Low-Intensity Transcranial Focused Ultrasound Engages Parvalbumin-Positive GABAergic Interneurons in a Humanized Mouse Model of Chronic Pain: From Electrophysiology to Cellular Investigation

Abstract

Objective.

Low-intensity transcranial focused ultrasound (tFUS) offers high spatial specificity and deep brain penetration, representing a promising non-invasive approach for modulating brain activity and behavior. Although emerging studies indicate that tFUS can modulate pain-related behaviors in rodents and humans, the underlying network-level and cellular mechanisms remain unknown. This study investigates the effects of tFUS neuromodulation on inhibitory neural circuits in a humanized mouse model of chronic pain, integrating electrophysiological, molecular, and histological analyses across a cohort of 50 animals, including wild-type controls.

Approach.

A 128-element random array transducer was used to precisely target the pain-processing brain circuit, while a non-invasive and flexible 30-channel electroencephalography (EEG) electrode was applied to record local evoked responses, topographical activity, and global excitation and inhibition dynamics, which were further validated by optogenetics experiments. Cellular-level modulation and safety outcomes were evaluated through blinded histological examination.

Main results.

We found that tFUS produced robust modulation of local and global brain activity, characterized by suppression of local theta oscillations and enhancement of network-level inhibitory dynamics. These electrophysiological patterns aligned with those observed during optogenetic activation of parvalbumin (PV) interneurons. Immunohistochemistry further showed significant increases in inhibitory neuronal markers, including elevated expressions of Glutamate Decarboxylase 67 and PV. Blinded histological assessment confirmed the absence of tissue damage, supporting the safety of the stimulation paradigm.

Significance.

These findings demonstrated that tFUS stimulation non-invasively engages PV GABAergic inhibitory circuits in a chronic pain mouse model, providing mechanistic insight and supporting its development as a precise and safe neuromodulation technology for clinical translation.

1. Introduction

Treating chronic pain by means of non-pharmacological approaches is urgently needed due to the severe side effects of drugs including abuse, overdose, and addiction, all of which contribute to a significant health pandemic [1–3]. To reduce reliance on pharmacological treatments, neuromodulation strategies targeting pain-processing circuits have been developed [4–9]. Deep brain stimulation (DBS) is effective for movement disorders and has shown promise in managing chronic pain [4,5], but it requires invasive surgical implantation [4,6,8]. Non-invasive neuromodulatory alternatives such as transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS) demonstrated diverse treatment efficacies [10–13]; however, they typically influence broad areas and have limited ability to reach deeper brain structures [6,7].

Transcranial focused ultrasound (tFUS) is an emerging neuromodulation technology with high spatial specificity and penetration depth compared to other non-surgical brain stimulation methods [14–17], enabling non-invasive modulation of targeted brain regions [14– 19] along with behavior control [20,21] and treatment of neuropsychological disorders [22–26]. Low-intensity tFUS has recently shown potential for pain treatment in both pre-clinical and human studies. In humans, tFUS targeting the anterior cingulate cortex modulated local activity [27] and alleviated both chronic pain symptoms and responses to acute heat stimuli [28]. Additional evidence demonstrated antinociceptive effects from anterior thalamus stimulation [29] and mood improvements with posterior frontal cortex tFUS in pain patients [30]. In animal models, stimulation of the periaqueductal gray (PAG) has been shown to attenuate formalin-induced nociceptive activity in the dorsal horn of the spinal cord [31], while stimulation of the primary somatosensory cortex (S1), insula, or both regions simultaneously significantly suppressed pain-related behaviors in a humanized sickle cell disease (SCD) mouse model [32,33]. These studies suggest that targeted neuromodulation in the brain can ameliorate pain; however, further investigation is needed to elucidate the underlying mechanisms driving the modulation of pain-related behaviors and neural circuits. A deeper understanding of both cellular and electrophysiological effects, including sustained modulation of neuronal expression and network-level dynamics, will be essential to bridge critical knowledge gaps.

In this study, we investigate the mechanisms and safety of low-intensity tFUS via integrated electrophysiological and cellular approaches in a humanized mouse model of chronic pain, building upon our prior behavioral observations of pulse repetition frequency (PRF)-dependent inhibitory (low PRF) and excitatory (high PRF) effects under the validated stimulation parameter combination [19,32,33]. We hypothesize that tFUS with a lower PRF can significantly activate specific GABAergic interneurons, such as the Parvalbumin (PV) cells, and enhance inhibitory effects in the mouse model of SCD which expresses >99% human sickle hemoglobin and exhibits both severe chronic pain and hyperalgesia [34,35]. We first examine local evoked responses, topographical brain activity, global network dynamics in excitation/inhibition (E/I) balance using flexible 30-channel electroencephalography (EEG) recordings paired with a highly-focused 128-element random array transducer. At the cellular level, we next incorporate immunohistochemistry (IHC) targeting inhibitory circuit markers, including PV and Glutamate Decarboxylase 67 (GAD67), a key GABA-synthesizing enzyme [36–39]. Additionally, we evaluate the safety profile of 14-day repeated tFUS through blinded histological analysis of the same brain samples examined by IHC. This comprehensive approach aims to uncover how low-intensity tFUS non-invasively modulates the inhibitory circuits at both the cellular and systems levels, thereby providing foundational insights into its translational potential for managing chronic pain.

