Animals

Experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Protocols were approved by the Local Animal Ethics Committee (Committee Charles Darwin no. 5, registration nos. 9529 and 26889) and conducted in agreement with Directive 2010/63/EU of the European Parliament. Long–Evans male rats aged between 2 and 12 months and WT male mice (C57BL/6J) aged 9 weeks were obtained from Janvier Laboratories; P23H (line 1) male transgenic rats (9–22 months) were raised locally.

Plasmid cloning and AAV production

Plasmids containing the E. coli mscL sequence in the WT form and with the G22S mutation were obtained from Francesco Difato (Addgene plasmids #107454 and #107455)²⁸. To target RGCs, the SNCG promoter³¹ was inserted into an AAV backbone plasmid containing the mscL sequence fused to the tdTomato gene and the Kir2.1 ER export signal, to drive expression at the plasma membrane. An AAV2.7m8 vector was used for intravitreous delivery. For targeting neurons in the V1 cortical layers, the SNCG promoter was replaced by the CamKII promoter and an AAV9.7m8 vector was chosen. Recombinant AAVs were produced by the plasmid co-transfection method, and the resulting lysates were purified by iodixanol purification³¹.

US stimulus

Three focused US transducers with different central frequencies were used: 0.50 MHz (diameter, Ø = 1.00″ = 25.4 mm; focal distance, f = 1.25″ = 31.7 mm) (V301-SU, Olympus), 2.25 MHz (Ø = 0.50″ = 12.7 mm, f = 1.00″ = 25.4 mm) (V306-SU, Olympus) and 15.00 MHz (Ø = 0.50″ = 12.7 mm, f = 1.00″ = 25.4 mm) (V319-SU, Olympus), corresponding to numerical apertures of F/Ø = 1.25 and 2.00. Acoustic fields radiated by those three focused transducers are presented in Fig. 1 (simulations) and Extended Data Fig. 3 (experimental measurements). A TiePie Handyscope (HS5, TiePie Engineering) was used to produce the stimulus waveform, which was then passed through an 80 dB RF power amplifier (VBA 230-80, Vectawave) connected to the transducer. Transducer pressure outputs (pressure at focus, three-dimensional (3D) pressure maps) were measured in a degassed water tank with a Royer–Dieulesaint heterodyne interferometer⁴⁷. US stimuli used for ex vivo and in vivo stimulation had the following characteristics: 1 kHz pulse repetition frequency with a 50% duty cycle, sonication duration between 10 and 200 ms and interstimulus interval between 0.01 and 2.00 s. Peak acoustic pressures ranged from 0.11 to 0.88 MPa, 0.30 to 1.60 MPa and 0.20 to 1.27 MPa for the 0.50, 2.25 and 15.00 MHz transducers, respectively. The corresponding estimated spatial peak pulse average intensity (Isppa) values were 0.39–25.14, 2.92–83.12 and 1.30–52.37 W cm^–2.

Intravitreous gene delivery and retinal imaging

Rats were anaesthetized⁴⁸ and AAV suspension (2 µl), containing between 8 and 14 × 10¹⁰ viral particles, was injected into the centre of the vitreous cavity. One month later, tdTomato fluorescence imaging was performed on the injected eyes, with a MICRON IV retinal imaging microscope (Phoenix Research Laboratories) and Micron Discover v.2.2.

MEA recordings

Retinal pieces were flattened on a filter membrane (Whatman, GE Healthcare Life Sciences) and placed on an MEA (electrode diameter, 30 µm; spacing, 200 µm; MEA256 200/30 iR-ITO, MultiChannel Systems) coated with poly-l-lysine (0.1%, Sigma), with RGCs facing the electrodes³¹. AMPA/kainate glutamate receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX, 25 μM, Sigma-Aldrich), the NMDA glutamate receptor antagonist [3H]3-(2-carboxypiperazin-4-yl) propyl-1-phosphonic acid (CPP, 10 μM, Sigma-Aldrich) and a selective group III metabotropic glutamate receptor agonist, l-(+)-2-amino-4-phosphonobutyric acid (LAP4, 50 μM, Tocris Bioscience), were bath applied through the perfusion line. Light stimuli were delivered with a digital micromirror display (Vialux; resolution, 1,024 × 768) coupled to a white light-emitting diode light source (MNWHL4, Thorlabs) focused on the photoreceptor plane (irradiance, 1 µW cm^–2). US transducers were coupled with a custom-made coupling cone filled with degassed water and mounted on a motorized stage (PT3/M-Z8, Thorlabs) placed orthogonally above the retina. The reflected signal of the MEA chip and the retina was detected with a US key device (Lecoeur Electronique). The distance between the retina and transducer was equal to the focal length of the transducer; this was verified with the flight time of the reflected signal. From RGC recordings with a 252 channel preamplifier and MC_Rack v. 4.6.2 (MultiChannel Systems), spikes were sorted with Spyking CIRCUS 0.5 software⁴⁹. RGC responses were analysed with custom scripts written in MATLAB (MathWorks 2018b) for classification as ON, ON–OFF or OFF, with the response dominance index⁵⁰. Latencies were calculated as the time between stimulus onset and the maximum of the derivative of the spike density function (SDF). Two classes of US-responding cells were identified on the basis of latency—SL and LL—by fixing a threshold equal to the minimum of the latency distribution of the responses of NT cells to US (45 ms). We determined the peak value A of the SDF for the calculation of response duration, which was defined as the time interval between the two time points for which the SDF was equal to A/e (where A is peak depolarization and e is Euler’s number). The Fano factor, quantifying spike count variability, was calculated as the ratio of the variance of the spike count to the mean. The Euclidean distance between two activated cells was weighted according to the maximum firing rate of the cells. The ratio of the number of activated cells to the size of the area stimulated on the MEA chip was calculated considering the size of the US focal spot for 2.25 and 15.00 MHz and the size of the MEA for 0.50 MHz, because the focal spot was larger than the MEA for this frequency. The centre of the response was estimated by weighting the maximum firing rate of each cell by its distance from other responding cells, and the displacement of the response was calculated as the Euclidean distance between two centre-of-response positions.

