Music-inspired acoustic-piezoelectric stimulation accelerates extracellular vesicle production and programs therapeutic function

Abstract

Macrophage small extracellular vesicles (sEVs) carry phenotype-linked cargo and bioactivity for immunomodulation and regeneration, but therapeutic translation is limited by low secretion and poor control of function. We introduce a music-activated piezoelectric nanofiber substrate (PES) that converted audible sound into programmable electrical stimulation to enhance sEV biogenesis while tuning macrophage polarization. Adjusting acoustic parameters increased sEV yield, while musically inspired “assemblies” biased macrophage phenotypes: dissonant, low-frequency stimuli promoted M1-like inflammation, whereas consonant, higher-frequency stimuli favored M2-like, regenerative states. These shifts produced distinct sEV cargo and bioactivities. We rationally designed customized music stimulus that maximized both vesicle production and M2 bias, yielding sEVs exhibited regeneration potentials. This work establishes a programmable acoustic–piezoelectric strategy to scale macrophage sEV production while tailoring their therapeutic potency.

Statement of Significance:

Macrophage small extracellular vesicles (sEVs) are promising as a multi-faceted therapeutic tool for regeneration and immunomodulation. Current methods are challenged by low production rate and poor functional control of sEVs. Here, we developed a piezoelectric nanofiber substrate (PES), as a highly controlled stimulative culturing platform for enhancing sEV production and programming macrophages based on the tuned acoustic input. This provides an important platform that allows for scaled sEV production, and tunable cellular phenotypes tailored for various therapeutic applications.

1. Introduction

Small extracellular vesicles (sEVs) are nature-derived, lipid-based nanoparticles that that mediate intercellular communication and are emerging as a versatile therapeutic modality for cancer therapy, immunomodulation, and tissue regeneration [1–5]. In particular, sEVs secreted by tissue-resident macrophages are attractive because their molecular cargo and resulting bioactivity are tightly coupled to the activation state of the parent cells, enabling phenotype-dependent modulation of immune responses and regenerative processes [6,7]. Yet, realizing the full therapeutic potential of macrophage-derived sEVs will require strategies that enhance secretion and enable tunable control over their native functions [8,9].

Piezoelectricity is a pervasive feature of biological tissues and extracellular matrices (ECM), providing cells with both mechanical and dynamic electrical cues that influence mechanotransduction and cell state [10,11]. Inspired by this feature, piezoelectric materials have been leveraged to stimulate cells and enhance sEV production by utilizing the electrical component of the piezoelectric stimulation [12–15]. This electrical component can be programmed through acoustic parameters such as frequency, amplitude, timbre, and harmonic structure, well established in audible acoustic sensing systems [16,17]. Hence, ECM–mimetic piezoelectric platforms may provide a unified strategy to both enhance sEV biogenesis and reprogram key cellular pathways that govern the composition and functional potency of the resulting sEVs.

Meanwhile, acoustic stimuli are biologically relevant signals for innate immune cells. Macrophages exhibit measurable responses to audible sound exposures, including shifts in cytokine expression associated with polarization state [18,19]. Prior studies report that music dominated by consonant, higher-frequency components (e.g., classical and opera) is associated with increased anti-inflammatory cytokines, aligning with regeneration phenotypes [20,21]. Conversely, intense or dysregulated stimulation can promote proinflammatory signaling, desirable in select immunotherapy contexts [19,22,23]. Related approaches, such as vibroacoustic therapies (VAT), use bulk low frequency waves (20–150 Hz) to modulate immune system through combined mechanical and bioelectrical pathways at the cellular level [24–27]. However, it remains poorly understood how complex music-theory features such as waveform structure, amplitude stability, dissonance, and consonance, beyond basic acoustic parameters (amplitude, frequency), are transmitted through the piezoelectric elements of native tissues and the ECM into bioelectrical cues to generate bioelectrical cues that regulate the innate immune system. This knowledge gap limits the rational design of acoustic-piezoelectric strategies for immunomodulation and regenerative medicine.

Here we introduce a music-activated piezoelectric nanofiber substrate (PES) to transduce audible acoustic waves to cellular electrical stimulation, enabling regulation of macrophage-derived sEV production and macrophage polarization (Fig. 1). We show that PES activation significantly accelerate sEV production and that musically inspired acoustic “assemblies” can bias macrophage polarization to tune sEV bioactivity. Finally, we engineered a customized regenerative musical stimulus that jointly optimizes sEV production and regeneration function, yielding high sEV concentration and enhanced osteoblast proliferation and gap closure. Together, this music-inspired acoustic–piezoelectric platform provides a programmable strategy to scale sEV production while simultaneously tailoring therapeutic properties through phenotype control.

Fig. 1.

Overview schematic of the piezoelectric nanofiber substrate (PES) platform with various acoustic stimuli. The PES platform mimics the piezoelectrically sensitive extracellular matrix that macrophages adhere to in a tissue. Using this phenomenon, this platform can take advantage of acoustic frequency power and harmonic density to differentiate the small extracellular vesicle (sEV) production and cellular phenotype. Electronic Dance Music (EDM) utilizes low frequency power referring to the acoustic power of a single frequency stimulus below 200 Hz frequency. EDM also possesses a high dissonance at low frequencies referring to two low frequencies (<200 Hz) superimposing each other, causing amplitude instability. These acoustic characteristics result in high sEV production, but with promoted inflammation. A selected classical music stimulus takes advantage of powerful singular high frequency stimulus greater than 200 Hz, and low dissonance in frequencies above 200 Hz, referring to two high frequency stimuli superimposed to one another. These characteristics promote regeneration but suffer from low sEV Production. The proposed mechanism for these phenomena shows dissonant music promoting Calcium ion influx through voltage gated calcium channels (VGCC) which results in ROS generation and M1 polarization while mechanically gated calcium channels (TRPM7) promote M2 polarization likely through STAT6 signaling. Custom-made regenerative music is a hybrid stimulus that takes advantage of both the power in low frequencies and the dissonance in high frequencies to promote sEV production while also promoting regeneration. Figure made in Biorender.

2. Methods

2.1. Materials and reagents

Polyacrylonitrile (PAN; 181315), and 4′,6-diamidino-2-phenylindole (DAPI, 28718-90-3) were purchased from Sigma Aldrich (MA). N,N-Dimethylformamide (DMF; D119–4), and NS 8593 hydrochloride (459710) was purchased from Fischer Scientific (MA). Brass material (3L494) was purchased from Grainger (IL). Copper conducting tape (05085A-AB) was purchased from SPI Supplies (PA). Polyimide tape (S-7596) was purchased from ULINE (WI). Ethyl alcohol (3791–10), 99% acetic acid (BDH3092), Dulbecco’s Modified Eagle Media (DMEM/High Glucose, 16750-074), Minimum Essential Medium (EMEM, 45000-380), PBS (45000-446), and 0.22 μm vacuum filters (76010-388) were purchased from VWR (PA). Chitosan powder (c1569) with a minimum degree of acetylation of 90%, a maximum moisture of 10%, and a maximum ASH of 5% was purchased from Spectrum Chemical (NJ). Corning SpinX centrifuge filters (431491) were purchased from Corning (NY). Fetal bovine serum (FBS; 26140079), 100X antibiotic–antimycotic (15240062), 0.25% trypsin–EDTA (25200072), CD86-PE (12-0862-82), CD206-PE (12-2061-82), IL6 ELISA kit (88-7064-88), IL10 ELISA kit (88-7105-88), LPS (00-4976-93), LIVE/DEAD staining kit (R37601), BCA assay (23227), Fura-2AM (F1221) were purchased from ThermoFisher (MA). 4% paraformaldehyde in 0.1 M phosphate buffer (15735), the TEM copper grids (CF200-CU-25), and the Uranyless negative staining (22409) were purchased from Electron Microscopy Science (PA). Cell counting kit-8 (CCK-8; 850-039-kl01) was purchased from Enzo Life Sciences (NY). The miRNeasy tissue/cells advanced kit (217684) was purchased from QIAGEN (NL). Nifedipine (B1988) was purchased from Apexbio Technology (TX). DCFH-DA (A439906) was purchased from Ambeed (IL). Radio-Immunoprecipitation Assay (RIPA) Buffer (98065) and secondary antibodies (anti-mouse HRP-linked antibody, 7076) were purchased from Cell Signaling Technologies (MA). Nitrocellulose membranes (1662807) were purchased from BioRad (CA). Anti-CD9 (C-4, sc–13118), Anti-CD63 (MX-49.129.5, sc-5275), Anti-CD47 (B–11, sc–166029), Anti-ALIX (1A12, sc-53540), Anti-TSG101 (51, sc–136111), and Anti-b-actin (C4, sc-47778) were purchased from Santa Cruz Biotechnology.

