Two-Photon Bidirectional Control and Imaging of Neuronal Excitability with High Spatial Resolution In Vivo

1Within brain circuits, information about sensory stimuli is encoded in complex spatial and temporal patterns of activity distributed across cells. For example, population recordings, combined with statistical analysis, showed that specific features of sensory stimuli elicit temporally structured responses in specific ensembles of neurons. However, using statistical analysis and correlative evidence to causally test which sensory features are encoded in neural circuits and how this information is used to drive behavior may prove difficult. To achieve this goal, we would ideally need a method to monitor and bidirectionally perturb the activity of multiple neurons maintaining single-cell resolution. With such a technique, it would be possible to study how the concerted activity of identified neurons contributes to network function by activating or inactivating populations of functionally characterized neurons with cellular resolution.

2Optical approaches, in particular two-photon microscopy, hold promise to achieve this goal. Moreover, wave-front engineering methods using digital holography largely extended the potential of two-photon microscopy for imaging and photostimulation applications. In parallel to these improvements in optics, the last decade witnessed the development of a large toolbox of light-sensitive molecules to monitor and manipulate the activity of neurons, including opsins such as channelrhodopsin-2 (ChR2), C1V1, and Guillardia theta Anion Channelrhodopsins (GtACRs) and functional indicators such as GCaMPs and RCaMPs. Combining advanced two-photon approaches with the use of these bioengineered molecules, it became possible to perform simultaneous functional imaging of GCaMP signals and stimulation of various opsins (e.g., C1V1) with high spatial resolution in the rodent brain in vivo and in other experimental systems. However, several limitations need to be overcome to efficiently apply these approaches. First, crosstalk between imaging and photostimulation needs to be minimized. For instance, the red-shifted channelrhodopsin C1V1 is maximally activated using 540 nm light, but it is still more than half-maximally activated by 470 nm light. This shoulder toward shorter wavelengths typical for red-shifted opsins is reflected in their two-photon absorption spectra and may lead to non-negligible neuronal depolarization during two-photon GCaMP imaging. This undesired effect worsens when opsins with slow off kinetics and high-amplitude photocurrents, which are the preferred choice for two-photon activation of neurons with the raster or spiral scanning approach, are used. Second, while published data demonstrated cellular resolution two-photon activation of neurons, evidence for efficient patterned two-photon inhibition, as well as all-optical imaging and high-resolution inhibitory manipulation in vivo, is still to be provided. Third, whether single-cell two-photon optogenetics can be efficiently applied across the various cell types that are engaged during sensory stimulation and that differ in morphology, biophysical properties, and cortical depth is unclear.

3Here we developed an experimental approach to address all of these challenges in the mouse cortex in vivo. We combined digital holography to stimulate blue light-sensitive opsins with two-photon imaging of a red-shifted functional indicator. We show that holographic illumination of ChR2 with extended shapes can be used to efficiently stimulate various cellular populations, including principal cells and different interneuron types, in cortical layer 2/3 and in layer 4, the main thalamorecipient lamina in sensory cortex. We then characterized the two-photon excitability of the chloride-permeable channelrhodopsin GtACR2 in slice preparation and showed that it efficiently decreases neuronal firing with high spatial resolution in vivo upon holographic illumination. Finally, combining soma-targeting of opsins (ChR2 and GtACR2), which improved the spatial resolution of stimulation, with the use of the red-shifted calcium indicator jRCaMP1a, we provide a proof-of-principle demonstration of simultaneous two-photon imaging and bidirectional holographic stimulation of cells with negligible effect of the imaging beam on opsin excitation.

