聚焦超声增加 Lucifer Yellow 跨 hCMEC/D3 内皮单层转运

Abstract

Introduction:

Focused ultrasound (FUS) is being investigated as a non-invasive approach to modulate vascular barriers and thereby facilitate drug delivery to the central nervous system and other protected tissues. However, interpretation of in vitro endothelial responses to FUS is complicated by culture-vessel acoustics and by the need to distinguish transport occurring during exposure from transport occurring afterward.

Methods:

We investigated Lucifer Yellow transport across hCMEC/D3 human cerebral microvascular endothelial monolayers cultured on ThinCert inserts during and after a 60-s continuous-wave FUS exposure at 1.2 MHz and 80 V_pp (peak-to-peak voltage) without added microbubbles. Hydrophone measurements at 100 V_pp identified 1.2 MHz as the frequency yielding the highest mean peak rarefactional pressure in the frequency scan at an axial distance of 5.0 mm from the lens apex. Separate axial scans at 1.2 MHz located the axial pressure maximum at the same distance; first-order voltage scaling of the corresponding peak rarefactional focal pressure (PRFP) yielded an estimated PRFP of 0.838 ± 0.054 MPa for the 80-V_pp condition in the calibration geometry. Lucifer Yellow accumulation was quantified immediately before exposure, at the end of exposure, and 9 min later in eight independent paired experiments using sampling- and volume-corrected receiver amounts.

Results:

During the 60-s exposure interval, the mean control-adjusted increase was 0.283 nmol (95% confidence interval, 0.185–0.380 nmol; Holm-adjusted ; Cohen’s ), and all eight experiments showed the same effect direction. During the subsequent 9-min interval, the additional control-adjusted accumulation was 0.035 nmol (95% confidence interval, −0.100–0.171 nmol; Holm-adjusted ). In a separate preliminary cell-free session, three within-session FUS/control pairs did not reproduce the positive exposure-interval response seen in the cellular series; these observations were descriptive only. In five separate paired experiments, acute Calcein-AM- and ethidium homodimer-1-derived image metrics showed mixed FUS-minus-control differences without a reproducible FUS-associated direction.

Discussion:

These findings demonstrate an exposure-associated increase in Lucifer Yellow transport across the hCMEC/D3 monolayer–insert system that arose predominantly during insonification, without statistically detectable continued differential accumulation during the observed post-exposure interval. Together, these results provide a controlled in vitro framework for temporally resolving FUS-associated transport in this monolayer–insert system.

1 Introduction

Neurovascular endothelial barriers regulate molecular exchange between the circulation and neural tissues. Two prominent endothelial examples are the blood–brain barrier (BBB) and the inner blood–retinal barrier (BRB), which maintain tightly regulated extracellular environments required for neuronal function while restricting the tissue penetration of many systemically administered therapeutics (; ). This protective function creates a major therapeutic challenge in neurological and retinal diseases, because many systemically administered drugs and emerging therapeutic modalities do not efficiently reach their targets in the brain or neural retina (; ; ). Consequently, considerable effort has focused on developing approaches that modulate the BBB or BRB in a spatially and temporally controlled manner to improve therapeutic access. Focused ultrasound (FUS), particularly in combination with microbubbles, has emerged as a promising non-invasive strategy for spatially controlled vascular-barrier modulation. FUS-mediated BBB opening has progressed from preclinical studies to early clinical investigations in neurological diseases (; ; ), while experimental studies have demonstrated targeted and reversible BRB opening and explored its use for enhancing retinal delivery (; ).

Human cerebral microvascular endothelial models provide experimentally controlled systems in which selected barrier functions, transport processes, and endothelial responses can be investigated. The hCMEC/D3 cell line retains several endothelial and blood–brain-barrier-associated characteristics, including contact-inhibited monolayer formation, expression of junctional proteins, and functional uptake and efflux transport systems, and it has consequently been used extensively for mechanistic and permeability studies (; ; ). Its practical advantages include human origin, reproducibility, and compatibility with conventional permeable-support assays. Under static monoculture conditions, however, hCMEC/D3 monolayers generally exhibit lower electrical resistance and greater small-molecule paracellular permeability than high-resistance primary or stem-cell-derived barrier models (; ). hCMEC/D3 monocultures should therefore be interpreted as models of selected human neurovascular endothelial properties rather than as complete reconstructions of the in vivo blood–brain barrier.

Although hCMEC/D3 cells are of cerebral rather than retinal endothelial origin, the broader motivation for the present work is the development of experimentally controlled approaches for focused-ultrasound modulation of neurovascular barriers, including the blood–retinal barrier. The inner blood–retinal barrier and the blood–brain barrier share fundamental endothelial barrier features, including continuous nonfenestrated microvascular endothelium, restrictive intercellular junctions, and regulated transcellular transport, while differing in their tissue-specific cellular environment and physiological specialization (; ). These shared endothelial barrier features provide a rationale for investigating selected FUS-associated endothelial transport responses in a cerebral endothelial model, while not implying interchangeability with retinal endothelial models. Accordingly, hCMEC/D3 monolayers were used here as a reductionist human neurovascular endothelial system rather than as a retinal endothelial or complete blood–retinal barrier model; evaluation in retinal endothelial models represents a subsequent step toward assessing relevance to ocular barrier modulation.

Most mechanistic and translational work on FUS-mediated neurovascular barrier modulation has focused on the blood–brain barrier, where the dominant strategy combines pulsed FUS with intravenously administered microbubbles. Acoustically driven microbubble oscillations can transiently increase blood–brain-barrier permeability through endothelial transport and junctional responses that depend on the cavitation regime and exposure level (; ). Acoustic-emissions-based feedback and related monitoring strategies were subsequently developed to improve exposure control and distinguish oscillatory behavior associated with effective opening from broadband emissions associated with higher-risk conditions ().

These advances also demonstrate that a biological response cannot be inferred from frequency or a nominal pressure value alone. It depends on waveform, pulse structure or continuous-wave operation, exposure duration, microbubble presence and activity, biological context, and the local acoustic field. In vitro barrier models are valuable because they can isolate endothelial transport responses from whole-organ variables and resolve early events during and after exposure. Primary porcine brain endothelial monolayers and multicellular permeable-support models have been used to investigate ultrasound–microbubble interactions under controlled in vitro exposure configurations (; ). More recently, used hCMEC/D3 monolayers cultured on permeable supports and showed microbubble-concentration-dependent increases in Lucifer Yellow transport after microbubble-assisted ultrasound, accompanied by decreased transendothelial electrical resistance, localized propidium iodide uptake, and reversible alterations in tight-junction organization. , also using hCMEC/D3 cells, demonstrated enhanced macromolecular transport under ultrasound–microbubble exposure and implicated calcium signaling and transcytotic pathways in the observed response. These studies provide particularly relevant endothelial comparators, but their use of acoustically driven microbubbles differs fundamentally from exposure without added microbubbles. Evidence for endothelial responses under such conditions is more limited. In hCMEC/D3 monolayers, vertically deployed 33-MHz surface acoustic waves elicited intercellular calcium responses and altered barrier measurements under a shear-dominant exposure regime (). Although this platform differs substantially from megahertz FUS in frequency, field geometry, and dominant mechanical conditions, these findings support the broader premise that cerebral endothelial monolayers can respond to acoustic stimulation in the absence of added microbubbles. Short continuous-wave FUS exposures of hCMEC/D3 monolayers without added microbubbles remain comparatively undercharacterized.

A persistent methodological challenge is that standard culture plates and permeable inserts are not acoustically equivalent to an unbounded calibration medium. Millimeter-scale liquid depths, plate and insert walls, porous membranes, liquid–air interfaces, and coupling layers introduce impedance transitions, reflections, and resonant conditions. Experimental and computational studies of in vitro ultrasound systems have demonstrated that vessel geometry can materially alter pressure distributions, promote standing waves and multiple reflections, and permit lateral field spread toward nominally untreated wells (; ; ; ). Consequently, a hydrophone-derived focal pressure measured in a simplified calibration geometry can define an operating condition without establishing the pressure amplitude or spatial distribution across the endothelial monolayer in the complete well-and-insert system. Time-resolved tracer-transport experiments introduce a second analytical requirement: repeated tracer sampling changes compartment volumes and removes tracer, so cumulative transport calculations must distinguish accumulation generated during sonication from accumulation occurring after exposure.

Here, we investigated Lucifer Yellow transport across hCMEC/D3 monolayers cultured on permeable ThinCert inserts during and after a 60-s continuous-wave FUS exposure at 1.2 MHz and 80 V_pp (peak-to-peak voltage), without added microbubbles. The acoustic operating condition was defined by hydrophone characterization in a simplified calibration geometry, while time-resolved, sampling-corrected tracer measurements were used to distinguish transport occurring during exposure from transport occurring afterward. Receiver-compartment Lucifer Yellow concentrations were measured immediately before exposure, at the end of the 60-s exposure, and 9 min after exposure in eight independent paired experiments. We hypothesized that FUS exposure would increase Lucifer Yellow transport during the 60-s exposure interval. Acute live/dead fluorescence imaging was evaluated exploratorily to determine whether the transport response was accompanied by a consistent FUS-associated pattern in the corresponding fluorescence metrics.

2 Materials and methods

2.1 Study design

The study comprised two main cellular experimental series investigating the effects of FUS on Lucifer Yellow transport and acute live/dead fluorescence metrics in hCMEC/D3 human cerebral microvascular endothelial monolayers. In addition, a separate preliminary paired cell-free FUS/control session was performed before the cellular transport series.

The transport series comprised eight independent paired experiments. For each experiment, one FUS-exposed monolayer and one plate-matched unexposed monolayer were prepared from the same independent culture preparation. Together, these matched monolayers constituted one paired experimental unit. Each experiment was performed using newly prepared ThinCert inserts, and each insert was used in one experiment only.

