2.1.

Animal Preparation

Male and female double-transgenic mice (20 to 30 g; aged $>7$ weeks) expressing GCaMP6s and $\mathrm{CaMKII}\alpha$-Cre were used for in vivo two-photon calcium imaging in three types of experiments: (1) neuronal activity evoked by two-photon optogenetics, (2) resting-state imaging with no stimulation, and (3) neuronal activity evoked by air-puff stimulation of the whiskers. The mice were obtained by mating GCaMP6s transgenic mice (B6J.Cg-Gt(ROSA)26Sor<tm96(CAG-GCaMP6s)Hze>/MwarJ, No. 028866, The Jackson Laboratory, Sacramento, California, United States) and CaMKIIα-Cre transgenic mice (B6.Cg-Tg(Camk2a-cre)T29-1Stl/J, No. 005359, The Jackson Laboratory, Sacramento, California, United States). The animals were housed with food and water provided ad libitum in cages maintained at 25°C under a 12-h light/dark cycle. All experiments were conducted in accordance with institutional guidelines for the humane care and use of laboratory animals and were approved by the Institutional Committee for Animal Experimentation of the National Institutes for Quantum Science and Technology.

Three weeks before in vivo two-photon microscopic imaging, a chronic cranial window was attached to the animal’s skull above the left hemisphere of the brain. The animals were anesthetized with a mixture of air, oxygen, and isoflurane (3% to 5% for induction and 2% for surgery) via a facemask. Rectal temperature was monitored, and the animal was heated to maintain body temperature at 36°C to 37°C. The cranial window (4 to 5 mm in diameter) was positioned over the left somatosensory cortex following the “Seylaz-Tomita method,” so that its center was 1.8 mm caudal and 2.5 mm lateral to the bregma.²⁰ During this surgery, adeno-associated virus (AAV) vectors carrying the C1V1 gene (pAAV-CamKIIa-C1V1(t/t)-mScarlet-KV2.1, No. 124650, $1\times {10}^{13}\mathrm{vg}/\mathrm{mL}$, Addgene, Watertown, Massachusetts, United States) were injected into the barrel cortex using a glass needle. The soma-targeted C1V1 allowed specific single target neurons to be stimulated.¹⁴ Plastic bars 14 mm long were also attached to the front and rear ends of the skull so that the animal’s head could be stereotactically fixed under the microscope.

The surgical procedure used for unilateral common carotid artery occlusion (CCAO) has been described in previous studies.^21,22 Briefly, a mixture of air, oxygen, and isoflurane anesthesia (3% to 5% for induction and 1.5% to 2% for surgery) was provided via a facemask. After a midline cervical incision was made, the left common carotid artery (ipsilateral to the cranial window) was isolated from the adjacent vagus nerve and double-ligated with 4–0 silk sutures.

2.2.

Two-Photon Microscope Setting

A two-photon microscope (Thorlabs, Newton, New Jersey, United States) equipped with two independent lasers (Spark Lasers, Martillac, France) was prepared. One laser emitted 920 nm wavelength light, and the other emitted 1064 nm wavelength light. The 920 nm laser was used for calcium imaging of neuronal activity with a resonant mirror in its optical path. The 1064 nm laser, also with a resonant mirror installed in the optical pathway, was used for mScarlet imaging when confirming C1V1 expression. During the two-photon optical stimulation, a galvano mirror was introduced into the optical path of the 1064 nm laser to irradiate the cell body of the target neuron in an inward spiral pattern. The optical path of the 1064 nm laser was switched for imaging and optical stimulation using a mirror that was rotated by an electric motor. Different dichroic mirrors were used for imaging and optical stimulation. Two independent photomultiplier tubes, one for 460 nm imaging and the other for 532 nm imaging, were installed for signal detection. Before the optical stimulation experiment on each mouse, fluorescent microscope slides (Thorlabs) were used to verify that the 1064 nm laser accurately irradiated the target cell. The two-photon microscope was controlled by the ThorImage software (Thorlabs). The calcium images had a resolution of $512\times 512\mathrm{pixels}$ and field of view (FOV) of $600\mu \mathrm{m}\times 600\mu \mathrm{m}$ and were acquired at a frame rate of 15 Hz.

2.3.

Selective Two-Photon Optogenetic Activation of Single Neurons

The experimental protocol used for measurements on awake mice has been reported previously.^23,24 Briefly, the plastic bars on the animal’s head were screwed into a custom-made stereotactic apparatus. The animal was then placed on a styrofoam ball floating on a stream of air, which allowed the animal to exercise freely on the ball while its head was fixed to the apparatus.

