What if you didn’t need surgery to implant a pacemaker on a faulty heart?
What if you could control your blood sugar levels without an injection of insulin, or mitigate the onset of a seizure without even pushing a button?
I and a team of scientists in my laboratory at the Salk Institute are tackling these challenges by developing a new technology known as sonogenetics, the ability to noninvasively control the activity of cells using sound.
From light to sound
I am a neuroscientist interested in understanding how the brain detects environmental changes and responds.
Neuroscientists are always looking for ways to influence neurons in living brains so that we can analyze the outcome and understand both how that brain works and how to better treat brain disorders.
Creating these specific changes requires the development of new tools.
For the last two decades the go-to tool for researchers in my field has been optogenetics, a technique in which engineered brain cells in animals are controlled with light.
This process involves inserting an optic fiber deep within the animal’s brain to deliver light to the target region.
When these nerve cells are exposed to blue light, the light-sensitive protein is activated, allowing those brain cells to communicate with each other and modify the animal’s behavior.
But the obvious drawback is that this procedure depends on surgically implanting a cable into the brain – a strategy that cannot be easily translated into people.
My goal had been to figure out how to manipulate the brain without using light.
I discovered that ultrasound – sound waves beyond the range of human hearing, which are noninvasive and safe – is a great way to control cells.
Since sound is a form of mechanical energy, I figured that if brain cells could be made mechanically sensitive, then we could modify them with ultrasound.
This research led us to the discovery of the first naturally occurring protein mechanical detector that made brain cells sensitive to ultrasound.
Our technology works in two stages.
First we introduce new genetic material into malfunctioning brain cells using a virus as a delivery device.
This provides the instructions for these cells to make the ultrasound-responsive proteins.
The next step is emitting ultrasound pulses from a device outside the animal’s body targeting the cells with the sound-sensitive proteins.
The ultrasound pulse remotely activates the cells.
Proof in worms
We were the first to show how sonogenetics can be used to activate neurons in a microscopic worm called Caenorhabditis elegans.
Using genetic techniques, we identified a naturally occurring protein called TRP-4 – which is present in some of the worm’s neurons – that was sensitive to ultrasound pressure changes.
Sound pressure waves that occur in the ultrasonic range are above the normal threshold for human hearing.
Some animals, including bats, whales and even moths, can communicate at these ultrasonic frequencies, but the frequencies used in our experiments go beyond what even these animals can detect.
My team and I demonstrated that neurons with the TRP-4 protein are sensitive to ultrasonic frequencies. Sound waves at these frequencies changed the worm’s behavior.
We genetically altered two of the worm’s 302 neurons and added the TRP-4 gene that we knew from previous studies was involved with mechanosensation.
Sound frequency ranges for infrasound, audible and ultrasound waves and the animals that can hear them. People are able to hear only between 20 Hz and 20,000 Hz. The image is adapted from The Conversation news release.
We showed how ultrasound pulses could make the worms change direction, as if we were using a worm remote control. These observations proved that we could use ultrasound as a tool to study brain function in living animals without inserting anything into the brain.
The advantages of sonogenetics
This initial finding marked the birth of a new technique that offers insights into how cells can be excited by sound. Additionally, I believe that our results suggest that sonogenetics can be applied to manipulate a wide variety of cell types and cellular functions.
C. elegans was a good starting point for developing this technology because the animal is relatively simple, with only 302 neurons. Of these, TRP-4 is in only eight neurons. So we can control other neurons by first adding TRP-4 to them and then directing the ultrasound precisely at these specific neurons.
But humans, unlike worms, do not have the have the TRP-4 gene.
So my plan is to introduce the sound-sensitive protein into the specific human cells that we want to control.
The advantage of this approach is that the ultrasound won’t interfere with any other cells in the human body.
It is currently not known if proteins other than TRP-4 are sensitive to ultrasound. Identifying such proteins, if there are any, is an area of intense study in my lab and the field.
The best part about sonogenetics is that it doesn’t require a brain implant. For sonogenetics, we use artificially engineered viruses – that are unable to replicate – to deliver genetic material to brain cells.