Intracranial injections

AAV suspensions were injected into the right hemisphere at two different locations in rats (2.6 mm ML, 6.8 mm AP and 3.1 mm ML, 7.2 mm AP from the bregma) or at one location in mice (2.5 mm ML, 3.5 mm AP from the bregma)⁴⁸. For rat injections, the suspension (200 nl containing 0.2–8.0 × 10¹⁵ viral particles) was injected at three different depths (1,100, 1,350 and 1,500 µm from the cortical surface) with a microsyringe pump controller (Micro4, World Precision Instruments) operating at a rate of 50 nl min^–1 and 10 µl Hamilton syringe. In mice, the AAV suspension (1 µl containing 0.2–8.0 × 10¹⁵ viral particles) was injected at 400 µm from the cortical surface at a rate of 100 nl min^–1.

In vivo extracellular recordings

One month after AAV injections, a small craniotomy (5 × 5 mm²) was performed above V1 in the right hemisphere⁴⁸. The tdTomato fluorescence was checked with a MICRON IV retinal imaging microscope and Micron Discover v. 2.2 (Phoenix Research Laboratories). A 32 site µEcog electrode array (electrode diameter, 30 µm; electrode spacing, 300 µm; FlexMEA36, MultiChannel Systems) was positioned over the transfected region or in a similar zone for control rats. MEA recordings were performed with a 16 site silicon microprobe tilted at 45° to the brain surface (electrode diameter, 30 µm; spacing, 50 µm; A1x16-5mm-50-703, NeuroNexus Technologies) and MC_Rack v. 4.6.2. The MEA was advanced 1,100 µm into the cortex with a three-axis micromanipulator (Sutter Instruments). US transducers were coupled to the brain with a custom-made coupling cone filled with degassed water and US gel on a motorized stage. The distance between the cortex and transducer was equal to the focal length of the transducer. Visual stimuli were generated by a white-light-collimated light-emitting diode (MNWHL4, Thorlabs) placed 15 cm away from the eye (4.5 mW cm^–2 at the cornea). Recordings were digitized with 32 channel and 16 channel amplifiers (model ME32/16-FAI-μPA, MultiChannel Systems). The µEcog recordings were analysed with custom-developed MATLAB scripts and the MEA recordings were analysed with Spyking CIRCUS software and custom-developed MATLAB scripts. The response duration was calculated as the interval between the two time points at which the cortical-evoked potential was equal to A/e. The activated area was defined as the area of the pseudocolour activation map over which peak depolarization exceeded the background-noise level calculated as 2 × s.d. of the signal. The response centre was estimated by weighting the peak depolarization of each electrode by its distance from the other electrodes. Its relative displacement when moving the US transducer was calculated as the Euclidean distance of the two positions. For intracortical recordings, cell latency was estimated as the time between stimulus onset and the maximum of the derivative of SDF.