2.2. Preparation of PAN solution and PAN nanofiber substrates

Solution property characteristics were performed similarly to prior work [12,28]. Solution viscosity was measured using a CPA-40 spindle connected to a Brookfield DV-I Prime viscometer (Brookfield, Toronto, Canada). The rotational speed of the spindle was ramped up from 0.5 rpm to whichever speed at which the torque reached closest to 100% (at least above 95%). After confirming that viscosity was independent of the shear rate, the viscosity value at maximum torque was recorded. Surface tension was measured using an automatic surface tensiometer (QBZY–1; Shanghai Fangrui Instrument, Shanghai, China), which had a platinum-coated plate connected to a hook. The force exerted on the hook as the plate came in contact with the solution was converted into surface tension values Arduino code (Atlas Scientific, NY) was used to take electrical conductivity measurement through a glass-body electrical conductivity probe (K = 0.1, Oakton) paired with an embedded conductivity circuit (EZO-EC; Atlas Scientific, NY) and an Arduino Uno Rev3 board. All solution property measurements were taken at room temperature immediately before or after electrospinning to correlate them most closely with the resulting nanofiber properties. Nanofibers with a diameter of 760 nm were produced through an electrospinning process. A solution of 10 wt% PAN was prepared in DMF. Electrospinning was carried out under specific conditions, namely an electrospinning distance of 10 cm, an applied voltage of 13 kV, and a solution feed rate of 1 mL h⁻¹. This process was conducted in a controlled environment of 23°C and 40% relative humidity. The resulting nanofibers were collected on a rotating collector drum covered with aluminum foil, operating at 400 rpm. The electrospinning duration was optimized to achieve nanofibers with the desired thickness of approximately 100 μm.

2.3. Piezoelectric property measurement

Nanofiber substrates were prepared in a cantilever setup, similar to prior work [12,28]. This setup allows for the controlled application of strain to the samples while simultaneously measuring their electrical output. The PAN nanofiber substrates were cut into strips of size 4 × 1.2 cm, and brass slabs of size 7.2 × 1.6 × 0.01 cm, electrically isolated with polyimide tape, were employed as electrodes to measure the voltage. One brass slab was in direct contact with the nanofiber sample, secured with double-sided copper tape, while the other slab remained unexposed. Two 24-gauge wires were soldered to these electrodes and sealed with polyimide tape. To avoid off target static electrical readings, the exposed end of the soldered wires were safely discharged before connecting them to a breadboard with inputs to a PicoScope 2204A (Pico Technology Ltd, Cambridgeshire, UK) for voltage measurement. To induce controlled strain, a 2.3 gram proof mass was placed on the cantilever’s end, driven by a custom-made oscillatory system, with the cantilever holder clamped atop a subwoofer diaphragm. A sinusoidal sine wave with a controlled amplitude and frequency was applied to the speaker system, and the voltage output was measured.

2.4. Scanning electron microscopy images of PES

Scanning electron microscopy (SEM; Thermo Prima environmental-SEM; ThermoFisher, MA) was used to image samples with and without cells. Cell-free samples were fixed on a metallic stud (75210; Electron Microscopy Science, PA) with double sided conductive tape and sputter-coated with gold before imaging under 10 kV. The sample thickness, average fiber diameter, and pore size were measured based on the SEM images using the ImageJ software.

2.5. 2D Cell culture and media collection

Mouse macrophage cell line (RAW264.7) was purchased from the American Type Culture Collection (ATCC) and were cultured in T75 flasks at a density of 2×10⁶ cells per well in DMEM/high glucose medium containing 10% FBS and 1% antibiotic-antimycotic. The medium was exchanged every 48 hours. Once the culture reached 70% confluency, the cells were washed with PBS 3 times, replaced with EV depleted medium, and incubated for 24 hours before collection for sEV isolation. Cells were then harvested after media collection with trypsin–EDTA for 5 minutes followed by spinning down the cells at 1100 rpm for 4 minutes. The cells were resuspended in fresh MEM and diluted appropriately to be counted using a hemocytometer.

2.6. PES Cell culture and media collection for sEV Isolation

The substrates were cut into 3 cm x 1.5 cm strips and processed for seeding, which included washing, coating with chitosan for cell adhesion, and sterilization. Each sample was rinsed in distilled water for 30 min and transferred to a solution of 1 mg mL⁻¹ chitosan in 0.1 M acetic acid for 30 minutes. The coated PES were then rinsed with fresh distilled water for 30 minutes and air dried. The processed samples were placed in a 60 mm diameter Petri dish and UV-sterilized for 30 minutes before seeding. RAW264.7 cells were passaged and seeded separately at 5×10⁵ cells per substrate. The seeded substrates were then incubated for 30 minutes at 37°C before the addition of serum-containing DMEM/high glucose medium and incubation at 37°C and 5% CO₂. After 48 hours, the conditioned medium was replaced with serum-free medium, and PES was stimulated with the appropriate acoustics and incubated for 24 hours before being collected for sEV isolation. All sEV production and concentration measurements, cytokine release, and flow cytometry tests performed with the PES cell culture were performed under 5 biological replicates (N=5) to ensure repeatability. For per cell sEV production rate measurements, cells were counted after media collection using a cell counting kit-8 assay: The culture medium was aspirated, and the substrates were washed with PBS 3 times before adding 3% CCK-8 reagent in the cell culture media and incubation for 2 hours at 37°C and 5% CO₂. The cell count was then measured through the absorbance at 460 nm wavelength.

2.7. Cell seeding efficiency for PES culture

A cell seeding efficiency measurement testing was performed based on a previous study[12]: RAW264.7 cells were seeded separately on 2 mm x 2 mm substrates (N = 3) with and without a chitosan coating at a density of 5×10³ cells per substrate and incubated in serum-containing MEM at 37°C and 5% CO₂ for 24 hours. The cultured medium was aspirated, and the substrates were washed with PBS thrice before the addition of 10% CCK-8 reagent in cell culture media and incubation for 4 h at 37 °C and 5% CO2. The cell count was then measured through absorbance at 460 nm wavelength. Chitosan coated substrates were used for future cell culture tests.

2.8. Cell proliferation test

RAW264.7 cells were seeded separately on 2 mm x 2 mm substrates (N = 3) at 1×10³ cells per substrate and incubated at different set time periods (0.5, 1, 2, 3, 5, 7, 9, 11, 13 days). Media was replaced every day, and confluency was monitored. The media was aspirated, and the substrates were washed with PBS thrice before the addition of 10% CCK-8 reagent in cell culture media and incubation for 4 hours at 37°C and 5% CO₂. The cell count was then measured through absorbance at 460 nm wavelength.

2.9. Acoustic environment characterization

The acoustic stimulation environment was characterized using a microphone coupled with the DecibelX application. The volume and pitch settings were adjusted to achieve an appropriate average amplitude and frequency according to the DecibelX application. The average amplitude and frequency was measured over a period of 5 minutes in an acoustic controlled environment blocking external noise. Acoustic environments were created with a range of 50–110 dB amplitude and 50–1000 Hz frequency using this process.

2.10. sEV isolation and concentration measurement

sEVs were isolated from 100 mL of cell culture media via size-based separation. The isolated media first went through a 0.22 μm filter to capture larger vesicles and cell debris. The flowthrough solution was then added to a 100 kDa centrifuge filter and centrifuged at 200g for 4×30 minutes time periods; the mixture was washed with PBS between each centrifuge session. The sEV solution was then concentrated down to 1 mL for sEV characterization. The sEV concentration and size distribution were measured through nanoparticle tracking analysis (NTA; Nanosight NS300; Malvern, Worcestershire, UK). Samples were diluted appropriately to 1×10⁸ EV/mL to maintain accurate particle counts for size distribution analysis. For each sample, five 60 ″ videos were acquired at a camera level of 10 and detection threshold of 3. The laser chamber was cleaned with milliQ water between each sample reading to ensure no sample contamination occurred. The videos were analyzed using the NTA3.0 software to obtain particle concentration, along with the mean and mode particle sizes of each sample. ExoELISA kit was used to measure the CD63 positive sEVs where the manufacturers protocol was followed thoroughly. Per cell sEV production rates of both NTA and ExoELISA measurements were then calculated at 5 biological replicates for each sample group (N=5).

2.11. Acoustic stimulation and media collection

PES cultures containing serum-free medium were stimulated using sinusoidal acoustic waves (3-Ω subwoofer; PS-EW1-2; Samsung Electronics, Suwon-si, Republic of Korea) in an acoustic controlled box at room temperature and atmospheric CO₂ levels. Samples were stimulated for 15 minutes at various amplitude levels (50–110 dB) and frequencies (50–1000 Hz) levels. To ensure that the non-stimulated control groups experienced the same environmental conditions as the stimulated groups, PES OFF and 2D CNTR samples were also placed in the acoustic controlled box for 15 min without acoustic stimulation. After treatment, all samples were returned to standard culture conditions and incubated at 37°C and 5% CO₂ for 24 hours before media collection for sEV isolation. For the music study, PES cultures were stimulated with the selected music assembly for 15 minutes followed by incubation at 37°C and 5% CO₂ for 24 hours before media collection and sEV isolation. All sEV production and concentration measurements, cytokine release, and flow cytometry tests performed with the PES cell culture were performed under 5 biological replicates (N=5) to ensure repeatability.

