of 15
Article
https://doi.org/10.1038/s41467-024-48748-6
An implantable piezoelectric ultrasound
stimulator (ImPULS) for deep brain
activation
Jason F. Hou
1,7
, Md Osman Goni Nayeem
1,7
,KianA.Caplan
2
,EvanA.Ruesch
3
,
Albit Caban-Murillo
3
, Ernesto Criado-Hidalgo
4
, Sarah B. Ornellas
1
,
Brandon Williams
5
, Ayeilla A. Pearce
2
, Huseyin E. Dagdeviren
6
,
Michelle Surets
3
, John A. White
5
, Mikhail G. Shapiro
4
, Fan Wang
2
,
Steve Ramirez
3
& Canan Dagdeviren
1
Precise neurostimulation can revolutionize therapies for neurological dis-
orders. Electrode-based st
imulation devices face chal
lenges in achieving pre-
cise and consistent targeting due to t
he immune response and the limited
penetration of electrical
fi
elds. Ultrasound can aid i
n energy propagation, but
transcranial ultrasound stimulatio
n in the deep brain has limited spatial
resolution caused by bone and tissue s
cattering. Here, we report an implan-
table piezoelectric ultras
ound stimulator (ImPULS) that generates an ultra-
sonic focal pressure of 100 kPa to modu
late the activity of neurons. ImPULS is
a fully-encapsulated,
fl
exible piezoelectric micro
machined ultrasound trans-
ducer that incorporates a biocompati
ble piezoceramic, potassium sodium
niobate [(K,Na)NbO
3
]. The absence of electroche
mically active elements poses
a new strategy for achieving long-ter
m stability. We demonstrated that
ImPULS can i) excite neurons in a mouse
hippocampal slice ex vivo, ii) activate
cells in the hippocampus of an anesthetized mouse to induce expression of
activity-dependent gene c-Fos, and iii)
stimulate dopaminergic neurons in the
substantia nigra pars compacta to elicit time-locked modulation of nigros-
triatal dopamine release. This work introduces a non-genetic ultrasound
platform for spatially-localized neur
al stimulation and exploration of basic
functions in the deep brain.
Precise and reversible spatiotemporal control of neural activity is the
ultimate goal of neurostimulation strategies both for therapeutic
applications and neuroscience research. Current neurostimulation
strategies can be broadly divided into two categories: (i) non-invasive
and (ii) invasive. Some existing non-invasive methods used in clinical
treatment are transcranial magnetic stimulation (TMS)
1
, transcranial
current stimulation (TCS)
2
, and transcranial-focused ultrasound
(tFUS)
3
. While these methods can avoid surgery and associated
recurrent risks
4
, TMS and TCS encounter scattering of electromagnetic
energy through bone and tissue attenuation
5
,
6
. Ultrasound is a
Received: 27 September 2023
Accepted: 13 May 2024
Check for updates
1
Media Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
2
Department of Brain and Cognitive Sciences, McGovern Institute for Brain
Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
3
Department of Psychological and Brain Sciences, The Center for Systems
Neuroscience, Boston University, Boston 02215 MA, USA.
4
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, CA
91125, USA.
5
Center for Systems Neuroscience, Neurophotonics Center, Department of Biomedical Engineering, Boston University, 610 Commonwealth Ave.,
Boston, MA 02215, USA.
6
Department of Neurosurgery, Faculty of Medicine, Istanbul University, Istanbul 34093, Turkey.
7
These authors contributed equally:
Jason F. Hou, Md Osman Goni Nayeem.
e-mail:
canand@media.mit.edu
Nature Communications
| (2024) 15:4601
1
1234567890():,;
1234567890():,;
modality that has been used by the medical community for half of a
century as a tissue-safe medium of energy transduction. Conformable
ultrasound electronics interface intimately with soft tissues to image,
deliver drugs to, or stimulate organs
7
9
. Unobstructed transcranial-
focused ultrasound (tFUS) beams can achieve millimeter-scale reso-
lution in neural tissue and penetra
te several centimeters to excite
neurons by affecting mechanoreceptive and other membrane-bound
ion channels
10
13
. Therefore, the ability to safely evaluate potential
stimulation targets, and with adjustable parameters such as frequency
and acoustic intensity, make it an advantageous approach
14
,
15
for
neurostimulation therapy in patients with conditions such as Alzhei-
mer
s disease, epilepsy, and depression. Ultrasound, when transmitted
from outside the human skull, faces signi
fi
cant scattering and re
fl
ec-
tion from the skull
s high acoustic impedance
16
, which can cause off-
target stimulation via conduction through bone and auditory
pathways
17
,
18
and even traumatic, irreversible brain injury
19
.Toachieve
a balance between skull transmission and spatial selectivity, most
signi
fi
cant modulations of neurons with ultrasound have been repor-
ted at frequencies less than 1 MHz
15
,
20
, and particularly with 500 kHz
with pressures at or above 100 kPa
14
,
21
23
.
