Towards an ion-channel-centric approach to ultrasound neuromodulation

Abstract

Ultrasound neuromodulation is a promising technology that could revolutionize study and treatment of brain conditions ranging from mood disorders to Alzheimer’s disease and stroke. An understanding of how ultrasound directly modulates specific ion channels could provide a roadmap for targeting specific neurological circuits and achieving desired neurophysiological outcomes. Although experimental challenges make it difficult to unambiguously identify which ion channels are sensitive to ultrasound in vivo, recent progress indicates that there are likely several different ion channels involved, including members of the K2P, Piezo, and TRP channel families. A recent result linking TRPM2 channels in the hypothalamus to induction of torpor by ultrasound in rodents demonstrates the feasibility of targeting a specific ion channel in a specific population of neurons.

Introduction:

Rationale for an ion-channel-centric approach to ultrasound neuromodulation

Ultrasound neuromodulation, a promising and potentially transformative new technology, compares favorably with other neuromodulation techniques, such as transcranial magnetic stimulation, direct electrical current stimulation, and deep brain stimulation in terms of its spatial resolution, depth penetration, and non-invasive nature [1,2]. Another important distinction between ultrasound neuromodulation and these more established techniques is that ultrasound does not directly induce electric fields and currents in the brain. Instead, ultrasound is thought to modulate neural activity by activating ion channels, the brain’s endogenous current-generating machinery. Ion channels are the basis for all electrical signaling in the brain [3], regulating the movement of charged ions across cell membranes in response to stimuli such as membrane voltage, binding of neurotransmitters, changes in local intracellular calcium concentration, changes in temperature, and mechanical force [4]. Together, these properties allow single neurons to accomplish incredibly complex feats of information processing. Differences in the expression patterns of ion channels across different neural cell types and brain regions likely contribute to cell-type- and brain-region-dependent effects of ultrasound, which need to be better understood as ultrasound neuromodulation progresses. If we can identify exactly how these ion channels are tuned to respond to ultrasound, and which ones are most sensitive, then it is possible to envision a future in which knowledge of the role of different ion channels in regulating neural activity in the context of large-scale brain dynamics allows us to target specific ion channels with ultrasound to achieve a desired neurophysiological outcome.

Experimental approaches for identifying ultrasound-responsive ion channels

Ultrasound could in theory modulate ion channel activity via different physical mechanisms. Mechanical effects (through acoustic pressure or acoustic radiation force) are generally considered the leading hypothesis, although thermal mechanisms (through ultrasound absorption) are also plausible in some contexts [1,5]. These hypotheses have motivated studies that evaluate the effect of ultrasound on ion channels known to be particularly sensitive to thermal and mechanical effects, as summarized in recent reviews [6–8]. Here, we discuss just a few examples of these studies, with the goal of providing context for evaluating past and future studies aimed at identifying ion channel targets of ultrasound and, where appropriate, highlighting ways in which our knowledge of the physiological functions of these channels can guide the design of ultrasound neuromodulation studies.

Approaches for identifying ion channel targets of ultrasound can be divided into two broad categories: direct testing in which the ion channel of interest is studied in a heterologous expression system, and indirect testing using genetic or pharmacological manipulation in a native context, which could range in complexity from in vitro cultured neurons to the in vivo brain in an animal model. Each approach has strengths and weaknesses. While in vivo systems are most relevant to clinical applications, it is challenging to resolve whether a given channel responds directly to ultrasound in an intact brain where neural activity is regulated by complex interactions among dozens of different types of ion channels. Since nearly all the ion channels expressed by a neuron play a role in regulating membrane potential dynamics, and interactions between voltage-dependent ion channels are highly non-linear, such that small perturbations can have large and sometimes counterintuitive effects, manipulation of an ion channel may influence ultrasound neuromodulation even if the channel itself is not directly affected by ultrasound. Heterologous expression systems allow ion channel physiologists to circumvent the complexities of native systems by studying ion channels in isolation. (While all cellular heterologous expression systems have some level of endogenous background ion channel expression, it is normally possibly to achieve current levels for heterologously expressed channels that are orders of magnitude greater than endogenous currents. However, if the heterologous channel expresses at very low level, confusion can occur between the channel of interest and endogenous channels. In this context, it is worth noting that the mechanonsensitive and—at least in some contexts—ultrasound-sensitive channel Piezo1 is endogenously present in the popular HEK cell heterologous expression system [9]). The main disadvantage of heterologous expression experiments is that the essentially non-physiological context (typically an isolated cell adhered to a rigid substrate and surrounded by a small volume of aqueous solution exposed to the atmosphere) may introduce ultrasound responsiveness that is not representative of effects in vivo. In general, any part of the experimental system that absorbs or reflects ultrasound may alter the response, and unless the investigator is attentive to all sources of artifact, results in heterologous systems may not be relevant to the in vivo situation. The review by Snehota et al. [10] provides a helpful summary of sources of experimental artifacts in in vitro studies of biological effects of ultrasound.