2.12. Acoustic characterization of music tracks

An audio input file was imported into a python code and the time-based frequency, waveform, and overall frequencies were plotted. Waveform was determined through plotting the wave amplitude as a function of time. The frequency was determined through applying a fast Fourier transform on the wave amplitude. Stimuli plotted in the study included basic sine waveforms at different frequencies, square waveforms, triangle waveforms, sawtooth waveforms, instruments played at 250 Hz Frequency, and tested music tracks.

2.13. Dissonance calculations of music tracks

Dissonance D_f was calculated and summed up over each frequency at a given time using the following equations:

$$D_f=\frac{1}{2}\sum _{i=1}^n\sum _{j=1}^{n}d(f_i,f_j,v_i,v_j)$$

$$d(f_i,f_j,v_i,v_j)=v_{ij}(e^{-as(f_j-f_i)}-e^{-bs(f_j-f_i)})$$

$$s=\frac{d^{*}}{(s_1{f}_i+s_2)}$$

$$v_{ij}=v_i{v}_j$$

Where f is frequency, ν is amplitude, and s is an interpolation parameter between two frequencies which are determined to be 0.021 and 19 based on least squared fit modeling. The constants a and b, and d* were deciphered constants from previous calculations of local dissonance [29]. The average dissonance was calculated over the time-length of the stimuli and normalized between 0 and 1. Consonance was then calculated using the following equation:

$$Consonance=1-Dissonance$$

These values were used as the horizontal portion of the radar plot.

2.14. Cytokine release measurement of cells on stimulated PES

IL6 and IL10 cytokines were detected using an ELISA kit from Thermofisher (IL6 – 88-7064-88, IL10 - 88-7105-88). Cytokine release tests were performed through 5 biological replicates to ensure repeatability (N=5). The manufacturer’s protocol was followed for the test: 96 well ELISA plates were coated with capture antibody in the coating buffer. The plates were sealed and incubated overnight at 4°C. The wells were aspirated and washed 3 times with PBS. Each sample was soaked for 1 ′ for each wash step followed by blotting the plate on absorbent paper to remove any residual buffer. Wells were then blocked with ELISA Diluent (1X) and incubated at room temperature for 1 h. The diluent was aspirated and washed once followed by the incubation of antigen standard and culture supernatant overnight at 4°C. The sample was aspirated and washed 3 times followed by adding the antibody detection to all wells and incubating at room temperature for 1 h. The wells were aspirated and washed 3 more times. Avidin-HRP (IL6) or Streptavidin-HRP (IL10) was added and incubated at room temperature for 1 h, followed by 5 more washes. TMB solution was added to each well and incubated at room temperature for 15 minutes. Stop solution (2M sulfuric acid) was added to each well and the absorbance at 450 nm was measured using a microplate reader. Protein concentration was calculated using the standard curve absorbance measurements.

2.15. Flow cytometry

Stimulated cells at various acoustic stimuli were detached using enzyme-free dissociation buffer and the stained with either CD86-PE (Clone – GL1, 1:100 dilution), or CD206-PE (Clone – MR6F3, 1:100 dilution) for 30 minutes at 4°C, in a dark environment. Cells were washed 5 times with FACS buffer and then counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (1 μM) for 15 minutes in room temperature in a dark environment. Cells were washed 5 times with FACS buffer and tested for fluorescence using a cytek Norther Lights (NL)-CLC flow cytometer. Positive controls for CD86 consisted of RAW264.7 cells that were stimulated with LPS for 12 hours before detachment. Positive controls for CD206 consisted of RAW264.7 cells that were stimulated with IL10 for 12 hours before detachment. The following gating strategy workflow was used to obtain the fluorescence results: (1) side scatter (SSC) vs. forward scatter (FSC) to remove cell debris and aggregates, (2) SSC vs DAPI to remove dead cells, (3) SSC vs CD86-PE or CD206-PE tested separately against negative and positive control samples, and (4) Count vs CD86-PE or CD206-PE histogram results. Fluorescence results were analyzed through FlowJo. Flow cytometry tests were performed under 5 biological replicates (N=5) to ensure repeatability.

2.16. Western blot

sEVs were lysed with 1 x RIPA buffer (9806; Cell Signaling Technology, USA), and the total protein concentration was quantified using pierce BCA protein assay kits (23225; Thermo Fisher, USA). The protein amount in lysed sEVs was estimated based on a calibration curve plotted by albumin (BSA) standards. Next, 12 μg of proteins from sEV lysates were denatured and loaded on sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE). The separated proteins were subsequently transferred onto a nitrocellulose membrane (1662807; Bio-Rad, USA) and blotted overnight with primary antibodies purchased from Santa Cruz Biotechnology, USA: anti-CD9 antibody (C-4), anti-CD63 antibody (MX-49.129.5), anti-CD47 antibody (B–11), anti-ALIX antibody, anti-TSG101 antibody (W27), The secondary antibodies (anti-mouse HRP-linked antibody, 7076; Cell Signaling Technology, USA) were then treated for blotting and the HRP on the immunoblots was detected by Clarity Max Western Enhanced Chemiluminescence (ECL) substrate (1705060; Bio-Rad, USA) using a ChemiDoc XRS+ system (Bio-Rad, USA).

2.17. Transmission electron microscopy of sEVs

The sEV solutions were negatively stained and imaged through transmission electron microscopy (TEM) using the Talios F200i (S)TEM (ThermoFisher, MA) at an 80 kV accelerating voltage. The TEM samples were prepared by adding 3 μL of 1×10⁹ particles per mL sEV solution to an ultrathin carbon film copper grid and incubating at room temperature for 2 minutes. The solution was then aspirated using filter paper and washed with 3 μL filtered distilled water for 10 seconds. After aspirating the distilled water, 3 μL of Uranyless negative staining solution (22409; Electron Microscopy Science, PA) was added to the sample grid (CF200-CU-25; Electron Microscopy Science, PA) and incubated for 1 ′. The Uranyless solution was aspirated, and the grid was left to dry for 2 minutes before imaging. The sEV particle diameter was measured using the default distance measurement tool in ImageJ.

2.18. Intracellular Ca²⁺ measurements

The intracellular Ca²⁺ concentration was measured using Fura 2-AM following a similar protocol as in previous studies [12]. Cells were seeded at a density of 3×10⁵ cells per PES sample and incubated overnight in 37°C and 5% CO₂. After incubation, they were treated with 10 μM Fura 2-AM in HEPES-buffered saline solution and pluronic f–127 for 1 h in 37°C and 5% CO₂. The cells were then washed to remove the extracellular dye and replenished with MEM. The PES samples were then exposed to VGCC (Nifedipine) and TRPM7 (NS8593) Ca²⁺ blockers (Nifedipine: 300 nM; NS8593: 2 μM), followed by exposure to the appropriate music assembly stimuli. The changes in the fluorescence intensity were measured with a spectrophotometric plate reader (TECAN, Mannedorf, Switzerland).

2.19. Intracellular ROS detection

RAW264.7 cells were seeded on PES samples and cultured for 24 hours. The cells were then blocked with the appropriate Calcium channel blocker and stimulated with the appropriate music assembly for 15 minutes followed by 24-hour incubation in 37°C and 5% CO₂. Each sample was washed three times with PBS and loaded with 2μM DCFH-DA and incubated for 30 minutes at 37°C in a dark environment. After probe loading, the cells were washed three times with PBS and imaged immediately after using a confocal microscope using the 495 nm excitation, and 530 nm emission wavelength fluorescence specifications.

2.20. RNA sequencing and analysis

RNA was isolated from both RAW264.7 cells and sEVs using the Qiagen miRNeasy Tissue/Cells Advanced Kit using the manufacturer’s protocol. Buffer AL was added to a pellet of cells or 700 μl sEV solution at 10¹⁰ sEVs/ml and incubated at room temperature for 3 minutes followed by transferring the lysate to a gDNA eliminator spin column and centrifuged for 30 seconds at 8000 g. The flow through was saved and 1.3x volume of isopropanol was added. The mixture was transferred to a RNeasy UCP MinElute spin column and centrifuged for 15 seconds at 8000 g, discarding the flowthrough. Buffer RWT was added and centrifuged through, discarding the flowthrough. Buffer RPE was added and centrifuged through, discarding the flowthrough. 80% ethanol was added and centrifuged, discarding the flowthrough. RNase-free water was added directly to the center of the spin column membrane and incubated for 1 ′ before centrifuging for 1 ′ at 12000 g. The RNA solution was tested for purity using a nanodrop absorbance spectrometer before RNA sequencing. RNA sequencing was performed (N=2) through a third-party company (HaploX, Hong Kong). Sample quality control tests, including RNA purity (OD260/280) and stability (RIN and 28S:18S), were conducted through HaploX before library construction and sequencing. Gene quantification was performed using HTSeq, and differential sequencing was conducted between each sample using DESeq2. The P_adj values in DESeq analysis was calculated using the BH calculation method (ref) and the differential gene screening criteria was: |log₂(Foldchange)| > 1 and P_adj > 0.05 (Table S1) [30–33].