Three unexposed cell-free inserts were additionally included as membrane-transport references. One cell-free insert was measured contemporaneously in well A3 during experiment 1 and served as the reference for experiments 1–3. The second was measured contemporaneously in well A3 during experiment 4 and served as the reference for experiments 4–6. The third was measured contemporaneously in well A3 during experiment 7 and served as the reference for experiments 7–8. These cell-free reference inserts were not considered independent experimental replicates and were not included in the sample size of the paired transport experiments. These unexposed cell-free references were included to provide descriptive membrane-transport context across the experimental blocks; because no paired FUS-exposed cell-free insert was included in the corresponding cellular transport-series sessions, they were not used to estimate a cell-free FUS effect.

The preliminary cell-free FUS/control session was performed on the same experimental day as, and before, the first block of the cellular transport series, comprising experiments 1–3. The session used three FUS-exposed cell-free ThinCert inserts and three paired unexposed cell-free controls. Each FUS-exposed insert was paired with an unexposed cell-free insert from the same session. Because all three pairs were obtained within this single experimental session, they were treated as within-session replicate pairs rather than as three independent experiments. The preliminary cell-free observations were not paired with the independent cellular transport experiments and were not included in the sample size of the cellular transport series.

All cellular inserts belonging to the same transport experiment were placed in a single six-well plate. Well positions were fixed across experiments and were not randomized. The FUS-exposed monolayer was positioned in well A1 and the unexposed cellular control in well A2. A cell-free reference insert was positioned in well A3 only during experiments 1, 4, and 7; otherwise, well A3 remained empty. The remaining wells were left empty. The fixed plate layout and well assignments are shown schematically in Figure 1A. Only the insert in well A1 was coupled to the FUS applicator and intentionally insonified. The inserts in wells A2 and A3 did not undergo applicator placement and were not intentionally insonified. Possible indirect acoustic transmission to non-target wells was not measured. The cellular control is therefore described as a plate-matched unexposed control rather than a sham-exposed control. Each of the three within-session cell-free FUS/control pairs was run on a separate six-well plate, with the FUS-exposed insert positioned in well A1 and its paired unexposed control in well A2, matching the A1–A2 arrangement used for the cellular transport experiments. Within each preliminary cell-free plate, only the A1 insert underwent applicator placement and intentional insonification; possible indirect acoustic transmission to the paired A2 control well was not measured.

FIGURE 1

Lucifer Yellow transport was assessed immediately before FUS exposure, at the end of the 60-s exposure, and 9 min after exposure. Repeated samples were collected from the same insert at these time points and were therefore treated as repeated measurements rather than as independent observations.

The separate viability series comprised five independent paired experiments, denoted V1–V5. Each experiment included one FUS-exposed monolayer and one unexposed monolayer prepared from the same independent culture preparation. The five experiments represented five independent culture preparations; some independent experiments were conducted on the same day. Inserts used for the viability experiments were separate from those used for the Lucifer Yellow transport measurements. Well positions were fixed, with the FUS-exposed insert in well A1 and the unexposed control in well A2. The viability experiments were not randomized, and assignment, exposure, fluorescence-image acquisition, and image-field selection were not blinded.

The paired experiment was defined as the biological and statistical unit. Measurements obtained from individual wells, repeated sampling time points, cells, microscopic fields, pixels, fluorescence channels, or image-derived objects were not treated as independent biological replicates. The independent sample sizes were paired experiments for the transport series and paired experiments for the viability series. The three paired observations from the preliminary cell-free FUS/control session were within-session replicates and therefore did not constitute an additional independent experimental sample size. No prospective sample-size calculation was performed; sample sizes were constrained by available experimental time and materials.

2.2 hCMEC/D3 cell culture

The human cerebral microvascular endothelial cell line hCMEC/D3 () was kindly provided by Pierre-Olivier Couraud (Institut Cochin, Paris, France) and was used at passages 5–30. Cells were maintained in EBM-2 Endothelial Cell Growth Basal Medium-2 (Lonza, CC-3156) supplemented with 5% fetal bovine serum, 1% penicillin–streptomycin, 5 ng/mL recombinant human basic fibroblast growth factor (bFGF/FGF-2; Merck, GF003), 2 ng/mL recombinant human vascular endothelial growth factor (VEGF; Merck, 01-185), 5 ng/mL recombinant human epidermal growth factor (EGF; Merck, GF316), 15 ng/mL recombinant human insulin-like growth factor (IGF; Merck, GF306), 0.90 μg/mL hydrocortisone (Sigma-Aldrich, H4001, Louis, MO, United States), and 51 μg/mL L-ascorbic acid (Sigma-Aldrich, A92902, Louis, MO, United States). The supplemented medium was brought to a final volume of 500 mL.

Cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂. The culture medium was replaced daily during preparation of the experimental monolayers. No documentation of mycoplasma testing was available for the cultures used in this study.

2.3 Preparation of the endothelial insert model

hCMEC/D3 cells were cultured on ThinCert cell-culture inserts designed for six-well plates (Greiner Bio-One, 657641, Frickenhausen, Germany). The inserts contained a transparent polyethylene terephthalate membrane with a nominal pore size of 0.4 µm and a pore density of 2 × 10⁶ pores cm^‒2. The measured inner diameter of the insert was 23.80 mm, corresponding to an available membrane area of approximately 4.449 cm².

Before cell seeding, the membranes of the cellular inserts were coated with growth-factor-reduced Matrigel basement membrane matrix (Corning, 354230, Corning, NY, United States) diluted in Dulbecco’s phosphate-buffered saline without calcium and magnesium (DPBS; Gibco, 14190-094, Grand Island, NY, United States) to a final protein concentration of approximately 0.70 mg/mL. One mL of the diluted Matrigel solution was added per insert and incubated for 1 h at 37 °C. The inserts were then washed once with 1 mL DPBS before cell seeding. All cell-free ThinCert inserts, including the three unexposed membrane-transport references and the six inserts used in the preliminary cell-free FUS/control session, were coated with Matrigel using the same coating protocol as the cellular inserts. The cell-free inserts were not seeded with cells or maintained for the 6-day culture period; for the transport measurements, they underwent the same 15-min pre-exposure Lucifer Yellow incubation as the cellular inserts.

A total of cells were seeded into each insert, corresponding to an initial seeding density of approximately cells/cm². Cells were cultured for 6 days with daily medium changes.

The 6-day culture period was selected based on preliminary optimization of the culture duration. Although continuous cellular coverage was visually apparent at earlier time points, consistently low baseline Lucifer Yellow transport was observed after 6 days, whereas longer culture periods were associated with increased numbers of detached cells.

Before FUS exposure, each insert was examined by phase-contrast microscopy. Inserts were included when the membrane showed continuous cellular coverage without visible cell-free gaps, atypical monolayer morphology, or increased numbers of detached cells. These criteria had been defined before the experiments were performed. No insert was excluded based on these criteria. Transendothelial electrical resistance (TEER) and junctional-protein organization were not measured in the present experiments.

2.4 Fluid geometry of the exposure system

Fluid volumes were selected based on the measured well and insert geometries to maintain matched inner and outer liquid levels and minimize hydrostatic pressure differences across the membrane.

For the transport experiments, the inner compartment initially contained 1.10 mL of solution. Based on the measured inner diameter of 23.80 mm, this volume corresponded to a liquid height of approximately 2.47 mm. Immediately before FUS exposure, a 100-µL sample was removed from each of the inner and outer compartments. The resulting liquid volume within the insert was 1.00 mL, corresponding to a liquid height of approximately 2.25 mm during exposure. This exceeded the minimum liquid height of 2.20 mm required to maintain a continuous fluid-coupling path between the FUS applicator and the insert contents without an intervening air gap.

The outer compartment initially contained 1.91 mL of solution. The immersed insert displaced approximately 1.43 mL. Thus, the combined volume of the outer-compartment liquid and the displaced volume of the insert was 3.34 mL. In the 35.00-mm-diameter well, this corresponded to an initial external liquid height of approximately 3.47 mm. Following removal of the 100-µL baseline sample, the outer compartment contained 1.81 mL, corresponding to an estimated external liquid height of approximately 3.37 mm during FUS exposure when insert displacement was taken into account. At each subsequent sampling time, equal volumes were removed from the inner and outer compartments. The sampled volumes were not replaced. The preliminary cell-free FUS/control session used the same ThinCert insert type and the same initial donor and receiver volumes and fluid-level geometry as the cellular Lucifer Yellow transport experiments.

For the viability experiments, no samples were removed before exposure. During the 60-s FUS exposure, the inner compartment contained 1.00 mL of the regular hCMEC/D3 culture medium described above, and the outer compartment contained 1.81 mL DPBS. These volumes produced the same estimated inner and outer liquid heights during insonification as in the transport experiments, although the solutions used in the two experimental series were not identical.

2.5 FUS applicator and electrical excitation

FUS was generated using a custom-built applicator comprising a spherically curved piezoelectric ring whose concave inner surface was bonded to a polymethyl methacrylate (PMMA) acoustic lens. The ring was custom-manufactured by PI Ceramic GmbH (Lederhose, Germany) from PIC255 piezoceramic material. The applicator was designed to produce an axial acoustic pressure maximum at a distance of 5.0 mm from the lens apex at the intended 1.2-MHz operating frequency. Dimensioned technical drawings of the ring, lens, and complete applicator assembly are provided in Supplementary Figure S1.

A continuous sinusoidal signal at 1.2 MHz was generated using an AFG3022C arbitrary/function generator (Tektronix, Beaverton, OR, United States) and amplified using an NF-HSA4101 power amplifier (Cosinus Messtechnik GmbH, Taufkirchen, Germany). The main output of the amplifier was connected directly to the piezoelectric ring using copper conductors, without an intermediate electrical matching network.

