2.1.

AOW Fabrication

Surface of glass coverslip was cleansed using a series of solutions of acetone (67-64-1, Samchun Pure Chemical), degenerative alcohol (64-17-5, Samchun Pure Chemical), 1% water dissolved detergent (1104-1, Alconox), and deionized water. After masking the glass coverslip with a Kapton masking tape (CT-10, iNexus) excluding the central region ($\sim 3\times 3{\mathrm{mm}}^2$), it was cleansed with an oxygen plasma cleaner (PDC-002-HP, Harrick Plasma) for 20 min at 20 sccm as shown in Fig. 1(a). An indium–tin–oxide (ITO) electrode (${\mathrm{In}}_2{\mathrm{O}}_2:{\mathrm{Sn}}_2{\mathrm{O}}_2$, 9:1 wt/wt) was loaded onto the glass coverslip using a DC magnetron sputtering system (Flexlab-50, A-Tech System) to form an optically transparent bottom electrode. The ITO-coated coverslip was heated up to 300°C on a hot plate for 1 h to increase the transparency and conductivity.¹⁷ Because the cable was soldered on the other side of the transducer, the coverslip was flipped, and two gold island electrodes for the cable connection were formed by sputtering a gold electrode extended to the edge of the coverslip to make connections to the ITO electrode. The polyvinylidene fluoride-trifluoroethylene (PVDF-TrFE) solution was prepared by dissolving 10 g of PVDF-TrFE powder (FC-20, Piezotech) in 15 g dimethylacetamide (271012, Sigma-Aldrich) and 36 g methyl ethyl ketone (63140-1230, Junsei). The solution was spin-coated on the ITO-loaded coverslip at 500 rpm for 30 s to obtain a uniform layer of $16\mu \mathrm{m}$ thickness. The sample was cured in a dry oven at 70°C for 30 min and further baked at 135°C for 2 h. The top electrode was also formed on the PVDF-TrFE layer by sputtering ITO, followed by masking the patterned electrode and roughening the surface with oxygen plasma to enhance the adhesion of ITO on the polymeric surface. A $38\text{-}\mu \mathrm{m}$-thick coaxial cable (38 AWG micro coax cable, Samtec) was bonded on the prepared gold electrode with the soldering temperature of 280°C. To obtain the piezoelectric characteristics of PVDF-TrFE, the application of a high DC voltage of $100\mathrm{MV}/\mathrm{m}$ along the transducer, called poling process, was performed in a dry oven at 105°C for 20 min. Subsequently, a layer of parylene (Parylene-N, Suzhou Chireach Biomedical Technology Co., Ltd.) was created on the transducer using a parylene coater (OBT-PC200, Obang Technology) to form an acoustic matching and waterproof layer. The AOW was connected to the Micro-Miniature Coaxial (MMCX) connector to prevent gnawing of the cable by the mouse. The cable length was maintained as short as 5 cm, and an MMCX connector (C-RMMF. WST0.1, Ace RF Comm) with a small form factor was selected [Figs. 1(b) and 1(c)].

Fig. 1

The AOW. (a) Schematic layout of the AOW. The AOW is fabricated on a circular glass coverslip with a diameter of 5 mm. The squared active area contains a piezoelectric material (PVDF-TrFE) and ITO electrodes for ultrasound generation. (b) Photograph of the fabricated AOW. (c) A step-by-step process for fabricating AOW. (d) Optical transmittance of the AOW active area. The visible spectral region (400 to 700 nm) was used for collecting fluorescence emission. The near-infrared spectral region (800 to 1000 nm) was used for two-photon excitation of a femtosecond laser.

2.2.

Measurement of Acoustic Performance of the AOW

The AOW acoustic performance was characterized by measuring its pulse-echo response, surface pressure distribution, and acoustic beam profile. The pulse-echo test was conducted using a pulser/receiver (UT 340, UTEX Scientific) with an electrical transmit pulse of 1 ns of $100{\mathrm{V}}_{\mathrm{pp}}$ and a gain of 20 dB. During the test, the transducer was placed at 12 mm above a flat crystal reflector with its natural focus.

The acoustic beam profile of the AOW was measured using a needle hydrophone (NH1000, Precision Acoustics) in a tank filled with deionized water. The needle hydrophone was installed in a custom-made motorized three-axes stage. The beam profile of the AOW was scanned with a resolution of spatial translation along the axial and lateral directions of 100 and $200\mu \mathrm{m}$. The similarity of the profile at this depth was confirmed by numerical simulation using the finite element method with the simulation tool (PZFlex 1.25.3.0, Onscale).

2.3.