Surgery for in vivo behavioural testing

C57BL6J mice were subcutaneously injected with buprenorphine (0.05 mg kg^–1) (Buprécare, Axience), and dexamethasone (0.7 mg kg^–1) (Dexazone, Virbac). Animals were anaesthetized with isoflurane (5% induction and 2% maintenance, in an air/oxygen mixture) and the head was shaved and cleaned with an antiseptic solution. Animals were head fixed on a stereotactic frame with an isoflurane delivery system and eye ointment, and a black tissue was applied over the eyes. The body temperature was maintained at 37 °C. After a local injection of lidocaïne (4 mg kg^–1) (Laocaïne, Centravet), an incision was made on the skin. Two screws were fixed in the skull, after a small craniotomy (approximately 5.0 × 5.0 mm²) was performed above V1 in the right hemisphere (0.5 mm steel drill) and a cortex buffer was applied. The cortex was covered with a TPX plastic sheet (125 µm thick) and sealed with dental acrylic cement (Tetric Evoflow). For behavioural experiments, a metallic headbar (PhenoSys) for head fixation was then glued to the skull on the left hemisphere with dental cement (FujiCEM 2). Animals were placed in a recovery chamber, with a subcutaneous injection of physiological serum and ointment on the eyes (Ophtalon, Centravet). Buprenorphine was injected during post-surgery monitoring.

Mouse behavioural tests

Mice were placed on a water restriction schedule until they reached approximately 80–85% of their weight. Following habituation to the test conditions³⁶, mice were trained to respond to an LS by performing a voluntary detection task: licking a waterspout (blunt 18 gauge needle, approximately 5 mm from the mouth) in response to white-light full-field stimulation (200 and 50 ms long) of the left eye (dilated with tropicamide, Mydriaticum Dispersa) over 35 trials per stimulation duration and therefore 70 trials per day. Water (~4 μl) was automatically dispensed 500 ms after the light was switched on, through a calibrated water system. The behavioural protocol and lick detection were controlled by a custom-made system³⁶. The next four days (two-day break during the weekend), US stimulations were delivered on V1 for 50 ms at three different pressure values (0.2, 0.7 and 1.2 MPa). These pressure values were delivered in a different order each day (35 trials each). The intertrial intervals randomly varied and ranged between 10 and 30 s. The 15 MHz US transducer was coupled to the brain with a custom-made coupling cone filled with water and US gel. The success rate was calculated by counting the number of trials in which the mice performed anticipatory licks (between stimulus onset and the opening of the water valve). The anticipatory lick rate (Fig. 6e) for the session was calculated by subtraction from the anticipatory lick rate of a trial, the spontaneous lick rate (calculated on all the 1 s time windows before each individual stimulus onset (Fig. 6a) for all the trials) and multiplication by the success rate. Lick latency was calculated by determining the time to the first anticipatory lick after stimulus onset. The mice retained for analysis presented a success rate superior or equal to 60% on the fourth day following LS. Then, light or US sessions showing a compulsive licking behaviour were excluded based on the outlier identification made using the ROUT method (Q = 1%) on the session’s spontaneous lick rate averaging the measurements on all the trials of the session in the 1 s time window before the stimulus onset of the trial.

Immunohistochemistry and confocal imaging

Samples were incubated overnight at 4 °C with a monoclonal anti-RBPMS antibody (1:500, rabbit; ABN1362, Merck Millipore) for the retina³¹, with a monoclonal anti-NeuN antibody (1:500, mouse, clone A60; MAB377, Merck Millipore) for brain sections⁴⁸. The sections were then incubated with secondary antibodies conjugated with Alexa Fluor 488 (1:500, donkey anti-mouse and donkey anti-rabbit IgG 488, polyclonal; A-21202 and A-21206, Invitrogen, respectively) and DAPI (1:1,000; D9542, Merck Millipore) for 1 h at room temperature. An Olympus FV1000 confocal microscope with ×20 objective (UPLSAPO 20XO with a numerical aperture of 0.85) was used to acquire the images of flat-mounted retinas and brain sections (FV10-ASW v. 4.2 software).

On the confocal images processed with Fiji (ImageJ v. 1.53q), RBPMS- and NeuN-positive cells were automatically counted with the ‘analyze particles’ plugin. The cells were manually counted by two different users, with the ‘cell counter’ plugin. Quantification was performed by acquiring confocal stacks in at least four randomly chosen transfected regions of 0.4 mm² (Extended Data Fig. 1). For V1 neurons, the sagittal brain slice with the largest tdTomato fluorescence zone was selected for each animal. A region of interest was manually defined in V1 and the quantifications were performed in at least six randomly chosen regions of 0.4 mm².

US-induced tissue-heating simulations

A three-fold process was used for the estimation of thermal effects: (1) simulation of the acoustic fields generated by the three transducers, with realistic acoustic parameters; (2) verification that nonlinear acoustics did not play an important role in heat transfer; and (3) realistic simulations of heat transfer and temperature rise induced at the focus by US in a linear regime for the parameters used in this study.