This allows the cells to manufacture sound-sensitive proteins.
Sonogenetics, though still in the very early stages of development, offers a novel therapeutic strategy for various movement-related disorders including Parkinson’s, epilepsy and dyskinesia.
In all of these diseases, certain brain cells stop working and prevent normal movements. Sonogenetics could enable doctors to turn on or turn off brain cells at a specific location or time and treat these movement disorders without brain surgery.
Credit: Salk Institute.
For this to work, the target region of the brain would need to be infected with the virus carrying the genes for the sound-sensitive protein.
This has been done in mice but not yet in humans. Gene therapy is getting better and more precise, and I am hoping that other researchers will have figured out how to do this by the time we are ready with our sonogenetic technology.
We have received substantial support to advance this technology, fuel the initial study and establish an interdisciplinary team.
With additional funding from Defense Advanced Research Projects Agency’s ElectRx program, we can focus on finding proteins that can help us “turn off” neurons.
We recently discovered proteins that can be manipulated to activate neurons (unpublished work).
This is crucial for developing a therapeutic strategy that can be used to treat central nervous system diseases like Parkinson’s.
Our team is also working on expanding the sonogenetic technology.
We have now observed that certain plants, such as the “touch me not” (Mimosa pudica), are sensitive to ultrasound. Just as the leaves of this plant are known to collapse and fold inward when touched or shaken, applying pulses of ultrasound to an isolated branch produces the same response.
Finally, we are developing a different method to test if ultrasound can influence metabolic processes such as insulin secretion from pancreatic cells.
Sonogenetics could one day circumvent medications, remove the need for invasive brain surgeries and be useful for conditions ranging from post-traumatic stress disorder and movement disorders to chronic pain.
The great potential for sonogenetics is that this technology could be applied to control nearly any type of cell: from an insulin-producing cell in the pancreas to pacing a heart.
Our hope is that sonogenetics revolutionizes the fields of neuroscience and medicine.
Funding: Sreekanth Chalasani receives funding from the National Institutes of Health, Kavli Institute of Brain and Mind, and the Defense Advanced Research Agency (DARPA)’s ElectRx program.
Understanding how neural circuits generate specific behaviours requires identifying the participating neurons and subsequently recording and perturbing their activity.
The best-understood motor circuit, the crab stomatogastric ganglion, has benefited from electrophysiological access to well-defined cell types as well as an ability to manipulate their activity1.
While these methods have revealed insights into circuit computations in a variety of model systems, they suffer from one drawback: difficulty in delivering stimulus to target neurons located in deeper brain regions4,5.
To address this issue, we have developed a new method that genetically sensitizes targeted neurons to low-pressure ultrasound, a stimulus that can be delivered to small regions of deep tissue throughout an animal, including its brain.
Ultrasound stimulation is non-invasive.
This is particularly important for manipulating vertebrate neurons, as it eliminates the need for invasive surgery to insert fibre optics (required for some current optogenetic methods6).
Interestingly, activating neurons in these cases requires exposure to continuous or repeated pulses of ultrasound between 690 kHz and 3 MHz16.
Ultrasound has also been shown to safely manipulate deep nerve structures in human hands to reduce chronic pain17.
Despite these observations and the development of theoretical models18, the mechanisms by which ultrasound stimulates these neurons remain poorly understood.
Moreover, an additional challenge lies in developing a method to target individual neurons to ultrasound stimulation, as the minimum focal zone of the ultrasound is larger than an individual cell.
To overcome these challenges we developed a new method to stimulate neurons that we call ‘sonogenetics’, using the nematode, Caenorhabditis elegans. C. elegans with its small nervous system consisting of just 302 neurons connected by identified synapses19 has well-characterized robust behaviours20 and reliable methods to monitor neural activity21. We identified a pore-forming subunit of a mechanotransduction channel, TRP-4, that is sensitive to low-pressure ultrasound.
Further, we show that individual neurons misexpressing TRP-4 show changes in neural activity upon low-pressure ultrasound stimulation. Finally, we correlate these neural activity changes with specific behaviours at the level of whole animals.Go to:
Imaging set-up delivers ultrasound waves to animals
To investigate the role of ultrasound in wild-type C. elegans neural activity and behaviour, we developed a set-up that aligns optical imaging with the ultrasound focal zone (Fig. 1a).