Mice

All mice were housed with littermates in groups of two to five in a reverse day/night cycle and provided ad libitum access to food and water. Male or female C57BL6J wild-type mice aged 8 to 16 weeks (Orient Bio) were used to image the cortical vasculature. Male or female Gad2-GCaMP6-tdTomato mice aged 8 to 16 weeks were obtained by crossing Gad2-IRES-Cre (010802, The Jackson Laboratory) and CAG-floxed-GCaMP6f-tdTomato (031968, The Jackson Laboratory). All animal experiments were performed in compliance with institutional guidelines and were approved by the Subcommittee on Research Animal Care of Seoul National University.

2.4.

Surgical Implantation of the AOW

The AOW was implanted over the somatosensory cortex following the procedures previously described for the cranial window model.¹⁸ A mouse received an intraperitoneal injection of dexamethasone ($1\mathrm{mg}/\mathrm{kg}$) 2 h before surgery. After anesthetizing the mouse with isoflurane (induction: 5%, maintenance: 1% to 1.5%, ${\mathrm{O}}_2$ at $1\mathrm{L}/\min$), the mouse was affixed to a stereotaxic frame. The body temperature was maintained at 37°C using a homeothermic blanket. After removing the scalp and periosteum, the cranium over the somatosensory area was removed using a dental drill (diameter $\sim 4\mathrm{mm}$) and the AOW was attached to the exposed cortex with the side of the ultrasound transducer facing the cortex. A customized metal frame was glued around the implanted AOW using a dental cement [Fig. 3(c)]. After the frame was loaded, the MMCX and coaxial cable were firmly fixed to the back of the mouse neck using dermal tape (1527-1, 3M) to relieve the weight of the MMCX connector (0.25 g). The mouse had an acute recovery time of 2 h without receiving anesthetic gas.¹⁹

2.5.

In Vivo Imaging

The imaging procedure was initiated 2 h after AOW implantation under isoflurane anesthesia. For the vascular imaging, a wild-type mouse received retro-orbital injection of tetramethylrhodamine isothiocyanate-dextran (2.5% w/v, $100\mu \mathrm{L}$ in saline; 52194, Sigma-Aldrich). For the imaging of ultrasound-evoked neuromodulation, the AOW implanted on a Gad2-GCaMP6-tdTomato mouse was hooked up to the control board²⁰ including a function generator and a power supply [Fig. 3(a)]. The ultrasound parameters used were as follows: frequency, 10 MHz; duty cycle, 50%; acoustic pressure, 0.035 MPa; and stimulation duration, 1 to 20 s. Imaging was performed on a galvanomirror-based multiphoton microscope (Bergamo II, Thorlabs) coupled to a 920-nm femtosecond fiber laser (FemtoFiber Ultra 920, Toptica Photonics). The imaging power was maintained at $<60\mathrm{mW}$ at the objective back aperture. For volumetric imaging, an image stack of $413\times 413\times 200\mu {\mathrm{m}}^3$ was acquired with $1024\times 1024\text{pixels}$, at an axial interval of $1.5\mu \mathrm{m}$. For functional imaging, time-series imaging of a $413\times 413\mu {\mathrm{m}}^2$ area was acquired with $256\times 256\text{pixels}$ at $\sim 4\mathrm{Hz}$.

2.6.

Data Analyses

Transmittance of the AOW was acquired using a fiber-coupled spectrometer (CCS200, Thorlabs) and high-power LED (SOLIS-3C, Thorlabs). Two-photon microscopy images were processed using ImageJ FIJI (NIH) and MATLAB (MathWorks). Image processing included adjustment of brightness and contrast, averaged projection of $z$-stacked images, and time-series analysis of neuronal calcium activity. The calcium activity of $F$ was calculated by dividing the GCaMP6 fluorescence by the tdTomato fluorescence signal. To calculate $\mathrm{\Delta }F$, the resting calcium activity during the 10 s prior to stimulation was averaged, and the averaged value was subtracted from the time-series fluorescent signal. The ultimate calcium signal level change of $\mathrm{\Delta }F/F$ was calculated by dividing $\mathrm{\Delta }F$ by $F$. The area under the curve was quantified by integrating the positive area of $\mathrm{\Delta }F/F$ for 40 s after the start of stimulation, because all the excited calcium signals returned to baseline before the period.

2.7.

Statistical Analysis

Statistical and regression analyses were performed using the Prism software (GraphPad). Data are expressed as mean ± Standard Error of the Mean (SEM), unless otherwise indicated. A one-way ANOVA (Dunnett’s multiple comparison test) is shown in Fig. 4(c). Statistical significance was set at $p<0.05$.

3.1.