2. Materials and methods

2.1. Animals

All animal experimental procedures were approved by the Institutional Animal Care and Use Committee of Carnegie Mellon University and the Long Beach VA Medical Center, and complied with the National Institutes of Health guidelines. Female wild-type mice (C57BL/6J, Jackson Laboratory, Bar Harbor, ME, USA), female and male humanized transgenic homozygous Berkley sickle cell mice (HbSS-BERK), and male Wistar outbred rats (Hsd:WI, Envigo Laboratory, Indianapolis, IN, USA) were used as subjects. HbSS-BERK mice were on a mixed genetic background (FVB/N, 129, DBA/2, C57BL/6, and Black Swiss) with murine α and β globin knockouts (α^−/− and β^−/−) expressing >99% human sickle hemoglobin S (HbS), showing features of SCD including chronic pain [34,40]. Mice were generated in-house by cross-breeding homozygous (SS) males with hemizygous (AS) females [40]. All pups were phenotyped and those homozygous for sickle Hb were used. Mice were bred and raised with ad libitum access to food and water on a 12-h light/dark cycle in conventional housing. Older mice were used in this study, including HbSS-BERK mice (13-17 months) and age-matched wild-type mice (11-14 months), as the hallmark feature chronic pain severity in SCD is known to worsen with age [33,35,41]. Animals were assigned to experimental groups as follows: Single-session tFUS experiments in HbSS-BERK mice (11 female, 6 male), multi-session tFUS experiments in female HbSS-BERK mice, control experiments in female wild-type mice, and optogenetic validation experiments in male Wistar rats aged 3-6 months.

2.2. Characterization of custom-developed 128-element random array transducer

To achieve high spatial specificity for precise stimulation within the small mouse brain, a previously developed spherically-concave and highly-focused 128-element random array ultrasound transducer at 1.5 MHz center frequency (H276, Sonic Concepts, Inc., Bothell, WA, USA) was used in this study [32]. The transducer has an aperture radius of 8.5 mm and exit-plane diameter of 15 mm, defined as the effective active diameter of the acoustic radiating surface. Sub-apertures (0.8 mm diameter) were arranged in a randomized layout with a 1 mm pitch.

To evaluate the pressure and intensity profiles of tFUS through the mouse skull, we utilized a customized ex-vivo 3D acoustic field mapping system with a needle hydrophone (HNR500, Onda Corporation, Sunnyvale, CA, USA) and a three-axis positioning stage (XSlide, Velmex, Inc., Bloomfield, NY, USA) [33,42]. A freshly harvested skull from a euthanized mouse was positioned between the array transducer and the hydrophone to closely mimic the in vivo tFUS configuration.

2.3. tFUS stimulation targeting procedures and parameters

Mice were anesthetized using 3% isoflurane with a 1 L/min oxygen flow for induction, and 1-1.5% isoflurane with a 0.3-0.5 L/min oxygen flow for maintenance. A heating pad was used to maintain body temperature, and ophthalmic ointment was applied to protect the eyes. Core body temperature was regularly monitored using a rectal temperature probe.

To estimate the targeted brain circuits using the 128-element random array transducer, we followed previously established targeting procedures [32]. Briefly, the Allen Mouse Brain Atlas was referenced to determine the coordinate of the ultrasound focus. The transducer was positioned by visually aligning the transducer’s X-axis with the posterior edge of zygomatic structure and the Y-axis with the mouse’s rostral caudal midline (figure 1(A) and 1(B) and figure S1). We applied single- or multi-session tFUS with specific parameters including tone-burst duration (TBD), ultrasound duration (UD), inter-sonication interval (ISoI), pulse repetition frequency (PRF), and total sonication time (TST) (figure 1(C)). All acoustic parameters were held constant across sessions, and transducer coupling and head positioning were consistently reproduced at the same or closely approximated location based on our established targeting procedures.

Figure 1. Single-session tFUS induced local brain circuit modulation revealed by EEG.

(A) Schematic of tFUS stimulation using a 128-element random array transducer. It features dynamic beam steering capabilities as demonstrated by red and blue beams when targeting specific brain circuits. (B) 128-element random array transducer and 30-channel flexible EEG electrode setup. Highly focused ultrasound stimulation beam represents S1HL stimulation with the location of the analyzed electrode representing S1HL activity. (C) Single-session tFUS stimulation with selected parameters: a PRF of 40 Hz or control value, a UD of 100-ms, an ISoI of 2-sec, and a TST of 20-min. Recording timeline below illustrates EEG acquisition during a 99-trial baseline period (TP1) followed by 600 trials of tFUS stimulation or control (TP2-TP7). (D) Representative EEG traces and topographic maps during tFUS with PRF of a 3 kHz related to excitatory effect show that a single-session tFUS can perturb local brain activity. The gray rectangle represents the tFUS stimulation or negative control period. (E) Significant suppression of low-frequency EEG oscillations, particularly within the theta and alpha bands, was observed in the tFUS-treated region when using a 40 Hz PRF in SCD mice (n=17; 11 female and 6 male). Normalized power represents averaged values across all 30 channels for a given frequency band. (F) Significant suppression of low-frequency EEG oscillations, especially in the theta, alpha, and beta bands, was similarly observed during optogenetic activation of PV-positive interneurons in S1 of rats (n=8). Normalized power represents averaged values across all 30 channels for a given frequency band. ^#p <0.05, ^##p <0.01 using t test with Mann-Whitney test; *p <0.05 using t test with Wilcoxon matched-pairs signed rank test. ns, not significant.

2.4. In vivo electrophysiology recordings and data analysis in mice

After securing the mouse head, cranial fur was removed using a depilatory cream under the isoflurane anesthesia. EEG recordings were acquired using a 30-channel mouse electrode array (HC32 Mouse EEG, NeuroNexus, Ann Arbor, MI, USA) placed on the scalp using a predefined geometric reference corresponding to bregma, without direct exposure of the skull landmark and stabilized with saline. As illustrated in figure 1(B) and figure S1, the intersection of the X- and Y-axes served as the geometric origin for positioning. The relative position of the electrode array with respect to bregma was estimated using atlas-derived anatomical coordinates from the Allen Mouse Brain Atlas. EEG data acquisition was performed through a ZIF-Clip headstage connected to a PZ5 NeuroDigitizer and Synapse software (Tucker-Davis Technologies, Alachua, FL, USA). Stimulation onset triggers were generated and logged within the same Synapse software control loop governing EEG acquisition, ensuring precise temporal alignment between tFUS stimulation and continuous EEG recording. All analyses were therefore time-locked to the actual hardware trigger for each trial, eliminating temporal uncertainty in oscillatory and time-resolved electrophysiological metrics. To mitigate potential confounding effects of anesthesia, tFUS-evoked neural responses were compared with their corresponding negative control or pre-stimulation baseline recordings obtained under stably maintained anesthetic levels. In addition, for within-subject experiments, the order of experimental conditions was randomized to minimize potential confounding effects related to anesthesia duration. Data was preprocessed using TDT-provided offline scripts and subsequently analyzed with custom MATLAB scripts.