Implantable devices allow electrical and chemical modulation of
the brain, leading to signi
fi
cant advancements in treating neurolo-
gical and psychiatric disorders
24
27
. Electrical deep brain stimulation
(DBS) can induce reversible activation of neurons but is limited by
anisotropic charge transfer across the brain
s ionic medium to
regions proportional to the size of the electrode
28
,
29
. Both the charge
provided by these electrodes and the sensitivity of surrounding tis-
sue can decrease signi
fi
cantly over time due to biofouling and cor-
rosion, which limits the longevity of the device
28
,
30
,
31
. Using various
frequencies of light, an optogenetics approach provides minimally
invasive neurostimulation with high spatiotemporal resolution and
cell-type speci
fi
city
32
. Development of safe, widespread non-
immunogenic delivery in the brain remains a challenge for clinical
translation
33
,
34
. A miniaturized, non-genetic platform for localized
neurostimulation is therefore needed to
fi
ll the gap for next-
generation neural interfaces to reach high standards of safety and
longevity. Recently, several reports of miniaturized ultrasonic neu-
rostimulation devices have shown that directed ultrasound energy
can activate cultured neurons
35
and neurons in brain slices
36
. How-
ever, the proposed platforms are not suitable for implantation in the
deep brain due to their rigid form factors, material composition, or
high-power requirements. A scalable implant system with no elec-
trochemically active elements that has the capability to non-
genetically and locally modulate neurons in deep subcortical brain
regions is needed to
fi
ll the translational gap.
Here, we report an implantable piezoelectric ultrasound stimu-
lator (ImPULS) that delivers acoustic energy directly and precisely to
populations of neurons in deep brain regions. Our key
fi
ndings include
the design and development of a low-power, micron-scale
fl
exible
piezoelectric micromachined ultrasound transducer (30
μ
m thick with
an outline width of 140
μ
m where the diameter of the active piezo
element is 100
μ
m) that can evoke neurons adjacent to the transducer
in the deep brain. The ImPULS (i) uses biocompatible piezoelectric thin
fi
lm of potassium sodium niobate (KNN)
37
as an active element sus-
pended over an air-
fi
lled cavity as an acoustic backing, (ii) generates
ultrasound at a pressure of 59.2 kPa at 15
μ
m away from device
(100 kPa adjacent to the transducer) for a single element, (iii) remains
functional after 7 days in an accelerated (75 °C) phosphate-buffered
saline (PBS) solution without incurring signi
fi
cant electrical and
mechanical degradation, and (iv) does not cause temperature rise
above safe tissue thresholds during ultrasound generation
10
.We
demonstrate the stimulation of neurons in a coronal hippocampal slice
ex vivo captured by two-photon microscopy and activation of hippo-
campal cells in anesthetized mice to induce expression of the activity-
dependent gene cFos across acute and 14-day timescales.
Furthermore, in vivo stimulation of dopaminergic neurons in the
substantia nigra pars compacta (SNc) with the ImPULS elicits time-
locked modulation of striatal dopamine release, highlighting the
ImPULS as a potent neuromodulatory tool.
Results
Design, fabrication, and characterization of the ImPULS
The implantable piezoelectric ultrasound stimulator, ImPULS, is a
fl
exible piezoelectric micromachined ultrasound transducer (pMUT),
that is surgically implanted into the brain. A schematic of the ImPULS
implanted into a subcortical region of a wild-type mouse is shown in
Fig.
1
a. Upon application of an alternating voltage, the ImPULS gen-
erates an ultrasound beam and excites nearby neurons as described in
detail in later sections. The ImPULS applies a transfer printing process
that enables the fabrication of implantable ultrasound stimulators
using biocompatible piezoceramics insulated with durable polymers.
This involves wet etch patterning and release of target piezoelectric
fi
lms from host Silicon (Si) wafers and integration onto polymer SU-8
by transfer printing (Supplementary Figs. 1 and 2). The substrates can
be engineered to serve the requirements of the application such as
chemical resistance, stiffness, and biostability. Figure
1
b is the peeled
view of the ImPULS revealing each constituent layer; comprising SU-8
as substrate (0.8
μ
m in thickness), encapsulation and backing layers
(0.5
μ
mand15
μ
m in thickness, respectively); piezoelectric KNN layer
(1
μ
m in thickness, and 100
μ
m in diameter); chromium/gold (Cr/Au,
10/250 nm in thickness) and platinum (Pt, 100 nm in thickness) serving
as top and bottom electrodes, respectively and Cr/Au as metal inter-
connects (10/250 nm in thickness). In order to maximize the vibration
amplitude of the active membrane, the dimensions of the element
geometries are designed to replicate a pMUT with a pinned boundary
device structure
38
. Compared to bulk piezoceramics and silicon-based
pMUTs, the ImPULS has a thinner pro
fi
le and lower Young
s Modulus
which better couples to soft brain tissue. Without a pMUT, a simple
thickness-mode resonance-dependence makes piezoceramic devices
at the 500 kHz range millimeters thick and unfavorable for minimally-
invasive neurostimulation. We chose to use a single-element pMUT for
this study to target adjacent neuron somas in the 50
μ
mhemispherical
radius of the transducer as will be presented in simulation later in
Fig.
2
d. However, the ImPULS fabrication process can be scaled to
produce
fl
exible arrays that target larger regions of the brain tissue. A
microfabricated pMUT array of 9 elements, where each element has
the same dimension as the ImPULS active piezo unit and the resonant
frequency is in the 500 kHz range is shown in Supplementary Fig. 3. As
the microfabrication of suspended free-standing membranes remains
a challenge using standard processes, we fabricated the devices in an
inverted manner, which included a
fi
nal bonding of a backing layer and
exposure through a thin transparent polyethylene terephthalate (PET)
layer to seal the air-
fi
lled cavity. The cavity and backing layers are also
designed to increase the stiffness of the ImPULS for acute implantation
without an insertion shuttle (see
Methods
section for a full descrip-
tion of the fabrication process).
