2.1.1.

Animals

A total of 34 C57BL/6J mice were used in this study (16 male, 18 female), obtained from Charles River (Wilmington, Massachusetts, United States). After arrival in the animal facility, mice had at least 4 days of rest. Mice had ad libitum access to food and water and were housed under a 12-h light/dark schedule. All procedures were approved by the Animal Research Ethics Committee of the Montreal Heart Institute and conformed to the regulations of the Canadian Council on Animal Care. A general overview of the procedures undergone by one mouse can be seen in Fig. 1(a). In all mentions of anesthesia, mice were anesthetized with isoflurane (3% induction; 1.5% surgery in pure ${\mathrm{O}}_2$), and their body temperature was controlled at 37°C with the help of an anal probe and a feedback-controlled heating pad. The heartrate, breathing, and body temperature were monitored continuously during this period (Small animal physiological monitoring system, Labeo Technologies Inc., Montréal, Canada).

Fig. 1

(a) Overview of the procedures that the mice were subjected to. (b) Schematic overview of the carotid artery calcification surgery. (c) Schematic overview of the imaging system. DM, dichroic mirror; LP, long pass filter. (d) Timecourses of a single pixel in the right sensory area of a single mouse. The GCaMP timecourse (in green) follows the $y$-axis on the left, whereas the hemodynamic timecourses (oxygenated hemoglobin—HbO—in red, deoxygenated hemoglobin—HbR—in blue, and total hemoglobin—HbT—in black) are on the right $y$-axis. The threshold for detecting spontaneous activation ($z\text{-}\text{score}>1.95$) for this pixel is depicted as a green dotted line for GCaMP, and a red dotted line for HbO. (e) Average images of the GCaMP, green, and red channels of a single acquisition. Black lines show the fitted Allen atlas. Lines in the top left corner depict the scale - lines are 1 mm. (f) Overview of the different regions of interest of the brain. VR, right visual area; SR, right sensory area; MR, right motor area; RR, right retrosplenial area; VL, left visual area; SL, left sensory area; ML, left motor area; RL, left retrosplenial area.

2.1.2.

Calcium rigidification and GCaMP injection

The procedure followed here was the same as in Ref. 40. A piece of parafilm was immersed in glutaraldehyde for at least 10 h, after which it was rinsed with sterile saline. The mouse was anesthetized, and an incision of $\sim 1.5\mathrm{cm}$ was made at the midline in the neck to allow for isolation of the right common carotid artery. A sheet of parafilm was inserted below the artery, and a sterile cotton gauze soaked in 0.3M ${\mathrm{CaCl}}_2$ solution (or 0.9% NaCl for Sham) was held to the carotid artery for 20 min [Fig. 1(b)]. After the carotid treatment, the incision was closed with sutures and tissue adhesive (VetBond). Mice were treated with the local anesthetic Bupivacaine (Marcaine, $4\mathrm{mg}/\mathrm{kg}$ SC infiltration at the incision site), the antibiotic Trimetoprim-Sulfa (Tribrissen; $30\mathrm{mg}/\mathrm{kg}$ SC), and the anti-inflammatory Carprofen (Rimadyl; $5\mathrm{mg}/\mathrm{kg}$ SC) immediately after surgery. Mice recovered for at least 2 days before GCaMP injection, for which mice were shortly anesthetized and injected with AAV2/php.eB-hSyn-GCaMP6s (CERVO molecular platform) in the tail vein with a catheter ($50\mu \mathrm{L}$, diluted with $150\mu \mathrm{L}$ of saline).

2.1.3.

Window surgery

Mice were anesthetized, ketoprofen ($5\mathrm{mg}/\mathrm{kg}$, SC) was injected subcutaneously, and local analgesia (Lidocaine 0.2 mL of $4\mathrm{mg}/\mathrm{mL}$, SC) was administered on the top of the head. The head was shaved and cleaned with ethanol. The skin on top of the skull was cut away, and the tissue covering the skull was cleaned with a cotton swab. Vetbond was applied to the edges of the skin. A titanium headbar was fixed on the head with dental cement (C&B Metabond). The skull was covered with thin layers of cyanoacrylate glue. Carprofen ($12.5\mathrm{mg}/\mathrm{kg}$, SC) was given right after the surgery, and Ketoprofen ($5\mathrm{mg}/\mathrm{kg}$, SC) was administered the following day. The mouse rested for at least 3 days after the surgery.

