = 8 for sham [3 mice], n = 6 for 2 Hz aVNS [3 mice], n = 6 for 40 Hz aVNS [2 mice], and n = 4 for 100 Hz aVNS [2 mice]). p-values were determined by one-way ANOVA followed by Dunnett’s multiple comparisons test. **p < 0.01; ns, not significant.

Low-frequency aVNS enhanced cortical CSF influx

To validate the effective frequency of aVNS for enhancing CSF influx most effectively, we utilized in vivo transcranial macroscopic imaging technique (Fig. 2A). In vivo transcranial macroscopic CSF imaging enables to visualize CSF dynamics on large imaging field of entire cortex with minimal invasiveness [19]. One hour of CSF imaging showed CSF tracer influx in PVS along both MCAs (Fig. 2B and Supplementary Video 3). Low-frequency (2 Hz) aVNS demonstrated a prominent increase in tracer mean intensity during the 30-min stimulation period, with the signal continuing to increase even after the aVNS had ended (Fig. 2C). To quantify the dynamics of CSF tracer influx, we analyzed the peak intensity, area under the curve (AUC), and the slope of the increasing trend over 30 min using simple linear regression for each individual signal trace. The results showed that, while there was no statistical significance in the peak intensity (Fig. 2D; p = 0.0717), the highest values were observed at 2 Hz. Both the AUC and the slope were significantly increased by the 2 Hz aVNS compared to the sham (Fig. 2E, F). At high-frequency (100 Hz) aVNS, there was a slight increasing trend in CSF influx, but it was not statistically significant compared to the sham group. Furthermore, at middle-frequency (40 Hz) aVNS, the CSF signal change was nearly identical to that of the sham, showing no significant difference.

Figure 2. Low-frequency (2 Hz) auricular vagus nerve stimulation (aVNS) enhances cortical cerebrospinal fluid (CSF) influx around the middle cerebral artery (MCA). (A) Schematic illustration showing in vivo transcranial macroscopic CSF imaging and the flow of CSF tracer after injected into the cisterna magna. (B) Influxed CSF tracers into the cortex along the MCAs. Scale bar = 1,000 µm. (C) Normalized CSF signal traces during and after 30 min of aVNS (n = 8 for each group). Based on each normalized trace, peak intensity (D), area under the curve (E), and increasing slope (F) during aVNS were calculated. p-values were determined by one-way ANOVA followed by Dunnett’s multiple comparisons test. **p < 0.01; *p < 0.05; ns, not significant.

We investigated the effective stimulation frequency of aVNS with respect to CSF modulation. To visualize the dynamics of cerebral vasculature and cortical CSF, we adopted in vivo imaging methods, including two photon and transcranial macroscopic imaging. As a result, continuous low-frequency aVNS of 2 Hz promoted arterial vasomotion, thereby enhancing cortical CSF influx and manifesting a brain clearance effect. It suggests that low-frequency aVNS could serve as a neuromodulatory intervention for neurodegenerative disorders characterized by impaired CSF circulation.

Impaired CSF circulation, or glymphatic dysfunction, has been demonstrated through delayed influx of CSF tracers into the brain parenchyma, or reduced efflux of tracers [28-30]. From the perspectives of impaired influx and efflux, the consequent accumulation of amyloid-beta [1], tau [2], or excessive fluid [8] leads to behavioral impairments. Also, it has been suggested that restoring the CSF dynamics in neurological diseases not only helps in removing metabolic wastes from the brain but also exerts beneficial effects via keeping the homeostasis of the brain and regulating neuroinflammation [31,32]. Several studies have suggested that non-pharmacological interventions including multisensory gamma stimulation [22], focused ultrasound stimulation [33,34], and aVNS [11] improved the glymphatic function and attenuated the pathological phenotypes. Such non-pharmacological approaches, particularly for managing neurodegenerative diseases with progressively worsening characteristics, have the potential to be applied more safely and effectively over the long term. Thus, they have been increasingly suggested as promising therapeutic strategies for clinical application.

Non-pharmacological neuromodulation techniques encompass various adjustable factors, unlike pharmacological methods where the factors such as dosage can be controlled. Determining the most effective stimulation parameters, including quantitative aspects like the number and duration of stimulations, as well as factors such as duty cycle, intensity or amplitude, and frequency of stimulation, has been relatively unexplored and primarily based on empirical approaches. Particularly, the frequency is a crucial factor involving electrical stimulation. The mechanisms of action vary significantly depending on the frequency used in the cases of electroacupuncture [35,36], electrical field stimulation [37], and transcutaneous electrical nerve stimulation [38].

Like cervical VNS, aVNS is most commonly used at 20 Hz, and a systematic review reported an average frequency of 22.5 Hz [39]. There have been reports suggesting a potential difference in the neuropeptides released during aVNS depending on the frequency, similar to the context observed with electroacupuncture [40]. A functional magnetic resonance imaging (fMRI) study involving healthy subjects reported that respiratory-gated aVNS at 2 Hz and 100 Hz significantly increased responses in the locus coeruleus and raphe nuclei, which are nuclei of monoamine neurotransmitters such as norepinephrine and serotonin. However, at the conventional frequency of 25 Hz, there was minimal signal change observed on fMRI [41]. Although numerous studies report on the parameters of aVNS, few have applied it continuously for extended periods of around 30 min as we have. Also, the possibility of interactions between frequency, amplitude, and duty cycle presents limitations in ensuring the effectiveness of low-frequency aVNS for disease treatment, especially since we have only addressed the frequency among the adjustable parameters of aVNS. For instance, it cannot be ruled out the possibility that the effective frequency may vary at certain intensities.

Therefore, further study should demonstrate that the increase in CSF flow induced by low-frequency aVNS and the resulting so-called brain clearance effect leads to substantial improvements in disease conditions. It should include behavioral improvements and a reduction in specific brain markers such as amyloid-beta. Additionally, future research is necessary on the modulatory effects resulting from changes in adjustable parameters of aVNS other than just stimulation frequency.

This research is funded by the BK21 FOUR program of Graduate School, Kyung Hee University (GS-1-JX-ON-20230363).

None.

Neurogrin Inc., founded by S.K.K., holds the patent applications related to the contents of this article (#10-2022-0072656 in Korea and #PCT/KR2022/013378), in which S.C. and S.K.K. are listed as inventors. The other author (I.S.B. and K.L.) declares no conflict of interest.

Supplementary data including three videos can be found with this article online at https://doi.org/10.4196/kjpp.24.266