Specific EEG preprocessing approaches were applied depending on the type of EEG analysis. For EEG traces and topographic mapping, 20-min responses (600 trials) per mouse during tFUS were averaged following the application of a band pass filter (1-30 Hz) and common average referencing (figure S2). For EEG power spectral analysis at the tFUS-targeted region during stimulation, only EEG signals recorded from the electrode positioned directly above and closest to the hindlimb region of the primary somatosensory cortex (S1HL), identified using anatomical landmarks (figure 1(B)), were included in the analysis. Although scalp EEG is influenced by volume conduction, it reliably represents cortical electrical dynamics [43]. Preprocessing included bandpass filtering from 1 to 40 Hz and application of common average referencing to reduce noise. The resulting signal was segmented into the ISoI (2-sec) for subsequent analysis. The spectral characteristics of the data were evaluated to assess the intrinsic brain response to tFUS. Each 2-sec epoch was processed using a short-time Fourier transform with 50% overlap and a Hamming window. The resulting power values were converted to decibel (dB) scale by applying a base-10 logarithm and multiplying by 10. Spectral power was then summed within predefined frequency bands: theta (4-8 Hz), alpha (8-12 Hz), and beta (12-29 Hz) [32]. For comparison across animals, power values were normalized to each subject’s baseline by subtracting the mean spectral power during the pre-stimulation (figure 1(c)). To assess the aperiodic exponent as an indicator of global E/I balance [44,45], all 30-channel EEG signals were bandpass filtered between 0.5 and 40 Hz and referenced using a common average. Independent component analysis was performed using EEGLAB [46], and visually detected artifactual components were removed for accurate calculation of E/I balance. Power spectral density (PSD) was computed for each 2-sec epoch, and PSDs were then averaged across trials for each mouse and parameter. To estimate the aperiodic exponent, oscillatory peak components were excluded, and a first-order polynomial was fit to the remaining spectrum. The resulting slope of this fit provides the aperiodic exponent, which serves as a marker of underlying excitation–inhibition (E/I) balance. The goodness of fit was evaluated using the coefficient of determination (R²) [44,45].

2.5. Histology, imaging, and analysis

After 14-days of tFUS stimulation or sham treatment, mice were euthanized and perfused intracardially with phosphate buffered saline (79383, MilliporeSigma, Burlington, MA, USA) and phosphate-buffered paraformaldehyde (1224SK, Electron Microscopy Sciences, Hatfield, PA, USA). Brain tissues were immediately collected and fixed in 10% neutral buffered formalin (HT501128, Sigma-Aldrich, St. Louis, MO, USA) or phosphate-buffered paraformaldehyde (1224SK, Electron Microscopy Sciences, Hatfield, PA, USA) following established protocol [32,47]. The fixed brain samples were embedded in paraffin blocks for sectioning and immunostaining at HistoWiz, Inc. (Long Island City, NY, USA) using an automated stainer (Leica Bond RX, Leica Microsystems, Nussloch, Germany).

Brain sections (4 μm) were dewaxed using xylene and alcohol-based dewaxing solutions. Epitope retrieval was performed via heat-induced epitope retrieval (HIER) of the formalin-fixed, paraffin-embedded tissue using citrate-based buffer at pH 6 (AR9961, Leica Biosystems, Nussloch, Germany) for 20 min at 100°C. Endogenous peroxidase activity was blocked using a peroxide block buffer (Leica Microsystems, Nussloch, Germany). Sections were immunostained with primary antibodies, including anti-GAD67 (1:300, MAB5406B, Sigma-Aldrich, St. Louis, MO, USA), anti-GABA (1:500, PA5-32241, Invitrogen, Waltham, MA, USA), PV (1:500, CST80561, Cell Signaling, Danvers, MA, USA), anti-CD31 (1:100, CST77699, Cell Signaling Technology, Danvers, MA, USA), and anti-Iba1 (1:1200, 019-19741, Wako, Richmond, VA, USA) for 30 minutes at room temperature (22.5 °C). For TUNEL staining, tissues were incubated in 10% Neutral Buffered Formalin fixative (28600-1, StatLab. McKinney, TX, USA) for 5 min, Equilibration buffer (G327C, Promega, Madison, WI, USA) for 10 min, TDT1 (G327C, G715A, and M828B, Promega, Madison, WI, USA) for 10 min at 37°C, and 20X SSC (SSC-G329A, Promega, Madison, WI, USA) for 15 min at room temperature (22.5 °C). Detection of anti-GAD67 antibody and TUNEL were carried out using DAB streptavidin secondary reagents including Strep HRP (RE7104, Leica, Nussloch, Germany), DAB refine, and hematoxylin (DS9800, Leica, Nussloch, Germany); detection of anti-GABA antibody, PV, anti-CD31 antibody, and anti-Iba1 antibody were carried out using DAB rabbit secondary reagents including polymer, DAB refine, and hematoxylin. The slides were dried, coverslipped using Sakura Tissue-Tek Film, and visualized using an Aperio AT2 slide scanner (Leica Microsystems, Nussloch, Germany) at 40× magnification. All procedures were performed according to the manufacture’s protocols.

For pathology assessment, H&E images were blindly evaluated by a senior board-certified research pathologist for immune infiltrates (e.g. macrophages, lymphocytes, and neutrophils), necrosis, and neuronal vacuolation, indicated as present or absent. Necrosis and apoptosis were assessed in 5 random regions of interest (ROI) within contralateral and ipsilateral S1HL brain regions, and assigned a severity score on TUNEL stained whole slide images (WSIs). On GABA stained WSIs, H-scores were determined for annotated regions in the left and right hemispheres focusing on astrocytic cells. The percentage of cells exhibiting negative, weak (1), moderate (2), or strong (3) staining intensity was recorded, and mean H-scores were subsequently calculated. In addition, the number of marker-positive cells within the entire S1HL region was quantified using atlas-based regions of interest defined by fixed stereotaxic coordinates referenced to the Allen Mouse Brain Atlas. Positive cells exceeding a predefined staining threshold, established by a senior board-certified research pathologist, were manually counted under blinded condition using consistent criteria across all sections. Quantification focused on brain regions receiving multi-session tFUS stimulation and was compared with corresponding contralateral hemisphere, which served as internal controls. Given the high spatial specificity of the ultrasound focus enabling unilateral stimulation, the effect was assessed by calculating the difference in the number of marker-positive cells between the ipsilateral and contralateral S1HL regions.