For nonlinear simulations, we used MATLAB’s k-Wave toolbox by defining the geometry of the transducer in three dimensions and using the following parameters for the propagation medium (water): sound speed, c = 1,500 m s^–1; volumetric mass, ρ = 1,000 kg m^–3; nonlinearity coefficient, B/A = 5; attenuation coefficient, α = 2.2 × 10^–3 dB cm^–1 MHz^–y; frequency power law of the attenuation coefficient, y = 2 (ref. ⁵¹). We simulated quasi-monochromatic 3D wavefields using long bursts of 50 cycles; this gave us the maximum pressure field in three dimensions as well as the waveform at the focus. Simulations were calibrated by adjusting the input pressure (excitation of the simulated transducer) to reach the pressure at the focus measured in the water tank with real transducers. The full-width at half-maximum (FWHM) focal-spot diameter in the x–y plane was 4.360, 1.610 and 0.276 mm, and the length of the major axis in the x–z plane was 32.3, 20.6 and 3.75 mm for the 0.50, 2.25 and 15.00 MHz transducers, respectively (Fig. 1b–d). Nonlinear effects were evaluated by estimating the relative harmonic content of the waveform at the focus. In the 15 MHz focus transducer example in Fig. 1d, the experimental and simulated signals at the focal spot were compared and found to be highly concordant (Extended Data Fig. 4a). Furthermore, the amplitude of the second harmonic is 19.8 dB below the fundamental (20.9 dB in the simulated case), meaning that if the fundamental energy is E, the second harmonic has energy E/95 (Extended Data Fig. 4b). Therefore, we can reasonably neglect the nonlinear effects in the calculations of the thermal effects, as they account for ~1% of the energy involved. The same conclusions were drawn at 0.5 MHz and 15.0 MHz. Linear wave propagation approximations considerably decreased the computing cost of the simulations. Linear propagation simulations were conducted with the Field II toolbox in MATLAB^52,53, in the monochromatic mode, with the same medium properties as k-Wave (water), to obtain the 3D maximum pressure fields. These maximum pressure fields were used to build a heating source term (Q_{mathrm{US}} = frac{{alpha _{mathrm{np}}p_{mathrm{max}}^2}}{{rho _mathrm{b}c_mathrm{b}}}), where α_np is the absorption coefficient of the brain at the considered frequency (59.04 Np m^–1 at 15 MHz, calculated from α_brain = 0.21 dB cm^–1 MHz^–y and y = 1.18), the brain volumetric mass ρ_brain = 1,046 kg m^–3, the brain sound speed c_brain = 154 s^–1 and p_max is the 3D maximum pressure field. This source term was then used in the resolution of a Penne’s bioheat equation (rho _{mathrm{brain}}C_{mathrm{brain}}timesfrac{{partial T}}{{partial t}} = mathrm{div}left( {K_mathrm{t}timesnabla T} right) – rho _{mathrm{blood}}C_{mathrm{blood}}P_{mathrm{blood}}left( {T – T_mathrm{a}} right) + Q) in k-Wave, where C_brain is the blood specific heat capacity (3,630 J kg^–1 °C^–1), K_t is the brain thermal conductivity (0.51 W m^–1 °C^–1), ρ_blood is the blood density (1,050 kg m^–3), C_blood is the blood specific heat capacity (3,617 J kg^–1 °C^–1), P_blood is the blood perfusion coefficient (9.7 × 10^–3 s^–1), T_a is the arterial temperature (37 °C), Q = Q_US + ρ_brainγ_brain and γ_brain is the heat generation of the brain tissue (11.37 W kg^–1) (refs. ^54,55). The initial condition for brain temperature was set to T₀ = 37 °C.

This simulation corresponds to the worst-case scenario regarding the given temperature rise. (1) The acoustic propagation is simulated in water only (non-derated value), with a lower attenuation coefficient (2.2 × 10^–3 dB cm MHz^–2.00) than the brain (0.59 dB cm MHz^–1.27), even if a part of the propagation occurs within the brain. The p_max maps are, therefore, overestimated. (2) Thermal absorption is simulated in the brain tissue only, with a higher absorption coefficient (0.21 dB cm MHz^–1.18) than water, even if a part of the maximum pressure field is actually located within the water of the acoustic coupling cone. Therefore, Q_US is slightly overestimated. We mapped the temperature in three spatial dimensions and time, and looked for the point of maximum temperature rise (Extended Data Fig. 4c–f).

Statistical analysis

Statistical analyses were carried out with Prism software (Prism 9, GraphPad). Values are expressed and represented as mean values ± standard error of the mean (s.e.m.) on figures and in the text, unless specified otherwise. Data were analysed in unpaired Welch’s t-tests (two tailed) or an unpaired multiple t-test with Sidak–Bonferroni correction for multiple comparisons. Statistical tests are provided in the figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

• SEO Powered Content & PR Distribution. Get Amplified Today.
• Platoblockchain. Web3 Metaverse Intelligence. Knowledge Amplified. Access Here.
• Source: https://www.nature.com/articles/s41565-023-01359-6