The amplifier monitor output reproduced the main-output waveform at a specified 1:100 voltage ratio and was connected to a Tektronix TDS 2002 oscilloscope using a passive 1:1 voltage probe. The measured monitor voltage was multiplied by 100 according to this ratio. The main-output voltage was set to 100 V_pp during hydrophone characterization and to 80 V_pp during biological exposure.

For all transport experiments, the preliminary cell-free FUS exposures, and the viability experiments, the piezoelectric ring was driven continuously at 1.2 MHz and 80 V_pp for 60 s, without pulse modulation or amplitude ramping.

2.6 Hydrophone characterization

The acoustic output of the FUS applicator was characterized using a PVDF hydrophone RP 39 s (RP acoustics, Germany) with a sensitive-element diameter of 1 mm, a stated bandwidth of 1–3 MHz, and frequency-dependent sensitivity values specified in the calibration certificate over the 0.1–2.0-MHz measurement range. The calibration-certificate sensitivity at 1.2 MHz was 62 mV/MPa. For the frequency-response analysis, the certificate sensitivity corresponding to each measurement frequency was used. The lateral focal width was not measured experimentally. No uncertainty for the hydrophone sensitivity was specified in the calibration information. No external hydrophone preamplifier was used.

The hydrophone signal was recorded using a Tektronix TDS 2002 oscilloscope at an actual sampling rate of 1 GS/s. The applicator was driven continuously for 5 s at each measurement position. No signal filtering was applied.

The PMMA lens was acoustically coupled to the sensitive surface of the hydrophone through a continuous cast volume of Aquasonic 100 Ultrasound Transmission Gel (Parker Laboratories Inc., Fairfield, NJ, United States). No additional membranes, films, or solid interfaces were present in the acoustic propagation path. The sensitive hydrophone surface was oriented normal to the acoustic axis.

The magnitude of the peak rarefactional pressure (PRP) at each measurement position was calculated from the most negative hydrophone voltage relative to the zero-pressure baseline:where is the minimum measured hydrophone voltage, is the zero-pressure baseline, and is the calibration-certificate hydrophone sensitivity at measurement frequency . For measurements at 1.2 MHz, . The maximum PRP identified along the acoustic axis was termed the peak rarefactional focal pressure (PRFP).

To characterize the frequency-dependent pressure response, PRP was measured at an axial distance of 5.0 mm from the lens apex from 0.1 to 2.0 MHz in increments of 0.1 MHz. Four repeated frequency-response measurements were performed. The highest mean PRP across the scanned frequencies occurred at 1.2 MHz. This frequency was selected for biological exposure because it combined the highest mean PRP with a shorter wavelength than lower-frequency alternatives, which was expected a priori to favor greater spatial confinement of the focused acoustic field.

The axial acoustic field was subsequently characterized at 1.2 MHz by translating the hydrophone along the acoustic axis over a distance of 0–7 mm. Measurements were obtained at 0.5-mm intervals using a manually adjustable holder with an approximate positioning resolution of 0.1 mm. Four complete independent axial scans were performed. The axial focal position was defined as the measurement location at which the maximum PRP was recorded.

Hydrophone measurements were acquired at an amplifier output voltage of 100 V_pp. No direct hydrophone measurement was performed at the biological exposure voltage of 80 V_pp. Assuming first-order proportionality between hydrophone-derived pressure amplitude and amplifier output voltage, the pressure values measured at 100 V_pp were scaled by a factor of 0.8 to estimate the corresponding values at 80 V_pp:

This scaling was treated as an engineering approximation rather than as an experimentally established voltage–pressure calibration. First-order proportionality between hydrophone-derived pressure amplitude and amplifier output voltage was assumed over this voltage range. Potential deviations associated with thermal effects, amplitude-dependent resonance behavior, or differences in the actual piezoelectric-terminal voltage were not quantified.

At an axial distance of 5.0 mm from the lens apex, corresponding to the measured axial pressure maximum at 1.2 MHz, the estimated PRFP at 80 V_pp was 0.838 ± 0.054 MPa ( complete independent axial scans; mean ± SD). The reported SD represents between-scan variation and does not include an uncertainty contribution for the hydrophone sensitivity. This exposure setting is therefore referred to as an approximately 0.8-MPa calibration condition.

Only the axial acoustic field was characterized experimentally. The lateral pressure distribution, possible indirect acoustic transmission to non-target wells, and the acoustic field within the complete well-and-insert geometry were not measured experimentally. The reported hydrophone-derived pressure was obtained in the calibration geometry and does not represent a direct pressure measurement at the endothelial monolayer within the complete well-and-insert system.

2.7 Positioning during biological exposure

The FUS applicator was mounted on a custom, manually adjustable three-axis positioning assembly constructed from Thorlabs optomechanical components. The assembly provided an approximate translational resolution of 0.1 mm.

The axial distance between the applicator and the monolayer was initially determined from the measured exposure-system geometry and the measured 5-mm axial pressure maximum at 1.2 MHz. After initial alignment, reproducible axial and lateral positioning was achieved using fixed mechanical reference stops. The six-well plate was positioned against the reference surfaces, and the applicator was advanced to the predefined stop such that the endothelial monolayer was positioned at the nominal 5.0-mm axial location corresponding to the pressure maximum measured in the hydrophone calibration geometry (Figure 1B).

The same mechanical reference configuration was used for the cellular transport experiments, the preliminary cell-free FUS exposures, and the viability experiments.

2.8 Conditions during biological exposure

Experimental plates were removed from the 37 °C incubator and insonified immediately. The actual liquid temperature at the time of exposure and any exposure-associated temperature change were not measured. The solutions were not degassed. Visual inspection before exposure did not identify visible air bubbles in the coupling path, and no visible bubble activity was observed during exposure. Cavitation was not assessed using an acoustic or other instrument-based method. Acoustic streaming was not measured.

2.9 Lucifer Yellow transport assay

Lucifer Yellow CH dipotassium salt (Sigma-Aldrich, L0144) was used as a low-molecular-weight tracer to assess transport across the hCMEC/D3 endothelial monolayers. A 100-µM Lucifer Yellow working solution was prepared in phenol-red-free Hanks’ balanced salt solution (HBSS; Sigma-Aldrich/Merck, H6648) supplemented with 10 mM HEPES (Sigma-Aldrich, H3375). Calibration standards were prepared from the same Lucifer Yellow stock solution and in the same assay matrix as the experimental samples.

At the beginning of the assay, the cell-culture medium was aspirated, and the endothelial monolayers were washed twice, each time with 1 mL of tracer-free assay buffer. Subsequently, 1.10 mL of the Lucifer Yellow working solution was added to the donor compartment above the endothelial monolayer, while the receiver compartment contained 1.91 mL of tracer-free assay buffer. The inserts were incubated for 15 min before initiation of the FUS exposure protocol.

Immediately before FUS exposure, defined as min, a 100-µL sample was removed from each of the donor and receiver compartments. The receiver-compartment sample represented the pre-exposure baseline after 15 min of tracer transport. FUS was then applied continuously for 60 s. A second 100-µL sample was collected from each compartment at min, immediately after the end of exposure. A third 100-µL sample was collected from each compartment at min, corresponding to 9 min after the end of FUS exposure.

The sampled volumes were not replaced. Only the receiver-compartment samples were analyzed fluorometrically. Donor-compartment samples were removed to maintain the predefined fluid-level relationship by applying the same sampling volumes to both compartments but were not analyzed. The preliminary cell-free FUS/control session followed the same Lucifer Yellow protocol as the cellular transport experiments, including the 100-µM donor concentration, 15-min pre-exposure incubation, donor and receiver volumes, 100-µL sampling volumes, sampling at 0, 1, and 10 min, and non-replacement of sampled volumes.

2.10 Fluorescence measurement and calibration

The entire 100-µL receiver sample was transferred to one well of a black, flat-bottom 96-well microplate with a clear bottom (Corning, 3904). The plate lid was removed, and the plate was manually swirled once in a brief horizontal circular motion before fluorescence measurement.

Fluorescence was measured in top-reading mode using a Tecan ULTRA microplate reader operated with XFLUOR4 software, version 4.40. The excitation and emission wavelengths were set to 485 and 530 nm, respectively. The detector gain was held constant across all experiments. FUS-exposed samples and their corresponding unexposed control samples were measured on the same microplate and within the same measurement run. The FUS-exposed and paired unexposed samples from the preliminary cell-free FUS/control session were likewise measured within the same measurement run.

Ten flashes were acquired at each measurement position using an integration time of 40 µs. Four spatially distributed positions within each well were measured in a 2 × 2 pattern and averaged internally by the reader software. Each sample occupied one well and was represented by the resulting internally averaged relative fluorescence unit (RFU) value.

One separate five-point calibration curve was generated for each of three experimental blocks: experiments 1–3, experiments 4–6, and experiments 7–8. Each calibration curve comprised Lucifer Yellow standards at concentrations of 0, 0.1, 0.3, 1, and 3 µM prepared in the same assay matrix as the experimental samples. The preliminary cell-free FUS/control session was performed on the same experimental day as the first experimental block, comprising experiments 1–3, and was quantified using the same five-point calibration curve.

For each experimental block, the relationship between fluorescence and Lucifer Yellow concentration was fitted using linear regression with an unconstrained intercept:where is the intercept, is the slope, and is the Lucifer Yellow concentration. Sample concentrations were calculated by inverse prediction:

2.11 Calculation of transported Lucifer Yellow amounts

Because receiver-compartment samples were repeatedly removed without volume replacement, cumulative receiver amounts were corrected for both the decreasing compartment volume and the amount of Lucifer Yellow removed at earlier sampling time points. With concentrations expressed in µM and volumes expressed in mL, the resulting Lucifer Yellow amounts were expressed in nmol.

Let , , and denote the Lucifer Yellow concentrations in the receiver compartment determined from the corresponding calibration curves at min, min, and min, respectively, and let , , and denote the corresponding cumulative receiver amounts.