4To test whether using a blue light-sensitive channelrhodopsin would lead to a reduction in the undesired cross-activation during activity reporter imaging, we expressed ChR2 and C1V1(T/T) in cultured hippocampal neurons and characterized the photocurrent evoked by two-photon scanning at the wavelengths typically used for calcium imaging of green and red calcium indicators (920 and 1,080 nm, respectively) (Figure S1). ChR2-expressing neurons showed lower relative peak current amplitudes when scanned at 1,080 nm than did C1V1-expressing neurons scanned at 920 nm (Figures S1C–S1E). The relative average photocurrent evoked during scanning at 1,080 nm was also higher for C1V1 (Figures S1C and S1F). Increasing the raster-scanning rate led to further elevation of C1V1 activation (Figures S1C and S1F) because of the slower closing kinetics of C1V1. Conversely, increasing the raster scanning rate when recording from ChR2-expressing cells did not increase photocurrents (Figures S1D and S1F), consistent with its faster off kinetics.

5To stimulate neurons with high spatial resolution in vivo, we used a liquid crystal spatial light modulator (SLM)-based holographic module, which was integrated in a commercial laser scanning two-photon microscope (Figure 1A), and we programmed the holographic module (see Experimental Procedures) to project on the sample plane elliptical shapes that were centered on the cell body of target neurons (Figure 1B; Figure S2). To validate our approach, we performed simultaneous two-photon targeted juxtasomal recordings and photostimulation experiments in anesthetized mice in layer 2/3 principal neurons co-expressing ChR2 and the red fluorescent protein tdTomato, which facilitated targeting neurons under the microscope (Figure 1C). Once a stable electrophysiological recording was achieved from an opsin-positive neuron (see Experimental Procedures for definition), a high-resolution image was acquired and an elliptical shape (ellipse axis: 7–16 μm) was projected on the cell body of the recorded neuron. Significant increase in action potential (AP) firing frequency was observed upon two-photon holographic illumination with extended elliptical shapes (stimulus duration: 500 ms; stimulus power: 30–92 mW/cell; λexc = 920 nm) (Figures 1D and 1E). To verify that the observed effect depended on opsin activation, not on membrane depolarization due to direct two-photon stimulation, we performed similar experiments in opsin-negative cells (Figure S3). We found that holographic illumination with extended shapes of the same spatial profile and light intensity did not modify the membrane potential or the AP firing rate of recorded opsin-negative neurons in vivo (Figures S3C and S3D).

6To evaluate the spatial resolution of our stimulation method, we measured the spiking response to holographic stimulation of opsin-positive neurons while incrementally shifting the stimulation shape in the radial and axial direction (Figure 1F). We found spatial constants (see Experimental Procedures for definition) of ∼20, ∼32, and ∼16 μm in the radial, axialup, and axialdown directions, respectively. Targeting ChR2 to the soma (Figure 2) increased the average spiking response in the illuminated neuron (ChR2: ΔAPFreq = 1.2 ± 0.3 Hz, N = 15 cells from 6 mice; soma-targeted ChR2: ΔAPFreq = 4.6 ± 1.0 Hz, p = 2.2E−2, Mann-Whitney test, N = 21 neurons from 6 mice; stimulus power: 30 mW for both ChR2 and soma-targeted ChR2). Somatic targeting of ChR2 improved the spatial resolution of holographic stimulation compared to non-soma-targeted opsins, decreasing the axialup (13 and 32 μm for soma-targeted and non-soma-targeted opsins, respectively; p = 1.4E−2, unpaired Student’s t test) spatial constant (Table S1). Confocal analysis of fixed sections from injected animals confirmed restricted expression in the somatic and perisomatic compartments with the soma-targeted ChR2 compared to the non-soma-targeted ChR2 (Figure S4). Table S1 also shows the density of opsin-expressing cells under our experimental conditions. The values of the radial and axial space constants of photostimulation resolution normalized to the soma diameter of the stimulated cells are shown in Table S2.