Ultrasound with different peak negative pressures was generated from a transducer and focused onto an agar plate, where animals were corralled into a small area using a copper solution (Fig. 1b).
The transducer focused the ultrasound wave to a 1-mm-diameter circular area at the agar surface (red circle in Fig. 1b).
The entire set-up was placed in a large tank filled with water to facilitate uniform transduction of the ultrasound wave (Fig. 1a).
Depending on solution or tissue gas concentrations, high ultrasound peak negative pressures (>2.5 MPa) can create inertial cavitation, with the resulting shockwaves compromising the integrity of cell membranes22,23.
Consistently, we observed that animals exposed to multiple pulses of high ultrasound pressures were unable to maintain their normal body posture (Supplementary Fig. 1). Therefore, we chose to use low-pressure ultrasound, which does not cause these damaging effects, to stimulate animal behaviour.
We found that a single 10-ms-duration ultrasound pulse of 2.25 MHz and peak negative pressures below 0.9 MPa had no effect on animal behaviour (Fig. 1c).
The mechanical disturbances24 of the fluid and tissue in the ultrasound focal zone take the form of compression and expansion deformations as well as bulk tissue distortions caused by acoustic radiation forces, but at low pressures they were not large enough to influence C. elegans locomotion. Previous studies have shown that ultrasound waves can cause temperature changes in the focal zone25.
We first estimated the temperature increase as a result of ultrasound exposure. In a previous study, a continuous 1.1-MHz-ultrasound pulse with a peak negative pressure of 2.6 MPa increased the temperature of the surrounding media at the rate of 35 °C s−1 (ref. 25).
Using these data, we estimated the temperature increase around the worms on the agar surface to be 0.04 °C for a single ultrasound pulse at 0.9 MPa.
Moreover, we also directly measured the magnitude of temperature change on the agar surface using a miniature thermocouple (Supplementary Fig. 2) and found that an ultrasound peak negative pressure of 0.7 MPa caused a temperature increase of less than 0.1 °C (see Methods).
Together, these results show that C. elegans is unlikely to respond to the temperature and mechanical changes induced by the low-pressure ultrasound wave.
Basic characteristics of ultrasound
Several factors influence the interaction of US with biological tissues. For in depth discussions of these principles, we refer the reader to some other sources covering the basic biophysics of US (reviews see Refs. [3, 4]).
Some critical factors affecting outcomes include the acoustic frequency and intensity profiles of US waveforms used to affect biological activity.
First, the spatial resolution that can be achieved with US is a function of both the acoustic frequency and transducer aperture used. It is possible to focus US fields in the brain using acoustic lenses or with phased array methods that can correct for aberrations or distortions caused by tissue interference .
In soft tissues like brain, muscle, and fat that have acoustic properties similar to water , the diffraction-limited spatial resolutions for 1.0 MHz and 100 MHz US beams in the far-field are 0.75 mm and 7.5 μm respectively.
While the spatial resolution of US increases with frequency, power loss due to absorption and scattering of US by tissues also becomes more significant as frequency increases.
In other words, higher frequencies of US can provide fine spatial resolutions, but are less capable of being transmitted through tissues.
For the transcranial modulation of human brain circuits, <0.7 MHz US has been deemed particularly useful [11••, 12, 13•] and up to 5 MHz has been used to modulate the intact brain circuits of mice with thinner skulls. When US transmission through a skull is not required, frequencies up to 43 MHz have been used to modulate neuronal activity .
While the acoustic frequency of US used determines the spatial resolution, the acoustic intensity and exposure times are major factors in determining the dominate bioeffects (thermal or mechanical) on cells and tissues.
Heating tissues or ablating diseased brain circuits for therapeutic purposes is generally performed using continuous wave HIFU at intensities greater than 200 W/cm2.
In contrast to HIFU, low-intensity US (0.5–100 W/cm2) delivered in a pulsed mode for brief periods of time are less likely to produce tissue heating, but can still be focused through the skull and other tissues to produce prominent mechanical bioeffects on cells [3, 11••].