We choose KNN as the lead-free piezoelectric layer due to (i) its
comparably high piezoelectric coef
fi
cients (e
31
,d
31
) and durability (DC
stress lifetime: >24 h at 200 °C and 30 kV/cm, and Curie temperature
of 350 °C) that exceeds commercially available doped lead zirconate
titanate (PZT) and polymer-based piezoelectric, (ii) its proven bio-
compatibility and non-toxicity, and (iii) its commercial availability. The
initial P-E characteristic of KNN before microprocessing is shown in
Supplementary Fig. 4.
The microfabricated unit is connected to a printed circuit board
(PCB) using an anisotropic conductive
fi
lm (ACF) based cabling to
complete the
fi
nal device fabrication (Fig.
1
c). A colorized scanning
electron microscopy (SEM) image of the cross-section of the ImPULS
shown in Fig.
1
ddepictstheair-
fi
lled cavity and encapsulation of the
active electrical elements.
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
2
To optimize the device parameters systematically, the electro-
mechanical properties of the ImPULS were investigated before surgical
implantation. Deionized water serves as a representative testing
medium due to its similar acoustic properties to brain tissue
39
and its
similar effects in terms of the resonant frequency shift of ImPULS, as
we validated in 0.6% agar gel phantoms mimicking brain tissue. The
electrical impedance and phase angle spectra of the ImPULS measured
in air and water medium are shown in Fig.
2
a. The resonance frequency
of the ImPULS when submerged in water or implanted in phantom/
mice brain has a pronounced but consistent shift due to the hydro-
static forces exerted on the
fl
exible material surrounding the entire
device and the piezoelectric thin
fi
lm itself
40
,
41
.Accordingly,we
observe consistent decreases in resonance frequency of 40.4%, from
840 kHz to 500 kHz, in air versus water and agarose gel mediums,
respectively. Similar resonance behavior of the ImPULS in air, water,
and agarose gel mediums was observed under a scanning laser Dop-
pler vibrometer (LDV) (Supplementary Fig. 5). We determined the
resonant electromechanical behavior of ImPULS with a LDV by
applying a periodic chirp excitation ranging from 100 kHz to 2 MHz to
the device and measuring the frequency spectrum of the resulting
vibrations (Fig.
2
b bottom). Once the resonance frequency was
determined, a pure sinusoidal signal was applied to the device to
determine the maximum displacement of the piezoelectric membrane.
A displacement of 137 nm at a resonance frequency of 840 kHz (air
medium) and displacement of 230 nm (water medium) was achieved
upon application of a 4 V (p-p) sinusoidal signal.
The pressure output of a pMUT device is proportional to the
center displacement of the vibration plate, and the center displace-
ment represents the point of maximum mechanical energy
transduction
38
. To optimize the device performance, we fabricated the
ImPULS with different cavity sizes. As shown in Supplementary Fig. 6a,
the resonance frequency decreases with the increase in cavity dia-
meter. It has been reported that the most signi
fi
cant modulation of
neurons occurs with the application of ultrasound at frequencies less
than 1 MHz
15
,
20
. We, therefore, choose to characterize devices with a
cavity size of 105
μ
m that has a resonance frequency of ~500 kHz. To
investigate the effect of applied voltage on the transducer, we varied
voltage from 2 V to 10 V (p-p) and recorded the peak displacement
usingalaserDopplervibrometer(LDV)inbothairandwatermediums
(Fig.
2
c). The detailed experimental setup is described in
ImPULS
Ultrasound
transducer
Activated
neurons
Subcortical
regions
ab
cd
SU-8
encapsulation
Top electrode
KNN thin film
Cavity layer
Backing layer
SU-8
membrane
SU-8
SU-8
KNN
Au
Pt
Air cavity
Backing layer
ImPULS assembled to PCB
Fig. 1 | An implantable piezoelectric ultrasound stimulator: ImPULS.
a
Schematic illustration of an implantable piezoelectric ultrasound stimulator
(ImPULS) implanted in a subcortical brain region of a wild-type mouse. A magni
fi
ed
view showing the activated neurons with ultrasound application. Schematic cre-
ated with BioRender.com, released under a Creative Commons Attribution-
NonCommercial-NoDerivs 4.0 International license.
b
Schematic of a peeled view
of the ImPULS, revealing each layer. The ImPULS is a piezoelectric micromachined
ultrasound transducer (pMUT) structure where biocompatible potassium sodium
niobate (KNN) is sandwiched between two thin SU-8 layers, and an air-
fi
lled cavity
and a backing layer is formed underneath the piezoelectric thin-
fi
lm membrane.
c
Optical image of the ImPULS assembled with
fl
exible ACF cable and a custom
printed circuit board (PCB) with a magni
fi
ed view of the ImPULS probe (right, top
inset) and further zoomed version of the tip of the probe (ultrasound unit) under a
microscope (bottom, left inset). Scale bars, 5 mm, 2 mm, and 100
μ
m, respectively.
d
Colorized cross-sectional scanning electron microscope (SEM) image of the
ImPULS. The device consists of (1) SU-8 encapsulation layer, (2) Top Pt electrode,
(3) KNN thin
fi
lm, (4) Bottom Au electrode, (5) SU-8 membrane layer, (6) Air-cavity,
and (7) SU-8 backing layer. Scale bars, 20
μ
m and 500 nm, respectively.
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
3
Supplementary Methods and Supplementary Fig. 7. As shown in
Fig.
2
c, the output peak displacement increased from 98 nm to 995 nm
when the applied voltage changed from 2 V to 10 V. A two-dimensional
(2-D) point scan of the displacement was used to characterize the
lateral resolution of the focal point, which reaches max intensity within
an 80
μ
mdiameter(Fig.