Fabrication of the AOW

To fabricate an AOW capable of being implanted on mouse cranium, we formed transparent acoustic stacks on a circular glass coverslip with a diameter of $\sim 5\mathrm{mm}$ [Fig. 1(a), refer to Sec. 2.1 for detailed procedures]. We chose PVDF-TrFE^21–23 as the piezoelectric material because it showed sufficient piezoelectricity ($d_{33}=25\mathrm{pC}/\mathrm{N}$²¹) to generate the 10-MHz signal to stimulate brain cells while being thin enough to achieve decent optical transparency to determine neuronal activities in vivo. Considering the small form factor of the glass coverslip, we designed the transducer with an area of $3\times 3{\mathrm{mm}}^2$. To obtain both optical transparency and acoustic output, we set the thickness of the PVDF-TrFE to $16\mu \mathrm{m}$ thick layer of parylene, which provided the best impedance matching the operational frequency (acoustic impedance of 2.5 MRayl²⁴) and waterproof characteristics.

As shown in Fig. 1(b), the active area of the AOW appeared to be translucent with a faint yellowish color. To quantitatively investigate the optical transparency, we measured the optical transmittance using a spectrometer. Transmittance was $\sim 40\%$ at 500 to 1000 nm [Fig. 1(d)]. In two-photon fluorescence imaging, an estimated 60% of attenuation for excitation (920 nm) and emission (546 nm) was compensated by increasing excitation power by approximately fourfold (2.5-fold for compensating two-photon excitation loss and additional 1.6-fold for compensating emission loss).

3.2.

Acoustic Characterization of the AOW

To characterize the acoustic performance of AOW, we first simulated the pulse-echo impulse response [center frequency of 24 MHz with $-6\mathrm{dB}$ bandwidth of 35.5%; Fig. 2(a)] in glass coverslips with a mean thickness of $170\mu \mathrm{m}$ coverslip detected a pulse-echo signal in both the time and frequency domains [center frequency and bandwidth were 24 MHz and 40%, respectively; Fig. 2(a)]. As predicted by the simulation results, a hump in the frequency domain at 17 MHz was also observed in the measured data. In the pulse-echo measurement, the lowest $-6\mathrm{dB}$ frequency was $\sim 10\mathrm{MHz}$, which might be the lowest frequency used for stimulating the brain with the proposed device.

Fig. 2

Acoustic characterization of the AOW. (a) The left panel displays the simulated pulse-echo impulse response result in the time and frequency domains in the $170\text{-}\mu \mathrm{m}$-thick glass coverslip. The right panel displays the measured pulse-echo impulse response result. (b) Measured impedance graph of the AOW. (c) Measured ultrasound beam profiles and pressure line plots in lateral ($XY$) and axial ($XZ$) views. Simulated pressure line plot in axial view ($XZ$) was visualized with a red solid line positioned $100\mu \mathrm{m}$ from the surface. The scale bar in $XY$ and $XZ$ is 1 mm. (d) Simulated ultrasound beam profile and averaged pressure in regions $A$ and $B$ in the lateral view with a 2-mm depth and 0.1-mm depth from the surface of AOW. Scale bar, 1 mm.

As shown in the impedance curve in Fig. 2(b), there were two resonance peaks in the magnitude curve with corresponding two peak frequencies at 14.9 and 23.5 MHz. Although the impedance of the device at 10 MHz was $750\mathrm{\Omega }$, which could cause impedance mismatch with the amplifier, the pressure generated from the device at 10 MHz was high enough to stimulate mouse brain.

Because the use of higher frequencies ($>1\mathrm{MHz}$) could limit ultrasonic wave penetration, we measured the acoustic beam profile of the AOW by scanning the acoustic field using a needle hydrophone installed in a three-axes motorized stage. We measured the pressure with three cycles of 10 MHz with $60{\mathrm{V}}_{\mathrm{pp}}$ from 2 to 22 mm axially and laterally [Fig. 2(c)]. The maximum pressure in axial direction was 0.11 MPa at a natural focus depth of 12 mm and measured average pressure in lateral direction at 2 mm was $0.09\pm 0.0009\mathrm{MPa}$ (standard deviation), corresponding to 1% of the mean value, which might be negligible changes for delivering ultrasound into tissue. Considering the optical penetration depth ($<2\mathrm{mm}$), we measured the pressure distribution within 2 mm. However, 2 mm was the closest approachable distance to the hydrophone. To overcome this limitation, we performed the PZFlex simulation of the pressure at 0.1 and 2 mm to confirm that there was a negligible difference in pressure between these pressures [Fig. 2(d)]. According to the numerical simulation shown in Fig. 2(d), we inferred that the pressure within an optical accessible depth might be $\sim 0.078\mathrm{MPa}$ at 0.1 mm, which was close to the pressure level of 0.075 MPa at 2 mm.