2.21. MS-based proteomics of sEVs

sEVs were lysed in 1x RIPA buffer. A total of 100 μg of protein from exosomes were trypsinized and eluted using an S-Trap^™ Micro Spin Column. The eluted protein was then dried down and resuspended in a mobile phase buffer. The suspension was loaded into an ultrahigh-resolution mass spectrometer coupled to a nano-ultrahigh-pressure liquid chromatography for bottom-up proteomics.

2.22. sEV efficacy test on MCF7 and 3T3 cell lines

A CCK-8 assay used to short term viability of sEVs from different music genres were tested on MCF-7 cells purchased from the American Type Culture Collection (ATCC). MCF-7 cells were seeded into a transparent 96-well plate at a concentration of 1×10⁴ cells/well. After 24 hours, sEVs were added at a dose of 1000 EVs/cell and transfected the cells for 48 hours at 37°C. 10% CCK-8 reagent was added, and the cells were incubated for 4 hours at 37°C. Cell viability was measured by absorbance at 460 nm and was normalized to the control group without sEVs. For long-term effects, a clonogenic assay was conducted. MCF-7 cells were seeded into transparent 6-well plates at a seeding density of 4×10⁴ cell/well. After 24 hours, sEVs were added at a dose of 1000 EVs/cell and transfected the cells for 48 hours at 37°C. Cell were then washed, trypsinized, and reseeded into 6 well plates at 500 cells/well. Medium was replaced every 48 hours. After 1 week, the medium was aspirated and the cells were washed with PBS three times. Colonies were fixed with 6% glutaraldehyde for 5 minutes at room temperature and stained with 0.5% (W/V) crystal violet for 15 minutes at room temperature. Each well was photographed, and colony numbers were quantified using ImageJ. The relative colony formation (%) was normalized by the control group without sEVs.

2.23. Testing efficacy of sEVs on the cell proliferation of MCF7 and 3T3 cells

MCF7 cells were seeded into a transparent 96 well plate and incubated at different set time periods (0.5, 1, 2, 3, 5, 7, 9, 11, 13 days). Media supplemented with sEVs at 1×10⁷ EVs/well was replaced every day, and confluency was monitored. The media was aspirated, and the substrates were washed with PBS thrice before the addition of 10% CCK-8 reagent in cell culture media and incubation for 4 hours at 37°C and 5% CO₂. The cell count was then measured through absorbance at 460 nm wavelength.

2.24. Testing the effect of sEVs on MCF7, 3T3, and hFOB1.19 cells (scratch assay)

In addition to the MCF7 and 3T3 cells, human osteoblast cells (hFOB1.19) were purchased from the American Type Culture Collection (ATCC). MCF7, 3T3, and hFOB1.19 cells were incubated separately with sEVs at 1000 sEVs/cell for 48 hours on a 6 well plate. Sterilized 20 μl pipet tips were used to make scratches. The scratched gaps were monitored at 0, 12, and 24 hours after the scratch under a light optical microscope. The closure of the gaps was quantified using ImageJ. The relatively scratched area was normalized to the control group.

2.25. Statistical analysis

The data presented in this paper are expressed as the mean ± the standard deviation (S.D.) or the standard error of replicate measurements (S.E.M.) and analyzed using one way ANOVA test using GraphPad Prism 9 software, where applicable.

3. Results

3.1. Design of PES, and acoustic-based optimization of sEV production

We successfully synthesized the PES, (Fig. 2A–B, and Fig. S1) exhibiting robust piezoelectric transduction, converting acoustic-driven mechanical stress into a measurable electrical output in PBS (Fig. 2C). The piezoelectric output of PES can be tuned by adjusting amplitude and frequency of the acoustic input (Fig. 2D). The amplitude of the acoustic input was directly proportional to the peak-to-peak output, plateauing at 191.7 ± 14.6 mV (Fig. 2E). In contrast, the frequency of the input was inversely proportional to the piezoelectric output, with a maximum voltage of 311.5 ± 15.5 mV (Fig. 2F). Long-term piezoelectric performance testing showed a fluctuation between 5.6% and 15.8% of the average peak-to-peak output. Batch-to-batch variation testing showed a variance in piezoelectric performance between 3.2% and 10.1% of the average piezoelectric performance across 5 batches (Fig. S2A). Standardized d₃₃ coefficient measurements for the pristine PES material ranged from 24.11 ± 3.48 pC/N to 27.21 ± 2.68 pC/N, which was consistent with reported values for PAN nanofiber-based materials (Fig. S2B–C) [34–36]. Varying the waveform shape of the acoustic input can alter the shape of the electric output, while possessing similar correlations between frequency and amplitude and electric output as the basic sine wave input (Fig. S3). As a control, the piezoelectric activity of PES can be reduced by up to ~90% through post-synthesis thermal treatment, providing a low-piezoelectric substrate for mechanistic studies (Fig. S4). To improve cell adhesion and biocompatibility, PES was coated with chitosan, which supported 91.2 ± 10.8% viability of RAW264.7 cells (Fig. S5). After chitosan coating, the standardized d₃₃ coefficient decreased to 9.99 ± 2.21 pC/N to 12.54 ± 1.34 pC/N. Although the coating dampened the piezoelectric response, the chitosan coated PES retained sufficient piezoelectric activity to stimulate the cell membrane and generated a similar pattern to that of pristine PES for subsequent studies. We measured and compared the sEV production rate of RAW264.7 cells in traditional 2D culture, PES without stimulation (PES OFF), and PES with acoustic stimuli (PES ON) under 100 Hz, and 80 dB acoustic waves (Fig.2G–H, and Fig. S6). PES ON group successfully produced sEVs 6.7-fold faster than 2D culture control, and 4.1-fold faster than PES OFF group, indicating that the PES stimulated by acoustics promoted sEV secretion. To note, deactivated PES with a stimulation reduced to 19.44 ± 1.99 mV led to a minimal increase in production rate (1.6-fold) compared to 2D culture, thus the change was solely due to the piezo effect rather than acoustics. PES ON derived sEVs showed the highest charges compared to the other sEVs (Fig. 2I), which indicated good dispersity in physiological solutions. Positive western blot signals for sEV surface markers (CD9, CD63, CD47), and cargo (ALIX, TSG101) confirmed the abundant existence of sEVs in all groups showing suitable purity between 2D culture and PES culturing conditions (Fig. 2J and Fig.S7) [37,38]. Lastly, transmission electron microscopy (TEM) imaging of all samples (Fig.S8) showed round morphology with similar size distributions to nanoparticle tracking analysis (NTA) data (Fig.2G). We also successfully tuned the sEV production rate through adjusting the acoustic parameters on the same PES material. We observed a linear increase in sEV production rate when increasing the amplitude of stimuli, whereas the acoustic frequency had an inverse relationship with sEV production rate (Fig. 2K–P). Additional CD63 positive sEV measurements through ExoELISA further confirmed the production rate trends observed in the NTA measurements (Fig. S9). Interestingly, this trend was observed universally when we adjusted the waveform of the stimuli (Fig. S10). When comparing sEV production rate with piezoelectric output, we saw a linearly positive trend with sEV production rate, suggesting the piezoelectric output was a driving force in sEV production (Fig. 2Q). This was supported by observing a linear decrease in sEV production when stimulated with gradually heat deactivated PES material under the same acoustic stimuli (Fig.S4). Interestingly, when comparing zeta potential with piezoelectric effect, we observed an inverse relationship (Fig. 2R). Overall, stimulated PES promoted sEV production from RAW264.7 cells by as high as 15.5-fold (50 Hz, 80 dB stimuli) compared to monolayer 2D culture systems and this sEV production can be tuned through controlling acoustic parameters and thus piezoelectric stimuli.

Fig. 2.