2.6. Virus injection and optical stimulation during recordings in rats

Wistar rats were used for optogenetic experiments to examine the phenomenological and mechanistic alignment with the electrophysiological signatures observed under tFUS stimulation. The targeted cortical regions and major classes of inhibitory interneurons are conserved across rodent species in both molecular identity and circuit organization [48,49], enabling cross-species comparison at the level of network-level electrophysiological signatures. The rats were anesthetized in an induction chamber with isoflurane and then secured in a stereotaxic apparatus with a nose cone attachment, using 2%-3% isoflurane delivered by inhalation at 0.4 L/min to maintain the anesthetic plane. The fur was removed with an electric razor, and the anti-inflammatory drug carprofen was injected subcutaneously. A small midline scalp incision was made using a scalpel, and four hemostats were used to retract the skin to expose the bregma and lambda. The skull was subsequently cleaned by removing the connective tissue. After identifying the coordinates of the bregma and lambda, stereotaxic coordinates of the left S1 were determined according to the rat brain atlas [50]. A small burr hole was drilled at the identified injection site.

A 10 μL Hamilton syringe was prepared and filled with PV-ChR2(H134R)/cherry (Duke Viral Vector Core, Durham, NC, USA or 213941-PHPeB, Addgene) or SST-ChR2/cherry virus (213945-PHPeB, Addgene) and mounted onto a microinjection system (Legato^™ 130, KD Scientific, Holliston, MA, USA), set at a rate of 0.1 μL/min for injection. Using the stereotaxic arm, the needle was lowered to the surface of the brain via the burr hole and slowly advanced to the desired depth of 1.5 mm in the S1 region. Following injection of 1μL virus, the needle was left in place for 10 minutes to allow diffusion and minimize backflow. The syringe was then slowly withdrawn, the burr hole sealed with Vetbond tissue adhesive (3M, Maplewood, MN, USA), and the incision sutured. Carprofen and enrofloxacin were administered postoperatively.

After recovery from anesthesia, rats were returned to the home cages in the housing facility. Recordings and stimulation experiments were conducted 4-5 weeks post-injection to allow for viral expression. If signs of infection were observed, additional antibiotic treatment or topical application of triple antibiotic ointment was administered as needed.

For neural recordings in the optogenetics stimulation setting, the skull was exposed by making a midline scalp incision and retracting the skin. A 32-channel rat EEG electrode (HC32 Rat Functional, NeuroNexus, Ann Arbor, MI, USA) was positioned on the surface of the skull to record brain-wide activity, and a 32-channel (A1x32-Poly3-10mm-50-177, NeuroNexus, Ann Arbor, MI, USA) or 64-channel silicon probe (A1x64-Edge-6mm-22.5-177-OXA64LP, NeuroNexus, Ann Arbor, MI, USA) with an integrated optical fiber was inserted into the left S1 to a depth of 1.5 – 1.8 mm, and pulses of blue light (465 nm, 50-ms, 300 mA intensity, 2-sec interval) were delivered using PlexBright LD-1 Single Channel LED Driver (Plexon Inc., Dallas, TX, USA) to selectively activate PV-positive or SST-positive interneurons expressing ChR2.

2.7. Statistical analysis

Data were presented as mean ± standard error of the mean (s.e.m.). Statistical analysis was performed using commercial software (GraphPad Prism, San Diego, CA, USA). To assess changes relative to the pre-stimulation baseline and differences in brain responses within the same subject, a non-parametric paired t-test (Wilcoxon matched-pairs signed rank test) was employed with significance thresholds: ns: not significant; *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Additionally, intergroup comparisons between different treatment conditions were evaluated using the non-parametric t-test (Mann-Whitney test) with significance levels: ns: not significant; ^#p<0.05, ^##p<0.01, ^###p<0.001, ^####p<0.0001.

3. RESULTS

3.1. Single-session tFUS suppressed local theta oscillation consistent with PV-driven inhibition

We previously demonstrated that a spherically concave, highly focused 128-element random-array ultrasound transducer operating at a center frequency of 1.5 MHz provides high spatial specificity for precise stimulation within the small mouse brain, as validated by computational simulations and in vivo intracranial electrophysiological recordings [32]. Additionally, the −3 dB focal dimensions were measured as 0.46 mm (lateral) and 1.36 mm (axial) in free water, and 0.84 mm (lateral) and 1.71 mm (axial) through a freshly excised mouse skull. We further characterized in situ ultrasound exposure behind the mouse skull for single- and multi-session tFUS at TBD of 200 μs with a peak negative pressure of 45.2 kPa, corresponding to mechanical index (MI) for transcranial application (MI_tc) of 0.037 [51], and a spatial peak pulse-average intensity (I_SPPA) of 0.35 W/cm² held constant across PRF conditions (40 Hz and 3 kHz), while adjusting PRF resulted in the spatial peak temporal-average intensity (I_SPTA) of 2.02 mW/cm² at 40 Hz and 208.46 mW/cm² at 3 kHz when the ultrasound focus was steered to the S1HL. To investigate how specific tFUS parameters modulate intrinsic brain activity in pain circuits, we integrated 30-channel scalp EEG recordings with a 128-element random array ultrasound transducer to evaluate tFUS-evoked brain responses in SCD mice (figure 1(A) and 1(B)).