2.1.4.

Imaging

Optical imaging was done on awake mice. To limit the stress that the mice could experience during imaging, they were habituated to head fixation and imaging over a period of 4 days, (15, 20, 30, and 45 min). Following habituation, intrinsic optical imaging (IOI) and GCaMP imaging were done simultaneously, at rest. IOI/GCaMP acquisitions lasted 10 min and were performed once per week for 3 weeks. These acquisitions will be referred to as A1, A2, and A3 and occurred on days 22 to 25, 29 to 32, and 35 to 39 after calcium rigidification surgery, respectively.

To excite the genetically encoded reporters, a blue light-emitting diode (LED) (472 nm) combined with a bandpass filter (FF02-472/30) was used to restrain the spectral excitation [Fig. 1(c)]. For the intrinsic optical imaging (IOI⁴¹), green (535 nm) and red (620 nm) LEDs were used intermittently with the blue LED. A two-camera design was chosen to simplify the dissociation of the three recorded signals. Camera 1 captures the green reflectance ($\sim 535\mathrm{nm}$) and the GCaMP emission ($\sim 515\mathrm{nm}$), whereas camera 2 captures the red reflectance ($\sim 620\mathrm{nm}$). To separate the red from the green/GCaMP, a dichroic (610 nm) was used. A long-pass filter (FF01-496/LP) was placed in front of camera 1 to filter out the fluorescence excitation of blue light. Two cameras were used (CS2100M, Thorlabs, Newton, New Jersey, United States), each with a framerate of 30 Hz. This yielded a final framerate of 15 Hz per color. Images were $512\times 512\text{pixels}$. Imaging was done with a LightTrack OiS200 system (Labeo Technologies Inc.). Analysis was done with the help of MATLAB (R2020b and R2023a). Code can be found on github.com/marl1bakker/CaCl_P2. The two cameras had a slight offset, which was corrected through an automatic coregistration of the first frame captured by each camera, followed by a manual verification. The system recorded images mirrored, and thus, a left–right flip was administered.

2.2.1.

Movement and outlier removal

Mice were fixed with their head bar during acquisition and placed on a treadmill, which registered locomotion (Fig. 7 in Appendix A). Several recordings were made, and the best session was chosen a posteriori with the help of the registered movements of the treadmill. Furthermore, during analysis, recordings of the movement of the treadmill were used to detect movement, and corresponding frames were marked, as were frames 1 s before and after the movement. These frames were excluded from the analysis. For a session to be included in the analysis, the threshold for the number of marked moved frames was set to 3000. All mice met this criterion for at least one session in all acquisitions. To detect frames with artifacts or movement that went undetected by the treadmill, an outlier detection was done on the processed data. All pixels over the brain were averaged per frame. Frames that diverted more than three standard deviations of the mean in either GCaMP or hemodynamic data were marked and excluded from the analysis. Last, an outlier detection of single pixels was done to remove pixels that were marked as an outlier on 1000 frames or more. This helped detect and exclude artifacts in the window, such as bubbles or scratches.

One mouse failed to show GCaMP expression and was thus excluded from all calculations that used GCaMP measurements. As the cranial window was of good quality and the hemodynamic measurements worked well for this mouse, it remained in the study for calculations that only included hemodynamic measurements. Two other mice showed window artifacts in acquisitions 2 and 3 that hindered the imaging of hemodynamic data. Because of this, the hemodynamic measurements for one mouse were excluded from acquisitions A2 and A3, and another mouse was only excluded from hemodynamic measurements of A3.

2.2.2.

