Direct activation of zebrafish neurons by ultrasonic stimulation revealed by whole CNS calcium imaging

The analysis was performed on zebrafish larvae at three different (3, 4 and 5) days post-fertilization (dpf). Unless otherwise stated, all results refer to 5 dpf larvae. Adult and larval zebrafish were maintained for breeding at 28 °C on a 14/10 h light/dark cycle according to standard procedure [43]. Larvae were raised up to 5 dpf in fish water (150 mg L⁻¹ Instant Ocean, 6.9 mg L⁻¹ NaH₂PO₄, 12.5 mg L⁻¹ Na₂HPO₄, pH 7.2) in a Petri dish kept at 28 °C. We used 3 to 5 dpf transgenic Tg(elavl3:GCaMP6s) zebrafish larvae [41, 42] in homozygous albino background to avoid the presence of skin pigments. Following [42], each larva was then transferred into an Eppendorf tube, containing in 1.5% w/v low gelling temperature agarose (A9414, Sigma) dissolved in fish water kept at 38 °C, and introduced into a glass capillary tube (O.D. 1.5 mm) with a pipette. After gel solidification, the larva was extruded from the capillary, laid in the center of the bottom of a small modified Petri dish (5 cm diameter with the central part of 2 cm diameter replaced by a 25 µm thin polystyrene membrane (Goodfellow, Hungtinton, Cambridge, UK)) and this position was blocked with a drop of melted agarose. After solidification the dish was filled with fish water. Fish rising and experiments were carried in accordance with European and Italian law on animal experimentation (D.L. 4 March 2014, n.26), under authorization No. 407/2015-PR from the Italian Ministry of Health.

US stimulation was applied by placing a 44 mm diameter PZT (Lead Zirconate Titanate) unfocused transducer (Precision Acoustics LTD, Dorchester, UK, S/N PA886, central frequency equal to 490 kHz, −3 dB bandwidth equal to 160 kHz, −6 dB X-profile beamwidth 19.88 mm, −6 dB Y-profile beamwidth 18.66 mm) at a distance of 165 mm from the animal specimen. The transducer was driven by a waveform generator (Agilent33220A Keysight Technologies, Santa Rosa, CA, USA) in series with a 50 dB gain radio frequency power amplifier (240 l, Electronics & Innovation, Rochester, NY, USA) and it was immersed in the polymethyl methacrylate (PMMA) tank. The US transducer was characterized in free field conditions both in terms of ultrasonic pressure field mapping and intensity vs. driven voltage calibration. The pressure signals were detected by a 2 mm polyvinylidene fluoride needle hydrophone (Precision Acoustics LTD, Dorchester, UK, model number: NH2000) at different locations, using a three-axis step-by-step motorized positioning frame (XYZ BiSlide, Velmex, Bloomfield, NY, USA), thereby producing maps that provide information on the geometry of the generated acoustic field. A dedicated LabVIEW program (National Instruments, Austin, TX, USA) allowed synchronization between the wave generator, motors and signal acquisition from an oscilloscope (7034 B, InfiniiVision, Agilent Technologies). The root mean square pressure (P_rms) and the spatial peak pulse average intensity (I_sppa = P_rms ²/(ρ c)) were evaluated at the experimental distance of 165 mm at different driving voltages, each measured at the output of the power amplifier. The medium density $\rho \, = \,1000{\text{ }}\text{Kg}\,{\text{m}^{ - 3}}$ and the wave speed $c\, = \,1484{\text{ }}\text{m}\,{\text{s}^{ - 1}}$ are approximated to the density and speed of sound of water, respectively. The spatial peak temporal average intensity (I_spta) was defined as

$$\begin{equation}{I_{spta}} = {\text{ }}{I_{sppa}}*DC = \frac{{P_{rms}^2}}{{\rho c}}*DC\end{equation} \tag{ 1 }$$

where DC is the duty cycle used in the stimulation protocol. Finally, in order to consider possible acoustic reflections and attenuations phenomena, additional intensity measurements were performed by positioning the hydrophone tip inside the Petri dish used during experiments (see figure S3 (available online at https://stacks.iop.org/JNE/17/056033/mmedia) of [44] for the results of the hydrophone measurements). These measurements yielded the I_spta calibration reported in figure 1.