The cumulative receiver amount immediately before FUS exposure was calculated as:

The cumulative receiver amount at the end of the 60-s exposure was calculated as:

The cumulative receiver amount at the final measurement time point was calculated as:

The net amount accumulated during the 60-s exposure interval was calculated as:

The net amount accumulated during the subsequent 9-min post-exposure interval was calculated as:

For descriptive cumulative analyses, the net amount accumulated from baseline to the final measurement was calculated as:

The primary endpoint was the paired, control-adjusted difference in Lucifer Yellow accumulation during the exposure interval:

The corresponding paired, control-adjusted difference during the post-exposure interval was treated as a secondary endpoint:

For descriptive cumulative analyses, the paired FUS-minus-control difference from baseline to the final measurement was defined as:

Baseline- and control-adjusted changes in receiver concentration at 1 and 10 min were retained as prespecified sensitivity analyses.

A conventional apparent permeability coefficient, , was not calculated because donor-compartment concentrations were not quantified over time under the repeated-sampling protocol. Interval-specific transported amounts were therefore used to quantify the time-resolved transport response.

2.12 Acute live/dead fluorescence imaging and quantitative image analysis

Acute live/dead imaging was performed in five independent paired experiments, denoted V1–V5. In each experiment, the FUS-exposed and unexposed control monolayers originated from the same independent culture preparation. FUS exposure used the same six-well-plate format, applicator-coupling geometry, mechanical positioning procedure, and acoustic settings as in the transport experiments, with continuous-wave excitation at 1.2 MHz and 80 V_pp for 60 s. During exposure, the inner compartment contained 1.00 mL of regular hCMEC/D3 culture medium and the outer compartment contained 1.81 mL DPBS, producing the same estimated liquid heights as in the transport experiments.

Immediately after exposure, the monolayers were washed once with 1 mL DPBS and subsequently incubated with 1 mL DPBS containing Calcein-AM (MedChemExpress, HY-D0041, Monmouth Junction, NJ, United States) and ethidium homodimer-1 (EthD-1; Sigma-Aldrich, E1903), each at a final concentration of 3 µM. Staining was performed for 30 min at 37 °C protected from light. The inserts were then washed three times with 1 mL DPBS. During the final wash, DPBS was removed only partially so that a small residual volume remained in the insert during image acquisition; this residual volume was not quantified. Fluorescence imaging was initiated approximately 2 min after completion of the final wash.

Fluorescence images were acquired using a ZEISS LSM 710 confocal laser-scanning microscope on an Axio Observer platform controlled with ZEISS ZEN black 2.3 SP1 FP3 (version 14.0.27.201). Calcein-AM and EthD-1 were selected using the predefined dye-selection functionality of the acquisition software and excited using the 488- and 561-nm laser lines, respectively, with fluorescence recorded in separate green and red detection channels. Acquisition settings were held constant between the FUS-exposed and corresponding unexposed control inserts within each paired experiment. One fluorescence field was acquired per insert from the nominal lateral acoustic target region centered on the acoustic axis in the endothelial monolayer plane. A pilot laser provided a visual reference for alignment of the nominal acoustic target region, and the position was verified visually before image acquisition. The images therefore represented the nominal acoustic target region rather than the complete insert area.

Each image set comprised a 488-channel image, a 561-channel image, and a merged image. The separate fluorescence-channel images were saved directly from the ZEISS ZEN software as 512 × 512-pixel, 8-bit RGB JPEG files. The field width used for quantitative calibration was 1,417 μm, corresponding to an effective pixel size of 2.768 µm/pixel and a field area of 2.008 mm². Quantification was performed on the separate fluorescence-channel images; merged images were used for visualization and segmentation quality control.

Image processing was performed using Python 3.13.5 with NumPy 2.3.5, SciPy 1.17.0, scikit-image 0.26.0, and matplotlib 3.10.8. The green component of each 488-channel image and the red component of each 561-channel image were Gaussian-smoothed ( pixels).

To accommodate between-experiment fluorescence-intensity variation, the primary segmentation used one condition-independent threshold per paired experiment, calculated from pixel intensities pooled across the corresponding FUS and control images. Calcein-positive area was segmented using paired pooled Otsu thresholding. Bright EthD-1-positive events were defined as the highest-intensity class from a paired pooled four-class multi-Otsu segmentation. The resulting experiment-specific green thresholds ranged from 84.17 to 116.03, and the red thresholds ranged from 61.65 to 77.19, on an 8-bit intensity scale of 0–255. Connected components with an area of fewer than 3 pixels were excluded, corresponding to an object-area threshold of approximately 23.0 µm².

The four exploratory image outcomes were Calcein-positive area percentage, mean green-channel intensity, bright EthD-1-positive area percentage, and bright EthD-1-positive object density per square millimeter. Mean green-channel intensity was calculated as the mean 8-bit intensity of the unsmoothed green channel across all pixels in the image field and was independent of segmentation thresholds. Object density was calculated as the number of retained bright EthD-1-positive connected components divided by the calibrated image-field area.

Segmentation sensitivity was evaluated using a single global threshold pooled across all ten inserts, with thresholds of 94.13 for the green channel and 70.93 for the red channel on the 0–255 intensity scale. A lower-threshold analysis used a fixed green threshold of 80 and a globally derived three-class multi-Otsu upper red threshold of 50.94. A higher-threshold analysis used fixed green and red thresholds of 120 and 100, respectively.

All images were reviewed for technical usability before quantitative analysis. All images passed this review, and no image was excluded. No deliberately damaged positive-death control was included.

2.13 Statistical analysis

Statistical analyses were performed using Python 3.13.5 with pandas 2.2.3, NumPy 2.3.5, SciPy 1.17.0, and statsmodels 0.14.6. Viability image processing used scikit-image 0.26.0. Quantitative plots were generated using matplotlib 3.10.8.

The independent paired experiment was treated as the biological and statistical unit. FUS-exposed and unexposed control values originating from the same independent culture preparation were analyzed as paired observations. Individual wells, repeated sampling time points, measurements, cells, microscopic fields, pixels, segmented objects, and fluorescence channels were not treated as independent biological replicates.

The primary transport endpoint, , was tested against zero using a two-sided one-sample -test applied to the within-experiment paired differences. The secondary post-exposure endpoint, , was analyzed using the same approach. For both endpoints, means, sample standard deviations, mean paired differences, -based 95% confidence intervals, and Cohen’s , calculated as the mean paired difference divided by the sample standard deviation of the paired differences, were reported. The two -test values were adjusted together using the Holm procedure and constituted the primary interval-specific inferential family. For the descriptive cumulative amount difference , the mean and -based 95% confidence interval were reported without hypothesis testing.

Exact two-sided Wilcoxon signed-rank tests were performed for the same two transport endpoints as nonparametric sensitivity analyses. The two Wilcoxon values were Holm-adjusted within a separate two-test sensitivity-analysis family. These sensitivity results were considered supportive and were not used to define additional confirmatory conclusions.

As an additional prespecified sensitivity analysis, transport effects were evaluated on the concentration scale using baseline- and control-adjusted difference-in-differences estimates at 1 and 10 min. For each time point , the estimate was calculated as:

Each concentration-scale estimate was tested against zero using a two-sided one-sample -test, and the two -test values were Holm-adjusted within this sensitivity-analysis family. For the concentration-scale estimates, mean differences, -based 95% confidence intervals, and Cohen’s were also reported. Exact two-sided Wilcoxon signed-rank tests were calculated as supportive nonparametric sensitivity analyses and were reported without a further multiplicity adjustment.

Receiver-compartment concentrations determined from the corresponding calibration curves at 0, 1, and 10 min were additionally evaluated using a two-factor repeated-measures analysis of variance, with exposure condition and time specified as within-experiment factors. The condition-by-time interaction was considered the relevant omnibus effect. Partial was reported for this interaction. A Greenhouse–Geisser correction was examined as a sensitivity analysis because sphericity could not be assessed reliably given the small number of independent experiments.

Sequential leave-one-experiment-out analyses were performed for the paired transport endpoints to assess whether the estimated effects depended disproportionately on any single experiment.

The three unexposed cell-free reference inserts measured during cellular transport experiments 1, 4, and 7 were summarized descriptively as membrane-transport references for the corresponding experimental blocks. They were not included in the primary inferential analyses. Because no paired FUS-exposed cell-free inserts were included in those cellular transport-series sessions, these reference measurements were not used to estimate a cell-free FUS effect.

The preliminary cell-free FUS/control session comprised three paired observations obtained within a single experimental session. For each pair, receiver concentrations and sampling- and volume-corrected amounts were calculated using the same equations as for the cellular transport series, and the paired FUS-minus-control differences , , and were derived. Because all three pairs originated from one experimental session and therefore did not constitute three independent experiments, the preliminary cell-free data were summarized descriptively only. No inferential statistical tests, confidence intervals, or standardized effect sizes were calculated, and these observations were not included in the primary inferential analyses of the cellular transport series.

For the four viability image endpoints, each endpoint was analyzed using the paired FUS-minus-control differences from the five independent experiments. Because only five independent paired experiments were available, all paired observations and paired differences were displayed. Two-sided exact Wilcoxon signed-rank tests were used for the primary exploratory inference. Means, sample standard deviations, mean paired differences, -based 95% confidence intervals, paired -tests, exact sign tests, exact paired sign-flip randomization tests of the mean, and Hedges’ standardized paired effect sizes were reported as sensitivity summaries. Hedges’ was obtained by applying the small-sample correction to Cohen’s . With five nonzero paired differences, the minimum attainable two-sided exact Wilcoxon value was 0.0625.

Tukey fences, robust modified scores based on the median absolute deviation, and leave-one-experiment-out analyses were used to assess the influence of individual viability experiments. These diagnostics were not used as post hoc exclusion criteria, and no image or experiment was excluded because of its observed outcome.

