Differential effects of amplitude-modulated transcranial focused ultrasound on excitatory and inhibitory neurons

Although stimulation with ultrasound has been shown to modulate brain activity at multiple scales, it remains unclear whether transcranial focused ultrasound stimulation (tFUS) exerts its influence on specific cell types. Here we propose a novel form of tFUS where a continuous waveform is amplitude modulated (AM) at a slow rate (i.e., 40 Hz) targeting the temporal range of electrophysiological activity: AM-tFUS. We stimulated the rat hippocampus while recording multi-unit activity (MUA) followed by classification of spike waveforms into putative excitatory pyramidal cells and inhibitory interneurons. At low acoustic intensity, AM-tFUS selectively reduced firing rates of inhibitory interneurons. On the other hand, higher intensity AM-tFUS increased firing of putative excitatory neurons with no effect on inhibitory firing. Interestingly, firing rate was unchanged during AM-tFUS at intermediate intensity. Consistent with the observed changes in firing rate, power in the theta band (3-10 Hz) of the local field potential (LFP) decreased at low-intensity, was unchanged at intermediate intensity, and increased at higher intensity. Temperature increases at the AM-tFUS target were limited to 0.2°C. Our findings indicate that inhibitory interneurons exhibit greater sensitivity to ultrasound, and that cell-type specific neuromodulation may be achieved by calibrating the intensity of AM-tFUS.

Introduction stimulation (Fig 4F; significant cluster from 6.0 Hz to 9.9 Hz, = 0.022). On the other hand, the 141 acute theta increase at high-intensity did not persist in the period after stimulation (Fig 4H; no 142 significant cluster in the theta range). Interestingly, we also found effects that only manifested in the AM-tFUS: -24.1 % ± 11.9%, = 0.023; means ± sem). At high-intensity, a significant increase in 158 theta power was identified during AM-tFUS (Fig 5C, 29.6 % ± 13.0%, = 0.048, = 20). There 159 was no significant change in theta power at intermediate intensity, either during or after stimulation. 160 Similar changes in theta power were observed in the CA1 and CA3 subregions (Figs S3 and S4). 161 To gain insight into the dynamics of the observed theta-band changes, we measured theta 162 power in a time-resolved fashion (i.e., 5 second windows). At low intensity, the sham stimulation 163 time course showed a steady increase throughout the 10-minute recording period, while active 164 7 AM-tFUS produced an immediate sharp reduction near stimulation onset that was sustained after stimulation ( Fig 5D). On the other hand, high-intensity AM-tFUS produced an immediate increase 166 at stimulation onset that was abolished following stimulation ( Fig 5F). 167 In the awake state, the rat hippocampus is known to alternate between "theta" and "non-theta" 168 states, which loosely correspond to active (locomotion and REM sleep) and idle states (53; 54). 169 Analogues of these states have been observed under anesthesia (55). We suspected that AM-tFUS 174 Temperature increases are limited to 0.2 • C Given the three-minute sonications employed in 175 this study, it is important to determine whether heating of the brain occurred, and if so, to assess 176 its contribution to the observed neural changes. We therefore conducted follow-up experiments 177 ( = 8) that replicated the main experiments, except that the recording electrode was replaced with a 178 thermocouple ( Fig 6A). As expected, temperature increased monotonically during stimulation, with 179 the largest increases observed during high-intensity AM-tFUS (Fig 6B-D). However, the magnitude 180 of the increases was relatively small for all intensities: at the end of the three-minute stimulation 181 period, the temperature change from baseline was Δ = 0.05 ± 0.14, 0.12 ± 0.19, and 0.20 ± 0.18 • , 182 for low, intermediate, and high-intensity stimulation, respectively (mean ± sem across = 8 animals).

183
The highest instantaneous temperature increases during AM-tFUS (relative to baseline) recorded 184 were 0.33, 0.59, and 0.67 • C for low, intermediate, and high-intensity, respectively. high-intensity. These findings provide the first experimental evidence for the notion that specific 194 cell types can be targeted by appropriate selection of AM-tFUS parameters. 195 We found congruence between the measured changes in neuronal firing rates and those in 196 the LFP theta power. Inhibitory interneurons have been previously reported to have a causal role 197 in the generation of theta rhythms, as mice bred with impaired inhibitory synaptic activity exhibit 198 lower theta, but not gamma, power (56). Moreover, inhibitory interneurons exhibit spike timing 199 that is phase locked with theta cycles (57; 58; 59; 60). Therefore, one interpretation of the findings 200 at low-intensity stimulation is that AM-tFUS impaired action potential generation in inhibitory 201 interneurons, consequently producing less field activity (or perhaps less coherence) in the theta range.   Coupled with ultrasound's capability of deep and focused stimulation of the intact human brain, tFUS 269 represents a technique with immense potential as a therapeutic tool and circuit mapping technique.

270
To realize this potential, it will be important to carry out comprehensive parameter sweeps in the 271 awake brain (35), as this will mitigate the presence of spontaneous fluctuations in brain state during 272 anesthesia, and the potential mismatch between tFUS effects on an anesthetized versus awake brain.