Characterization of the Piezoelectric substrate (PES) and evaluation of sEV production. A. Photographic representation of the PES mat. Scale bar is 1 cm B. Scanning electron micrograph of the microstructure of the PES mat. (inset) The average fiber diameter of the fiber-mat used in future experiments are 225 ± 51 nm. Scale bar is 5 μm. C. Graphical representation of the piezoelectric effect converting acoustic mechanical forces to electrical output for piezoelectric stimulation. D. The relationship of acoustic parameters in wave frequency and amplitude on the piezo-responsive behavior of the fiber-mat. Overall, the frequency is inversely related to the peak-to-peak output (E), and the amplitude is directly proportional to the peak-to-peak output (F). G-J. Production calculations and characterization of the RAW264.7 derived sEVs from various culturing conditions including standard monolayer 2D culture (2D Cntr) which was the 2D culture control samples, PES platform without acoustic stimulation (PES OFF) which was the PES control samples, and the PES platform with acoustic stimulation. G. Size distribution and concentration calculations of sEVs derived from each condition using nanoparticle tracking analysis (NTA). H. Production rate calculation measured through concentration with NTA, and normalized by the culturing time, and cell count through manual cell counting. Units of production rate are EVs/cell/day. I. Zeta potential measurements of sEVs derived from each condition using DLS. J. Validation of the production of sEVs through sEV-derived protein marker characterization using western blot. K-M. Relationship between amplitude of acoustic input on K. size distribution, L sEV production rate, and M. zeta potential. N-P. Relationship between Frequency of acoustic input on N. size distribution, O sEV production rate, and P zeta potential. Q. Linear correlation between sEV production and the piezoelectric output. R. Inverse correlation between zeta potential and the piezoelectric output.

3.2. Music-inspired control of sEV production and cell programming

We used music assemblies spanning common listening genres (e.g. Electronic Dance Music (EDM), classical, rock) as standardized audible acoustic waves to investigate how musically distinct stimuli influence sEV production and macrophage phenotypes. Before choosing the individual assemblies, we first screened assemblies with a fixed tempo of 140 BPM since tempo may influence sEV production and possibly affect cell phenotype. After controlling for tempo, we selected examples that represented three distinct acoustic profiles: (1) a distorted, highly dissonant waveform resembling square and sawtooth waves, (2) a high frequency, low dissonant waveform resembling triangle waves, and (3) a balanced assembly containing features of both waveform profiles. Based on these criteria, we chose the following music assemblies chosen for each genre were (1) On the floor ultrafunk by KPHK to represent Electronic Dance Music (EDM), (2) Don’t Go Breaking My Heart by Elton John and Kiki Dee to represent rock, and Bach’s Suite in B minor, BWV1067: V. Polonaise by Johann Sebastian to represent classical music. To understand how the acoustic features differentially engage piezoelectric substrates and cellular responses, we extracted each track’s waveform, time-resolved spectral content, bulk frequency distribution, and acoustic dissonance (Fig. 3A). These three music assemblies possessed significant differences in sound pressure (power), spectral composition (frequency), and harmonic dissonance. First, the EDM assembly possessed strong single frequency energy concentrated in bass and sub-bass bands (20-200 Hz) at high sound pressures resulting in low-frequency power. In addition to the powerful sub bass frequencies, EDM displayed near-maximal normalized dissonance (0.98 ± 0.05) (see Methods), referring to groups of low frequency stimuli interacting with each other resulting in considerable amplitude instability, previously reported to promote systemic proinflammatory responses under short term exposures (Fig. 3A.i) [39–41].

Fig. 3.

A Schematic representation of the acoustic characterization process starting with an audio input, and obtaining the time-based waveform, the fast Fourier transform frequency-based information, and a radar plot characterization the (top) total acoustic power at frequencies below 200 Hz, (left) Frequency density at frequencies lower than 200 Hz, (bottom) Overall power at frequencies above 200 Hz, and (right) frequency density at frequencies above 200 Hz. Acoustic characterization was performed for 3 musical samples. (i) An electronic music sample containing powerful low frequency waves, and high density at <100 Hz frequencies, resulting in a radar plot at the top-left position. (ii) A rock music sample containing a balance both power and density in low and high frequencies resulting in a radar plot at the center. (iii) A classical music sample containing a powerful high frequency wave, and high density in >200 Hz frequencies, resulting in a radar plot at the bottom-right position. B-F. Testing the effects of the acoustic stimuli input on the sEV production and macrophage phenotype of RAW264.7 cells through sEV production calculation, cytokine release measurements, and Cell marker characterization. B. A flow chart representation of the characterization process to observe the immune response and sEV production output based on a given music input. C-D. sEV production measurements of the 3 musical stimuli compared to the 2D culture control sample (2D Cntr). E. Surface marker characterization of CD86 (M1 polarization) and CD206 (M2 polarization). Negative controls and the positive control samples were measured to compare the signal with the musical stimuli. The negative control was unstained cells, the positive control for CD86 was LPS stimulated macrophages, and the positive control for CD206 was IL–10 stimulated macrophages. F. Measurement of cytokine concentration released into the cell culture supernatant of the control groups, and the various musical stimuli groups. Cytokine concentrations were measured using ELISA (N=5, mean ± s.d).

In contrast, the classical assembly was characterized by powerful single frequency signals at high frequencies, outside the primary piezoelectric activation range of the PES, greater than 200 Hz. Classical music also exhibited minimal dissonance (0.02 ± 0.01), describing a stable, harmonious construction of frequencies, with minimal amplitude instability at high frequencies, previously reported to promote systemic anti-inflammation outcome (Fig. 3A.ii) [20,21]. The rock assembly fell between these extremes, with broader spectral balance and intermediate dissonance (0.51 ± 0.05) (Fig. 3A.iii). When comparing bulk frequency information with dissonance calculations, we observed a consistent trend: high signal density occurring at sub-bass frequencies was a driving force in promoting acoustic dissonance. To test whether this relationship generalizes beyond music, we performed parallel analyses using canonical waveforms (triangle, square, and sawtooth). As expected, waveforms with higher harmonic density exhibited higher dissonance, with sawtooth > square > triangle (Fig. S11–13).

To understand how music-driven activation of the PES influences sEV biogenesis and macrophage polarization, RAW264.7 cells were cultured on PES and applied defined acoustic stimuli (waveforms and music assemblies) and then quantified functional outputs at the vesicle and cellular levels. Specifically, we measured (i) sEV production rate, (ii) polarization-associated cytokine release (IL-6 and IL–10), and (iii) macrophage surface marker expression (CD86 and CD206) (Fig. 3B–D). As a result, all music stimuli promoted sEV production compared to 2D culture system (Fig. 3E). EDM promoted the highest sEV production enhancement (991.5 ± 21.4 EVs/cell/day) followed by Rock (502.7 ± 6.7 EVs/cell/day) and classical (429.4 ± 7.9 EVs/cell/day) (Fig. 3B–D). These trends were observed similarly when measuring CD63 positive sEVs through ExoELISA, further confirming the sEV production trend observed in NTA (Fig. S14). Interestingly, the power at low frequencies (<200 Hz) followed the sEV production trend (EDM > rock > classical) (Fig. 3A). EDM’s stronger power at low frequencies increased piezoelectric stimulation, inducing greater sEV production [13]. Control experiments were conducted using macrophage cells on PES without stimulation (PES OFF) to determine whether the substrate and chitosan coating alone influenced macrophage polarization. Analysis of macrophage surface marker expression and cytokine release showed no evidence of polarization under the PES OFF condition. These results indicate that the chitosan coating did not measurably affect macrophage polarization in the absence of acoustic stimulation. When we assessed macrophage polarization under music stimuli, highly dissonant EDM stimuli (Dissonance = 0.98 ± 0.05) induced a significant increase in CD86 expression and IL6 cytokine release, suggesting M1 polarization of macrophages (Fig. 3E–F). In contrast, highly consonant classical music stimuli (dissonant = 0.02 ± 0.01) promotes CD206 expression and IL10 cytokine release, suggesting M2 polarization (Fig. 3E–F). Lastly, we observed no significant change in surface markers or cytokine release under a balanced dissonant rock stimulus (dissonance = 0.51 ± 0.05) (Fig. 3E–F). To support this correlation between dissonance and polarization, we compared different levels of dissonant waveforms (sawtooth > square > triangle) as controls and found highly dissonant sawtooth (0.86 ± 0.12), and square waves (0.74 ± 0.22) promoted M1 polarization, while low dissonant triangle waves (0.24 ± 0.19) promoted M2 polarization (Fig. S15–18). We confirmed the morphological stability and size distribution of sEVs from all groups through TEM and NTA (Fig. S19–20). Overall, these results showed that macrophage polarization can be controlled by tuning the dissonance of the acoustic stimuli, highlighting how musically distinct inputs, and the piezoelectric simulation they elicit, can program macrophage state.