The viability analyses were considered exploratory. All reported viability values were unadjusted for multiplicity and were interpreted according to the exploratory hierarchy specified above, together with the paired observations, confidence intervals, effect sizes, and sensitivity analyses. No biologically justified equivalence margin was specified prospectively, and no formal equivalence or noninferiority testing was performed. Consequently, nonsignificant viability results were not interpreted as evidence of equivalence, preserved whole-insert viability, biological safety, or absence of cellular injury.

Missing measurements were not imputed. Analyses were based on the available complete paired observations for the corresponding endpoint. Placeholder zero values used in the original transport-series cell-free reference data table to denote measurements that had not been performed were treated as missing values rather than as observed concentrations of zero. Genuine measured zero values, if present, were retained as observations.

All -tests, Wilcoxon signed-rank tests, sign tests, and sign-flip randomization tests were two-sided. Statistical significance for the two interval-specific transport endpoints was defined as a Holm-adjusted . Both unadjusted and adjusted values were reported where applicable. All other analyses were interpreted as supportive, sensitivity, descriptive, or exploratory analyses according to the hierarchy specified above.

3 Results

3.1 Hydrophone characterization identifies the highest mean PRP at 1.2 MHz and an axial pressure maximum at 5.0 mm

Across four repeated frequency-response measurements obtained at an axial distance of 5.0 mm from the lens apex, the highest mean PRP across the scanned frequencies occurred at 1.2 MHz (Figure 2A). At 100 V_pp, the mean PRP at this position and frequency was 0.940 ± 0.088 MPa. Application of the first-order voltage-scaling approximation described in the Methods yielded a corresponding estimate of 0.752 ± 0.070 MPa at the biological exposure amplifier output voltage of 80 V_pp.

FIGURE 2

At 1.2 MHz, all four complete independent axial scans showed their maximum PRP at an axial distance of 5.0 mm from the lens apex (Figure 2B). The mean PRFP at 5.0 mm was 1.048 ± 0.067 MPa at 100 V_pp and was estimated as 0.838 ± 0.054 MPa at 80 V_pp after scaling (Figures 2C,D). The latter value is referred to as the approximately 0.8-MPa calibration condition. These pressure estimates were obtained in the hydrophone calibration geometry and do not represent direct measurements at the endothelial monolayer within the complete well-and-insert system. The lateral pressure distribution and possible indirect acoustic transmission to non-target wells were not determined experimentally.

3.2 FUS exposure is associated with greater Lucifer Yellow accumulation during the 60-s exposure interval

Three separate five-point calibration curves were used to convert receiver-compartment fluorescence into Lucifer Yellow concentrations (Supplementary Figure S2A). The calibration relationships were highly linear across the investigated range of 0–3 μM, with values of 0.99948, 0.99993, and 0.99963, respectively. Calibration residuals are shown in Supplementary Figure S2B, and all Lucifer Yellow sample RFU values fell within the corresponding calibration ranges.

The exposure and sampling sequence is shown in Figure 3A. After the initial 15-min tracer-incubation period and immediately before FUS exposure, the mean receiver-compartment Lucifer Yellow concentration was 0.900 ± 0.102 µM in the inserts assigned to FUS exposure and 0.951 ± 0.085 µM in the paired unexposed controls (Figure 3B). Because each insert had already undergone 15 min of tracer transport before FUS exposure, interval-specific effects were calculated relative to the corresponding pre-exposure value.

FIGURE 3

During the 60-s exposure interval, the calculated Lucifer Yellow amount in the receiver compartment increased more strongly for FUS-exposed monolayers than for their paired unexposed controls (Figure 3C). After correction for the decreasing receiver volume and for tracer removed during baseline sampling, the mean control-adjusted increase was 0.283 nmol (95% confidence interval [CI], 0.185–0.380 nmol; ; unadjusted ; Holm-adjusted ; Cohen’s ). The corresponding mean interval-specific changes were +0.242 nmol in the FUS-exposed inserts and −0.040 nmol in the paired controls. All eight independent experiments showed an effect in the same direction, with individual control-adjusted estimates ranging from 0.082 to 0.410 nmol.

The primary result was supported by the exact Wilcoxon signed-rank sensitivity analysis (unadjusted ; Holm-adjusted ). On the concentration scale, the corresponding baseline- and control-adjusted difference at min () was 0.156 µM (95% CI, 0.102–0.210 µM; ; unadjusted ; Holm-adjusted ; Cohen’s ; exact Wilcoxon ; Supplementary Figure S2C).

Sequential omission of each independent experiment yielded primary-effect estimates between 0.264 and 0.311 nmol, and every leave-one-experiment-out 95% confidence interval remained above zero (Supplementary Figure S2D). The primary effect was therefore not dependent on a single experimental pair.

A repeated-measures analysis of the calibrated receiver concentrations also supported a condition-dependent difference in the temporal profiles (; uncorrected ; partial ). The condition-by-time interaction remained evident in the Greenhouse–Geisser sensitivity analysis (corrected ).

3.3 No statistically detectable continued differential Lucifer Yellow accumulation during the post-exposure interval

During the subsequent interval from 1 to 10 min, corresponding to the 9 min following termination of FUS, the mean additional control-adjusted receiver amount was 0.035 nmol (95% CI, −0.100–0.171 nmol; ; unadjusted ; Holm-adjusted ; Cohen’s ; Figure 3C). The exact Wilcoxon signed-rank sensitivity analysis likewise provided no evidence of a systematic post-exposure difference (unadjusted ; Holm-adjusted ).

The cumulative control-adjusted receiver-amount difference from 0 to 10 min () remained positive, with a mean of 0.318 nmol (95% CI, 0.151–0.485 nmol); seven of eight experiments showed a positive value (Figure 3D). On the concentration scale, the corresponding baseline- and control-adjusted difference at min () was 0.177 µM (95% CI, 0.081–0.273 µM; ; unadjusted and Holm-adjusted ; Cohen’s ; exact Wilcoxon ; Supplementary Figure S2C).

The positive cumulative differences at 10 min predominantly reflected the additional Lucifer Yellow transport that occurred during the 60-s exposure. Between min and min, the mean cumulative receiver-amount difference increased by only 0.035 nmol, and this additional post-exposure difference was not statistically detectable.

The absence of a statistically detectable additional differential accumulation during the post-exposure interval does not establish complete recovery or restoration of barrier function, because no formal equivalence test or later functional measurements were performed.

3.4 Exploratory cell-free insert time series

Three unexposed cell-free inserts were included as exploratory membrane-transport references and were analyzed descriptively. The inserts were measured contemporaneously during cellular transport experiments 1, 4, and 7 and served as references for experimental blocks 1–3, 4–6, and 7–8, respectively. The three cell-free time series are presented in Supplementary Table S1. Their 0–1-min changes in sampling- and volume-corrected receiver amount () were +0.345, +0.445, and −0.141 nmol for the references associated with blocks 1, 2, and 3, respectively. These measurements provided descriptive membrane-transport context across the experimental blocks but did not estimate a cell-free FUS effect because no paired FUS-exposed cell-free insert was included in the corresponding cellular transport-series sessions.

3.5 Preliminary paired cell-free FUS/control observations showed no comparable positive 0–1-min response in a single experimental session

The preliminary cell-free FUS/control session, performed before the cellular transport series, comprised three FUS-exposed cell-free ThinCert inserts and three paired unexposed cell-free controls. Because all three pairs were obtained within a single experimental session, they were treated as within-session replicate pairs and evaluated descriptively without inferential statistical testing (Supplementary Table S3).

During the 0–1-min exposure interval, the paired FUS-minus-control differences in sampling- and volume-corrected Lucifer Yellow accumulation () were −0.026, −0.196, and −0.498 nmol, with a descriptive mean of −0.240 nmol. The corresponding mean interval-specific changes were −0.131 nmol in the FUS-exposed cell-free inserts and +0.109 nmol in the paired cell-free controls. The paired estimates were heterogeneous, spanning from −0.498 to −0.026 nmol, indicating substantial short-interval variability within this single preliminary session. Thus, none of the three preliminary cell-free pairs showed a positive FUS-minus-control difference during the exposure interval. By contrast, all eight independent hCMEC/D3 experiments showed positive values, with a mean of +0.283 nmol.

During the subsequent 1–10-min interval, the paired cell-free values were +0.078, +0.053, and +0.448 nmol (descriptive mean, +0.193 nmol). The corresponding cumulative paired differences from 0 to 10 min were +0.052, −0.143, and −0.050 nmol (descriptive mean, −0.047 nmol). Accordingly, the preliminary cell-free FUS/control session did not show a comparable consistently positive FUS-associated 0–1-min response to that observed in the cellular series, while the cumulative paired differences at 10 min were mixed in direction. Because these observations originated from a single experimental session, they do not provide an independently replicated estimate of the cell-free FUS effect.

3.6 Acute live/dead image metrics show no consistent FUS-associated direction across five independent experiments

Five independent paired viability experiments were analyzed (Figure 4). The primary condition-independent segmentation used a common threshold derived from pooled pixel intensities of the FUS and control images within each experiment, thereby accommodating between-experiment fluorescence-intensity variation without selecting thresholds separately by treatment condition. All merged images are shown in Figure 4A, and the experiment-specific segmentation thresholds are shown in Supplementary Figure S3B.

FIGURE 4

Calcein-positive area percentage was 37.4% ± 6.4% after FUS and 40.3% ± 13.9% in controls (Figure 4B). The paired FUS-minus-control differences were +0.42, +4.65, −16.45, −23.73, and +20.90 percentage points. The mean paired difference was −2.84 percentage points (95% CI, −24.81 to 19.13; exact Wilcoxon ; ).