273
For example, despite its desirable generation of a long-lasting stable plane of anesthesia, urethane 274 has been previously found to affect multiple neurotransmitter systems (78). Furthermore, in order to 275 disambiguate the influence of mechanical and thermal forces in shaping neuronal changes by tFUS, 276 it will be necessary to conduct experiments that combine electrophysiology with measurements of 277 either mechanical displacement (for example, harmonic motion imaging (79)) or brain temperature  well as with a model rat skull placed between the transducer and hydrophone. From this, it was 306 determined that the skull attenuates the acoustic pressure to a value that is 2/3 of the free-field 307 pressure. We then determined input voltages on the function generator such that the spatial-peak 308 temporal-average intensity ( spta ) for a continuous waveform traversing the skull was equal to 2.5 309 W/cm 2 , 5.0 W/cm 2 and 10.0W/ cm 2 , for the "low", "intermediate", and "high" intensities employed 310 here. After taking into account the AM nature of the waveform, these values are reduced to 1.6, 3.2, 311 and 6.4 W/cm 2 , yielding the final intensity values reported in the main text. In order to generate the 312 acoustic beampattern on a fine spatial grid, we employed the k-Wave toolbox (82) to simulate our 313 transducer's pressure distribution (Fig 1B).    show results for the dentate gyrus. The dependent variable was the change in spectral power from 372 baseline. When depicting time-resolved spectral power (Fig 5D-F analysis. Detection of a spike required a negative deflection of at least 6 standard deviations from 388 the channel mean. In a few recordings (i.e., 6), a very low number of spikes was observed. In these 389 cases, the spike detection threshold was lowered to 5 standard deviations, and the procedure was 390 re-run. Due to limitations on the maximum file size, spike sorting was performed independently on 391 the data of each intensity/condition. Thus, six independent runs of the procedure were performed 392 for each animal.

393
The results of the spike sorting were post-processed to retain only those units that exhibited 394 physiological spiking behavior. For each unit, we measured the basal firing rate and the trough-395 to-peak duration of the unit's template waveform. Any unit with a firing rate below 0.1 Hz was 396 excluded from analysis. Units with a trough-to-peak duration shorter than 100 s or longer than 397 1500 s were also excluded from the analysis. To remove units with clearly artifactual waveforms, 398 we excluded units that did not exhibit their trough within ± 250 s of the threshold crossing, or 399 those that exhibited a positive peak within ± 167 s of the threshold crossing.

400
In order to classify units into excitatory and inhibitory, we employed a priori knowledge of 401 both the spike duration and the basal firing rate. A unit was classified as excitatory if its trough-402 to-peak duration exceeded 900 s and if its basal firing rate was less than 7.75 Hz. Conversely, a 403 unit was classified as inhibitory if its trough-to-peak duration was less than 900 s and its basal   To control for multiple comparisons in the analysis of spectral power changes (Fig 4), we    To determine the temperature increase produced by AM-tFUS, we stimulated the hippocampus but now with a thermocouple inserted into the focus of the ultrasonic beam. Temperature was recorded before, during, and after AM-tFUS. As expected, temperature increased monotonically during stimulation. However, the observed temperature increases were low, with mean changes at the end of stimulation of (B) 0.05 ± 0.14 • C, (C) 0.12 ± 0.19 • C, and (D) 0.20 ± 0.18 • C at low, intermediate, and high intensity, respectively (means ± sem, = 8). Shaded error bars denote sem. Figure S1: AM-tFUS modulates LFP spectral power in an intensity and frequency-dependent manner in area CA1. Panels replicate Fig 4 of the main text, except that here we show results from the CA1 region of the hippocampus. As in the dentate gyrus, a bidirectional effect on theta power was observed. On the other hand, the significant high-frequency clusters after AM-tFUS were not resolved here. Note that the presence of large increases during high-intensity AM-tFUS at 40 and 80 Hz likely indicates physical displacement of the probe. 31 Figure S2: AM-tFUS modulates LFP spectral power in an intensity and frequency-dependent manner in area CA3. Panels replicate Fig 4 of the main text, except that here we show results from area CA3. Unlike in the dentate gyrus and CA1, a significant reduction in the theta band was not resolved after low-intensity AM-tFUS. Note also the significant reductions in gamma power (>30 Hz) during and after high-intensity AM-tFUS. Figure S3: AM-tFUS produces a bidirectional modulation of theta power in area CA1. Panels replicate Fig 5 of the main text but are here shown for CA1. Similar to what was found in the dentate gyrus, the reduction in theta power during low-intensity AM-tFUS outlasted the stimulation. On the other hand, the increased theta power during high-intensity AM-tFUS was not resolved following stimulation. Figure S4: AM-tFUS produces a bidirectional modulation of theta power in area CA3. Panels replicate Fig 5 of the main text but now show data from hippocampal region CA3. The results are qualitatively similar to those found in both the dentate gyrus as well as area CA1.