7We investigated whether holographic stimulation could be efficiently applied to cell types other than layer 2/3 excitatory neurons. To this end, we first expressed ChR2 in two major subpopulations of cortical interneurons in layer 2/3, the somatostatin-positive (SST+) and the parvalbumin-positive (PV+) cells. Using simultaneous photostimulation and two-photon targeted juxtasomal recordings in vivo, we found that illumination with an extended shape (stimulus power: 30 mW/cell) increased the firing rate of targeted interneurons (Figures 3A–3C, left and middle). We then expressed ChR2 selectively in sodium channel, non-voltage-gated 1 alpha-positive (Scnn+) excitatory neurons of layer 4, the main thalamorecipient cortical population of the sensory cortex. We found that holographic illumination (stimulus power: 50 mW/cell) increased the spike rate of layer 4 Scnn+ neurons (Figures 3A–3C, right). In all excitatory neurons recorded in layer 2/3 (Figures 1 and 2) and layer 4 (Figure 3), the spontaneous firing rates before and after photostimulation were not significantly different (Table S3). In all cell types the response to photostimulation depended upon the illumination power (Figure S5).

8Our previous data demonstrate that holographic stimulation with extended shapes can be used for activation of neurons with high spatial resolution in vivo. We determined whether holographic illumination could also be used for efficient two-photon optogenetic inhibition with similar spatial precision. To address this question, we focused on GtACR2, a chloride-permeable channelrhodopsin. Although the two-photon excitability of GtACR2 has not yet been reported, we reasoned that its large photocurrent, high light sensitivity, and blue light-sensitive, single-photon absorption spectrum made it a good candidate for two-photon holographic stimulation at λ < 1,000 nm. In addition, expression of GtACR2 has been reported to be well tolerated by neurons. We first expressed this inhibitory opsin in the cortex and recorded GtACR2-mediated photocurrents in a patch-clamp, voltage-clamp configuration from opsin-positive cells (see Experimental Procedures for definition) in acute brain slices (Figure 4A). We found that holographic illumination of GtACR2-expressing neurons with an elliptical shape targeted to the cell body of the recorded cell (λexc = 920 nm; stimulus power: 30 mW; stimulus duration: 500 ms) triggered clear outward currents (range: 6–112 pA; holding potential: −50 mV; chloride equilibrium potential: −68 mV). Peak amplitude of photocurrents increased with power (Figure 4B) and showed a nearly power-squared dependence for low power values (Figure 4B, inset). Moreover, while keeping light power density constant, we performed holographic two-photon illumination at different light wavelengths (range: 740–1,040 nm). We found that GtACR2 photocurrents had large peak amplitude at 920 nm and decreased for longer and shorter wavelengths (Figure 4C). We thus concluded that GtACR2 can be efficiently stimulated through a two-photon absorption process, that holographic illumination triggers clear inhibitory photocurrents in opsin-expressing neurons, and that the two-photon absorption spectrum of GtACR2 shows a clear peak ∼920 nm.

9We then asked whether holographic stimulation of GtACR2 could be used to decrease neural excitability with high spatial resolution in vivo. To this end, we performed whole-cell, current-clamp recordings from layer 2/3 cortical neurons expressing GtACR2 in anesthetized mice (Figures 4D–4I). We found that illumination with an extended shape (stimulus power: 10–80 mW/cell; stimulus duration: 500 ms) while a small depolarizing current was injected (current amplitude: 74 pA) led to a significant hyperpolarization of the cell (average membrane potential before [Pre]: −46.2 ± 1.1 mV, during [Stim]: −50.1 ± 1.1 mV, after [Post]: −44.4 ± 1.3 mV; p = 2E−15, ANOVA test with Bonferroni’s correction, N = 14 from 7 mice). Moreover, we found that holographic illumination decreased cellular firing induced by a small current injection (Figures 4D and 4E). We measured the spiking response of GtACR2-positive neurons to holographic illumination while incrementally shifting the stimulation shape in the radial and axial directions (Figure 4F). We found spatial constants of ∼11, ∼33, and ∼29 μm in the radial, axialup, and axialdown directions, respectively (Table S1). Prolonged illumination (stimulus duration: 10 s) of GtACR2-expressing neurons decreased the firing rate (Figures 4G and 4H), and it hyperpolarized the membrane potential of the illuminated cell for the duration of the light stimulus (Figure 4I). The values of the cell resting membrane potential Pre- and Post-photostimulation were not significantly different (average resting membrane potential Pre: −61.2 ± 2.2 mV, Post: −62.2 ± 1.5 mV; p = 0.43, Student’s t test, N = 5 from 2 mice).