3.3. Music-inspired stimuli controls EV production and cell programming through voltage gated Ca²⁺ channels (VGCC) and mechanically gated Ca²⁺ channel (TRPM7) pathways

After observing the music-driven phenomenon, we investigated the underlying mechanism by decoupling the contributions of stimulus types (mechanical and electrical), key acoustic parameters (frequency and dissonance), and the Ca²⁺ influx pathways (VGCC and TRPM7). Intracellular Ca²⁺ measurements showed that deactivating the electrical component of the PES decreases intracellular Ca²⁺ levels by ~46.3 ± 3.1% under EDM music assembly stimuli (Fig. 4A). This reduction was less pronounced under Rock and Classical assembly stimulation, with decreases of 18.9 ± 2.8% and 4.2 ± 1.2%, respectively (Fig. 4B–C). These results suggested that the electrical component of PES contributes most strongly to Ca²⁺ influx under EDM stimulation. Furthermore, blocking the VGCC Ca²⁺ pathway produced a comparable decrease in intracellular Ca²⁺ to that observed after deactivating the electrical component, indicating that the electrical component and VGCC activation are functionally linked.

Fig. 4.

The effect of calcium channel activity on sEV production and macrophage phenotype. A-C. The effect of deactivating the electrical component of the piezoelectric stimulation (V_off), blocking voltage gated calcium channel (VGCC_off) and mechanically gated calcium channel (TRPM7_off) on calcium influx for A. EDM, B. Rock, and C. Classical stimuli. PES sample with no piezoelectric stimulation was the negative control. D. The calculated percent calcium influx after each musical stimuli. E-G The effect of calcium concentration on sEV production through V_off, VGCC_off, and TRPM7_off conditions for EDM, Rock, and Classical stimuli. PES OFF was the negative control. (F) sEV production rate calculation through nanoparticle tracking analysis and CD63 ELISA assay. (G) The effect of intracellular calcium concentration on the sEV production rate. (H-I) Effect of calcium influx pathway on surface expression of CD86 and CD206 through flow cytometry, and (I) comparison of the mean fluorescence intensity of CD86 and CD206 for each condition and stimuli. (J) (i) IL6 and (ii) IL10 cytokine release measurements through V_off, VGCC_off, and TRPM7_off conditions for EDM, Rock, and Classical stimuli. Measurement of intracellular ROS through V_off, VGCC_off, and TRPM7_off conditions for EDM, Rock, and Classical stimuli. Scale bar is 100 μm. (L) Schematic representation of the mechanism where EDM stimuli is activating the VGCC pathway which is promoting ROS generation and thus enhancing M1 polarization and inflammation. In contrast, Classical stimuli is promoting the TRPM7 calcium influx pathway which through literature search, upregulates STAT6 signalling, promoting M2 polarization and regeneration.

Analysis of Ca²⁺ influx pathways under each musical assembly revealed distinct pathway distributions across stimulation conditions (Fig. 4D). EDM stimuli primarily promoted Ca²⁺ influx through the VGCC pathway (69.5 ± 4.7%) rather than the TRPM7 pathway (21.2 ± 3.9%). In contrast, Rock stimuli induced a more balanced contribution from VGCC and TRPM7 pathways (45.2 ± 4.8% vs 45.8 ± 4.5%). Classical stimuli predominantly activated the TRPM7 pathway (79.2 ± 6.8%) compared with the VGCC pathway (12.74 ± 1.8%). These findings indicate that different musical stimulation conditions promote Ca²⁺ influx through distinct piezoelectric mechanisms: electrical-potential-dominated stimulation in the EDM assembly, mechanically dominated stimulation in the Classical assembly, and a balanced electrical/mechanical contribution in the Rock assembly.

NTA and CD63 ELISA measurements revealed a linear relationship between intracellular Ca²⁺ concentration and sEV production (Fig. 4E–G). Under the EDM assembly stimulation, sEV production was most strongly reduced when the electrical component or VGCC pathway was blocked, resulting in a 61.7 ± 5.1% decrease in sEV production rate. Rock and Classical assemblies showed similar reduction in sEV production with decreases of 46.9 ± 3.7% and 36.5 ± 2.5%, respectively. These results suggests that intracellular Ca²⁺ concentration is a major regulator of sEV production, with EDM stimulation producing the highest sEV output among the three musical assemblies. In addition to exploring the effect of music-driven piezoelectric stimuli on Ca²⁺ influx pathways. We also decoupled the effects of frequency and dissonance acoustic parameters on Ca²⁺ influx. Decoupling was achieved by adjusting the dissonance through playing various overtone frequencies (100-200 Hz) of a waveform while keeping the carrier frequency the same at 100 Hz, and we named these stimuli conditions based on the ratio of the overtone frequency with the carrier frequency. We also looked at different carrier frequencies with the same dissonance level. Dissonance calculations show the peak dissonance is at an overtone-to-carrier frequency ratio of 1.3 with a log normal distribution at higher frequency ratios (Fig. S21A) Ca²⁺ concentration measurements demonstrate that dissonance is the dominant force that promotes Ca²⁺ influx compared to the carrier frequency (Fig. S21B).

After looking at the distribution of Ca²⁺ influx and its effect on sEV production rate, we explored its effects on macrophage polarization. Flow cytometry showed the effect of the PES’s electrical potential and pathway of Ca²⁺ influx on macrophage polarization. Under the EDM assembly, we observed a loss in CD86 expression and an increase in CD206 expression when the electrical potential or VGCC pathway was blocked, while the control PES, and the TRPM7 deactivated condition promoted CD86 expression. This suggests that these components have influence on the M1 polarization process (Fig. 4H–I). Under the Rock assembly, CD206 expression increased when electrical potential or VGCC pathway was blocked, and CD86 expression increased when TRPM7 pathway was blocked, signifying that there was a balance between both competing Ca²⁺ influx pathways (Fig. 4H–I). Under the Classical assembly, VGCC pathway or electrical deactivation did not affect the surface expression, while the TRPM7 pathway significantly reduced the CD206 expression, demonstrating the influence of the TRPM7 pathway on the M2 polarization process (Fig. 4H–I). Cytokine release measurements showed a similar trend to the surface expression where we observed a significant drop in IL6 release across all samples when blocking both M1 polarization driving components (Electrical stimuli + VGCC Ca²⁺ pathway). On the other hand, we observed decreases in IL10 secretion when we deactivated the M2 polarization driving component (TRPM7 Ca²⁺ pathway). Lastly, we observed the effect of the stimuli on ROS generation due to ROS’s ability to promote M1 polarization if over generated. We observed significantly higher accumulation in the ROS under the EDM assembly. ROS generation reduced completely when deactivating the electricity and VGCC channels, thus linking M1 driving components towards excessive ROS generation (Fig. 4K). Overall, we observed that the electrical component of piezoelectric stimulation and the VGCC Ca²⁺ pathway were responsible for M1 polarization in macrophages through ROS generation while TRPM7 Ca²⁺ pathway influx promotes M2 polarization which is previously known to activate cell programming through the STAT6 pathway (Fig. 4L). From our music assemblies, we observed EDM’s dissonance, and low frequency stimuli promote VGCC Ca²⁺ pathway influx enhancing M1 polarization through ROS generation. This was further supported by the effect of dissonance on macrophage polarization, cytokine release, and ROS generation, with dissonance coupled with the electrical component being the dominant effector of CD86 expression, IL6 secretion, and ROS generation (Fig. S21C–E). In contrast, Classical’s consonant, high frequency stimuli promoted TRPM7 Ca²⁺ pathway influx enhancing M2 polarization. Lastly, Rock’s balance of frequencies and dissonance promotes a balance of Ca²⁺ pathway influx resulting in M0 polarization.