Zoom In Zoom Out Reset image size

Pulsed ultrasound stimulations of 490 kHz were delivered in blocks of three stimuli of 100 ms each, with a duty cycle of 25%, 50%, 75% or 100%. A stimulus duration of 300 ms with a pulse repetition frequency of 10 Hz was chosen. Pressure amplitude varied from 8 to 40 kPa in nine discrete steps. The I_spta, therefore, varied from 1.5 to 120 mW cm⁻². Each recording session consisted of three identical US stimulation windows delivered every 15 s. The overall acquisition time for every recording session was fixed at 60 s. We tested 12 3 dpf, 10 4 dpf, and 8 5 dpf larvae. Each larva has undergone two sets of nine recording sessions, one for each pressure amplitude. For each pressure amplitude, the duty cycle was first set at either 25% or 50% and then increased to 100% or 75%.

Following the results of a previous work of our group [44], we analysed the responses as function of the US spatial-peak temporal-average intensity I_spta, (see equation (1) in section 2.2). To facilitate reading, we divided the stimulation acoustic intensity (I_spta, measured in mW/cm²) into five balanced groups ranging from very low to very high (figure 1(c)).

Imaging of the agarose restrained larvae was performed using a custom-made upright fluorescence microscope in epi configuration. Fluorescence was excited in wide-field approach using an LED with emission centered at 470 nm (M470L3, Thorlabs), followed by a tube lens of 200 mm focal length (AC508-200-A, Thorlabs) and by an excitation band-pass filter (FF01-469/35 nm, Semrock). The fluorescence signal was collected by an air objective (Nikon, 10x, NA 0.3, WD 16 mm). The detection path was implemented with a dichroic mirror (DC FF495-DI02, Semrock), an emission band-pass filter (FF03-525/50 nm, Semrock) and a tube lens of 200 mm focal length (AC508-200-A, Thorlabs) that created an image on a sCMOS camera (OrcaFlash 4.0 v3, Hamamatsu Photonics) with 16-bit dynamic range at a frame rate of 20 Hz.

2.4.1. X-Y drifts correction of calcium imaging videos

2.5.1. US shielded experiments

2.6.1. Relative fluorescence variation

As common practice in calcium imaging video analysis, the fluorescence dynamics was expressed as:

$$\begin{equation}\frac{{\Delta F}}{{{F_0}}}\left( t \right) = \frac{{F\left( t \right) - {F_0}\left( t \right)}}{{{F_0}\left( t \right)}}\end{equation} \tag{ 2 }$$

where F(t) is the fluorescence intensity and F₀(t) is the baseline fluorescence intensity at a given time instant t. To compute the baseline fluorescence, we used a running-window average of the 8th percentile of the fluorescence intensity F(t) [48]. We used a time window of 5 s, approximately 25 times larger than the decay time constant of the adopted calcium reporter [49]. (ΔF/F₀)_max was defined as the difference between the maximal value of ΔF/F₀ traces following US stimulation and the ΔF/F₀ value at stimulation onset.

Significant statistical differences in this value are assessed with Kruskal–Wallis test (p = 0.05) and multiple comparison have been accounted for by the Dunn approach. The dependence of the response by multiple factors and their interaction was assessed by two-way analysis of variance (ANOVA) [50]. The same analysis was performed for the latency of the response (i.e. the time from US stimulus onset to maximal amplitude of ΔF/F₀).

2.6.2. Spots of maximal and minimal activation

We identified, separately for every recording, the spots of maximal and minimal activation in the zebrafish encephalon following US stimulation considering the pixels (1.6 × 1.6 μm²) whose summed fluorescence variation ΔF/F₀ after the stimulus trigger was greater/lower than the 99-/1- percentile of the global relative fluorescence. The global relative fluorescence was defined as:

$$\begin{equation}{\left( {\frac{{\Delta F}}{{{F_0}}}\left( t \right)} \right)_{global}} = \mathop \sum \limits_{k = 1}^3 \mathop \sum \limits_{t = t_k^*}^{t_k^* + 5s} \frac{{\Delta F}}{{{F_0}}}\left( t \right)\end{equation} \tag{ 3 }$$

where k is the US stimulus number (every recording consisted in three US stimuli), $t_k^*$ is the corresponding US stimulus trigger instant and $\frac{{{\mathrm{\Delta }}F}}{{{F_0}}}\left( t \right)$ is the relative fluorescence variation. After an imaging registration onto a common brain orientation step employing a roto-translation transformation, the binary maps of the spots obtained for different larvae with the same US intensity I_spta were summed and normalized to compute the spatial probability maps.