10Red-shifted channelrhodopsins have been used for two-photon stimulation of single neurons simultaneously with genetically encoded calcium indicators (GECI)-based calcium imaging. One potential drawback of this approach is the remaining absorption by all red-shifted channelrhodopsins in the blue range of their action spectrum (Figure S1). As shown earlier, our data demonstrate that holographic illumination of blue light-sensitive opsins (ChR2 and GtACR2) at λ = 920 nm can be used to bidirectionally control the excitability of cortical neurons with high spatial resolution in vivo. We therefore asked whether this stimulation approach could be coupled with imaging of red-shifted functional indicators (e.g., jRCaMP), which are typically best excited in the two-photon regime at longer wavelength (λ = ∼1,100 nm), a wavelength at which blue light-sensitive channelrhodopsins show no detectable activity (Figure S1). To test this possibility, we first evaluated whether raster scanning at these long wavelengths caused opsin activation, leading to significant alteration of neuronal spiking activity in vivo. We expressed the soma-targeted ChR2 in layer 2/3 cortical neurons and performed juxtasomal electrophysiological recordings from ChR2+ neurons while raster scanning the field of view (FOV) containing the recorded cell at λ = 1,100 nm and scan rate of 11 Hz (scan resolution: 0.58 μm/pixel; dwell time: 4 μs) (Figure 5). We found that raster scanning did not significantly affect the firing activity of layer 2/3 soma-targeted ChR2+ neurons at both 30 and 50 mW imaging power (Figures 5B and 5C, left). As an important control, we found that raster scanning the same FOV at shorter wavelength (λ = 920 nm) and moderate power (laser power: 30 mW) increased the spiking activity of soma-targeted ChR2-expressing neurons (Figure 5C, right). Moreover, whole-cell, current-clamp recordings performed on layer 2/3 neurons expressing the inhibitory opsin GtACR2 showed that raster scanning at λ = 1,100 nm did not significantly modify the average resting membrane potential of opsin-positive neurons (average membrane potential: −62.1 ± 3.3 and −62.1 ± 3.3 mV in the absence and presence of raster scanning [laser power: 30 mW], respectively; −63.7 ± 3.2 and −63.9 ± 3.6 mV in the absence and presence of raster scanning [laser power: 50 mW], respectively; p = 1, Wilcoxon signed rank test at both 30 and 50 mW, N = 4 from 2 mice), as expected from the two-photon absorption spectrum of GtACR2 (Figure 4C).

11We combined holographic stimulation with raster scanning imaging to perform simultaneous two-photon imaging of jRCaMP1a and two-photon activation of ChR2 in vivo. Using two laser sources (Figure 6A) tuned at 1,100 nm (for imaging) and 920 nm (for holographic stimulation), we performed concurrent imaging and holographic photostimulation experiments in mice that co-expressed jRCaMP1a and soma-targeted ChR2 in layer 2/3 cortical neurons (Figure 6) or jRCaMP1a and ChR2 in SST+ interneurons (Figure 6B). We found that successive holographic stimulation of the cell body of ChR2-expressing neurons (stimulus power: 50 mW/cell) reliably evoked fluorescent transients in the stimulated cell (Figures 6B–6D). Combined juxtasomal electrophysiological and imaging recordings from jRCaMP1a-expressing neurons in vivo (Figure S6) confirmed that we could detect single and trains of APs with good accuracy. We combined this all-optical approach with the photostimulation of multiple specified neurons. We controlled the SLM to generate extended shapes covering the cell bodies of a group of four neurons (Figures 6C and 6D). We photostimulated the selected neurons (stimulus power: 56 mW/cell) while simultaneously imaging these and the surrounding neurons at 11 Hz. Targeted neurons that displayed clear ChR2-expression (Figures 6C and 6D, neuron 1–3) showed strong and reliable responses to photostimulation. Neighboring neurons responded weakly to holographic stimulation of target neurons, as expected from previous work.