We chose target cells and defined regions of interest (ROIs) for optical stimulation based on the presence of the red fluorescent protein mScarlet expressed in neurons. Layer 2/3 neurons 80 to $100\mu \mathrm{m}$ from the brain surface were targeted for optical stimulation. The laser pulse used for optical stimulation followed an inward “spiral” pattern as previous studies have demonstrated its effectiveness in inducing neuronal activity.¹¹ The diameter of the optically stimulated area was set between 20 and $30\mu \mathrm{m}$, which is large enough to completely cover the soma of the target neuron, and a stimulation duration of 30 to 50 ms was used. The laser intensity during stimulation was 88 mW, or 0.12 to $0.28\mathrm{mW}/{\mu \mathrm{m}}^2$.

2.4.

Magnetic Resonance Imaging of the Hypoperfusion Model Mice

Magnetic resonance imaging (MRI) measurements were performed 4 weeks after CCAO to evaluate reduction in cerebral blood flow. The experimental protocol was identical to that used in a previous study.²¹ Briefly, the MRI measurements were performed on mice under isoflurane anesthesia (3% to 5% for induction and 1.5% to 2% for surgery) using a 7-Tesla animal MRI system (Bruker, Billerica, Massachusetts, United States). Rectal temperature was continuously monitored and maintained at 36°C to 37°C using a heating pad. T2-weighted and CBF-ASL images were obtained using the same protocol as in our previous research.²¹ The CBF-ASL image was analyzed in NIH ImageJ.

2.5.

Analytical Methods for Evaluating Connectivity between Neurons

The analysis of neuronal activity was performed in NIH ImageJ. We used “moco,” an optional function within ImageJ, to remove motion artifacts in serial images obtained during awake imaging. Moco is a template-based motion correction technique that can correct misalignment in the $x-y$ direction but cannot correct motion artifacts in the z direction.²⁵ Motion artifacts in the z direction were instead treated by removing frames where the fluorescence intensity of the image changed unnaturally (the number of frames removed was less than 10% of the total video length). ROIs were manually placed on the cell bodies of neurons in the image. To reduce the risk of missing neurons with poor baseline fluorescence, the maximum intensity was extracted with $x-y-t$ projection. ROIs overlapping multiple neurons were excluded. The extracted signal was the average fluorescence intensity of all pixels belonging to the ROI. Connectivity between the target and surrounding neurons was assessed in three ways:

2.6.

Statistical Analysis

Statistical comparisons were performed using a Friedman test. A significance level of 0.05 was selected.

3.1.

Two-Photon Optogenetics-Based Neuronal Connectivity

We prepared mice co-expressing GCaMP6s and C1V1-mScarlet, which were excited by 920 and 1064 nm light, respectively [Fig. 1(a)]. C1V1-expressing cells were identified based on the presence of mScarlet, and target cells for optical stimulation were selected from these cells. Neurons expressing C1V1 (ROI 1) showed induced neuronal activity upon optical stimulation, whereas neurons without C1V1 (ROI 2) did not exhibit neuronal activity under optical stimulation [Fig. 1(b)]. An example of a correlation coefficient map is shown in Fig. 1(c). Each of the dots on the map represents a neuron observed within the $600\mu \mathrm{m}\times 600\mu \mathrm{m}$ FOV. There were usually 80 to 160 measurable neurons per FOV. Red dots indicate cells with high positive correlation coefficients with respect to the target cell (yellow), whereas dots indicate cells with low correlation coefficients with respect to the target cell. Blue dots indicate cells with negative correlation coefficients with respect to the target cell. In Fig. 1(d), $\mathrm{\Delta }F/F$ increased after optical stimulation for neurons exhibiting high correlation with the target cell (cell 1). Conversely, neurons appearing white on the map showed little response after optical stimulation (cell 3). Scatter plots of the $\mathrm{\Delta }\mathrm{Activity}$ for the target cell versus $\mathrm{\Delta }\mathrm{Activity}$ for the labeled cells (cells 1 to 3) are presented in Fig. 1(e). The correlation coefficients calculated for these scatter plots are used to characterize the two-photon optogenetics-based connectivity between the target neuron and all surrounding neurons.