We observed simultaneously the response to US stimulation, measured as fluorescence variation (see Methods), in four well characterized regions of the CNS (figure 2(a)). Both region and intensity played a significant role in determining the response (two-way ANOVA, p < 0.001 for each factor, p < 0.001 for their interaction). The cerebellum displayed the strongest response ((ΔF/F₀)_max = 7.1% for very high stimulation, figure 2(b) bottom left) and the highest sensitivity to ultrasound stimulation, with significant differences between the response to most stimulus levels (figure S1). The optic tectum and the medulla displayed a lesser response ((ΔF/F₀)_max = 5.8% and 5.45% respectively for very high stimulation, figure 2(b) top and bottom right). The telencephalon displayed the weakest response ((ΔF/F₀)_max = 2.1% for very high stimulation, figure 2(b) top left) with significant differences only between high/very high and low/very low intensity (figure S1). We also identified for each stimulation pattern the spots of maximal and minimal activation. Briefly, we defined as spots of maximal (minimal) activation those calcium imaging video pixels whose fluorescence variation after stimulus trigger was greater (lower) than the 99- (1-) percentile of global relative fluorescence (see Methods for details). Coherently with previous results, maximal and minimal activation spots were located for all animals in the cerebellum and in the telencephalon, respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Animals were positioned in a Petri dish filled with fish water during imaging. The US stimulation produced non negligible movements in such media. Consequently, the observed response could be due to a combination of direct stimulation on the ionic channels [32], on the neuronal membranes [33, 34] and/or of the animal sensitivity to the US-induced water movements and sound perturbation at few Hz.

To elucidate how the calcium fluorescence activity patterns induced by the sonication could be accounted for by these two factors, we performed then US-shielded control experiments (see Methods) in which the larval brain was not directly hit by the US, but the surrounding watery media moved due to the incoming stimulation sound waves. We used the same set of US stimulations used in the unshielded animals and then we compared responses associated to the same level of sensory stimulation (see Methods). This allowed to estimate the response due to neuronal activation, while removing the contribution of sensory stimulation.

Responses to US stimulation were found to be much weaker in the US-shielded experiments as compared to the US-unshielded scenario and were elicited for a shorter time by a narrower set of stimulations (compare figures 3(a) and (b)). In both experimental conditions, however, the cerebellum was the most responsive area, and we therefore compared the maximal amplitude of the two responses for the same levels of water movement (see Methods). We found that for all levels the response was significantly stronger in the unshielded condition (p < 0.01) with an average of +199 ±94% (mean ± standard deviation) relative increase with respect to the shielded condition (figure 3(c)).

We analysed then the temporal evolution of the response over the whole CNS in the US-unshielded experiments. We observed qualitatively that the first response in every I_spta group arose close to the maximal activation spot within the cerebellum, to later spread to other areas while increasing in intensity within the cerebellum (figure 3(a) and supplementary movie 1). We hypothesized then that at least for low stimulation intensities only these spots are activated by US stimulation, and all other responses are due to propagation of the activity. To test quantitatively this hypothesis, we calculated the latency of the response (i.e. the time from US stimulus onset to maximal amplitude of ΔF/F₀) in each region, observing that it depended on both intensity and region (p ≪ 0.001, two-way ANOVA), but not on their interaction (p = 0.13, two-way ANOVA). We found that the spots of maximal activation were triggered significantly before all the other regions at very low intensity (figure 4(a) top left and S2). This suggests that activity in all other regions at least for this intensity is not directly elicited by US, but by inputs from cerebellar afferences. The latency for the maximal activation spots did not depend on the intensity of the stimulation and was on average 1.5 s (figure 4(b)). For the other areas the latency varied with the stimulation intensity. For higher intensities cerebellum, optic tectum and medulla displayed latencies closer to the one at the maximal activation spots (figures 4(a), (c), and S2), which is compatible with direct US activation at these intensities. The telencephalon, instead, displayed an average latency which was always significantly higher than that of the other areas (figure S2) and reached a minimum value of 3.7 s for the highest stimulation intensity (figure 4(a) bottom right). In particular, the region of minimal activation within the telencephalon showed for all intensities a low-amplitude response starting more than one second after the end of US stimulation (figure S3). This strongly suggest that the telencephalon is not activated directly by the US stimulation, but only by the propagation of activity from other areas (figure 4(c)).

Zoom In Zoom Out Reset image size

All results in the previous sections refer to 5 dpf larvae. As both single neuron properties and connectivity are expected to change during development [51, 52], progressively expanding the larval cognitive and processing skills, we probed whether the sensitivity to US was present since the first tested larval stage and if it showed an increasing trend along dpf. To this end, we performed a finer analysis of the spots of maximal and minimal activation across the three tested days post fertilization.