12Moreover, in cells expressing only jRCaMP1a, we controlled for potential artifacts induced by holographic stimulation on jRCaMP1a fluorescence. We recorded jRCaMP1a signals at 1,100 nm while performing repetitive short (Figures S7A and S7B) or prolonged (Figures S7C and S7D) holographic stimulation at 920 nm. We found that stimulation of the imaged cell (stimulation power: 30 mW) generated an artifact in the jRCaMP1a signal that could be removed by background subtraction. Increasing stimulation power from 30 to 50 mW resulted in similar effects (Figure S7E). Repetitive stimulation did not decrease jRCaMP1a baseline, and it did not induce evident signs of jRCaMP1a photobleaching (Figures S7F–S7H). Similarly, in background-subtracted traces, photostimulation did not significantly affect the amplitude and the off kinetics of the responses to whisker deflection (Figures S7I–S7M).

13Finally, we performed simultaneous two-photon imaging and patterned photoinhibition in PV+ cells co-expressing jRCaMP1a and the soma-targeted GtACR2 in vivo. These neurons display a high spontaneous firing rate under our experimental conditions (Figures 3A, 3C, 7A, and 7D). We found that patterned illumination decreased the baseline jRCaMP1a signal in the stimulated cell (Figures 7A–7C). Simultaneous electrophysiological recording of the stimulated neuron confirmed that the baseline reduction in jRCaMP1a signal was associated with a decrease in the spike rate of the stimulated neuron (Figures 7A and 7D).

14Simultaneous two-photon imaging and manipulation is increasingly recognized as a crucial tool for the causal investigation of brain networks. Such a technique allows perturbing the activity of functionally identified ensembles of neurons and testing of the role of specific activity patterns in the regulation of network dynamics and behavior. Here we developed an all-optical approach for simultaneous two-photon imaging of a red-shifted functional indicator and bidirectional perturbation of neural activity using blue light-sensitive opsins in vivo. We validated our approach across different cell types and layers of the mouse neocortex. This is a fundamental step to apply all-optical methods to investigate the role of precise spatiotemporal activity patterns in driving higher cortical functions, because activity patterns are distributed in space and time across cellular subtypes.

15Our method expands the potential of simultaneous imaging and perturbation for the functional dissection of brain circuits. Previous work in the mammalian brain in vivo demonstrated that the blue light-sensitive calcium indicator GCaMP can be coupled to the red-shifted excitatory opsin C1V1 for simultaneous two-photon imaging and perturbation (see also Supplemental Information). However, red-shifted opsins generally display a blue-shifted tail in their absorption spectrum that may complicate spectral separation and lead to crosstalk between GCaMP imaging and opsin activation under certain conditions. This is especially true for red-shifted excitatory opsins with long off kinetics that are often the preferred choice for two-photon activation using scanning approaches. We showed (Figures 5, 6, and 7) that the use of jRCaMP1a, a red-shifted functional indicator that is excited using two-photon stimulation between 1,050 and 1,150 nm, in combination with blue light-sensitive opsins (e.g., ChR2) that display maximal two-photon excitability around 920 nm, minimizes this form of crosstalk. Using combined imaging and electrophysiological recordings (Figure 5), we found that the activity of ChR2-expressing cells was not changed by raster scanning in vivo (λimaging = 1,100 nm; laser intensity: 30–50 mW; frame rate: 11 Hz; scan resolution: 0.58 μm/pixel; FOV dimension: 58 × 58 μm2), in agreement with what observed in cultured neurons (Figure S1). The absence of crosstalk likely stems not only from the spectral separation of the two light-sensitive molecules that we have used (i.e., jRCaMP1a and ChR2) but also from the fast closing kinetics of ChR2 and the low power required for imaging. The experimental configuration that we presented provides other advantages. For example, the use of red-shifted indicators may facilitate deep imaging by using longer wavelengths for fluorescence excitation and emission, which are less sensitive to tissue scattering. In addition, because of their stability for long-term expression, they can be efficiently used for chronic experiments. Moreover, the stimulation of blue light-sensitive opsins at 920 nm may decrease tissue heating that is higher at the longer wavelengths (λ = 1,040 nm) used to stimulate red-shifted opsins (e.g., C1V1). Although photocurrents generated by ChR2 are generally smaller than those generated by C1V1, the two-photon cross section of ChR2 is high, and cells responded efficiently to stimulation (Figures 1, 2, and 3).