Fig. 1

Two-photon optogenetics-based neuronal connectivity: (a) a typical two-photon imaging FOV of layer 2/3 in the somatosensory cortex. Neurons co-expressing GCaMP6s (green) and C1V1-mScarlet (red) can be observed with 920 and 1064 nm imaging, respectively. The area surrounded by the white square is shown in panel (b). (b) Neuronal activity induced using two-photon optical stimulation. A neuron expressing C1V1 (ROI 1) responds to optical stimulation, whereas a neuron without C1V1 (ROI 2) does not respond. (c) An example of a correlation coefficient map. The target neuron for optical stimulation is shown in yellow. The correlation coefficient of each cell with respect to the target cell is indicated according to the color scale in the top right of the image. (d) Percent change in GCaMP6s fluorescence ($\mathrm{\Delta }F/F$) of the target cell and the labeled cells (cells 1 to 3) induced after optical stimulation. The graph shows representative data from 10 of the 20 to 25 stimulation periods during the experiments. (e) Scatter plots of ΔActivity for the target cell and the labeled cells (cells 1 to 3). The r-value in the top right corresponds to the correlation coefficient calculated for the scattered data points.

3.2.

Comparison of Neuronal Connectivity Evaluated Using Two-Photon Optogenetics and Resting-State Synchrony

A representative FOV where neuronal activity was assessed is shown in Fig. 2(a), where resting-state neural synchrony (6000 frames; 400 s) was calculated with respect to the target cell [Fig. 2(b)]. Neurons with high correlation coefficients in the resting state (cell 1) exhibit $\mathrm{\Delta }F/F$ changes resembling those of the target cell, whereas neurons with low correlation coefficients (cell 3) exhibit less similarity with the target cell [Fig. 2(b)]. Figure 2(c) shows two-photon optogenetics and resting-state synchrony correlation coefficient maps obtained for the same FOV. The map constructed using the two-photon optogenetics method [Fig. 2(c), left] is visually similar to the map obtained with resting-state neural synchrony [Fig. 2(c), middle]. Due to the difference in the number of data points acquired for each method (25 trials versus 6000 frames), the correlation coefficients were converted from r-values to t-values, via $t=r(\sqrt{N-2}/\sqrt{1-r^2})$ so that the results of the two methods could be quantitatively compared. Note that the $t$-distributions corresponding to $N=25$ and $N=6000$ are very similar. The right part of Fig. 2(c) presents a scatter plot comparing the two sets of t-values. In this example, the two sets of $t$-values were positively correlated ($r=0.3445$, $p<0.05$), but there were cases where the two sets of t-values were poorly correlated [Fig. 2(d)]. Two target neurons were selected for each of the six mice, making a total of 12 data sets. For eight out of the 12 cases, the r-values were above 0.2630, and the corresponding p-values were less than 0.05 [Figs. 2(e) and 2(f)], indicating that the neuronal connectivity measured with the two-photon optogenetics-based method for these eight cases was highly consistent with the results of the resting-state synchrony method. On the other hand, there were four cases where the r-value was below 0.1135, meaning that the p-values were above 0.05 and the correlation between the two methods was poor. A possible reason for this is that living animals experience spontaneous sensory inputs. As these experimental data were measured from the barrel region of the somatosensory cortex, it is possible that sensory inputs may have affected the neural activity during rest. The evaluation of neuronal connectivity may become less accurate when there is an external sensory input that disturbs resting-state neuronal synchrony, leading to lower agreement between the two methods. To test this hypothesis, we evaluated the synchrony of neural activity while artificially introducing sensory inputs.

Fig. 2

Comparison of neuronal connectivity evaluated using two-photon optogenetics and resting-state synchrony: (a) a representative FOV where the neuronal activity was assessed. Cells with green fluorescence are GCaMP6s-expressing neurons. (b) Percent changes in GCaMP6s fluorescence ($\mathrm{\Delta }F/F$) during the resting state (upper) and corresponding scatter plots of $\mathrm{\Delta }F/F$ (lower) for the target and labeled cells in panel (a). Correlation coefficients characterizing the connectivity based on resting-state synchrony (r-values) are shown in the top right of the plots. (c) Comparison of correlation coefficient maps based on two-photon optogenetics (left) and resting-state synchrony (middle). The t-values calculated for the two methods are compared in a scatter plot (right). The correlation between these two sets of t-values is $r=0.3445$ ($p<0.05$) in this example. (d) An example where the two sets of t-values are poorly correlated ($r=0.1134$, $p>0.05$). Two target cells were selected for each mouse, making a total of 12 experiments. The distribution of the corresponding r-values is shown in panel (e), and a histogram of the p-values is in panel (f). The black dots in panel (e) correspond to cases where $p<0.05$ in panel (f). A similar relationship exists for the green dots.