Mean green-channel intensity was 84.1 ± 13.2 intensity units after FUS and 88.1 ± 28.5 intensity units in controls (Supplementary Figure S3A). The paired differences were −0.97, +9.98, −20.37, −29.43, and +20.70 intensity units, yielding a mean difference of −4.02 intensity units (95% CI, −29.84 to 21.80; ; ).

Bright EthD-1-positive area percentage was 0.335% ± 0.176% after FUS and 0.321% ± 0.076% in controls (Figure 4D). The paired differences were −0.207, +0.299, +0.031, +0.122, and −0.178 percentage points, yielding a mean difference of 0.014 percentage points (95% CI, −0.249 to 0.276; ; ).

Bright EthD-1 object density was 57.8 ± 26.2 mm⁻² after FUS and 56.5 ± 16.6 mm⁻² in controls (Figure 4C). The paired differences were −31.87, +62.25, −4.98, +3.98, and −22.91 mm⁻². The mean paired difference was 1.29 mm⁻² (95% CI, −44.53 to 47.12; ; ).

Discrete high-intensity red-channel events consistent with EthD-1 labeling were detected in all ten 561-channel images (five FUS and five control), demonstrating that high-intensity red events were detectable under the applied acquisition settings. Their presence did not constitute formal validation of assay sensitivity in the absence of a deliberately damaged positive-death control.

The standardized paired mean differences were small: Hedges’ was −0.13 for Calcein-positive area, −0.15 for mean green-channel intensity, 0.05 for bright EthD-1-positive area, and 0.03 for bright EthD-1 object density. One EthD-1 object-density difference exceeded the upper Tukey fence (V2: +62.25 mm⁻², corresponding to +125 objects per field) and had a modified score based on the median absolute deviation of 2.53. No associated technical image failure was identified, and the experiment was retained. No other paired difference among the four image outcomes exceeded the Tukey fences. Leave-one-experiment-out analyses altered the numerical means but did not yield a consistent FUS-associated direction across the image endpoints.

Using a single global threshold and lower- and higher-threshold sensitivity analyses did not change the central interpretation: both positive and negative FUS-control differences remained present for Calcein- and EthD-1-derived metrics (Supplementary Figures S3C,D). Complete exploratory statistical summaries for all four image endpoints are provided in Supplementary Table S2. Thus, the acute live/dead image metrics showed no reproducible FUS-associated direction across the five independent experiments. These findings do not establish equivalence, preserved whole-insert viability, biological safety, or absence of delayed or spatially heterogeneous injury.

4 Discussion

In this study, a 60-s continuous-wave focused-ultrasound exposure was associated with increased Lucifer Yellow transport across the hCMEC/D3 monolayer–insert system during the exposure interval. The effect was consistent across all eight independent paired experiments and was supported by multiple sensitivity and robustness analyses. By contrast, no statistically detectable additional differential accumulation was observed during the subsequent 9-min post-exposure interval. The cumulative control-adjusted receiver-amount difference from 0 to 10 min remained positive, but this predominantly reflected transport generated during the 60-s exposure. In a separate series of five paired experiments, four acute Calcein-AM- and EthD-1-derived image outcomes showed mixed FUS-minus-control differences and no reproducible FUS-associated direction. Taken together, the data support an exposure-associated increase in Lucifer Yellow transport that arose predominantly during the 60-s insonification interval under the conditions of this specific in vitro system, but they do not establish opening of either the blood–brain barrier or blood–retinal barrier in vivo, complete post-exposure recovery, absence of cellular injury, or biological safety.

The temporal analysis is important for interpreting the transport result. The between-condition difference arose predominantly during insonification, whereas no statistically detectable additional differential accumulation was observed during the subsequent 9-min interval. Accordingly, the positive cumulative difference at 10 min should not be interpreted as evidence of continued differential transport throughout the post-exposure period; it primarily reflects tracer accumulated during exposure. Conversely, the absence of a statistically detectable 1–10-min difference does not demonstrate restoration of barrier function. Because only one post-exposure interval was assessed and no formal equivalence or later functional assessment was performed, complete recovery cannot be inferred. The data support the narrower conclusion that continued differential accumulation was not detected during the observed 9-min post-exposure interval.

Lucifer Yellow is a hydrophilic tracer commonly used to assess paracellular permeability and monolayer integrity in hCMEC/D3 models (; ). Its increased receiver accumulation is therefore compatible with increased tracer transport across the hCMEC/D3 monolayer–insert system. However, the present measurements do not identify the underlying route. No junctional proteins, cytoskeletal structures, vesicular transport processes, membrane-poration markers, or intracellular signaling pathways were measured. The result cannot therefore be assigned specifically to tight-junction remodeling, transcytosis, transient membrane permeabilization, or another transport mechanism. In addition, the outcome was the sampling-corrected receiver amount rather than an apparent permeability coefficient. Because donor concentrations were not quantified over time, the assumptions required for a conventional calculation were not met. The amount-based interval analysis was appropriate for distinguishing accumulation during and after exposure, but the reported effect should be understood as a change in tracer transport within the complete monolayer–membrane–fluid system.

The present findings occupy a distinct experimental space within the broader literature on ultrasound-mediated modulation of endothelial barriers. Most mechanistic and translational blood–brain-barrier studies have used pulsed FUS in combination with microbubbles. Electron-microscopy and cell-based studies under those conditions have implicated both paracellular and transcellular responses, including altered junctional organization, increased vesicular transport, calcium signaling, membrane permeabilization, and cytoskeletal changes (; ). Controlled in vitro blood–brain-barrier models have further shown that ultrasound–microbubble interactions can be studied together with barrier measurements, live-cell calcium imaging, and membrane-integrity assays (; ). Of particular relevance to the present study, investigated hCMEC/D3 monolayers cultured on permeable supports and reported microbubble-concentration-dependent increases in Lucifer Yellow permeability after microbubble-assisted ultrasound, accompanied by decreased transendothelial electrical resistance, localized propidium iodide uptake, and reversible alterations in ZO-1 and claudin-5 organization, while complementary cytotoxicity assays showed no detectable cytotoxic effects. This study provides a close biological and assay-level comparator because it likewise combines hCMEC/D3 monolayers with Lucifer Yellow transport, while differing fundamentally in its use of added microbubbles and additional functional, structural, and cytotoxicity measurements. , also using hCMEC/D3 cells, demonstrated enhanced macromolecular transport under ultrasound–microbubble exposure and implicated calcium signaling and transcytotic pathways in the observed response. Collectively, these microbubble-based studies establish biological plausibility for endothelial transport responses, but their mechanisms cannot be transferred directly to the present experiment because the current exposure used continuous-wave excitation and no microbubbles were added.

Evidence that cerebral endothelial monolayers can respond to acoustic stimulation without added microbubbles is more limited. reported intercellular calcium waves and changes in electrical barrier measurements in hCMEC/D3 monolayers exposed to vertically deployed 33-MHz surface acoustic waves. The frequency, field geometry, substrate coupling, and shear-dominant mechanical environment of that platform differ substantially from the 1.2-MHz focused applicator used here. Nevertheless, the study supports the broader premise that hCMEC/D3 monolayers can respond to acoustic stimulation in the absence of added microbubbles. Complementary mechanistic work in a brain-endothelial model showed that acoustically driven microbubbles or integrin-targeted microbeads could evoke calcium responses and increase paracellular transport, with membrane poration, reactive oxygen species, intracellular calcium release, and mechanosensitive channels contributing differently depending on the acoustic mediator (; ). Because both microbubbles and surface-bound particles materially alter local mechanical coupling, that work provides mechanistic hypotheses rather than direct evidence for the mechanism operating under the present exposure without added microbubbles or targeted particles. Together, these studies support the broader biological plausibility of acoustically induced endothelial responses, while emphasizing that the underlying mechanisms depend strongly on the acoustic exposure and the presence or absence of mechanical mediators.

Several nonexclusive processes could account for the transport change observed here. Acoustic radiation force, oscillatory deformation, fluid motion, or local shear could activate mechanosensitive signaling, alter cortical actin tension, or transiently reorganize cell–cell junctions. A contribution from membrane permeabilization or vesicular transport also cannot be excluded. The 60-s continuous-wave exposure additionally raises the possibility of a thermal component. Neither the liquid temperature during exposure nor an exposure-associated temperature rise was measured, and the effect should therefore not be described as nonthermal. Likewise, the absence of added microbubbles is not equivalent to proof that cavitation was absent. The liquids were not degassed, and no passive cavitation detection or other instrument-based cavitation measurement was performed. The absence of visible bubbles or visible bubble activity provides only a macroscopic observation under the applied conditions and does not exclude microscopic gas nuclei or acoustically induced bubble activity. Acoustic streaming was also not quantified. Thus, the present data establish an exposure-associated transport response but do not discriminate among mechanical, thermal, or bubble-related contributions.

Interpretation of the acoustic dose is further constrained by the exposure geometry. Hydrophone characterization showed that the highest mean PRP across the scanned frequencies occurred at 1.2 MHz and that, at this frequency, the axial pressure maximum occurred at a distance of 5.0 mm from the lens apex (Figure 2). The PRFP estimated for the 80-V_pp biological exposure condition was 0.838 ± 0.054 MPa, but this value was obtained by first-order scaling of measurements acquired at 100 V_pp and describes the hydrophone calibration geometry. It was not measured at the endothelial monolayer. Cell-culture wells and permeable inserts can substantially modify ultrasound fields through transmission losses, liquid-column resonances, standing waves, multiple reflections, surface-wave propagation, and lateral spread (; ; ; ). In particular, experimental studies have shown that nominally untreated wells can receive laterally propagated acoustic energy in multiwell plates (). The biological result should therefore be attributed to the complete applicator–coupling–well–insert configuration and the stated electrical operating condition, rather than interpreted as establishing an endothelial response threshold at 0.838 MPa.