16Previous work in vivo demonstrated high spatial resolution two-photon activation of excitatory opsins. Here we show that holographic two-photon illumination can be used for efficient suppression of neural activity with high spatial resolution and can be coupled with functional imaging for all-optical readout and inhibitory optogenetic manipulation in vivo. Although previous evidence in vitro showed that some light-sensitive proton pumps are excitable with a two-photon process, here we focused on the use of chloride-permeable anion channelrhodopsins. We reasoned that the increased flow of ions per photocycle that characterizes light-sensitive channels would allow generation of larger photocurrents and more efficient hyperpolarization of neurons in vivo compared to the use of light-sensitive pumps. Among the various chloride-permeable opsins, we focused on GtACR2 because of its higher single-channel conductance and its blue light-sensitive, single-photon absorption spectrum. We first demonstrated that GtACR2 was efficiently stimulated through an absorption process that is compatible with two-photon excitation. Significant photocurrents were generated through holographic illumination of the cell body of GtACR2-expressing neurons in brain slice preparation (Figures 4A and 4B). The photocurrent had maximal peak amplitude for λ = 920 nm, similar to ChR2. Moreover, holographic stimulation of GtACR2 significantly hyperpolarized principal neurons in vivo and efficiently reduced their firing rate while maintaining high spatial resolution of the optogenetic perturbation (Figures 4D–4F). Most importantly, holographic stimulation of GtACR2 could be efficiently coupled with jRCaMP1a imaging for simultaneous functional imaging and optogenetic inhibitory manipulation with high spatial resolution (Figure 7). We observed a decrease in the baseline jRCaMP1a fluorescence in PV+ cells expressing GtACR2 upon patterned illumination at 920 nm (Figures 7A and 7B). Activation of GtACRs may change the intracellular chloride concentration and may lead to pH variations. These modifications might interfere with the fluorescence activity reporter. However, two lines of evidence suggest that the decrease in jRCaMP1a baseline activity upon patterned stimulation of GtACR2 is mainly due to a decrease in the cell’s firing rate. First, in simultaneous imaging and electrophysiological recordings, the baseline fluorescence decrease of jRCaMP1a was always associated with the decrease in the cell’s spiking rate (Figures 7A and 7D). Second, our observation is consistent with the high spontaneous firing rate of PV+ cells (Figure 3) being integrated by the slow activity reporter jRCaMP1a and with previous reports showing decreased baseline of the fluorescence reporter upon sensory stimulation in PV+ interneurons corresponding to inhibited activity of these cells. A long recovery tail toward baseline level of the fluorescence reporter similar to the one observed in our experiments (Figures 7A and 7B) was also reported in that study. These results demonstrate that patterned two-photon optogenetics can be applied for high spatial precision optical inhibition of brain networks in vivo, making it possible to silence endogenous activity patterns triggered by sensory stimulation with very high cellular specificity.