3.3.

Evaluating Connectivity during External Sensory Stimulation: Comparison with Two-Photon Optogenetics-Based Connectivity

In this experiment, we introduced air-puff stimulation of the whiskers as a more potent sensory input. Because the measurements were conducted in the barrel region of the somatosensory cortex, it is expected that there will be increased neuronal activity in response to sensory input from the whiskers. If the target neurons respond to sensory stimulation, the neuronal activity induced in the target neurons and the surrounding neurons may synchronize. This synchrony poses a risk of overestimating the connectivity between neurons. We therefore analyzed measurements where the target neuron did not respond to sensory stimulation. Out of the 12 experiments performed in Sec. 3.2, there were eight cases where this situation occurred.

In the representative FOV shown in Fig. 3(a), neuronal activity was measured during sensory input in addition to two-photon optogenetics and resting-state measurements. The upper part of Fig. 3(b) shows normalized $\mathrm{\Delta }F/F$ changes for a target cell that did not respond to the sensory stimulation. A neuron with a large correlation coefficient (cell 4) exhibited similarities with the $\mathrm{\Delta }F/F$ of the target neuron, and a neuron with a small correlation coefficient (cell 5) exhibited different $\mathrm{\Delta }F/F$ changes from the target neuron. Correlation analysis was performed on the ΔActivity obtained by averaging $\mathrm{\Delta }F/F$ during sensory stimulation [Fig. 3(b) lower]. Figure 3(c) presents correlation coefficient maps constructed using four methods: two-photon optogenetics, resting-state synchrony, sensory-stimulation synchrony, and resting-state synchrony after sensory stimulation. As in Sec. 3.2, the map constructed using the two-photon optogenetics-based method is visually similar to the map obtained with resting-state neural synchrony, and the two sets of t-values were significantly correlated [$r=0.3917$, $p<0.05$; Fig. 3(c), upper right]. However, the correlation coefficient map based on synchrony during sensory input is quite different from the map based on two-photon optogenetics, which is reflected by the two sets of t-values being poorly correlated [$r=0.0541$, $p>0.05$; Fig. 3(c), bottom right]. In addition, when the correlation coefficient maps calculated using resting-state synchrony after sensory stimulation were compared with those obtained with two-photon optogenetics, the two sets of t-values were positively correlated ($r=0.3173$, $p<0.05$; scatter plot not shown). Figures 3(d) and 3(e) show the results for the eight experiments where the target neuron did not respond to the sensory stimulation. In comparison to the two-photon optogenetics-based method, the correlation coefficients for resting-state neuronal synchrony had substantially higher significance (or lower p-values) than those measured to evaluate synchrony during sensory stimulation. The r-values (or p-values) for two-photon optogenetics versus resting state after sensory stimulation were also significantly higher (or lower) than those for two-photon optogenetics versus sensory stimulation [Figs. 3(d) and 3(e)]. A Friedman test applied to compare the three sets of r-values and p-values found a significant difference ($p=0.0076$ for r-values, $p=0.0098$ for p-values). The two-photon optogenetics-based connectivity was more consistent with resting-state synchrony and resting-state synchrony after sensory stimulation than the synchrony measured during sensory stimulation. This suggests that the evaluation of neuronal connectivity using resting-state synchrony is compromised when the resting-state neuronal activity is disturbed by spontaneous external stimulation.

Fig. 3

Comparison of two-photon optogenetics-based connectivity and synchrony during sensory stimulation: (a) another representative FOV different from that in Fig. 2 where the neuronal activity was assessed. (b) Percent change in GCaMP6s fluorescence ($\mathrm{\Delta }F/F$) during sensory stimulation (upper) and scatter plots of $\mathrm{\Delta }\mathrm{Activity}$ for the target and labeled cells (lower). The upper part shows representative data from seven out of 20 stimulation periods during the experiments. Correlation coefficients characterizing the connectivity based on synchrony during sensory stimulation (r-values) are shown in the top right of the plots. (c) On the left is a comparison of the correlation coefficient maps obtained with (1) two-photon optogenetics, (2) resting-state synchrony, (3) synchrony during sensory stimulation, and (4) resting-state synchrony after sensory stimulation. To the right of these maps are scatter plots comparing the t-values obtained for two-photon optogenetics with those obtained for resting-state synchrony and synchrony during sensory stimulation. (d) Comparison of the set of r-values obtained for two-photon optogenetics versus resting-state synchrony [same analysis as for Fig. 2(c)], the set of r-values obtained for two-photon optogenetics versus sensory stimulation [shown in panel (c), bottom right], and the set of r-values obtained for two-photon optogenetics versus resting-state after sensory stimulation. The results shown are for the eight experiments where the target neuron did not respond to the external sensory stimulation. The three sets of r-values were compared with the Friedman tests. (e) Comparison of the three sets of p-values accompanying the r-values. The three sets of p-values were also compared with Friedman tests.