The peak acoustic intensity recommended for diagnostic US imaging is 190 W/cm2, which is higher than what has been shown capable of non-thermally modulating neural activity over the past decade [7•].
Ultrasonic modulation of central nervous system activity
Although thermal neuromodulation by HIFU provides immense clinical opportunities, the margins of safety are too narrow for widespread use in neuroscience.
To address these thermal limitations, the influence of low-intensity US on brain circuits began to draw consideration. Subsequently, it was shown that low-intensity US can directly stimulate action potentials and synaptic transmission in hippocampal slices .
Further, evidence showed these effects were partially mediated by the activation of voltage-gated sodium and calcium channels .
These observations inspired numerous studies over the past decade that have explored the effects of low-intensity US on CNS activity [7•].
Following the in vitro findings that low-intensity US could modulate CNS circuits, Tufail and colleagues (2010, 2011) described methods using low-intensity transcranial US for conducting in vivo stimulation of mouse motor cortex and hippocampus , as well as for rapidly attenuating seizure activity in mice .
Several others have since shown that the mechanical (non-thermal) bioeffects of low-intensity US (<100 W/cm2) can safely modulate the activity of intact cortical, thalamic, and hippocampal circuits in mice and rats [19, 20, 21, 22], rabbits , sheep , and pigs [25••].
Other studies have continued to provide insights into the acute effects of US on brain activity across a range of parameters while accumulating various safety observations.
The acute modulation of CNS activity by low-intensity US was shown safe in numerous animal models, which helped pave the way for conducting more recent human research studies.
Since it had become clear from clinical HIFU studies that US can be transmitted and focused across the skull into discrete brain regions, investigators began to further study whether low-intensity transcranial focused ultrasound could be used for non-invasive neuromodulation in non-human primates and humans.
In non-human primates, it was first demonstrated that transcranial low-intensity focused ultrasound (LIFU) can evoke visuomotor behaviors when targeted to frontal eye field regions of cortex .
More recently this work has been expanded to show that transcranial LIFU can stimulate individual cortical neurons in awake behaving macaques .
The authors demonstrated that a 0.5 MHz transcranial LIFU beam, having a lateral spatial resolution of about 5 mm and an axial resolution of about 18 mm, targeted to the somatosensory cortex at S1 can focally suppressed evoked EEG activity and produced a functional enhancement in somatosensory discrimination thresholds [11••].
While these studies point to a promising future, more studies are required to fully understand the safety and efficacy of focused US for acute applications in the brain, as well as to define the safety envelop for emerging chronic applications.
Technological advances in the field have been made by developing and demonstrating LIFU targeting methods that account for individualized variations in anatomy.
For example, one recent study developed realistic models using individualized measures of skull density and brain anatomy, so LIFU beams could be accurately delivered to specific regions of somatosensory cortex.
Using these methods, Lee and colleagues (2015) first showed that transcranial LIFU (0.25 MHz, 3 W/cm2) targeted to S1 of human volunteers can directly stimulate and evoke somatosensory potentials .
A particularly unique observation in these studies was that LIFU targeted to S1 could elicit different thermal/mechanical/pain sensations in the hands and fingers of volunteers in the absence of peripheral stimuli .
More recently the authors extended these by showing that transcranial LIFU (0.27 MHz, 16.6 W/cm2) targeted the primary visual cortex can stimulate visual sensations and evoke sensory potentials in different visual fields of humans as indicated by fMRI BOLD responses [13•].
Efforts to replicate these findings are under way and will help determine optimal parameters for stimulating or suppressing activity in various brain regions of humans.
With the continued safety observations and the further refinement of focusing methods aiming to decrease the costs of equipment and complexity of procedures required, transcranial LIFU can support several unique approaches to functional brain mapping as discussed further below.
Because US is compatible with EEG, MRI and other standard neurophysiological assessments, the use of focused US for high resolution, non-invasive brain mapping represents a potentially transformative opportunity.