2
c, inset). The movement and shape of
membrane deformation can be visualized from a 3-D reconstruction of
the 2-D scan as shown in Supplementary Fig. 8. Notably, the increase in
cavity size leads to higher displacement as shown in Supplementary
Fig. 6b, which can be attributed to the reduced piezoelectric dia-
phragm stiffness with the increased cavity size. Although pMUT
devices are commercially fabricated in Si substrates, our simulations
indicate that it would be inef
fi
cient to create a KNN-based pMUT on a
silicon substrate that is able to target the 500 kHz range without
compromising its maximum width dimension and/or output pressure
(Supplementary Note 2 and Supplementary Fig. 9).
To understand the generated pressure pro
fi
le, we performed
fi
nite element analysis (FEA) using the COMSOL Multiphysics simulator
(see
Methods
section and Supplementary Note 5 for parameters used
in the simulation). As shown in Fig.
2
d, the maximum pressure adjacent
to the transducer can be as high as 100 kPa and decreases gradually
following a spherical pressure distribution. We measured the pressure
ab
def
c
500
1000
1500
2000
0.0
0.5
1.0
1.5
Air
Water
Frequency (kHz)
Input: periodic chirp
0
75
150
225
Displacement (nm)
Input: sinusoidal
20
30
40
50
air medium
-86
-84
-82
-80
300
400
500
600
700
800
900
20
30
40
50
Frequency (kHz)
Impedance (kΩ)
-86
-84
-82
-80
Phase ( ̊)
water medium
Impedance
Phase
0.2
2
5
10
20
50
100
Pressure
(kPa)
-80
-40
0
40
80
r (μm)
20
60
100
140
180
Z (μm)
220
gh
0102030
0.00
0.25
0.50
0.75
1.00
Temperature change, ΔC ( ̊C)
Time (min)
US ON
US OFF
0.34
0.29
0.24
0.19
ΔT ( ̊C)
x
y
59.2
49.2
39.2
29.2
19.2
Pressure (kPa)
x
y
Day 0
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7
i
20 μs
20 kPa
0255075100
0
20
40
60
80
100
Pressure (kPa)
simulation
experiment
Distance from transducer (μm)
246810
0
200
400
600
800
1000
Displacement (nm)
Voltage (V)
Displacement
(nm)
0
400
800
0.0
0.2
0.4
0.6
0.8
1.0
Normalized displacement (a.u
.)
Fig. 2 | Characterization of implantable piezoelectric ultrasound stimulator
(ImPULS). a
The impedance and phase angle spectra of the ImPULS in air and
water, showing the resonance frequency in both mediums.
b
Displacement of
ImPULS in air and water mediums measured using laser Doppler vibrometer (LDV)
at 4 V (p-p) when the inputs are a periodic chirp (bottom) and a sinusoidal signal
(top).
c
Displacement of ImPULS as a function of input voltage (p-p) with inset
showing two-dimensional (2-D) point scan of displacement indicating the lateral
resolution of the beam of the device. Error bar represents the standard deviation in
measurement,
N
= 3 devices. Data represents mean values ± SD.
d
Simulated
acoustic pressure pro
fi
le of the ImPULS showing a spherical pressure distribution.
e
Comparison of simulated and experimentally measured pressure using a
fi
ber-
optic hydrophone at different distances. Error bar represents the standard
deviation in measurement,
N
= 3 devices. Data represents mean values ± SD in the x
and y directions.
f
2-D mapping of pressure generated by the ImPULS measured at
z=15
μ
m. Scale bar, 25
μ
m.
g
Microscopic image of the ImPULS taken each 24 h
apart during the aging test (left), and normalized displacement of ImPULS before
the start of the test and after 7 days. Scale bar, 100
μ
m. Error bar represents a
standard deviation in measurement,
N
= 3 devices. Data is normalized to the dis-
placement measured on day 0 per device. Day 8 data represents the mean dis-
placement ± SD. (a.u. arbitrary units).
h
Temperature change in water medium
when a continuous sinusoidal signal of 500 kHz at 20 V (p-p) applied to the ImPULS.
Ultrasound was
off
for 10 min,
on
for 10 min and
off
for 10 min.
i
2-D mapping of
temperature generated by the ImPULS measured at z = 15
μ
m. Scale bar, 25
μ
m.
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
4
generated by the ImPULS using a
fi
ber-optic hydrophone positioned
with its sensing element facing the device. ImPULS, with a diameter of
100
μ
mand1
μ
m thick piezoelectric thin
fi
lm, creates a near-
fi
eld
pressure within 10
μ
mandafar-
fi
eld region around 100
μ
minlength.
Measurements with an optically-coupled sensor for pressure is
imperative because electromagnetic interference (EMI) generated by
the wires and piezoelectric at distances less than 100
μ
mdegradesthe
sensitivity of conventional piezoceramic hydrophones and over-
powers the signal of interest. Therefore, we utilized a
fi
ber-optic
hydrophone mounted on a 3-axis stage that can bypass the EMI effect
as an orthogonal measurement modality (see
Methods
section for
the detailed experimental setup and Supplementary Fig. 10). As shown
in Fig.
2
e, we measured the pressure generated by the ImPULS at four
different distances. The experimentally measured pressure matched
well with the simulated pressure values. The slight discrepancy can be
attributed to the minimum resolution of the 3-axis stage and the
unknown true position of the sensing element in the
fi
ber hydrophone
tip. Next, we scanned the
fi
ber-optic hydrophone in lateral x and y
directions with a step size of 25
μ
m, keeping the axial z-distance con-
stant at 15
μ
m. Figure
2
f shows the pressure pro
fi
le mapping, where a
maximum pressure of 59.2 kPa is achieved at the center of the ImPULS.