We first tested the hypothesis that a single session tFUS could perturb local brain activity when precisely targeted to specific pain-processing brain regions. During EEG recordings, following a baseline period, ultrasound pulses were delivered to the right hemisphere’s S1HL region (200-μs TBD, 40 Hz or 3 kHz PRF, 100-ms UD, 2-sec ISoI, 20-min TST; figure 1(C)). Using the excitation parameter with 3 kHz PRF, we found significant activation in the targeted S1HL compared with both the negative control and the 40 Hz PRF condition related to the inhibition parameter (figure 1(D)).

We next investigated whether intrinsic brain rhythms associated with chronic pain could be modulated by single-session tFUS with a PRF of 40 Hz, previously shown to suppress pain-related behaviors in SCD mice (11 female and 6 male). We observed a significant suppression of low frequency brain oscillations, particularly in the theta and alpha bands during 40 Hz tFUS targeting S1HL region compared to the negative control (figure 1(E)). In contrast, 3 kHz tFUS produced excitatory effects, showing a bidirectional pattern when comparing the low versus high PRF responses relative to the negative control (figure S3). To further investigate the mechanisms underlying tFUS neuromodulation, we conducted optogenetic experiments targeting specific interneuron populations and confirmed stimulation-induced activation and identification of the labelled PV neurons (figure S4). We performed optogenetic activation of PV-positive interneurons in rats, which significantly reduced theta, alpha, and beta band power (figure 1(F)), suggesting that PV activation suppresses low-frequency oscillations. These findings align with the previous report showing that silencing PV interneurons increases low-frequency activity [52], indicating that active PV interneurons constrain these rhythms and play a central inhibitory role in shaping cortical dynamics. The spectral outcomes of tFUS and optogenetics generally align yet not fully congruent, suggesting that tFUS may engage broader or complementary inhibitory networks compared to the specificity of optogenetic stimulation. Nevertheless, the shared suppression of the theta band suggests that the fast-spiking PV interneurons are key contributors to tFUS effects. While GABAergic interneurons such as PV and somatostatin-positive (SST) cells both influence theta oscillations, prior studies demonstrated that optogenetic suppression of SST reduces spontaneous low-frequency (<30 Hz) activity, whereas suppression of PV increases it [52,53], underscoring their distinct but complementary roles in modulating cortical oscillations. Thus, the observed suppression of theta power following tFUS likely reflect PV-driven inhibition rather than SST mechanisms.

3.2. Multi-session tFUS enhanced network-level inhibitory activity consistent with PV-mediated inhibition

In the local brain circuit, single-session tFUS stimulation with a PRF of 40 Hz resulted in a marked suppression of theta and alpha band power at the targeted region. However, oscillatory changes alone may not fully capture the complex dynamics of brain networks, particularly the balance between excitation and inhibition. We therefore analyzed the aperiodic exponent across all 30-channel EEG signals, which provide broad coverage of the brain and allow assessment of global effects. The aperiodic exponent was computed by fitting the slope of the signal’s power spectrum as a marker of E/I balance (figure 2(A)). The E/I balance is a crucial aspect of neural circuit function, influencing the network’s ability to process information [45]. Shifts in this balance can indicate whether cortical activity becomes more excitatory reflected by a lower aperiodic exponent, or more inhibitory indicated by a higher aperiodic exponent [44,54,55]. We found a significant reduction in the aperiodic exponent in SCD mice (11 female and 6 male) relative to female wild-type mice (figure 2(B)), reflecting a disrupted E/I balance commonly associated with chronic pain features.

Figure 2. Multi-session tFUS induced global cortical inhibitory activity measured by EEG aperiodic exponent analysis.

(A) Illustration of slope of the power spectrum showing the aperiodic exponent (blue) fit to the original power spectrum (green), used as a marker of E/I balance. The aperiodic exponent was calculated from all 30-channel EEG signals to assess global changes in neural activity. Aperiodic exponents shown in (B), (C), (E), and (F) are averaged values across all 30-channel EEG. (B) Aperiodic exponent was significantly reduced in SCD mice (n=17; 11 female and 6 male) compared to wild-type controls (n=8), suggesting a disrupted E/I balance. (C) Compared to negative control group, single-session tFUS with a PRF of 40 Hz did not produce a significant change in aperiodic exponent in SCD mice (11 female and 6 male) exhibiting chronic pain phenotypes. (D) Multi-session tFUS stimulation paradigm: 14 consecutive days of 1-hour daily tFUS stimulation with a PRF of 40 Hz, a UD of 400-ms, and an IsoI of 4-sec. (E) Aperiodic exponent significantly increased in SCD mice following multi-session tFUS at S1HL (n=5), reaching levels comparable to wild-type mice. (F) An additional single-session tFUS delivered after the 14-day stimulation significantly increased the aperiodic exponent compared to 20-min single-session tFUS in sickle mice, indicating amplified inhibitory effect. (G) Scalp EEG topographic maps of aperiodic exponent values showed a progressive increase from untreated SCD mice and post-multi-session tFUS. (H) Potential long-term synaptic plastic changes along with multi-session tFUS were investigated by comparing responses to single-session 20-min tFUS in SCD mice and an additional single-session tFUS in 14-day treated SCD mice. ^###p<0.001 using t test with Mann-Whitney test; ns: not significant.