A model to strive for

The reader may feel like we have presented a catch-22: given the combined limitations of experiments in heterologous expression systems and native contexts, how can we hope to make progress? We suggest that confidence in studies in the native context can be increased if the observation of an effect on ultrasound neuromodulation is supplemented by insight connecting the effects of ultrasound with the known physiology of the ion channel in question. An impressive example of this was presented in the invertebrate model C. elegans [11]. The animal’s behavioral response to ultrasound was eliminated in the absence of MEC-4, a member of the eNaC/ASIC ion channel family expressed in C. elegans touch receptor neurons. In addition, the dependence of the ultrasound response on the stimulus parameters (varying either pulse repetition frequency or duty cycle while keeping the total acoustic energy constant) could be accurately predicted by a previously developed biophysical model of the frequency response of the C. elegans touch receptor neuron [12]. Critically, the touch receptor model had been developed prior to the ultrasound study and no ad hoc adjustment of its parameters was needed to get an excellent fit to the experimental data. While our understanding of the mechanics of mammalian neurons and neural tissue is not sufficiently advanced for this feat to be reproduced in the mammalian brain, it is a model to strive for. Another obstacle to achieving a similar integration of ultrasound neuromodulation and existing physiological knowledge in mammalian systems is that for many candidate ultrasound-sensitive channels, knowledge of the channel’s physiological function in the brain is limited. This is not meant to imply that channels without known physiological functions therefore lack important physiological functions. Lack of known function may reflect a lack of appropriate experimental tools, or the channel may be relatively recently discovered or simply understudied. However, limited knowledge of a channel’s physiological role does complicate the interpretation of knockout and knockdown studies.

Despite the limitations in our understanding of the mammalian brain and the caveats regarding experimental approaches outlined above, recent results have yielded insights into the role of various ion channels in ultrasound neuromodulation. In the next sections, we highlight a few of these.

TREK and TRAAK K2P channels – mechanosensitive channels that modulate action-potential firing

K2P channels are a family of relatively voltage-independent potassium channels that are regulated by diverse chemical and physical factors [13]. The K2P subfamily members TREK and TRAAK are attractive targets for ultrasound neuromodulation due to their exquisite sensitivity to mechanical and thermal stimuli. Another reason these channels are a compelling target for ultrasound neuromodulation is that their physiological role in neurons is relatively well understood [14]. TREK and TRAAK are broadly expressed at the nodes of Ranvier in the mammalian central nervous system (CNS), where they play a critical role in regulating action potential regeneration [15,16] and therefore synaptic connectivity between distant neurons. On the scale of the whole brain, myelinated axons where K2P channels are expressed form white matter tracts, which can be visualized in human subjects using MRI tractography [17]. As many applications of transcranial ultrasound in human subjects are currently performed in the context of an MRI scanner, a logical extension of current ultrasound technology would be to target white matter tracts for neuromodulation.

Independent studies using patch-clamp recording indicate that TREK and TRAAK channels are potential targets for an ion-channel-centric approach to ultrasound neuromodulation [18–20]. Sorum et al. showed that TRAAK channels can be activated by ultrasound in excised membrane patches from Xenopus oocytes and lipid vesicles (a simplified experimental preparation where purified protein is reconstituted into pure lipid membranes), and that ultrasound can activate TRAAK channels overexpressed in neurons in brain slices [19,20]. Prieto et al. used patch-clamp recording of pyramidal neurons in acute hippocampal brain slices to demonstrate that the effects of ultrasound on action potential firing could be entirely explained by activation of native TREK and TRAAK channels [18].