This characterization is important for the veri
fi
cation of precise and
localized stimulation of neurons residing within the pressure
fi
eld of
the device.
Durability, temperature stability, and biocompatibility
To test that an ImPULS remains functional over a long period of use
within a harsh biological environment, we tested the durability of the
ImPULS by performing an accelerated aging test in a phosphate-
buffered saline (PBS) solution at an elevated temperature of 75 °C for 7
days (Supplementary Fig. 11). Figure
2
g shows microscopic images of
an ImPULS taken each 24 h apart, where there is minimal damage
visually to the device after 7 days of continuous exposure to PBS at
75 °C. To con
fi
rm the ImPULS remains fully functional, we measured
the displacement of the ImPULS before and after 7 days of aging test.
As shown in Fig.
2
g (right), the displacement of the device degrades
only 2.4% in 7 days. In order to simulate the potential use of the ImPULS
as a chronic device, we repeated the accelerated aging test with an
additional application of stressor where a daily period of 10 min the
device was turned ON. We recorded the voltage necessary each day to
maintain the performance of the device as recorded on day 0. As seen
in Supplementary Fig. 12, this adaptive voltage strategy enabled
compensation for the loss of vibrational amplitude during the accel-
erated aging test, and the device sustained the displacement recorded
at day 0 across 7 days.
To test the durability of the ImPULS further, we performed a
fatigue test, during which a continuous sinusoidal signal of 500 kHz at
10 V (p-p) was applied continuously for 7 days and corresponding
output displacement was recorded in a water medium. As shown in
Supplementary Fig. 13, exposure to 302.4 billion cycles of a sine wave
in 7 days results in a 40% lower amplitude of initial displacement, as
degradation of the piezoelectric layer after extended electric cycling is
a common phenomenon
42
. Compensation for performance loss due to
piezoelectric degradation can be achieved by the application of
adaptive voltage, the same strategy used in the accelerated aging test.
We repeated the electromechanical fatigue test keeping the device ON
for 5 days, and readjusted the driving voltage daily to achieve the
target displacement equivalent to day 0 performance, as recorded by
Laser Doppler Vibrometer (LDV). As seen in Supplementary Fig. 14, an
adaptive voltage strategy was able to compensate for the expected
displacement loss and could stabilize the performance of the ImPULS
device.
Ultrasound generation from a transducer results in a temperature
rise in the surrounding medium due to the intrinsic heating of the
piezoelectric material and resistive losses
43
. We measured changes in
temperature in a water medium (22.5 °C baseline temperature, Onda
Aquas-10 tank) during ultrasound application using a dual sensing
fi
ber-optic hydrophone capable of simultaneous measurements of
acoustic pressure and temperature at the same location
44
.The
fi
ber-
optic thermometer positioned 15
μ
mawayfromthetransducer
recorded a temperature rise upon application of continuous ultra-
sound waves. As shown in Fig.
2
h, 10 min application of continuous
sinusoidal signal at an input voltage of 20 V (p-p) gave rise to only
0.6 °C which is much less than the t
hreshold of temperature-evoked
neuromodulation
10
,
45
. In practical neurostimulation applications,
thermogenic effects are further reduced without affecting peak pres-
sures generated due to the application of a pulsed ultrasound signal
instead of a continuous signal. Further, we measured the temperature
change upon application of pulsed signals with duty cycles of 50% and
5%. As shown in Supplementary Fig. 15a, b application of 50% and 5%
duty cycle pulsed waves gave rise to only 0.15 °C and 0.03 °C,
respectively. The dependence of temperature change on input voltage
is shown in Supplementary Fig. 15c, indicating a maximum tempera-
ture change of 0.46 °C with an 18 V (p-p) input, which decreases to
0.08 °C with a 10 V (p-p) input. We also measured the temperature
change in a 1.5% agarose gel (heat capacity of 3.9 J/kg/°C which is
similar to brain tissue
46
) at the application of 20 V (p-p) continuous
sinusoidal signal of 483 kHz, using a miniature beaded thermocouples
(Evolution Sensors, Type K with bead diameter ~300
μ
m) for sensing.
As shown in Supplementary Fig. 15d, e, the maximum temperature
rises by 0.95 °C with the application of 20 V signal, which is less than
the threshold of temperature-evoked neuromodulation
45
.Inour
ex vivo and in vivo neurostimulation experiments, a maximum of 10 V
(p-p) is applied as will be described in later sections. Figure
2
ishows
the temperature pro
fi
le mapping upon application of continuous
sinusoidal signal where a maximum temperature change of 0.34 °C
occurred, at 15
μ
m away in the z-direction from the probe center.
The ImPULS was surgically implanted in the deep brain for neu-
rostimulation. To con
fi
rm that surgical implantation did not deterio-
rate its performance, we tested the ImPULS performance before and
after insertion into the brain tissue-mimicking phantom. We prepared
0.6% agar gel to mimic the similar stiffness properties of brain tissue
47
and measured the displacement of ImPULS before and after insertion
into the gel. As shown in Supplementary Fig. 16, the change in dis-
placement is less than 1.5% con
fi
rming the device
s stability after
implantation into the brain tissue.