Dallapiazza and colleagues (2017) showed that pulsed LIFU (0.22–1.14 MHz, 25–30 W/cm2) can be transmitted to subnuclei of the pig thalamus to functionally modulate somatosensory evoked potentials induced by the stimulation of different peripheral nerves [25••].
Data from MR-thermometry combined with these neurophysiological observations confirmed LIFU can focally (1.14 MHz focal volume = 1 mm × 1 mm × 3 mm) modulate deep-brain activity for functional circuit mapping without causing tissue heating (Figure 2) [25••].
These observations are a critical step towards realizing the full potential of using LIFU to clinically map and functionally validate brain targets prior to DBS and other neurosurgical interventions.
Excitingly, such non-invasive deep-brain mapping methods for basic research applications is also becoming a realistic possibility since it was recently demonstrated that transcranial LIFU can modulate the thalamus of healthy humans [28••].
Other important technical advances highlight how the physics of US provide unique capabilities for interfacing with central nervous system circuits.
As mentioned previously, <0.7 MHz US is used for transcranial applications, but higher frequencies can be readily used when skull transmission is not required.
The highest spatial resolution achieved for ultrasonic neurostimulation to date is 90 μm, which was demonstrated using 43 MHz US to stimulate single neurons in salamander retina .
Interestingly, LIFU stimulated responses in retinal neurons faster than light stimulation because endogenous phototransduction cascades are bypassed when acoustic pressure was used as a stimulus (Figure 3) . In advanced embodiments using LIFU for retinal stimulation, efforts to develop acoustic retinal prosthetics that project ultrasonic holograms onto the retina for neurostimulation have begun (Figure 3) .
Engineering an acoustic retinal prosthetic presents several technical and intellectual challenges ahead.
However, proving out such an application would clearly validate the utility of ultrasound neuromodulation for advanced neural interfaces.
Modulation of peripheral nerve activity by ultrasound
In addition to the modulation brain activity, US has also been shown capable of differentially modulating the activity of peripheral nerves in a variety of in vitro and in vivo experimental models (for reviews see Refs. [7•, 16, 30, 31]).
Studies spanning several decades have shown US can modulate peripheral nerve activity for time periods lasting from milliseconds to days through different mechanisms depending on the intensity, frequency, and exposure times implemented. Some of the most heroic studies investigating the influence of focused US on peripheral activity have been conducted by Gavrilov and colleagues [32, 33].
In other investigations, electrophysiological recordings from frog sciatic nerves , crab leg nerves , earthworm giant axons  and others [7•, 31] have shown that US can reversibly modulate neural activity by exerting non-thermal actions.
There are several considerations that need to be highlighted when discussing peripheral modulation by US.
For example, somatosensory receptors naturally encoding mechanical stimuli are responsive to US.
In fact, studies have shown low-intensity US delivered to the hands of humans can differentially activate peripheral nerve structures and produce EEG, as well as fMRI BOLD activity patterns similar to those obtained using more conventional somatosensory stimulation methods .
Thus, one must be able to distinguish direct effects on peripheral nerve fibers from those on somatosensory system receptors when transmitting US through the skin. Clear demonstrations that peripheral LIFU (1.1 MHz, 14–93 W/cm2) can modulate the rat cervical vagus nerve has opened several therapeutic possibilities for exploration .
In other circuits, modulation of the tibial nerve activity by low-intensity US has been shown to affect rat micturition reflexes paving the way towards the development of ultrasonic devices for controlling bladder function .
Other studies directly measuring nerve responsivity to US indicate peripheral UNMOD will have clinical applications in neuromodulation and bioelectronic medicine .
Other potential clinical applications may support neuro-rehabilitation therapies or have implications for advanced prosthetics because low-intensity US has been shown to enhance nerve regeneration following nerve injury  and nerve grafts in rats .
Increased research into the utility and application of UNMOD for modulating the activity and plasticity of the peripheral nerves will continue to reveal expanded options where US can have significant impacts on medicine and neurotechnology beyond imaging.
Some recent studies have begun to more carefully examine how US may be acting on peripheral nerves to modulate activity.
A recent study demonstrating the modulation of crab leg nerves by US seem to indicate that cavitation may be a mechanism in the periphery .