3.3.

AOW-Based Cranial Window Model

For in vivo applications, we implanted the AOW in the mouse brain following a well-established open cranial window preparation.¹⁸ To generate ultrasound during in vivo imaging, a customized ultrasound transmitter²⁰ was used to generate patterned stimulation [Figs. 3(a) and 3(b)] with the trigger signal generated by an arbitrary waveform generator (AFG3252, Tektronix). The overall schematic of AOW implantation is shown in Fig. 3(a). After implantation, the cortical surfaces of the pial blood vessels were clearly visible [Fig. 3(c)]. To evaluate the compatibility of the AOW with two-photon microscopic imaging, we imaged the cortical vasculature from the pial surface to $200\mu \mathrm{m}$ in vivo using Fluorescein Isothiocyanate (FITC)-dextran [Fig. 3(d)]. Owing to the light attenuation of AOW [Fig. 1(d)], the input light power was increased to 60 mW, which had little possibility of a photothermal effect.²⁵

Fig. 3

AOW-based cranial window model. (a) Schematic illustration of the overall setup. The AOW is implanted after open craniotomy and connected to the printed circuit board (PCB) driver board fed by a function generator and a DC power supply. The PCB driver board amplifies the input trigger signal by approximately twofold. (b) A schematic illustration of the ultrasound parameter used in the experiment. Frequency, PRF, duty cycle, and pulse length are 10 MHz, 1.5 kHz, 50%, and 1 to 20 s, respectively. (c) Illurstration presenting the structural configurations of AOW-based cranial window implant in vivo (left), and a bright field image of the cortical surface visualized through the implanted AOW transducer (right). Note that the scale bars in the left and right figures indicate 10 and 0.5 mm, respectively. (d) A representative $z$-stacked images of cortical blood vessels up to $200\mu \mathrm{m}$ from the pia. Scale bar, $40\mu \mathrm{m}$.

3.4.

Microscopic Observation of Ultrasound-Mediated Neuromodulation In Vivo

Finally, we tested whether the AOW allows the microscopic observation of ultrasound-mediated neuromodulation. We used a Gad2-GCaMP6-tdTomato mouse line to measure functional calcium activity in a subset of inhibitory interneurons. Two-photon imaging through AOW provided near-artifact-free microscopic imaging of subcellular neuronal structures in Gad2-GCaMP6-tdTomato mice [Fig. 4(a)].

Fig. 4

Microscopic observation of ultrasound-mediated neuromodulation using the AOW-based cranial window model. (a) A representative two-photon fluorescence image on the somatosensory cortex of a Gad2-GCaMP6-tdTomato mouse (left) and a pseudocolored image of the change in GCaMP6 fluorescence ($\mathrm{\Delta }F$) by the ultrasound stimulation (right). Scale bars, $15\mu \mathrm{m}$. (b) GCaMP6-based neuronal ${\mathrm{Ca}}^{2+}$ traces represented as $\mathrm{\Delta }F/F$ ($n=25$ neurons). Ultrasound stimulation was exerted through the AOW during the blue shaded region (input voltage: 20 V). (c) The integrated neuronal activation in response to varying stimulation periods from 1 to 20 s ($n=34$, 17, 84, and 48 neurons for 1, 5, 10, and 20 s, respectively; green dot: mean, whisker: SEM, acquired from three mice). The dotted curve is fitted to the sigmoidal curve ($R^2=0.11$). ***, $p<0.0001$ (one-way ANOVA). (d) Average ${\mathrm{Ca}}^{2+}$ signal time course of PV neurons in response to varying stimulation periods in (c) (green line: mean, gray shade: SEM, acquired from three mice).

The relatively bright tdTomato signal helped us to confirm the minimal influence of movement artifacts from ultrasonic waves or physiologic motions. Upon ultrasound stimulation at an operation frequency of 10 MHz, a subset of fluorescent interneurons showed a reliable increase in calcium signal with the 10-s stimulation [Fig. 4(b)]. To further test the effect of stimulus duration on neuronal activity, we varied the stimulation period from 1 to 20 s and measured integrated neuronal activity of all neurons detected in different fields of view from three mice [Figs. 4(c) and 4(d)]. With increasing stimulation period, the integrated neural response increased in a sigmoidal relationship. Statistically significant activity was observed for stimulation periods longer than 10 s. During the increasing stimulation period, no significant temperature rise ($<0.1°\mathrm{C}$) was measured by the thermometer with tissue phantom at the focus region. These results suggest that the AOW can be used to screen the influence of ultrasonic parameters on neuronal ensemble activity.