Next, the ImPULS device was assessed for the biocompatibility of
its constituent materials with a cell viability test. Cortical tissues from
embryonic mice were harvested and dissociated cells were seeded on
glass-bottomed dished containing a
fi
xed ImPULS device. The dis-
sociated cells were allowed to differentiate into cultured primary
neurons for a period of 10 days in the presence of the device. Cell
densities across 6 plates (3 for control and 3 for ImPULS) were assessed
and found to stabilize after a cell medium change on day 3 (Supple-
mentary Fig. 17). Neurons on both Control and ImPULS plates differ-
entiated neurites normally and developed neurites stably after the 10-
day period.
Hippocampal neuronal stimulation ex vivo
The ImPULS was evaluated for its potential to stimulate healthy neu-
rons in a coronal hippocampal slice with two-photon imaging (Sup-
plementary Fig. 18a). Hippocampal neurons expressed the genetically
encoded calcium indicator GCaMP7F to report neural activity during
ultrasonic stimulation
48
. Neurons in the dentate gyrus were targeted
for stimulation (Supplementary Fig. 18b). After a 60 s baseline period, a
sinusoidal pulse (500 kHz, 10 V(p-p) with 1.5 kHz pulse repetition fre-
quency (PRF) and 50% duty factor) is used to stimulate neurons for
50 s. After stimulation ends, populat
ion activity is captured for another
60 s. Several neurons in the
fi
eld of view were activated during ultra-
sound stimulation (Supplementary Fig. 18c and Supplementary
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
5
Video 1). Region of interest 1 (ROI 1) exhibits a 30% change in
fl
uor-
escence approximately 15 s after stimulation begins and reaches
maximum intensity 9 s after the initial rise. The other marked neurons
belong to the same local cluster and showed smaller changes in activity
during stimulation. The delay in neural activation could be due to high
amounts of dissolved gasses in the arti
fi
cial cerebrospinal
fl
uid (aCSF),
which may lower cavitation thresholds and cause ultrasound energy
from small sources to be absorbed into the medium rather than the
tissue
49
. In the absence of other stimuli, including the effects on active
vasculature
50
,
51
, these results con
fi
rm that ImPULS can activate local
hippocampal neurons.
Stimulation of CA1 in anesthetized mice induces cFos expression
We next sought to test the potency of
the ImPULS to activate cellular
ensembles in vivo in mice. To do so, we surgically implanted the
ImPULS in the hippocampus
a subcortical brain region that is essential
for learning and memory across mammalian species
52
(Fig.
3
a). Spe-
ci
fi
cally, we targeted the ImPULS to the dorsal CA1 (dCA1) layer of the
hippocampal formation to test the ef
fi
cacy of different stimulation
protocols. We quanti
fi
ed the relative levels of cFos in dCA1, which is a
widely used marker of recent neuronal activity, to measure the extent
of neural activation resulting from different bouts of the ImPULS sti-
mulation under anesthesia
53
. Compared to a no-stimulation (control)
group, all stimulation parameters increased the levels of the activity-
dependent gene cFos on average. Speci
fi
cally, we observed an
approximately 2-fold increase in cFos expression following stimulation
with 500 kHz 10% duty factor (0.777 cFos+/mm
2
stim vs. 0.497 cFos
+/mm
2
in control, one-way ANOVA), suggesting that this stimulation
parameter was suf
fi
cient to activate large populations of cells in the
dCA1 layer of the mouse hippocampus (Fig.
3
b, c and Supplementary
Figs. 19 and 20). We quanti
fi
ed signi
fi
cant increases in cFos expression
along the entirety of dCA1. Furthermore, we visualized auxiliary cFos
expression in CA3 and DG that could be due to backward propagating
action potentials, local circuit increases in cFos resulting from ImPULS
stimulation spread, and intra-hippocampal communication in
general
54
, which nonetheless underscore the potency of the ImPULS
a
c
Hippocampus
ImPULS
1.0
0.8
0.6
0.4
0.2
cFos+ cell/mm
2
No-stim
500 kHz
500 kHz, 10% duty
factor
*
Implantation
Stimulation
90 min
Perfusion
cFOS
staining
Cell counts
48 h
Post-fix in 4%
PFA
72 h
Tissue
Processing
b
DAPI
cFOS
DAPI
DAPI
cFOS
Control
DAPI
cFOS
DAPI
cFOS
DAPI
cFOS
DAPI
cFOS
500 kHz
500 kHz 10% DF
Fig. 3 | Robust stimulation of the dCA1 in anesthetized mice. a
Experimental
design and schematic diagram of surgical procedure. Schematic created with
BioRender.com, released under a Creative Commons Attribution-NonCommercial-
NoDerivs 4.0 International license.
b
Representative images of the hippocampus
across experimental conditions: No-stim (top), and 500 kHz, 500 kHz, continuous
wave for 60 s (middle), and 10% duty factor for 60 s (bottom). Yellow dashed boxed
area approximates the dorsal CA1 (dCA1) area used for cell counts. Scale bar,
450
μ
m. Color-coded dashed box indicates the area shown in the magni
fi
ed section.
Scale bar, 100
μ
m.
c
cFos+ cells in dCA1 normalized by area across experimental
conditions (One-way ANOVA;
N
= 4 mice control implant no-stim;
N
=3micefor
500 kHz condition;
N
= 3 mice for 500 kHz, 10% duty factor; No-stim vs. 500 kHz:
p
= 0.0506; No-stim vs. 500 kHz, 10% duty factor:
p
=0.0184).