These observations may not reflect natural conditions since it can be incredibly difficult to control cavitation in vitro.
In fact, evidence from others indicates cavitation is not a predominant mechanism when stimulating mammalian peripheral nerves in vivo . This elegant study recently demonstrated that focused US can robustly stimulate rat sciatic nerves in a manner similar to electrical stimulation .
Matthew and colleagues (2018) further demonstrated that brief (0.8–10 ms) pulses of 3.5 MHz focused US stimulate peripheral nerve activity in a manner that indicates acoustic radiation force is a likely mechanism of action.
As discussed below, other recent evidence also indicates the influence of radiation forces are likely involved mechanisms.
While it may not be necessary to fully understand the mechanisms of action before implementing the basic methods, it will certainly advance our ability to use UNMOD once we uncover how it works.
Mechanisms of action
Considering the physical properties of neurons and their circuits, there are several ways in which US may act to influence their electrical activity.
Further complicating matters, the interactions of US with fluids including biological tissues is complex.
One straight-forward possibility however is that mechanical forces exerted by the acoustic pressure of US act on mechano-sensitive ion channels to alter neuronal activity (Figure 1).
Initial data in support of this hypothesis came from observations that US can stimulate brain activity through a non-thermal mechanism involving the activation of voltage-gated sodium channels and calcium transients as previously mentioned .
Numerous studies have confirmed that LIFU can modulate neuronal activity without causing significant tissue heating [7•].
Due to the experimental approaches used in these studies to assay activity however, whether the effects of US involve the direct mechanical modulation of ion channels has remained obscure until recently.
Mechanistic investigations have indeed shown that LIFU (10 MHz, <10 W/cm2) can modulate the activity of voltage-gated sodium channels (NaV1.5) and two-pore-domain potassium channels (TREK-1, TREK-2, and TRAAK) in xenopus oocytes .
Although demonstrated under unique conditions containing exogenous microbubbles, sonogenetic methods of activating TRP-4 channels in Caenorhabditis elegans has been shown a viable of regulating neuronal activity .
More convincing empirical evidence has recently shown that LIFU acts in a mechanical manner to modulate the activity of ion channels and neuronal activity .
Kubanek and colleagues (2018) conducted an insightful study in which they knocked out thermosensitive ion channels in C. elegans and found this did not affect behavioral responses to LIFU.
When the authors knocked out mechanosensitive ion channels however, LIFU responses were abolished.
Further, altering LIFU parameters to accentuate acoustic radiation forces produced by US elicited more robust responses .
These observations provide additional support to the hypothesis that LIFU acts, in part, by exerting mechanical actions through radiation forces on native ion channels.
Other complex physical mechanisms may also be involved in UNMOD. For example, it has been hypothesized that mechano-electric effects underlie the influence of LIFU on neuronal activity.
More specifically it is believed that LIFU can induce the formation of bilayer sonophores, which are small regions of phospholipid membrane that experience expansions and contractions .
These bilayer sonophores (microscopic membrane deformations) produced by US could theoretically generate capacitive displacement currents leading to charge build-up occurring over the course of tens of milliseconds (Figure 1) [47, 48].
Computational models incorporating these basic mechanisms have describes in some cases how US differentially affects neural activity depending on several factors including the types of ion channels expressed, the targeted neurons, and the duty cycle of the UNMOD waveform used .
Whether or not these models continue to hold up to empirically obtained physiological observations remains to be determined.
The possibility that such mechanisms may remain to be uncovered however, does indeed indicate that we should not exclude non-damaging cavitation as a putative mechanism of action.
Beyond direct effects on electrical activity, low-intensity US has been shown to modulate the activity of neurotrophic factors that could produce secondary effects on neural activity and plasticity .
Therefore, more studies are needed to refine our working models of how low-intensity US affects neuronal function versus activity.
Large cross-disciplinary efforts aimed at solving these issues are likely to reveal some completely novel information about how mechanical forces act to regulate neuronal activity and plasticity.
Sreekanth Chalasani – The Conversation
The image is adapted from The Conversation news release.