The spots of minimal activation were confined to the larval telencephalon for every applied stimulus parameter and animal development stage (figure 5(a), right columns). The location of the spots of maximal activation (see Methods) confirmed instead that the strongest sensitivity to the US was confined to the cerebellum for every dpf (figure 5(a), left columns). More precisely, maximal activation was stereotypically (across animals and recordings) located near the border between the cerebellum and the upper portion of the medulla. Of note, the sensitivity to US of these subregions grew with the dpf (figure 5(b)), confirming our expectation.

Zoom In Zoom Out Reset image size

The main aim of this work was to observe, isolate and characterize a direct neural response to US stimulation. By comparing neural responses in US-unshielded and US-shielded conditions, we were able to evaluate the extent of sensory contribution to such responses, and hence highlight the prominent role of direct response to US stimulation (figure 3). The entanglement between direct and sensory mediated neural response to US stimulation has been debated in several recent works, examining, in particular, the neuromodulation activity triggered by transcranial US brain stimulation. Wide-field calcium imaging of transgenic mice brains was used to observe the neuronal activation due to US stimulation on the visual cortex, at a central frequency of 500 kHz [28]. While the US focus was on the visual area, the first strong activation was observed on the auditory cortex, starting from the contralateral hemisphere, with a response similar to the one elicited by audible sounds. The excitation effect increased with the stimulation intensity in this region, but the activation spread onto a larger cortical area. This suggested that the observed neural activation was due to acoustic stimulation of the inner ear, rather than direct neural activation. Moreover, the response to both US and sound was attenuated in chemically deafened animals and disappeared when the transducer was not coupled to the head.

The companion study [20] investigated the effect of 220 kHz US stimulation on guinea pig brain, recording local field potentials from auditory cortex and inferior colliculus. Responses occurred whenever the transducer was in contact with the animal head or the eyeball. However, the first spike latency was shorter in the inferior colliculus, suggesting that the recorded signal in the auditory cortex should be the consequence of sensory stimulation. The response of the somatosensory cortex was largely independent from the transducer focus, suggesting that motor cortex elicitation might be caused by startle reflex. This US-induced cortical activation was completely abolished by auditory nerve transection and removal of cochlear fluids. Such results are in contrast with other works reporting the activation of a functional circuit associated to tactile sensation due to US stimulation in the somatosensory area [8, 12]. Moreover, other studies have found a significant effect of the ultrasound stimulation target location [53, 54].

These results are far from being conclusive. A work focusing on confounding activation of auditory pathways [55] observed auditory brainstem response that elicited EMG in forelimbs triceps in transgenic deaf mice. Another recent work observed US response in deep brain areas of macaques and their propagation to other brain regions [56]. They ruled out auditory stimulation confounding by an analysis of the correlation of such responses with auditory cortex activity. For zebrafish larvae there was no possible direct auditory confounding as the animal at this stage is not sensitive to such high frequency [57]. Nevertheless, auditory stimulation within its hearing range [58] may be produced by the pulsed stimulation sequence envelope that can be assimilated to a weak noisy sound source centered at 10 Hz. A recent hearing assessment study in larval zebrafish [47] has found that 20 Hz frequencies can be perceived with an approximately 100 times lower sensitivity than above 50 Hz. Consequently, we cannot exclude that zebrafish larvae are able to perceive sound even below 20 Hz, but with a further reduced sensitivity. Furthermore, we still had to investigate the possibility of mechanosensory confounding to tactile perception of US-induced waves in the water (figure 3(a)). Moreover, despite our efforts into immobilizing the animals during recording, we could not discard the possible contribution of a vestibular response in the unshielded experiments [46]. Responses due to mechanical (and possibly vestibular and auditory) stimulation in the US-shielded condition were found to be significantly smaller than those induced in the US-unshielded case for all intensities (figure 3(b)).

We have shown selective activation of zebrafish brain areas. Such a selectivity is not a complete novelty: for instance, in their seminal work [19] Yoo and colleagues observed that the US stimulation effect was stronger in motor than in visual cortex. In a more recent work [59], the mice awake cortex was stimulated while monitoring the hemodynamic change of the whole cortex due to the different stimulation protocols and observed local responses. However, to the best of our knowledge, our study is the first to exploit the simultaneous monitoring of the whole CNS to discriminate between direct, primary activation and secondary activation. In a much narrower context, propagation of US response was observed in experiments with stimulation of primary somatosensory cortex on monkeys [12]. It was indeed found that US stimulation not only activated the cortical area corresponding to the focus, but also spatially distant regions corresponding to the touch processing circuitry. The observed increase in response strength with dpf is consistent with the progressive development of the larval CNS with age.