We then applied a single-session tFUS in SCD mice (11 female and 6 male); however, it did not significantly alter E/I balance, suggesting a brief 20-min stimulation, consisting of 600 sonications with a 2-sec ISoI, may be insufficient to induce an enhancement of global inhibitory activity (figure 2(C)). Notably, single-session tFUS produced a tendency toward stronger neuromodulatory effects in male SCD mice, including greater suppression of low-frequency EEG oscillation, compared with female SCD mice (figure S5). These effects were comparable to those observed during optogenetic activation of PV neurons. Based on these findings, we hypothesized that female SCD mice may require repeated or more effective stimulation to achieve comparable and durable neuromodulatory effects. Accordingly, we focused subsequent multi-session tFUS paradigms on female mice to address the greater pain burden observed in females [35,41] and to determine whether repeated tFUS stimulation can overcome sex-dependent differences in responsiveness. A 14-day multi-session tFUS (1 h/day), consisting of 12,600 sonications with a 4-sec ISoI, was then implemented, and the change in E/I balance was assessed by resting-state EEG recordings obtained approximately 24 hours after the final stimulation session under the same anesthesia condition as the single-session tFUS experiments (figure 2(D)). The multi-session tFUS resulted in a significant increase in the aperiodic exponent in female SCD mice, indicating a shift toward increased inhibitory activity, approaching levels observed in wild-type controls (figure 2(E)). Interestingly, an additional 20-min tFUS following the 14-day stimulation further significantly increased the aperiodic exponent compared to single-session tFUS in SCD mice, suggesting an enhanced network-level inhibitory effect potentially driven by synaptic plastic changes (figure 2(F)). Topographical mapping revealed elevated aperiodic exponent across the cortex (figure 2(G)), with further enhancement especially in the targeted region following the additional 20-min tFUS (figure 2(H)). For comparison, we further found that optogenetic activation of PV interneurons leads to a robust increase in the aperiodic exponent, mirroring the strengthened inhibitory function induced by multi-session tFUS, whereas activating SST interneurons significantly decreases the aperiodic exponent (figure S6). These findings support the hypothesis that PV-mediated, rather than SST-driven, inhibition underlines the electrophysiological effects of tFUS.

3.3. Multi-session tFUS enhanced local expressions of GAD67 and PV interneurons

As repeated tFUS stimulations induced a significant and sustained increase in global inhibitory electrophysiological activity, we investigated cellular changes after multi-session 40 Hz PRF tFUS using IHC analysis and blinded assessment (figure 3(A)). Assessment revealed significantly higher GAD67-positive cells in the stimulated region compared to the contralateral side (figure 3(B) and 3(C)), whereas no such difference was observed in the sham-treated group (figure S7(A)). These findings suggest that tFUS enhances inhibitory GABAergic signaling by increasing the capacity for GABA synthesis, as reflected by elevated levels of GAD67.

Figure 3. Multi-session tFUS increased neuronal GABAergic capacity assessed by IHC.

(A) Schematic of histology processing workflow: mice underwent 14 days of tFUS, followed by brain perfusion, fixation, paraffin embedding, and blind IHC analysis comparing ipsilateral (stimulated) and contralateral (unstimulated) brain regions in SCD mice. (B) Representative IHC image of GAD67. (C) Quantification of GAD67-positive cells revealed a significant increase in the treated hemisphere following multi-session tFUS, suggesting enhanced capacity for GABA synthesis (n=5). (D) Representative IHC image of PV-positive neurons. (E) Quantification of PV-positive cells showed a significant increase in the stimulated hemisphere, indicating enhanced engagement of fast-spiking inhibitory interneurons (n=5). (F) Representative IHC image of GABA expression. Yellow, green, red, and white circles represent neuron, strong astrocyte, weaker oligodendrocyte, and microglial cells, respectively. (G) H-score quantification of GABA immunoreactivity in microglia and astrocytes showed no significant difference between ipsilateral and contralateral hemispheres, indicating that tFUS did not alter non-neuronal inhibitory signaling (n=5). *p <0.05 using t test with Wilcoxon matched-pairs signed rank test; ns, not significant.

Building on the observed upregulation of GAD67, we next addressed the question of whether tFUS could modulate specific subtypes of GABAergic interneurons, particularly PV-expressing neurons, based on our findings related to theta oscillations, distinct E/I balance patterns, and prior evidence involving optogenetic inactivation of PV neurons [52]. Given their important role in maintaining E/I balance and known vulnerability in various neurological conditions [45], we sought to determine whether PV interneurons are upregulated following multi-session tFUS stimulation. Stimulation of left S1HL significantly increased the number of PV-positive cells at the site relative to the contralateral region (figure 3(D) and 3(E)), but not from sham treatment (figure S7(B)). These findings suggest that tFUS strengthens local inhibitory networks by promoting activity or expression of the PV GABAergic neurons.

We further examined whether non-neuronal cell types such as microglia and astrocytes might also contribute. Since GABA is widely expressed in the brain, we alternatively used the H-score, a semi-quantitative method for evaluating protein expression [56,57], to compare GABA immunoreactivity between ipsilateral and contralateral regions. H-score analysis revealed that neither tFUS nor sham treatment led to significant changes in GABA expression within microglia and astrocytes across hemispheres (figure 3(F), 3(G), S7(C), and S7(D)). Additionally, Iba1 levels remained unchanged following tFUS stimulation (figure S7(E) and S7(F)), indicating a lack of glial activation. These results suggest that tFUS at 40 Hz modulates GABAergic signaling primarily through neuron-specific pathways in the humanized SCD mouse model.

3.4. Multi-dimensional histological analyses confirmed safety of multi-session tFUS

To facilitate potential clinical translation, we conducted a thorough safety evaluation by multi-dimensional histological analyses including H&E, TUNEL, and CD31 immunostaining, which are well-established approaches to evaluate tissue morphology [58], cellular integrity [59], and vascular architecture [60], respectively. Following multi-session tFUS stimulation, H&E staining revealed no gross morphological abnormalities, such as significant infiltration of macrophages or neutrophils, inflammation, necrosis, apoptosis, hemorrhage, fibrosis, vacuolation, or brain damage in either hemisphere under tFUS relative to the sham-treated group (figure 4(A) and S8(A)), reinforcing the structural safety of the protocol. Additionally, TUNEL staining showed no evidence of apoptotic cell death as no TUNEL-positive nuclei were detected in either brain region of tFUS-treated or sham-treated animals (figure 4(B) and S8(B)). Given that vascular structures in the SCD mouse model may exhibit increased vulnerability to mechanical perturbations [61], we further examined vascular integrity using CD31, a widely accepted marker of endothelial cells. CD31 immunostaining demonstrated similar vascular distribution without signs of pathological angiogenesis, vessel disruption, or hemorrhage in both hemispheres from tFUS-treated and sham-treated groups (figure 4(C) and S8(C)). Taken together, these results suggest that multi-session tFUS with a PRF of 40 Hz does not induce detectable cytotoxic effects or vascular damage in the brain samples used in the IHC analysis.