The results by Prieto et al. are striking because they provide a mechanistic basis for contrasting (excitatory vs. inhibitory) effects of ultrasound [18]. Several previous studies have shown that specific sets of ultrasound stimulation parameters yield excitatory effects while others produce inhibitory effects [21–23], but the underlying mechanisms are unknown. In Prieto et al., it was observed that identical ultrasound parameters can yield either excitatory or inhibitory effects, inhibiting firing when a neuron is firing at a relatively low rate but promoting firing when it is firing at a relatively high rate [18]. Activation of K2P channels provides a mechanistic basis for these contrasting effects: when neurons are in a low firing-rate state the most pertinent effect of K2P activation is to increase the amount of time it takes for a neuron to reach action potential threshold, thus causing inhibition; in contrast, at high firing rates neurons are constantly in a near-threshold state and the firing rate is limited by the recovery of sodium channels from inactivation, which is accelerated by K2P channel activation, causing excitation. This result exemplifies the challenges of interpreting experiments that manipulate ion channels in the native context, as illustrated in Fig. 1. Clearly, any ion channel manipulation that shifts basal firing rate towards the left or right could produce a change in the response to ultrasound. If the experimenter is particularly “lucky”, the manipulation might even land at the precise inflection point between inhibitory and excitatory effects (vertical purple line), leading to the incorrect conclusion that the channel in question is entirely responsible for the effects of ultrasound. Moving beyond single neurons to large populations and brain-wide networks, the difficulty of interpretation increases.

Figure 1. Experimental data and hypothetical in vivo knockout experiments illustrating the challenge in interpreting effects of ion channel manipulations on ultrasound neuromodulation in the native context.

The open and filled circles connected by dashed lines show data from an ultrasound neuromodulation experiment in rat hippocampal neurons. At low input current, ultrasound is inhibitory, while at high input current it is excitatory. (Data from Prieto et al. [18].) Varying input current in the experiment can be taken as an approximation for channel knockouts that modulate firing rate in vivo. Now consider a hypothetical in vivo experiment involving knockout of ion channels that do not respond to ultrasound but do regulate firing rate. In the WT condition (vertical orange line) ultrasound increases the measured neural activity. Knockout of “channel-1” (vertical green line) has an effect analogous to reducing the input current. Here, ultrasound still increases neural activity, but less so than for WT. Knockout of “channel-2” (vertical purple line) is analogous to a larger reduction in input current, resulting in an apparent loss of sensitivity to ultrasound. Considering the channel-2 knockout experiment in isolation, it could be incorrectly concluded that ultrasound neuromodulation is wholly mediated via channel-2.

The state-dependent effects of ultrasound mediated by K2P channels at the single-neuron level may be related to state-dependent effects of ultrasound at the population level [24–28]. The presence of different patterns of large-scale, synchronized, rhythmic neural activity including sharp wave ripples, theta rhythm, and gamma rhythm, as well as the timing of the ultrasound stimulus relative to the phase of these rhythms has been shown to affect the response to ultrasound [24,27,28]. Since large-scale rhythmic activity can be recorded in human subjects using non-invasive techniques such as EEG, and the excitability of different subtypes of neurons is known to increase and decrease relative to the phase of this rhythmic activity, this presents a possibility for “closed-loop” ion-channel-centric neuromodulation strategies.

Piezo channels – widespread mechanotransduction channels

The Piezo channels Piezo1 and Piezo2 are mechanically activated cation channels that regulate cellular mechanotransduction throughout non-brain mammalian tissues [29], so it is logical to hypothesize that they may also be involved in ultrasound neuromodulation. Indeed, several heterologous expression studies have demonstrated activation of Piezo channels by ultrasound [29–32]. In the more native-like setting of cultured neurons, one study found that knockdown of Piezo1 significantly reduced (but did not eliminate) sensitivity to ultrasound [33] while another reported no effect of Piezo1 knockdown [34], though the sample size in the latter study was relatively small (n=3). In vivo, a recent study used calcium imaging to provide the first evidence that Piezo1 channels contribute to ultrasound neuromodulation in the CNS [35], with knockout of Piezo1 channels diminishing but not eliminating the response to ultrasound.

How might the conclusion that Piezo channels are activated by ultrasound be used to guide ultrasound neuromodulation? Surprisingly little is known about the physiological function of Piezo channels in the CNS given the intense interest in these channels, although they are known to be important in brain development and in calcium signaling in glia, and they may play a role in Alzheimer’s disease [36]. Nonetheless, Zhu et al. [35] found that Piezo1 expression was enhanced in the central amygdala and other brain regions relative to the cortex, and they suggest that regions showing higher Piezo1 expression may be more sensitive to ultrasound, although additional data are needed to support this hypothesis. If the correlation between Piezo1 expression and sensitivity to ultrasound is robust, it may be possible to target ultrasound to Piezo1-enriched regions to achieve specific outcomes. Piezo2, the Piezo subtype predominant in sensory neurons might also be involved in ultrasound neuromodulation in the CNS, although this has not to our knowledge been investigated in vivo. Wang and Hamill [37] reported that Piezo2 is expressed in neocortical and hippocampal pyramidal neurons and cerebellar Purkinje cells and proposed that activation of Piezo2 by changes in intracranial pressure associated with cardiac cycles and respiration can synchronize neural activity. Although highly speculative, this effect could provide a non-synaptic mechanism for synchronizing activity across brain regions, and a potential means for modulating this synchrony by timing ultrasound with intracranial pressure changes. This idea could of course also apply to Piezo1 and other mechanosensitive channels.