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
6
stimulation parameters. Furthermore, we tested ImPULS stimulation in
the dCA1 layer on mice 14 days after ImPULS was implanted. The
500 kHz 10% duty factor stimulation group showed a signi
fi
cant
increase in average cFos+/DAPI+ in the dCA1 layer using an unpaired
t
-
test (
p
= 0.0258). Our results indicate that ImPULS stimulation was
suf
fi
cient to elicit neuronal activation following chronic implantation.
This suggests the ImPULS device is functionally viable on more chronic
timescales. Immunohistological staining of GFAP demonstrated mini-
mal microglial activation in response to chronic implantation in the
area of tissue most proximal to the ImPULS transducer (Supplemen-
tary Fig. 21). Together, these results demonstrate the applicability of
in vivo neuronal stimulation using an ImPULS.
Ultrasound stimulation of the SNc modulates nigrostriatal
dopamine release in an anesthetized mouse
Next, we tested the utility of an ImPULS stimulation for functional
modulation of neurotransmission in vivo using an anesthetized pre-
paration. Dopaminergic (DA) neurons of the SNc innervate the dorsal
striatum (DS) to form the canonical nigrostriatal dopamine pathway,
a circuit crucial for movement and reinforcement in the mammalian
brain. Furthermore, excitatory stimulation of dopaminergic trans-
mission has therapeutic implications in the treatment of Parkinson
s
Disease (PD) and memory disorders
55
,
56
. Therefore, we sought to
modulate nigrostriatal DA release through the ImPULS stimulation of
the SNc using an anesthetized preparation (Fig.
4
a). We targeted the
Fig. 4 | Stimulation of nigrostriatal dopamine release in anesthetized mice.
a
Schematic diagram of the experimental approach to stimulate substantia nigra
pars compacta (SNc) dopaminergic (DA) neurons, including post-hoc histological
validation of on-target implantation and DA2m sensor expression. Scale bar,
500
μ
m. Schematic created with BioRender.com, released under a Creative Com-
mons Attribution-NonCommercial-NoDerivs 4.0 International license.
b
Averaged
DA2m
fl
uorescence responses for control (top right) and SNc (bottom right) sti-
mulation trials. Average heatmaps of
fl
uorescence across trials for control (top left,
3 mice) and SNc (bottom left, 3 mice) stimulation trials. Solid line data represents
mean data and bands represent ± SD.
c
, Full recording trace of Z-score DA2m
fl
uorescence across stimulation trials, with onset and offset of stimulation
(5 s,1500 Hz, 50% duty factor) indicated by solid and dashed red lines, respectively.
d
, Area under the curve (AUC) analysis for 5 s pre-stimulation versus 5 s during
stimulation for average control and SNc stimulation trials (3 mice/group).
Repeated-measures 2-way ANOVA with Sidak
s multiple comparisons test
(***
p
= 0.0009, ****
p
< 0.0001).
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
7
ImPULS implantation to the anterior SNc with a lateral stimulation
direction and performed
fi
ber photometry recordings of extra-
cellular DA release in the DS using a genetically encoded DA sensor,
GRAB-DA2m
57
. Further details describing the surgical approach,
recording parameters, and data analysis can be found in the Sup-
plementary Information. Pulsed (PRF 1500 Hz, 50% duty factor) sti-
mulation of the SNc for 5 s (514 kHz, 10 V(p-p)) elicited robust, time-
locked increases in striatal DA release (Fig.
4
b). Notably, control sti-
mulation trials, in which we stimulated tissue approximately 200
μ
m
dorsal to the SNc, failed to alter DA2m
fl
uorescence (Fig.
4
c, top and
Supplementary Fig. 22). Therefore, at least in the areas of tissue
inferior to the device, stimulation does not reach beyond 100
μ
m. In
contrast, we observed a mean increase in DA2m
fl
uorescence
between 2 and 3 Z-scores throughout the duration of ImPULS-
mediated SNc stimulation (Fig.
4
c, bottom). As seen in the raw traces,
reduced magnitude and variability of DA signaling is pronounced
between stimulation and control trials. We calculated the mean area
under the curve (AUC) across stimulation trials from the 5 s baseline
period before stimulation and 5 s period during stimulation of both
SNc and control tissue (Fig.
4
d). Mean AUC of DA
fl
uorescence during
SNc ImPULS stimulation was signi
fi
cantly different from baseline
(F1,4 = 65.20,
p
= 0.0013) using a two-way repeated measures
ANOVA. Furthermore, SNc stimulation was signi
fi
cantly different
from control stimulation (F1,4 = 39.46,
p
= 0.0033), with an interac-
tion present between these two conditions (F1,4 = 47.00,
p
= 0.0024).
Š
ídák
s test for multiple comparisons showed that DA AUC during
SNc stimulation dramatically increased compared to pre-stimulation
baseline (
p
= 0.0009), with no difference between baseline and sti-
mulation AUC for control tissue (
p
= 0.6834). Furthermore, DA AUC
during SNc stimulation was signi
fi
cantly greater than control tissue
stimulation (
p
< 0.0001) with no signi
fi
cant difference in baseline
AUC between SNc and control tissue stimulation. Taken together,
these data suggest that the ImPULS evokes robust nigrostriatal DA
release in a spatially localized manner, given that the stimulation of
tissue only 200
μ
m above the SNc failed to alter DA release. In four
mice, we observed a mean increase in DA2m
fl
uorescence between 3
and 4 Z-scores throughout ImPULS-mediated SNc stimulation for
1.5 s with these same parameters (Supplementary Fig. 23a), indicating
that stimulation of shorter durations is suf
fi
cient to induce DA
release. In a
fi
fth mouse, however, we observed a mean decrease in
DA2m
fl
uorescence between 2 and 3 Z-scores throughout ImPULS-
mediated stimulation for 5 s with these parameters (Supplementary
Fig. 23b). Sections of the SNc were stained for tyrosine hydroxylase
(TH), a widely used marker of dopamine-producing cells, to validate
that the ImPULS probe tract contacted dopaminergic neurons in the
SNc (Fig.