Figures S1(a) and S1(b) in the Supplementary Material show the results for the four experiments where the target neuron responded to sensory stimulation. In three out of the four cases, the two sets of r-values and p-values exhibited the same trend. However, in one case, the r-value for two-photon optogenetics versus resting state was lower than that for two-photon optogenetics versus sensory stimulation.

3.4.

Evaluation of Neuronal Connectivity in the Hypoperfusion Model Using the Two-Photon Optogenetics-Based Method

The chronic hypoperfusion mouse model used in this study involves the ligation of one side of the common carotid artery, resulting in an $\sim 30\%$ sustained reduction in cerebral blood flow. However, no significant infarct lesions are formed [Fig. 4(a); Urushihata et al.²¹]. This model has been associated with pathological conditions such as white matter lesions, impairments in behavioral learning, and neuroinflammation.^26–28 We have observed neurofibrillary impairments in the cerebral cortex and cerebral vascular dysfunction in this mouse model.^21,22,29 Our experience with the CCAO model therefore made it an appropriate candidate for testing the two-photon optogenetics-based method in a pathological situation.

Fig. 4

Evaluation of neuronal connectivity in the hypoperfusion model using the two-photon optogenetics-based method: (a) a mouse model of common carotid artery occlusion (CCAO). The left common carotid artery (ipsilateral to the cranial window) was separated from the jugular vein and the vagus nerve (not shown) to be ligated. In this model, there is no significant change in T2WI to indicate cerebral infarction despite reduced blood flow (CBF-ASL). (b) The upper row of images shows a representative FOV before (pre) and 1, 2, and 4 weeks after CCAO. The lower images show corresponding correlation coefficient maps. The target neurons are indicated by the yellow arrows in the upper images and by yellow dots in the lower maps. (c) Mean of the r-values for each animal (dashed lines) and the average over all five animals (solid line) before (pre) and 1, 2, and 4 weeks after CCAO. Differences between the means of the r-values at each of the four timepoints were evaluated with the Friedman test. (d) Mean of the r-values for each animal (dashed lines) and the average (solid line) for five control animals without CCAO measured at the same timepoints as for (c).

Two-photon optogenetics-based connectivity was measured by targeting the same neuron at four timepoints (i.e., just before CCAO, as well as at 1, 2, and 4 weeks after CCAO) in five animals. The left CCA was ligated during CCAO surgery [Fig. 4(a)] and all measurements were performed in the barrel cortex of the left hemisphere. As shown by the series of microscopic images in Fig. 4(b) (upper row), measurements were performed in the same FOV at the four timepoints. There was no decrease in neuronal activity and the frequency of neuronal firing for at least 1 week after CCAO surgery.³⁰ For the target neurons, it was confirmed that the mScarlet fluorescence intensity, the location relative to the surrounding cells, the location relative to capillaries stained with sulforhodamine 101 (Adipogen Life Sciences, San Diego, California, United States), and the distance from the brain surface were not significantly altered over the 4 weeks of the experiments, which ensured that the target neurons for optical stimulation were identical at all timepoints. The visual impression from the maps is that hypoperfusion reduced the correlation coefficients over the imaged FOV [Fig. 4(b), lower row]. A Friedman test found that the average of the correlation coefficients for the five animals decreased significantly after CCAO [Fig. 4(c), $p=0.026$]. As a control experiment, we measured two-photon optogenetics-based connectivity at the same timepoints for five animals without CCAO and found no obvious change in the mean correlation coefficient [Fig. 4(d)].

The connectivity based on resting-state neural synchrony was also measured before and after CCAO in the same FOV. There was no significant change in the average of the correlation coefficients before and after CCAO [Fig. S2(a) in the Supplementary Material]. In addition, so as not to limit the analysis to the results for a single cell (i.e., the target neuron), a set of correlation coefficient maps was created using a different neuron as the target for each map in the set. The correlation coefficient was averaged over each map and then over the set to produce Fig. S2(b) in the Supplementary Material. The average correlation coefficients calculated in this way did not change significantly after CCAO.