Piezo channels are also an attractive modulation target for non-neuronal cells and for neurons outside of the central nervous system. A recent study highlighted a role for Piezo1 channels in glia in the maintenance of long-term potentiation and in neurogenesis in the dentate gyrus [38]. This raises the exciting possibility that ultrasound can activate cellular signaling pathways involved in neural plasticity leading to long-lasting changes in neural connectivity and potential “rewiring” of dysfunctional neural networks. (While Zhu et al. [35] reported that glial Piezo channels played no role in ultrasound neuromodulation, their experiments were designed to detect acute activation rather than long-lasting synaptic plasticity.) It is interesting to speculate that Piezo1-dependent synaptic plasticity may play a role in the “offline” effects of ultrasound that have been shown to persist for up to 30 minutes after ultrasound application in non-human primates [39] and humans [40]. Piezo1 channels have also been shown to function as mechanoreceptors in capillary endothelial cells in the brain, suggesting a potential role for Piezo1 in neurovascular coupling [41]. In peripheral mechanoreceptor neurons of mice, where Piezo2 channels are known to play a major role in mechanotransduction, knockout of Piezo2 decreased the response to sub-millisecond, high-intensity ultrasound pulses [42].

TRP channels

TRP channels are calcium-permeable cation channels that, like K2P channels, respond to a wide variety of stimuli including thermal and mechanical stimuli [43]. The precise function of TRP channels in the brain is not known but their role in calcium ion homeostasis is important for neurological activity[44,45]. In a heterologous expression system, Oh et al. [46] showed that TRPA1 channels can be activated by ultrasound. In a physiological context, Yoo et al. [34] found that knockdown or inhibition of the TRPM7, TRPP1, TRPP2, and TRPM4, but not TRPC1, subtypes of TRP channel all significantly reduced the response to ultrasound in cultured neurons. The finding that several TRP channel subtypes all make significant contributions to ultrasound neuromodulation is consistent with the complexity of ion channel interactions in the native context. In vivo, ultrasound stimulation of rodent tail movement was reduced in TRPA1 knockout animals [46]. Combining this result with additional results in vitro, Oh et al. developed a model for ultrasound neuromodulation in which calcium influx trough TRPA1 channels in astrocytes triggers glutamate release through BEST1 channels, leading to activation of glutamate receptor ion channels in neurons [46]. Yang et al. [47] demonstrated that targeting the hypothalamus with ultrasound can induce a torpor-like state in rodents, and that this effect could be attributed to activation of TRPM2 channels in a specific subpopulation of ultrasound responsive cells in the preoptic area of the hypothalamus. This work is especially significant as it represents perhaps the first instance where the expression of a specific ion channel subtype in a specific brain region can be linked to a specific physiological response to ultrasound. Finally, TRPV4 channels could be a potential non-neural ion channel target for ultrasound based on their potential physiological role as baroreceptor channels (as proposed for Piezo2 channels [37]) in astrocytes [48].

Conclusion and outlook

Although there are several outstanding challenges to developing an ion-channel-centric perspective on ultrasound neuromodulation, collectively there is strong evidence to support the idea that numerous mechanosensitive ion channels can be activated by ultrasound in vivo. It is striking that, in all the ion channel knockout/knockdown studies in mammalian systems reviewed here, the effect of deleting the channel was a reduction rather than a complete elimination of the response to ultrasound. This indicates that ultrasound neuromodulation likely depends on several channels that are sensitive to ultrasound to varying degrees, with the relative importance of these channels depending on cell type, brain region, activity state, and other variables. An outstanding problem that we have not had space to address here is understanding how acoustic pressure or radiation force are translated into the nanoscale forces that regulate mechanosensitive ion channels [49]. Ultimately, developing an ion channel-centric perspective on ultrasound neuromodulation may require studying ultrasound neuromodulation from the perspective of mechanobiology [50–54]. As we stated at the start of this review, a unique feature of ultrasound among neuromodulation techniques is that it is able to directly activate ion channels. This feature may turn out to be one of its greatest strengths. The ability to target various ion channels, with all of their diverse physiological functions in both neural and non-neural cells, may give ultrasound a level of versatility and flexibility unmatched by other neuromodulation techniques.