Figure 4. Multi-session tFUS did not induce pathological changes as assessed by histological analysis.

(A) Representative H&E-stained section showed preserved tissue architecture in both the ipsilateral and contralateral brain regions following 14-day tFUS. No signs of macrophages, neutrophils, inflammation, necrosis, apoptosis, hemorrhage, fibrosis, vacuolation were detected across hemispheres. (B) Representative image of TUNEL staining revealed an absence of TUNEL-positive nuclei, indicating no evidence of necrosis or apoptotic cell death in either hemisphere. (C) Representative image of CD31 immunostaining indicated normal vascular architecture and distribution, with no signs of hemorrhage or abnormal angiogenesis.

4. DISCUSSION

Prior studies have highlighted the potential of tFUS in pain management. In particular, we recently demonstrated that both single- and multi-session tFUS significantly suppress pain-related behaviors and modulate intrinsic brain rhythms associated with chronic pain in a humanized SCD mouse model [32]. In addition, tFUS targeting multiple pain-processing circuits more effectively reduced heat hyperalgesia in female SCD mice compared with single-target stimulation in the same animals [33]. However, how tFUS mediates its effects is still poorly understood. In this study, we have demonstrated that low-intensity tFUS with a PRF of 40 Hz engaged PV GABAergic interneurons in a humanized mouse model of chronic pain using combined electrophysiological and molecular approaches. These included enhanced local and global inhibitory activity in the pain-processing brain circuit, as indicated by suppression of low-frequency oscillations at S1HL, elevated aperiodic exponent derived from resting-state EEG, and increased GAD67-immunoreactivity (i.r.) and PV interneuron-i.r. in the stimulated region. Compared to high-intensity tFUS used in ablative protocols, which eliminate cellular activity in the targeted area but risk collateral damage to adjacent healthy circuits potentially leading to adverse side effects [62], this low-intensity tFUS enables targeted neuromodulation while preserving structural and functional integrity. Importantly, we showed tFUS effects without evidence of glial activation, cellular apoptosis, or vascular disruption, confirming safety and neuron-specificity. This specificity underscores its potential as a non-invasive and precise neuromodulation approach for chronic pain management, paving the way for promising therapeutic applications.

Our results demonstrated that 40 Hz tFUS led to upregulation of both GAD67 and PV expressions, two molecular hallmarks of the GABAergic inhibitory system. Given prior evidence that GAD67 deficiency in PV interneurons resulted in impaired inhibitory transmission and cortical disinhibition [39,63], the observed post-tFUS upregulation supports the restoration of inhibitory activity. GAD67 is the enzyme that catalyzes synthesis of the inhibitory neurotransmitter GABA from glutamate [36], and is currently being explored as a therapeutic target for treating pain, including via gene therapy approaches [64,65]. Recent study in SCD patients shown elevated brain glutamate, potentially suggesting the correlation between the excitatory neurotransmitter and pain hypersensitivity [66]; thus, increased GAD67 expression may help counterbalance this excitatory drive by enhancing the glutamate to GABA conversion. Our approach utilizing non-invasive tFUS led to upregulation of GAD67, supporting its potential as a convenient and cell-targeted strategy for pain relief. In parallel, GABAergic neurons in S1 primarily consist of PV, SST, and vasoactive intestinal peptide (VIP) interneurons [67]. These inhibitory neuron subtypes differ in distribution, firing properties, and synaptic targets [68]. Among them, PV interneurons are the dominant GABAergic population, characterized by rapid, high-frequency spiking and strong perisomatic inhibition. Their loss significantly heightens pain sensitivity [69–71], highlighting the relevance of PV GABAergic neurons for effective pain treatment. In our study, we observed a significant increase in PV-positive cell expression in the tFUS-stimulated area, positioning tFUS as a cell-type specific modality in this context.

A holistic interpretation of both electrophysiological and IHC findings revealed two complementary clinical applications for the use of tFUS with a PRF of 40 Hz. A single-session tFUS induced suppression of low-frequency EEG rhythms, consistent with short-term circuit inhibition, but did not alter global E/I balance. Thus, it appeared effective for providing rapid relief from acute or tonic pain, which typically arose from inflammation, healing, or mild injury and is generally reversible. It modulated cortical excitability locally and temporarily without inducing detectable morphological changes. As a non-pharmacological alternative, tFUS holds particular promise for patients who may be contraindicated for pharmacological analgesics, which are associated with risks such as gastrointestinal distress, organ toxicity, dependency, and overdose [1–3,72,73]. In contrast, multi-session tFUS accumulated over time, engaging inhibitory circuits through potential synaptic plasticity rather than short-term modulation. These findings aligned with prior results from repetitive TMS, where multiple TMS stimulation induced inhibitory plasticity in cortical networks [74]. This adaptive reorganization of a previously dysregulated inhibitory circuit highlighted a novel therapeutic mechanism of low-PRF tFUS, complementing previously studied pulse-locked neuromodulation or mechanosensitive ion channel effects [75,76]. By showing elevated PV-i.r. and GAD67-i.r. alongside persistent electrophysiological changes, our study provided post-stimulation mechanistic evidence of functional network remodeling. While previous reports have focused on acute electrophysiological responses, behavioral outcomes, or self-reported pain ratings [28–33], our findings expanded the mechanistic landscape to include molecular and cellular adaptations in inhibitory circuits driven by repeated subthreshold entrainment of PV neurons. Additionally, when single-session tFUS was applied to 14-day-treated SCD mice, the stimulation effect was significantly greater compared to that observed with single-session tFUS in naïve SCD mice, indicating that non-invasively repetitive stimulations can drive synaptic plasticity even when each session is subtle and transient. Importantly, we observed no signs of cellular or vascular damage across the 14-day stimulation compared to controls, which is an essential safety consideration for any long-term neuromodulation therapy. These results positioned low-intensity tFUS with 40 Hz PRF as a versatile and precision intervention, with flexible dosing paradigms tailored for different clinical needs: single-session use for episodic pain control, and multi-session delivery for chronic pain alleviation through inhibitory circuit restoration. Moreover, the combined use of electrophysiological and immunohistochemical endpoints provided cross-validated evidence for a reliable, non-invasive biomarker with translational relevance. The power spectral changes and the aperiodic exponent derived from EEG, a clinically accessible modality, closely reflected shifts in cortical inhibitory activity and mirrored molecular findings of GAD67 and PV-positive interneurons upregulation. This convergence of readouts suggested that EEG-based measurements may serve as practical surrogate biomarkers for monitoring neuromodulatory engagement in clinical settings.