The aim of our analysis of temporal evolution was to assess whether neurons in different regions were activated directly by US stimulation or were rather triggered by functionally connected regions. We noted that, as a consequence of US stimulation, the cerebellar area was the first one to respond. Right after cerebellar response, we observed a consistent activation of striped nuclei spanning the entire medulla down to the spinal cord (see figures 3(a), 4, S2 and supplementary movie 1), presumably signs of functional connection with a motor circuit. The telencephalon was the last activated area and showed a response so late (figure S3) to be consistent with an activity exclusively due to the propagation of the response induced by US in other areas of the CNS (figure 4(c)). Crucially, our work aims at discriminating between primary and secondary activation looking at the relative times of activation of the different areas rather than determining the absolute latency of the response, which is difficult to assess given the inherently slow dynamics of calcium imaging responses [28]. Indeed the latencies reported here are one order of magnitude larger than the ones previously reported for cellular response [44] or even motor activity following stimulation [31, 60].

US are known to interact with biological tissues through either thermal or non-thermal physical mechanisms. US-induced heat generation is dependent on several factors that can be controlled through proper choice of the stimulation parameters. The main factors contributing to the conversion of ultrasonic energy to heat are [61]: (i) the tissue properties (most and foremost the absorption coefficients and the impedance mismatch encountered by the ultrasonic waves); (ii) the US exposure parameters (mainly the frequency, the pressure amplitude, the pulse repetition frequency, and the pulse duration); (iii) the beam configuration. Considering these three factors in our experiments, we can rule out the recorded responses to be ascribable to significant thermal effects.

A recent modeling study [62], revealed, through retrospective temperature simulation, that thermal effects can be neglected in a large set of rodent setups. Only in one study [63], it was estimated a temperature increase as high as 7 °C due to thermal diffusion from the skull bone into the brain tissue; this was mainly due to the skull absorption coefficient [62], which in rodents is ten times higher than the soft brain tissue (α = 2.7 dB/cm/MHz^−1.18 in the skull, α = 0,21 dB/cm/MHz^−1.18). In our experiments, however, skull-induced heating is not present as in zebrafish skull bones develop and ossify long after the embryonic period. At larval stages, indeed, the craniofacial skeleton is primarily composed of cartilage elements often one or a few cell layers thick [64].

Furthermore, we choose to place the US transducer in water in order to minimize the heat produced by the probe itself when it is operated in air [65] and to provide thermal stabilization and water coupling of the ultrasound probe to the Petri dish.

As for the US stimulation parameters employed in our study, many studies [2, 61, 66–68] have pointed out that US exposure protocols consisting of low pressures (<0.6 MPa at focus), low frequencies (sub-MHz), short pulse durations (≤300 ms), and low acoustic intensities (in the 30–500 mW cm⁻² range) do not lead to thermal effects playing a prominent role in neuromodulation, with estimated temperature rise as low as hundredths of a degree. Additionally, we have adopted a pulsed exposure delivery mode of ultrasonic stimuli, that is known to further minimize the probability of tissue heating [38, 61, 69, 70].

Finally, we choose to employ an unfocused US transducer to stimulate the whole zebrafish encephalon. The absence of a small focal area furtherly minimizes the energy concentration and consequent heat production in the zebrafish encephalon.

Altogether these considerations can therefore rule out possible significant thermal effects in the fluorescence increase we recorded in the zebrafish encephalon.

We found that specific areas within the zebrafish cerebellum are highly sensitive to direct US stimulation, as pointed out by the spatial and temporal comparison between the neural responses with US-unshielded and US-shielded stimulation. Still, within the adopted experimental paradigm, it is not possible to completely rule out that the observed response contains a sensory-mediated contribution, originated from mechanical water motion and possibly vestibular and auditory perception. The current knowledge of the properties of zebrafish neurons does not allow us to understand what are the specific characteristics of the cerebellum that make it markedly sensitive to US. A full understanding of the biophysical dynamics leading to direct activation goes indeed beyond the scope of this work. Understanding these dynamics would be of paramount importance for a proper design of the stimulation, and even more important to assess the likelihood of the possibility to directly stimulate neurons also in other animals (see [44]) and in particular in humans for translational applications.