This uncertainty also affects interpretation of the paired control. Only the A1 insert was coupled to the applicator and intentionally insonified, whereas the A2 control did not undergo applicator placement and was not sham-exposed. Possible indirect acoustic transmission to the control well was not measured. Any such transmission would complicate the exposure contrast and could attenuate or otherwise alter the observed between-condition difference. Fixed well positions, absence of randomization, and absence of blinding further leave open the possibility of position- or handling-associated effects. The consistency of the primary paired differences and the within-preparation comparison reduce some sources of biological variation, but they do not remove these design limitations. Future experiments should randomize or alternate well positions, include an applicator-coupled sham condition, and directly characterize the acoustic field in both target and control wells.

The cell-free observations provide two complementary but distinct forms of descriptive control information. First, the three unexposed cell-free inserts collected during cellular transport experiments 1, 4, and 7 provided cell-free membrane-transport references across the experimental blocks but, because they lacked paired FUS-exposed cell-free counterparts within those sessions, could not estimate a cell-free FUS effect. Second, the separate preliminary paired cell-free FUS/control session provided a direct comparison in the absence of cells. The preliminary cell-free FUS/control session did not show a comparable consistently positive 0–1-min response to that observed in the cellular series: all three cell-free FUS-minus-control estimates were negative. However, their magnitudes were heterogeneous (−0.026 to −0.498 nmol), indicating substantial within-session variability on the scale of the cellular effect. Considerable short-interval variability was also evident in the unexposed cell-free reference measurements obtained across the experimental blocks. The preliminary cell-free observations therefore provide descriptive evidence that a comparable positive FUS-associated response was not observed in that session, but they do not reliably quantify or exclude non-cellular contributions to the cellular-series effect. All three paired cell-free observations in the present study originated from a single preliminary experimental session and were not paired with the independently replicated cellular experiments. They therefore cannot establish that the cellular-series response was specifically endothelial or support a formal cell-by-FUS interaction analysis. Independent replication of paired cell-free FUS/control experiments would be required to distinguish endothelial, membrane, and fluid-system contributions more rigorously.

The acute live/dead imaging data require similarly bounded interpretation. Across the five independent paired experiments, Calcein-positive area, mean green-channel intensity, bright EthD-1-positive area, and bright EthD-1-positive object density all showed both positive and negative FUS-minus-control differences. The standardized paired mean differences were small, and the qualitative interpretation was stable under pair-specific, global, lower, and higher segmentation thresholds. Together, these analyses showed no reproducible FUS-associated direction in the measured acute live/dead fluorescence metrics.

They also do not establish equivalence or preserved viability. With only five independent paired experiments, exact nonparametric inference had coarse resolution; when all five differences were nonzero, the minimum attainable two-sided exact Wilcoxon value was 0.0625. No prospective equivalence margin was defined, and the wide confidence intervals permit effects that could be biologically relevant. Only one field from the nominal acoustic target region was imaged per insert, so spatially heterogeneous or off-center injury could have been missed. The assay lacked a deliberately damaged positive-death control, and the detection of discrete EthD-1-positive events in all images confirmed only that bright red events were observable, not that assay sensitivity for a defined injury level had been validated. Imaging was acute, with no assessment of delayed apoptosis, metabolic dysfunction, detachment, proliferative capacity, or barrier recovery. Moreover, Lucifer Yellow transport and live/dead fluorescence imaging were examined in separate experimental series, using separate inserts and nonidentical exposure solutions. Thus, the viability data cannot establish that the monolayers exhibiting increased Lucifer Yellow transport would have been free from acute or delayed injury.

The hCMEC/D3 model provides a reproducible human endothelial system suited to paired barrier-response experiments, but its biological scope also limits generalization. Static hCMEC/D3 monocultures commonly exhibit lower electrical resistance and higher paracellular permeability than high-resistance primary or stem-cell-derived models and do not reproduce the complete cellular and hemodynamic environment of a neurovascular barrier (; ; ). In the present experiments, monolayers were screened before exposure using predefined phase-contrast criteria, and the pre-exposure Lucifer Yellow measurement provided an insert-specific baseline for the interval-specific transport analysis. However, TEER and junctional-protein organization were not measured. Baseline electrical barrier tightness was therefore not independently established, and post-exposure restoration of barrier function could not be assessed using an orthogonal functional or structural barrier measure. The baseline-adjusted paired transport analysis should therefore not be interpreted as demonstrating equivalence of baseline barrier properties between the paired monolayers. Pericytes, glial interactions, basement-membrane organization, physiological flow, and tissue-level mechanical constraints may alter both acoustic coupling and endothelial responses. This limitation is particularly relevant to the broader ophthalmic motivation of the present work. Although the inner blood–retinal barrier and the blood–brain barrier share several fundamental microvascular endothelial barrier properties, they differ in tissue-specific endothelial phenotypes and supporting cellular environments (; ). hCMEC/D3 cells are of cerebral origin () and therefore should not be assumed to reproduce the molecular phenotype or retinal microenvironment of the blood–retinal barrier. The present experiments should therefore be interpreted as establishing a controlled neurovascular endothelial test system for FUS-associated tracer transport, rather than as demonstrating blood–retinal barrier opening in vitro. Whether comparable responses occur in retinal microvascular endothelial cells or more physiologically complex neurovascular barrier models requires direct investigation. Accordingly, the observed effect should be regarded as specific to the investigated hCMEC/D3–ThinCert system and exposure configuration and not as evidence that the same exposure would produce comparable transport responses in the human BBB or BRB in vivo.

Several features strengthen the internal interpretation of the transport result. The independent culture preparation, rather than the well, measurement, or fluorescence reading, was treated as the biological unit. FUS and control inserts were paired within each preparation. The primary exposure interval and the post-exposure interval were analyzed separately, and cumulative amounts were corrected for both decreasing receiver volume and tracer removed during earlier sampling. All eight primary paired differences had the same direction, the estimated effect was large relative to between-experiment dispersion, and the conclusion was unchanged by nonparametric analysis, concentration-scale analysis, repeated-measures modeling, and leave-one-experiment-out assessment. The viability analysis similarly avoided treating cells, pixels, objects, or fields as independent replicates, used condition-independent paired thresholding, displayed all experimental pairs, and evaluated segmentation sensitivity. These choices reduce pseudoreplication and ensure that inference reflects the number of independent experiments rather than the much larger number of technical measurements.

Future work should combine improved acoustic dosimetry with mechanistic and longitudinal biological measurements. The complete well-and-insert geometry should be modeled numerically and validated against hydrophone measurements, with particular attention to the lateral field at the membrane plane and possible transmission to adjacent wells. Direct experimental characterization near the monolayer, a measured voltage–pressure response, temperature monitoring, passive cavitation detection, and quantification of acoustic streaming would substantially narrow uncertainty about the delivered stimulus. Experimentally, randomized positions, applicator-coupled sham controls, and independent replication of paired cell-free FUS/control experiments remain priorities. Biological follow-up should include time-resolved electrical resistance, multiple tracer sizes, junctional and cytoskeletal imaging, live calcium measurements, membrane-integrity assays, and measurements of transcytotic pathways. Whole-insert or systematically sampled imaging, validated positive controls, and follow-up over hours to days are needed to assess delayed or spatially heterogeneous injury. Prior rodent studies have demonstrated FUS-mediated BRB modulation and delivery across the BRB (; ), providing an in vivo rationale for evaluating whether FUS-associated endothelial responses can be reproduced in retinal barrier models and assessing their relevance to controlled retinal drug delivery. A particularly important translational next step is therefore replication in human retinal microvascular endothelial cells and, subsequently, multicellular blood–retinal barrier models incorporating retinal pericytes and other relevant retinal cellular components. Comparative studies using retinal and cerebral endothelial models would determine which FUS-associated transport responses reflect shared neurovascular endothelial properties and which are tissue-specific. More physiologically complex primary or induced-pluripotent-stem-cell-derived endothelial models, co-cultures, and flow-based systems could further clarify how barrier tightness and physiological context modify the response.

In conclusion, a 60-s continuous-wave FUS exposure was associated with a consistent increase in Lucifer Yellow transport during the exposure interval in the investigated hCMEC/D3–ThinCert system. No statistically detectable additional differential accumulation was observed during the following 9 min, although this does not demonstrate complete barrier recovery. Acute live/dead image metrics showed no reproducible FUS-associated direction across five independent experiments, but the data do not establish equivalence, preserved whole-insert viability, or safety. The physical mechanisms underlying the transport response remain to be resolved because the local acoustic field at the endothelial monolayer, temperature change, cavitation activity, and acoustic streaming were not directly characterized. The study provides a reproducible functional observation and a framework for distinguishing transport occurring during FUS exposure from subsequent post-exposure transport. Within these limits, the study establishes a controlled in vitro framework for temporally resolving FUS-associated transport in an hCMEC/D3 monolayer–insert system and for subsequent evaluation in retinal endothelial barrier models.

Statements

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement

Ethical approval was not required for the studies on humans in accordance with the local legislation and institutional requirements because only commercially available established cell lines were used.

Author contributions

SB: Investigation, Writing – original draft, Conceptualization, Writing – review and editing, Methodology, Visualization, Formal Analysis. K-ER: Writing – review and editing, Investigation, Conceptualization, Methodology. GM: Writing – review and editing, Methodology. RB: Funding acquisition, Writing – review and editing, Supervision. JB: Writing – review and editing, Supervision, Methodology, Conceptualization, Resources. CR: Funding acquisition, Project administration, Writing – review and editing, Supervision, Conceptualization.

Funding

The author(s) declared that financial support was received for this work and/or its publication. This work was supported by the German Federal Ministry of Research, Technology and Space (BMFTR) under grant 13N16466.

Acknowledgments

The authors thank Karoline Wenzel for valuable technical assistance with cell culture and experimental preparation.

Conflict of interest

GM, RB were employed by Medical Laser Center Lübeck.