17Although our method efficiently decreased crosstalk between the imaging laser and the opsin activation, stimulation with extended shapes induced artifacts in the fluorescence detection, as observed by previous investigators. This artifactual signal may be due to unwanted stimulation by holographic illumination of the fluorescence protein that is tagged to the opsin (e.g., eGFP), the fluorescence of which may leak into the red fluorescence detection channel despite the barrier filter positioned in front of the photomultiplier tube (PMT). Alternatively, because jRCaMP1a has low, but not negligible, absorption at 920 nm (i.e., the wavelength used for stimulation), the artifactual signal may originate from direct activation of jRCaMP1a by holographic stimulation (Figure S7). If jRCaMP1a is expressed at high levels and the area covered by stimulated neuronal somata represents a significant portion of the FOV (e.g., when many neurons are stimulated at the same time), the integrated emission of dim jRCaMP1a fluorescence generated by holographic stimulation at 920 nm may generate significant artifacts in the red detection channel, as suggested by our experiments (Figure S7). This artifactual signal could be removed using background subtraction (Figure 6B; Figure S7) or required a blanking period (Figure 6D). A solution to this problem could be to synchronize photostimulation with imaging so that stimulation is performed when the portions of the FOV that are of no interest are being scanned. For stimulus duration longer than frame duration, optimization of protein expression levels or further developments in red-shifted indicators with reduced absorption at the wavelength used for patterned illumination will be needed.

18In conclusion, we provide an experimental approach to image and bidirectionally manipulate brain networks with high spatial resolution in living animals. This all-optical approach will likely represent a powerful tool to dissect how activity patterns in specified ensembles of neurons determine brain function and animal behavior.

19All experiments were carried out according to the guidelines of the European Communities Council Directive and approved by the Instituto Italiano di Tecnologia (IIT) Animal Health Regulatory Committee and by the National Council on Animal Care of the Italian Ministry of Health (authorization 29-2011-A, 34/2015-PR). Animals were housed in individually ventilated cages under a 12-hr light:dark cycle. A maximum of 5 animals per cage was allowed. Access to food and water was ad libitum. Experiments were performed on young-adult animals (5–16 weeks old for in vivo experiments, 4–7 weeks old for in vitro experiments, either sex). Details about animal strains and viral injections are described in the Supplemental Experimental Procedures. For in vivo experiments, mice were anesthetized with intraperitoneal urethane (16.5%, 1.65 g/kg). The scalp was removed while infiltrating all incisions with lidocaine. A chamber with a central hole (hole diameter: 4 mm) was attached with dental cement to the animal’s skull for head-fixation. A craniotomy (∼700 × 700 μm2) was opened over the somatosensory (or visual cortex, in the case of experiments in Scnn mice) cortex, and the dura was carefully removed (unless otherwise stated). The location of the craniotomy was guided by the intensity of the fluorescence signal of the expressed transgene. The surface of the brain was kept moist with normal HEPES-buffered artificial cerebrospinal fluid (ACSF) (composed of 127 mM NaCl, 3.2 mM KCl, 2 mM CaCl2, and 10 mM HEPES [pH 7.4]). Body temperature was maintained at 37°C with a heating pad. Respiration rate, heartbeat, eyelid reflex, vibrissae movements, and reactions to tail pinching were typically monitored throughout the surgery and the experiment.

20See Supplemental Experimental Procedures.

21Two-photon targeted juxtasomal electrophysiological recordings were performed as described in and. Borosilicate glass pipettes were pulled with a resistance of 4–9 MΩ and were filled with ACSF solution mixed with Alexa Fluor 488 or 594 (20 μM). Neurons were targeted by imaging the fluorescent reporter with the two-photon microscope while monitoring the pipette electrical resistance by applying brief voltage pulses. Additional details are reported in the Supplemental Experimental Procedures.