While our study provided multi-level evidence that repeated 40 Hz tFUS strengthened cortical inhibitory activity and promoted long term circuit reorganization, several directions remained to be explored. First, although the use of PV and GAD67 markers was deliberate and mechanistically informative, future studies could incorporate a broader panel of IHC markers to map the GABAergic and Glutamatergic responses more comprehensively. For example, the inclusion of SST, VIP, CaMKII, and GABA receptors could allow for a more detailed dissection of cell type and synapse specific changes. Our results aligned with the overall low-frequency power suppression observed during optogenetic activation of PV neurons, the statistical significance inconsistency in the beta band could suggest other potential mechanisms contributing to the inhibition, such as the activation of other GABAergic interneurons or the inhibition of excitatory glutamatergic neurons. Mapping the involvement of additional GABAergic subtypes, including SST, VIP, and their synaptic interactions with excitatory neurons would greatly enhance our understanding of circuit-level reorganization and the precision of tFUS neuromodulation, given the distinct roles these subtypes contribute to the overall sensory perception [77]. Second, when grand-averaging EEG responses across 17 mice, we observed a relatively broader spatial distribution of activity (figure S9), likely reflecting inter-animal anatomical variability including slight differences in electrode placement, skull thickness, and cortical geometry, which can influence the precise localization of evoked responses. Notably, the flexible EEG substrate itself is unlikely to account for this spatial spread. The 1.5 MHz ultrasound beam traverses a 15 μm polyimide layer containing platinum electrode sites; based on representative acoustic impedances of water (approximately 1.5 MRayl) [78,79] and polyimide (about 3 MRayl) [79], the estimated reflection at a single interface is approximately 11%, corresponding to an overall transmission of approximately 80-85% intensity. Because the substrate thickness represents 1.5% of the acoustic wavelength (1 mm at 1.5 MHz), the expected phase perturbation is minimal, and even under a worst-case thickness gradient across the 15 mm exit-plane diameter, the predicted beam tilt (about 0.02°) would produce only 1 μm lateral focus shift at the 2.78 mm focal depth. Although the effects of the flexible EEG electrode and inter-animal anatomical variability are less critical given the relatively thin mouse skull, future studies should consider individualized targeting strategies or improved anatomical registration approaches to better align responses across subjects. Third, EEG recordings were conducted under light anesthesia, and multiple methodological controls were implemented to mitigate potential confounding effects. These included comparisons of tFUS-evoked neural responses with corresponding negative control and pre-stimulation baseline recordings, as well as randomization of experimental condition order. Consistent with this approach, we found no significant differences in pre-stimulation baseline PSD between the tFUS and negative control groups (figure S10), supporting the conclusion that anesthesia-related confounding effects were effectively minimized in the electrophysiological analyses. Nevertheless, anesthesia could alter neural excitability and affect circuit-level dynamics, which may complicate interpretation of direct neuromodulatory effects [80,81]. Future studies establishing a mechanistic framework for tFUS neuromodulation in awake, freely moving animals will be important to further delineate state-dependent effects and enhance translation relevance. Lastly, the translational pathway from rodent models to human application requires careful calibration of stimulation parameters including acoustic pressures and stimulation pulse paradigm due to differences in cortical organization and interneuron diversity [82]. Computational modeling may further help to determine the scaling effects between rodents and humans in terms of their biophysical properties, to optimize the ultrasound parameters for human translation. In addition, in this study, we used a humanized SCD mouse model with thin skull bone and low ultrasound attenuation compared with lager animals and humans. Although skull-induced aberration is less pronounced in mice, skull-based acoustic correction framework using a phased-array transducer [83,84] will be necessary to achieve accurate and reproducible tFUS targeting in translational and clinical applications, as demonstrated in recent ultrasound time-reversal-based phased-array control and modulated focusing to improve targeting precision [85–87].

5. CONCLUSION

We have investigated the mechanisms and safety of low-intensity tFUS through integrated electrophysiological and cellular approaches in a humanized mouse model of chronic pain. tFUS-induced suppression of local theta oscillations and alterations in E/I dynamics indicated engagement of PV-driven inhibitory circuits at both local and network levels. At the cellular level, enhanced expressions of GAD67 and PV interneurons further supported strengthened GABAergic signaling through neuron-specific pathways. The safety of repeated 14-day stimulation was confirmed by blinded histological evaluation of the same brain samples analyzed by IHC approach. Collectively, these findings highlighted that tFUS engages PV GABAergic inhibitory circuits to counteract chronic pain features, supported by electrophysiological and cellular analyses. These insights position tFUS neuromodulation as a clinically meaningful avenue for restoring inhibitory balance in the brain, with the potential to redefine therapeutic strategies for chronic pain and disorders involving circuit dysfunction.

Table 1.

Summary of administered tFUS parameters.

tFUS paradigm	PRF (Hz)	TBD (μs) (Duty Cycle (%))	UD (ms)	ISoI (sec)	TST (min)	Sonication numbers (TST/ISoI)	In situ pressure (kPa)	In situ ISPTA (mW/cm2)
Single-session	40	200 (0.8)	100	2	20	600	90.4	2.02
Single-session	3000	200 (60)	100	2	20	600	90.4	208.46
Multi-session	40	200 (0.8)	400	4	60	900	90.4	2.02

Data availability

The data supporting the conclusions of this study are provided in the paper and supplementary materials. Additional data can be found in Figshare repository at: EEG data (https://figshare.com/s/0320bc79a33aab073614); Histological analysis and images (https://figshare.com/s/d56f9ec581d6fa0ee24c).