The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The author JB declared that they were an editorial board member of Frontiers at the time of submission. This had no impact on the peer review process and the final decision.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The authors declare that ChatGPT using the GPT-5.6 Sol model (OpenAI) was used as a supporting tool for linguistic refinement, formal cleaning of the analysis code, and limited conceptual brainstorming during preparation of the Discussion. All AI-generated suggestions, substantive statements, scientific findings, and analysis results were thoroughly reviewed and verified by the authors, who bear full responsibility for the final manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/facou.2026.1968185/full#supplementary-material

References

• 1

  AbbottN. J.PatabendigeA. A. K.DolmanD. E. M.YusofS. R.BegleyD. J. (2010). Structure and function of the blood–brain barrier. Neurobiol. Dis.37, 13–25. 10.1016/j.nbd.2009.07.030

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  AbrahaoA.MengY.LlinasM.HuangY.HamaniC.MainprizeT.et al (2019). First-in-human trial of blood–brain barrier opening in amyotrophic lateral sclerosis using MR-guided focused ultrasound. Nat. Commun.10, 4373. 10.1038/s41467-019-12426-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  ChenJ.EscoffreJ.-M.RomitoO.IazoureneT.PressetA.RoyM.et al (2024). Enhanced macromolecular substance extravasation through the blood–brain barrier via acoustic bubble-cell interactions. Ultrason. Sonochem.103, 106768. 10.1016/j.ultsonch.2024.106768

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  ConwayG. E.ParanjapeA. N.ChenX.VillanuevaF. S. (2024). Development of an in vitro model to study mechanisms of ultrasound-targeted microbubble cavitation-mediated blood–brain barrier opening. Ultrasound Med. Biol.50, 425–433. 10.1016/j.ultrasmedbio.2023.12.005

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  DanemanR.PratA. (2015). The blood–brain barrier. Cold Spring Harb. Perspect. Biol.7, a020412. 10.1101/cshperspect.a020412

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  Díaz-CoránguezM.RamosC.AntonettiD. A. (2017). The inner blood-retinal barrier: cellular basis and development. Vis. Res.139, 123–137. 10.1016/j.visres.2017.05.009

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  EaK.PinaudM.FouanD.PressetA.TartuJ.-Y.SasakiN.et al (2026). Reversible and microbubble concentration-dependent permeabilization of an in vitro human endothelial barrier to small molecules using ultrasound: implications for neurodegenerative diseases therapy. Mol. Pharm.23, 3710–3721. 10.1021/acs.molpharmaceut.6c00168

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  HelmsH. C.AbbottN. J.BurekM.CecchelliR.CouraudP.-O.DeliM. A.et al (2016). In vitro models of the blood–brain barrier: an overview of commonly used brain endothelial cell culture models and guidelines for their use.J. Cereb. Blood Flow. Metab.36, 862–890. 10.1177/0271678X16630991

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  HenselK.MienkinaM. P.SchmitzG. (2011). Analysis of ultrasound fields in cell culture wells for in vitro ultrasound therapy experiments. Ultrasound Med. Biol.37, 2105–2115. 10.1016/j.ultrasmedbio.2011.09.007

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  HinkelS.MatternK.DietzelA.ReichlS.Müller-GoymannC. C. (2019). Parametric investigation of static and dynamic cell culture conditions and their impact on hCMEC/D3 barrier properties. Int. J. Pharm.566, 434–444. 10.1016/j.ijpharm.2019.05.074

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  HsiaoM.-Y.LiaoD.XiangG.ZhongP. (2023). Intercellular calcium waves and permeability change induced by vertically deployed surface acoustic waves in a human cerebral microvascular endothelial cell line (hCMEC/D3) monolayer. Ultrasound Med. Biol.49, 1153–1163. 10.1016/j.ultrasmedbio.2022.12.019

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  HynynenK.McDannoldN.VykhodtsevaN.JoleszF. A. (2001). Noninvasive MR imaging-guided focal opening of the blood–brain barrier in rabbits. Radiology220, 640–646. 10.1148/radiol.2202001804

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  JuffermansL. J. M.van DijkA.JongenelenC. A. M.DrukarchB.ReijerkerkA.de VriesH. E.et al (2009). Ultrasound and microbubble-induced intra- and intercellular bioeffects in primary endothelial cells. Ultrasound Med. Biol.35, 1917–1927. 10.1016/j.ultrasmedbio.2009.06.1091

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  LeluS.AfadziM.BergS.ÅslundA. K. O.TorpS. H.SattlerW.et al (2017). Primary porcine brain endothelial cells as in vitro model to study effects of ultrasound and microbubbles on blood–brain barrier function. IEEE Trans. Ultrason. Ferroelectr. Freq. Control64, 281–290. 10.1109/TUFFC.2016.2597004

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  LeskinenJ. J.HynynenK. (2012). Study of factors affecting the magnitude and nature of ultrasound exposure with in vitro set-ups. Ultrasound Med. Biol.38, 777–794. 10.1016/j.ultrasmedbio.2012.01.019

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  LinJ.QiaoC.JiangH.LiuZ.HuY.LiuW.et al (2025a). Reversible Ca²⁺ signaling and enhanced paracellular transport in endothelial monolayer induced by acoustic bubbles and targeted microbeads. Ultrason. Sonochem.112, 107181. 10.1016/j.ultsonch.2024.107181

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  LinJ.QiaoC.JiangH.LiuZ.HuY.LiuW.et al (2025b). Corrigendum to “Reversible Ca2+ signaling and enhanced paracellular transport in endothelial monolayer induced by acoustic bubbles and targeted microbeads” [Ultrason. Sonochemistry 112 (2025) 107181]. Ultrason. Sonochem.119, 107366. 10.1016/j.ultsonch.2025.107366

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  LipsmanN.MengY.BethuneA. J.HuangY.LamB.MasellisM.et al (2018). Blood–brain barrier opening in Alzheimer’s disease using MR-guided focused ultrasound. Nat. Commun.9, 2336. 10.1038/s41467-018-04529-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  O’ReillyM. A.HynynenK. (2012). Blood–brain barrier: real-time feedback-controlled focused ultrasound disruption by using an acoustic emissions-based controller. Radiology263, 96–106. 10.1148/radiol.11111417

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  ParkJ.ZhangY.VykhodtsevaN.AkulaJ. D.McDannoldN. J. (2012). Targeted and reversible blood-retinal barrier disruption via focused ultrasound and microbubbles. PLoS One7, e42754. 10.1371/journal.pone.0042754

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  ParkY.ShinJ.ParkJ.KimS.ParkJ. H.KimJ.et al (2024). Focused ultrasound as a novel non-invasive method for the delivery of gold nanoparticles to retinal ganglion cells. Transl. Vis. Sci. Technol.13, 5. 10.1167/tvst.13.5.5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  PatelU. S.GhorayebS. R.YamashitaY.AtandaF.WalmsleyA. D.SchevenB. A. (2015). Ultrasound field characterization and bioeffects in multiwell culture plates. J. Ther. Ultrasound3, 8. 10.1186/s40349-015-0028-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  PollerB.GutmannH.KrähenbühlS.WekslerB.RomeroI.CouraudP.-O.et al (2008). The human brain endothelial cell line hCMEC/D3 as a human blood–brain barrier model for drug transport studies. J. Neurochem.107, 1358–1368. 10.1111/j.1471-4159.2008.05730.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  RamsayE.LajunenT.BhattacharyaM.ReinisaloM.RillaK.KidronH.et al (2023). Selective drug delivery to the retinal cells: biological barriers and avenues. J. Control. Release361, 1–19. 10.1016/j.jconrel.2023.07.028

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  SecomskiW.BilminK.KujawskaT.NowickiA.GriebP.LewinP. A. (2017). In vitro ultrasound experiments: standing wave and multiple reflections influence on the outcome. Ultrasonics77, 203–213. 10.1016/j.ultras.2017.02.008

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  SheikovN.McDannoldN.VykhodtsevaN.JoleszF.HynynenK. (2004). Cellular mechanisms of the blood–brain barrier opening induced by ultrasound in presence of microbubbles. Ultrasound Med. Biol.30, 979–989. 10.1016/j.ultrasmedbio.2004.04.010

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  SweeneyM. D.ZhaoZ.MontagneA.NelsonA. R.ZlokovicB. V. (2019). Blood–brain barrier: from physiology to disease and back. Physiol. Rev.99, 21–78. 10.1152/physrev.00050.2017

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  WekslerB. B.SubileauE. A.PerrièreN.CharneauP.HollowayK.LevequeM.et al (2005). Blood–brain barrier-specific properties of a human adult brain endothelial cell line. FASEB J.19, 1872–1874. 10.1096/fj.04-3458fje

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  WekslerB.RomeroI. A.CouraudP.-O. (2013). The hCMEC/D3 cell line as a model of the human blood brain barrier. Fluids Barriers CNS10, 16. 10.1186/2045-8118-10-16

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  ZhaoW.HanL.BaeY.ManickamD. S. (2019). Lucifer Yellow—A robust paracellular permeability marker in a cell model of the human blood–brain barrier. J. Vis. Exp.150, e58900. 10.3791/58900

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

Keywords

acoustic bioeffects, continuous-wave ultrasound, endothelial barrier, focused ultrasound, hCMEC/D3, Lucifer yellow, endothelial transport

Citation

Bleker S, Redeker K-EM, Meußler G, Brinkmann R, Brockmöller J and Russmann C (2026) Focused ultrasound increases Lucifer Yellow transport across hCMEC/D3 endothelial monolayers. Front. Acoust. 4:1968185. doi: 10.3389/facou.2026.1968185

Received

14 August 2026

Revised

11 September 2026

Accepted

18 September 2026

Published

02 October 2026

Volume

4 - 2026

Updates

Copyright

© 2026 Bleker, Redeker, Meußler, Brinkmann, Brockmöller and Russmann.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Christoph Russmann, [email protected]

Disclaimer

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.