22Two-photon imaging (λexc = 1,050 nm) was performed to assess the expression pattern of the opsin (ChR2-eYFP or ChR2-eYFP soma targeted) and the calcium indicator (jRCaMP1a) at the same time. A reference image of the selected FOV was acquired, and shapes covering the soma of target neurons were generated by the SLM (λexc = 920 nm) and projected at the sample. Temporal series were acquired in raster scanning configuration with the imaging beam (100 × 100 pixels; frame rate: 11 Hz; pixel dwell time: 4 μs; λexc = 1,100 nm). Holographic photostimulation duration was 500 ms and was repeated at 0.08 Hz for 7–9 repetitions. For analysis, temporal series acquired in vivo were imported into the ImageJ/Fiji software to identify regions of interest (ROIs). For each ROI, the change in fluorescence relative to the baseline (ΔF/F0) was computed as a function of time with the fluorescence baseline (F0) calculated in ten frames at the beginning of the recorded session. Artifactual fluorescence signals due to holographic stimulation were removed by background subtraction (Figure 6B; Figure S7) or required blanking (Figure 6D).

23See Supplemental Experimental Procedures.

24For juxtasomal recordings, traces were high-pass filtered (cutoff frequency: 10 Hz) and spikes were detected with a threshold criterion. The threshold value was adjusted for each recorded sweep and set >3 times the SD of the trace. For experiments in Figures 1, 6, and 3 and in Figures S3 and S5, AP firing frequency was calculated in a time window Pre (window duration: 1 s), Stim (duration: 0.5 s) and Post (duration: 1.5 s) holographic stimulation over 15–20 stimulation trials. ΔAPFreq was calculated as the difference between the firing frequencies of the Stim and Pre time windows. Opsin-positive cells (for definition, see In vivo electrophysiological recordings in the Supplemental Experimental Procedures) were considered responsive to holographic stimulation when ΔAPFreq was >1 times the firing frequency in the Pre period at stimulation power ≤ 92 mW per shape. The fraction of opsin-positive neurons responding to holographic illumination was 14/16 for Ca2+/calmodulin-dependent protein kinase II-positive (CaMKII+) cells expressing ChR2 (Figures 1C–1E), 15/17 for layer 2/3 cells expressing ChR2 under the human synapsin promoter (Figures 1E and 1F), 31/33 for SST+ cells expressing ChR2 (Figure 3), 25/26 for PV+ cells expressing ChR2 (Figure 3), 19/20 for Scnn+ cells expressing ChR2 (Figure 3), and 21/21 for cells expressing the soma-targeted ChR2 under the human synapsin promoter (Figure 2). To compute the spatial resolution, neuronal responses (quantified as ΔAPFreq) were recorded first with the stimulation shape centered on the cell body and then during successive shifts of the excitation volume in the radial (20 μm steps) and in the axial (±25 μm steps) directions. ΔAPFreq as a function of the shift was then plotted for every recorded neuron in the three conditions (radial, axialup, and axialdown) and fitted with a mono-exponential function (ΔAPFreq(x) = A ∗ exp(−l ∗ x)). Fitting curves with l < 0 or with values of A that were different by more than 25% compared to ΔAPFreq at position x = 0 were not considered. The spatial resolution, l1/2, was defined as the distance at which the evoked response (calculated from fit) was equal to A/2. For the analysis of the recordings from GtACR2 expressing neurons, see Supplemental Experimental Procedures.

25All values are expressed as mean ± SEM unless otherwise stated. For each experimental group, sample size was chosen based on previous studies. No statistical methods were used to predetermine sample size. All recordings with no technical issues were included in the analysis. For N ≥ 10, a Kolmogorov-Smirnov normality test was used to test for normality. For N < 10, a Saphiro-Wilk normality test was adopted. In case of normal distribution, Student’s t test was used to calculate statistical significance when comparing two populations of data. For non-normal distributions, the non-parametric Mann-Whitney test or Wilcoxon signed-rank test (for unpaired or paired comparison, respectively) was used unless otherwise stated. When multiple (>2) populations of data were compared, one-way ANOVA with Bonferroni or Tukey’s honestly significant difference (HSD) post hoc test was used in case of Gaussian distribution. For non-normal distribution and multiple comparisons, the non-parametric Friedman test with Dunn’s post hoc correction was used. All tests were two sided. Statistical analysis was performed using Prism (GraphPad, La Jolla, CA) and OriginPro 9.1 (OriginLab).