3.4. Cargo analysis and In Vitro application of reprogrammed sEVs

We next performed RNA sequencing and proteomics profiling to validate music-driven cell phenotype shifts and determine whether they were reflected in sEVs. First, sample quality control reports confirmed high RNA purity (OD 260/280 > 2.0), and high stability (RIN > 9.8, 28S:18S > 1.7), suitable for reliable RNA-sequencing tests (Table S2). In addition, all samples achieved sequencing depths greater than 39 million reads per sample, with total mapping rates above 92% and unique mapping rates above 86%, providing sufficient depth and quality for reliable differential sequencing analysis (Table S3). When comparing the sEV and cell samples across all conditions, the global RNA signatures of sEVs broadly tracked those of their parental cells, consistent with phenotype-dependent vesicle biogenesis. Differential expression analysis revealed robust, stimulus-specific divergence across music conditions (Fig. 5A). Pathway enrichment analysis further highlighted distinct biological programs. In EDM-stimulated cells, adaptive immune processes and inflammatory pathways were significantly upregulated, whereas gene sets associated with cell cycle progression, cell division, and proliferation were downregulated. Interestingly, the EDM-sEVs possessed enhanced miRNA cargo for anti-cancer effects such as miR-34a, miR–146, miR-200, miR-206, and miR-31, making EDM-sEVs a good candidate as a tumor suppressive therapeutic (Fig. S22). In contrast, classical stimulation produced the opposite trend, with enrichment of proliferation-associated pathways and suppression of inflammation signaling. Rock-stimulated cells possessed similar expressions to the 2D control, exhibiting modest changes across these pathways (Fig. 5B and Fig. S23–25). Under all music conditions, the expression of CACNA (voltage-gated Ca²⁺ channel) gene family were upregulated relative to the 2D controls, and positively correlated with sEV production rate, suggesting Ca²⁺ influx as a driver of enhanced vesicle release (Fig. 5C). We next profiled sEV proteomic cargos and surface markers across music-stimulated conditions and benchmarked them against 2D culture and PES OFF controls (Fig. S26). All groups exhibited sEV surface markers (CD63, CD9, CD47) and cargos (ALIX, TSG101). Consistent with the observed cellular polarization, EDM-sEVs possessed M1 cargos in the form of proinflammatory cytokines (IL6, IL7, IL16), immune activation markers (CD86, CD84, CD27, etc.), and tumor growth-inhibitor proteins (SCAI, MIIP). In contrast, classical-sEVs possessed M2 cargos in the form of anti-inflammatory markers (CD206), matrix metalloproteinases (MMP3, MMP13) and collagen subunits (COL27A1, COL18A1, COL11A2) for bone and cartilage regeneration. To further investigate the potential therapeutic implications, we investigated the intrinsic therapeutic effects of EDM-sEVs and classical-sEVs in vitro. Interestingly, EDM-sEVs promoted cytotoxicity in breast cancer cells compared to the 2D-control-sEVs, with a decrease in cell density (Fig. 5D–F). We also observed inhibition of cell proliferation and thus insignificant gap closure from breast cancer cells incubated with EDM-sEVs, further supporting their anti-cancer potential (Fig. 5G–I). In contrast, classical-sEV promoted cell growth as observed in the enhanced cell density compared to 2D-control-sEVs (Fig. 5D–F), while accelerating gap closure, and long-term proliferation, demonstrating regeneration potentials (Fig. 5G–I, and Fig. S27).

Fig. 5.

Cellular and sEV RNA vary with music stimuli. A. Differential sequencing between EDM, Rock, and Classical music stimuli compared to the 2D controlled sample (CNTR). B. Clustered heatmap quantifying expression correlated to inflammation (top) and cell growth (bottom) C. Relative expression of CACNA1 family RNA responsible for voltage-gated calcium channel protein expression compared to the 2D controlled sample (CNTR) D-F characterization of anti-cancer properties of music derived sEVs when incubated with MCF7 cells. All music samples are compared to the 2D controlled sample (CNTR) D. measurements of the natural sEV toxicity to MCF7 cells using a Live/Dead assay. Scale bar is 50um. E. calculated cell viability from the live/dead assay. (N=3, mean ± s.d.) F. calculated cell density G-H characterization of regenerative properties of music derived sEVs when incubated with MCF7 cells. The control sample was a 2D controlled sample (CNTR) G. Scratch assay with 1000 EVs/cell dosage of sEV. H. Quantitative measurements of the scratch area (N=5, mean ± s.d). I. long-term proliferation study characterizing sEVs effect on MCF7 cell proliferation compared to the 2D controlled sample (CNTR) (N=3, mean ± s.d) One-way ANOVA with Tukey’s post-test. Ns: not significant; *p>0.05; **p<0.01; ***p<0.001; ****p<0.0001.

3.5. Engineering regenerative stimuli for enhanced production of anti-inflammatory sEVs

To integrate the high yield sEV production elicited by EDM with the pro-regenerative bioactivity associated with the classical stimuli, we designed a composite music assembly (“Engineered Regeneration”) that combined the key acoustic elements underlying each effect. The assembly incorporated (i) a sub-frequency bass within the resonance frequency of PES to promote sEV production, (ii) a piano chord progression that promotes M2 polarization suitable for regeneration applications, and (iii) water sound as a negative control that does not affect the cells. Among the 13 screened instrument samples, three instruments (Piano, Harp, and Clarinet) consistently promoted M2 polarization, which we attribute in part to their low dissonance timbre profiles (Fig. S28–31). Among these instruments, Piano was chosen because it provided the broadest frequency range, offering greater flexibility to optimize high-frequency power while maintaining high frequency consonance (Piano: 27.5-4186 Hz vs. Harp: 30.9-3322.4 Hz vs. Clarinet: 146.8 – 1864.7 Hz) [42,43]. This broader spectral coverage made piano the most suitable instrument for subsequent acoustic optimization. Acoustic analysis confirmed that the Engineered Regeneration music assembly combined high power within the PES-optimal frequency window, driven by the bass component (peak-to-peak voltage = 211.12 ± 18.44 mV), with a harmonically stable, low-dissonance musical structure provided by the piano chords (Fig. 6A, and Fig. S32). We then deconvoluted the contributions of each module by stimulating RAW264.7 cells with the individual components and defined combinations, and quantifying both vesicle output and polarization markers. Stimuli that contain the bass component (Fig. 6B–D: 1, 4, 6, 7) promote sEV production up to 8.2-fold higher than 2D culture, indicating sub-bass–driven piezoelectric activation as the dominant determinant of vesicle yield (Fig. 6B and Fig. S33). These sEV production trends observed through NTA were cross referenced with CD63 positive sEV production measurements using ExoELISA where the trends were similar (Fig. S34). In parallel, stimuli containing the piano component (Fig. 6B–D: 2, 4, 5, 7) promoted M2 polarization characterized by CD206 expression and IL10 release, identifying that the consonant piano progression is sufficient to promote an M2-like program under these conditions (Fig. 6C–D). Across all engineered conditions, sEVs maintained round morphology and comparable size distribution (Fig. S35–36), supporting that the stimulus primarily reprogrammed output and bioactivity rather than gross vesicle structure. We also observed enhancements in bone regenerative and osteogenesis-based miRNA such as miR-21, miR-375, miR-410, miR–127, and miR-27, making them good candidates for bone regenerative therapeutics (Fig. S22). Finally, we evaluated the functional consequence of the engineered regimen in a regenerative context. Engineered-Regeneration-sEVs increased hFOB1.19 osteoblast cell density, ~3-fold enhancements relative to 3D control sEVs and accelerated scratch closure ~2-fold faster, achieving 100% closure within 24 hours (Fig. 6E–G). These results demonstrated a rational, modular approach to “composing” acoustic–piezoelectric stimuli that jointly maximize sEV manufacturing and program therapeutic function, highlighting the PES platform as a tunable route to application-specific, phenotype-imprinted vesicle therapies.

Fig. 6.

Characterization of custom music optimized for promoting sEV production and M2 polarization for regeneration potentials. A. acoustic characterization of the regeneration music. B. A component analysis showing the impact of each acoustic component on piezoresponsive behavior (solid) and sEV production rate (striped), (N=5, mean ± s.d.) C. FACS characterization of CD86 and CD206 markers of each component. D. IL6 and IL10 cytokine measurements through ELISA assays (N=5, mean). (E-G) Characterization of tissue regeneration potential of regeneration music derived sEVs on hFOB1.19 cells. E. Scratch assay with a range of the sEV dosages. The negative control was 0 EVs/cell dosage F. Quantification of the scratch area with respect to time (N=3, mean ± s.d.) G. Cell density measurements of hFOB1.19 after incubation of regeneration music derived sEVs. (N=3, mean ± s.d.) Quantitative measurements of the scratch area (N=5, mean ± s.d). One-way ANOVA with Tukey’s post-test. Ns: not significant; *p>0.05; **p<0.01; ***p<0.001; ****p<0.0001.

4. Discussion

In this study, we leveraged our PES platform to connect audible acoustic inputs to bioelectrical cell stimulation and thereby control three coupled outputs: (1) sEV production rate, (2) macrophage phenotype and the resulting sEV cargo, and (3) the intrinsic therapeutic activity of the produced sEVs in vitro. Building on these, we further demonstrated a rational “compositional” strategy, an engineered music stimulus, that simultaneously maximizes vesicle yield and tunes vesicle function toward tissue regeneration, providing a tunable route to fast, controlled production of phenotype-imprinted sEVs.

A key mechanistic question is what drives the increase in sEV output under PES stimulation. Prior approaches typically enhance vesicle production using either high-voltage electrical stimulation to trigger Ca²⁺ influx or high-frequency mechanical stimulation (e.g., surface acoustic waves) to engage mechanotransduction [15,44–46]. In our platform, acoustic excitation generates both mechanical vibration and piezoelectric charge; however, two observations point to the electrical component of the piezoelectricity as the dominant contributor. First, thermal deactivation of PES’s electrical component of the piezoelectricity produced a proportional, linear reduction in sEV production rate, implicating the electrical transduction pathway as necessary for the yield gains (Fig. S4). Second, the low-frequency regime that most strongly boosted production (<200 Hz) differs from the previously reported MHz-GHz frequency range obtained through purely mechanical pathways [46]. This suggests that there is a different acting mechanism that is driving the boosted production which we uncovered in Fig. 4. Thus, PES enables a sound-to-electric route for scalable sEV production in an easily programmable frequency band. Importantly, this platform adds a capability largely absent from existing stimulation-based manufacturing: by shaping the acoustic “timbre” (including harmonic structure/dissonance), we can tune macrophage polarization and thereby tune the functional subtype of macrophage-derived sEVs illustrated by the engineered regenerative stimulus (Fig. 6).