4
a). There are multiple plausible explanations for the inhi-
bition of nigrostriatal DA release. First, ImPULS may elicit both
excitatory and inhibitory effects through an unknown underlying
mechanism. Second, this may result from the surgical targeting of the
SNc with a small and
fl
exible probe. The substantia nigra pars reti-
culata (SNr), which is predominantly composed of GABAergic neu-
rons, is situated directly ventral to the SNc and provides
monosynaptic inhibitory input to DA neurons
58
. Given that the SNc
DA neurons are angled on the medial-lateral axis and the direction of
ImPULS stimulation was lateral, it is feasible that stimulation reached
a signi
fi
cant number of SNr inhibitory neurons in this particular case,
effectively silencing DA neurons. Additionally, the tip of the probe
tract for this case reached the ventralmost portion of the SNc. Finally,
a third explanation for these different effects of ImPULS stimulation
is the possibility that deep brain ultrasound stimulation may differ-
entially modulate neural activity by indirect modulation, affecting
the activity of astrocytes
13
. Nevertheless, these data demonstrate that
spatially localized deep brain ultrasound stimulation is capable of
modulating neurotransmission in vivo, even through long-range
projections.
Discussion
This work presents a micron-sized, implantable ultrasound stimulator
capable of modulating neuronal activity in deep subcortical regions
and nigrostriatal dopamine production across long-range projections.
We demonstrated the scalable microfabrication of an ImPULS,
including the use of biocompatible materials such as the active pie-
zoelectric element (KNN), interconnects, and encapsulation, as well as
control over resonance frequency within the range of 0.2
1MHz
through manipulation of cavity size. We demonstrated that ImPULS
has consistent resonant behavior (Supplementary Fig. 5) and minimal
heating <1 °C (Supplementary Fig. 15) while placed in different acoustic
and thermal tissue-mimicking mediums such as 1.5%, 0.6% agar gel,
and air. When operating at its resonant frequency, the ImPULS drew
0.2 mA current with 10 V (p-p) of applied voltage. At a power con-
sumption of 0.2 mW, the ImPULS generated ultrasound pressures of
100 kPa at resonance frequency in pulsed and continuous waves within
its stimulation region and evoked modulation of cell activity without
causing thermogenic effects on nearby cells. The modulation of brain
circuitry with a pressure of 100 kPa around 500 kHz is consistent with
prior literature of in vivo experiments with mice using similar ultra-
sound driving parameters
21
. The ImPULS elicited neuronal excitation in
the hippocampus ex vivo, induced activity-dependent gene expression
in hippocampal cells of an anesthetized mouse, and modulated
dopaminergic neurons in the SNc to elicit precise timing of striatal
dopamine release. This presents the ImPULS as a potent
neuromodulatory tool.
Cells within the hippocampus have been shown to be sensitive to
ultrasound stimulation from transcranial and slice preparations
48
.
Within our immunochemistry studies of the ImPULS stimulation in the
hippocampus, the ImPULS activated large populations of neurons in
upstream canonical pathways relative to where stimulation occurred.
For instance, cells in the dentate gyrus are active accompanying sti-
mulation in the CA1 region. Contained
within the intrinsic circuitry of
the hippocampal formation are several parallel processing and feed-
back networks mediating excitation, inhibition, and disinhibition
59
.
Previous research has shown that inducing hypersynchrony in CA1
using optogenetic stimulation was suf
fi
cient to activate the entire
hippocampal formation
60
, which was consistent with our work show-
ing increased cFos expression in CA1, the DG, and CA3.
We achieved robust stimulation using an ultrasound-driving pro-
tocol that is known to excite neurons
22
,
61
, but we cannot discard the
possibility that there is an indirect pathway for neuronal excitation via
astrocytic modulation. Oh et al.
13
demonstrated in co-cultures of
neurons and astrocytes that TRPA-channels from astrocytes can be
ultrasonically activated and are suf
fi
cient to indirectly excite neurons
via glutamate release
13
. Demonstrations that illustrate the spatial
resolution of an ImPULS would be better replicated in regions of the
brain where fewer interconnections between neural circuits exist and
further studies investigating different stimulation parameters could
potentially be used to achieve a degree of cell selectivity as demon-
strated by other groups
62
.
Implantable ultrasound neurostimulation offers several advan-
tages compared to other neurosti
mulation modalities. Unlike
electrode-based deep brain stimulation, the ImPULS has no exposed
electrochemical area, thus avoiding biofouling and corrosion, and
compared to the DBS cross-sectional footprint (in the mm range) and
rigid form factor, the ImPULS is much smaller yet still capable of
precise and potent simulation. Furthermore, as glial scar tissue forms
in response to the implant, the acou
stic properties of the encapsulat-
ing tissue change negligibly. Therefore, it retains desirable properties
as the propagating medium for the effective delivery of ultrasound
energy. The ImPULS is implanted adjacent to target neurons, the
ultrasound generated bypasses the skull and is focused to a volume
<100
μ
m
3
, thereby avoiding off-target activation from scattering and
re
fl
ections. Finally, the use of ultrasound for neurostimulation offers
Article
https://doi.org/10.1038/s41467-024-48748-6
Nature Communications
| (2024) 15:4601
8