Another key mechanistic question is which components of music-driven piezoelectric stimulation regulate macrophage reprogramming between pro-inflammatory and anti-inflammatory states. To address this question, our mechanistic study (Fig. 4) was designed to decouple three major contributing factors: (1) the electrical and mechanical components of piezoelectric stimulation, (2) distinct Ca²⁺ influx pathways, including voltage-gated calcium channels (VGCCs) and the mechanically sensitive TRPM7 channel, and (3) acoustic parameters, particularly frequency and dissonance, that may influence macrophage polarization. To better understand why different music stimuli induce distinct macrophage phenotypes on the PES, we systematically evaluated these parameters under each music stimulation condition.

First, we found that EDM and classical music stimuli preferentially activated different Ca²⁺ influx pathways. EDM stimuli predominantly promoted through VGCCs, whereas classical music stimulation preferentially activated the TRPM7-mediated Ca²⁺ influx pathway. Second, inhibition studies revealed that VGCC-mediated Ca²⁺ influx was closely associated with M1 polarization, while TRPM7-mediated Ca²⁺ influx contributed to M2 polarization. These findings are consistent with the phenotypic observations shown in Fig. 3. Previous studies have reported that TRPM7 Ca²⁺ influx pathways are necessary for M2 polarization, and can activate STAT6 signaling, thereby promoting M2 polarization [47].

Furthermore, EDM stimuli, together with the voltage component of piezoelectricity and active VGCC-mediated Ca²⁺ influx, induced intracellular ROS accumulation. When either the VGCC-mediated Ca²⁺ influx or the voltage component of piezoelectric stimulation was inhibited, ROS accumulation was significantly reduced. Because intracellular ROS generation is known to promote M1 polarization through the activation of MapK and NFkB signaling pathways [48], these results suggested that EDM stimulation drove macrophages toward a pro-inflammatory phenotype by activating VGCC-mediated Ca²⁺ influx, increasing ROS generation, and subsequently engaging MAPK/NF-κB signaling. In contrast, classical music stimuli promoted an anti-inflammatory M2 phenotype by activating TRPM7-mediated Ca²⁺ influx and downstream STAT6 signaling. For Rock stimuli, the relatively balanced activation of VGCC- and TRPM7-mediated Ca²⁺ influx pathways may explain why macrophages largely remain in an unpolarized M0-like state, rather than shifting strongly toward either M1 or M2 polarization.

We further investigated whether dissonance vs frequency played a more dominant role in macrophage polarization. By varying the dissonance range while maintaining the same carrier frequency (Fig. S21), we observed graded changes in CD86 expression, IL-6 cytokine secretion, and ROS generation. These results indicate that dissonance is a key acoustic parameter governing music-induced M1 polarization.

Lastly, we found a distinct shift in macrophage polarization when the stimulation frequency exceeded 200 Hz. As shown in Fig. 4, M1 polarization appears to be primarily associated with electrical stimulation, whereas M2 polarization is more closely linked to mechanical stimulation. Our frequency-dependent piezoelectric measurements showed an exponential decrease in voltage output, with the generated voltage becoming undetectable above 200 Hz (Fig. 2D–E). These results suggest that the change in cellular reprogramming observed near 200 Hz reflects a transition in the dominant stimulation mechanism, from electrically driven M1 polarization at lower frequencies to mechanically driven M2 polarization at higher frequencies.

Multi-omic profiling further indicates that music-dependent shifts in cell state are imprinted onto vesicle composition. EDM stimulation produced a proinflammatory, immune-activation–enriched cellular program, supported by sEV cargo enriched for inflammatory cytokines and activation markers [49–51], together with proteins linked to suppression of tumor growth and invasion (e.g., SCAI, MIIP) [52–55]. These signatures align with the observed anti-proliferative and anti-migratory activity of EDM-sEVs in vitro. Therefore, this music-guided acoustic piezoelectric stimuli platform can boost the sEV production with natural anti-cancer properties which can and have been used in combination therapy applications with chemotherapy agents resulting in synergistic efficacy [56]. In contrast, classical stimulation drives cells towards proliferation-associated programs while dampening inflammatory pathways; correspondingly, classical-sEVs promoted proliferation and migration and were enriched for ECM remodeling and osteochondral-associated proteins (including MMPs and collagen subunits), consistent with regenerative function [57–60]. In addition to MMPs and collagen subunits, specific immunometabolic miRNA cargo (miR-21, miR-99a, miR–146b, miR-378a, miR-23a/b, and suppression of miR-33) was enriched in Classical and Regeneration-derived sEVs. These miRNAs have been associated with regenerative functions, including modulation of inflammatory signaling, tissue repair, recovery from heart failure, and attenuation of cardiac functional decline. (Fig. S22) [61]. These findings suggest that the Classical sEVs may possess regeneration-associated bioactivity. Most importantly, the Regeneration-derived sEV production platform can boost the production of reprogrammed sEVs with similar cargo to clinically relevant sEVs with natural cardiac repair applications [61,62]. However, because Classical sEVs also enhanced cancer cell proliferation and migration in vitro, their potential pro-proliferative effects should be carefully evaluated in future studies, particularly for applications involving patients with existing or early-stage pathological tissues. To assess whether cross-species incubation contributed to the observed effects, we treated murine 3T3 cells with murine sEVs as a control (Fig. S37). Similar migration and proliferation trends were observed across 3T3, MCF7, and hFOB1.19 cells, suggesting that these effects are driven mainly by sEV cargo rather than xenogeneic interactions. Overall, these data support a model in which fundamental acoustic structure, transduced through PES, acts upstream of macrophage polarization, which then determines sEV bioactivity.

This work also has practical implications for sEV biomanufacturing. The ability to independently and modularly control yield and function suggest a generalizable framework for designing acoustic–piezoelectric regimens matched to therapeutic intent without genetic engineering or extensive biochemical conditioning. Meanwhile, several limitations remain. Although production is accelerated, the downstream workflow remains similar to standard adherent culture and would benefit from integration with automated, closed-loop collection and purification to reduce labor and improve throughput. In addition, the current implementation favors adherent cells; extending these concepts to suspension systems will be important to broaden compatibility to other clinically relevant producers. Third, routine cell culture procedures, such as changing media, and transportation of the material may slightly alter the position of the PES material within the culture dish. Such positional changes could introduce unintended variation in piezoelectric output outside the defined acoustic stimulation period, potentially affecting the stability and repeatability of the stimuli. Nevertheless, our repeated biological experiments demonstrate that the overall stimulus-input/phenotype-output workflow remains reproducible despite possible minor positional changes. Although the music assemblies were normalized by tempo during the initial screening process, they differed in acoustic properties including frequency spectrum, amplitude distribution/stability, waveform profile, and dissonance. These parameters may influence PES-mediated mechanical/electrical stimulation of macrophages, potentially affecting calcium ion influx, macrophage polarization, and sEV production. However, because music assemblies contain coupled acoustic features, future studies using defined synthetic waveforms will be needed to determine the independent contribution of each acoustic parameter. Lastly, although the music-programmed sEVs were effective in anti-tumor, and regenerative activities in vitro, their efficacy and safety in vivo remain to be determined. Future studies using live animal models will be essential to evaluate their biodistribution, therapeutic efficacy, immunogenicity, and potential toxicity in a physiologically relevant environment. Addressing these limitations and related engineering constraints will further support the development of music-guided acoustic–piezoelectric stimulation as a scalable and programmable strategy for producing therapeutic sEVs.

5. Conclusion

Our work establishes a programmable, noninvasive strategy to jointly control sEV yield and function through precisely defined acoustic “music assemblies.” By tuning frequency and amplitude, we significantly increased sEV production while simultaneously directing sEV bioactivities. Importantly, music assemblies revealed a mechanistic link between waveform characteristics and macrophage fate: dissonant, powerful low-frequency stimuli biased RAW264.7 cells toward an M1-like inflammatory state, whereas consonant higher-frequency stimuli promoted M2-like polarization and a regenerative milieu. These polarization shifts were reflected in sEV cargos, generating vesicles with distinct, cell-mimetic bioactivities. Leveraging this insight, we engineered a custom stimulus that maximized both sEV output and M2 polarization, producing vesicles that enhanced osteoblast growth. Together, these findings position PES as a tunable, universal piezoresponsive platform that bridges acoustic-based interventions with sEV engineering, enabling scalable, cargo-tailored, and application-ready sEV therapeutics.

Supplementary Material

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.actbio.2026.07.005.

Data Availability:

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