ROI selection and image alignment

Regions of interest were determined by fitting the Allen atlas⁴² on top of the mouse brain, based on the location of bregma and lambda. Regions of the atlas were pooled into larger ROIs, resulting in left and right versions of motor (M), retrosplenial (R), sensory (S), visual (V), and auditory (A) regions [Figs. 1(e) and 1(f)]. Auditory regions were left out of the analysis because they often fell outside of the cranial window. To ensure the stability of the regions of interest over the distinct acquisitions over time, the Allen atlas was fitted on the first acquisition, and the second and third acquisitions were coregistered to the first. Coregistration was done using an affine 2D transformation where the mean squared error was used as an optimization parameter. Image registrations were checked manually. If automatic coregistration was off by more than 3 pixels, the coregistration was done by hand.

2.2.3.

Hemodynamic correction

The influx of blood that typically accompanies an increase in neuronal activation can cause a weakened recording of the GCaMP signal by absorbing both light that activates the GCaMP molecule and by absorbing the fluorescence that this molecule emits due to neuronal firing.⁴³ Having information on hemodynamics allows for a correction for this alteration. Here, we used the same approach described by Valley et al.,⁴³ in which a linear regression of the reflectance is used to remove the bulk of the interference.

2.2.4.

Normalization and filtering

A lowpass filter (cutting frequency at 3 Hz) is first applied to remove high-frequency noise. To remove slow drifts in the signal, we then normalized it by its own low-frequency content (lowpass filter with cutting frequency at 0.08 Hz), to get the fluorescence fluctuations ($\mathrm{\Delta }F/F$).

2.2.5.

Hemodynamic calculations

A lowpass filter of 1 Hz was applied to the green and red channels, after which it was normalized by a lowpass filter at 0.08 Hz to remove any non-functional variations from the signal. The light absorption of oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) varies as a function of wavelength. The difference in reflectance of the green and red light thus allows us to calculate the changes in HbO and HbR in the same manner as Ref. 43, using the modified Beer–Lambert equation. Baseline values of HbO and HbR were set to 60 and $40\mu \mathrm{M}$, respectively.^44–46 Total hemoglobin levels (HbT) are calculated by adding HbO and HbR values.

2.2.6.

Neurovascular coupling

Spontaneous spikes in GCaMP activation, which were not provoked by any type of stimulation, were defined as fluorescence peaks that had a $z$-score of 1.95 or over [Fig. 1(d)]. $Z$-score was calculated as follows:

2.2.7.

Ultrasound imaging

Ultrasound imaging (US) was done on a subset of mice ($n=7$ CaCl, $n=6$ Sham), using a transducer of 30 MHz. Mice were anesthetized and placed on a heating pad in a supine position. The hair on the neck was removed. Heartrate was measured with ECG. The right carotid artery was located with the help of the B-mode and the color Doppler.

Images were required in M-mode. The diameter of the vessel was calculated as the space between the brightly appearing vessel walls [Fig. 9(a) in Appendix C, for example frame]. Changes in vessel diameter were averaged over several (2 to 3) heartbeat cycles [green line in Fig. 9(a) in Appendix C]. The difference between the minimum and maximum diameters over this average curve was then used as a measurement. Anesthetized mice were sacrificed at the end of the ultrasound acquisition.

2.3.1.

Movement

To verify that the CaCl and Sham groups did not differ in the amount of movement during acquisition, we compared the number of frames where movement was detected between the two groups. As the data were not normally distributed, a Kruskal–Wallis test was performed for each acquisition, none of which showed significance.

2.3.2.

NVC in Sham

To compare NVC between brain regions and groups, nine parameters were calculated and compared between regions of interest. First, the peak values of HbO, HbR, and GCaMP were compared [black triangles in Fig. 3(b), and top row of Fig. 3(c)]. Then, the increase in HbO and GCaMP was calculated as the difference between the peak and the preceding dip. For HbR, the decrease (difference between the peak and proceeding dip) was calculated [Fig. 3(b) and middle row of Fig. 3(c)]. Next, the time between the GCaMP peak and HbO peak was calculated and labeled “delay” [Fig. 3(b) and bottom left of Fig. 3(c)]. Last, the strength of the neurovascular coupling was calculated by dividing the HbO peak by the GCaMP peak [Fig. 3(c), bottom row, middle column] and by dividing the HbO increase by the GCaMP increase [Fig. 3(c), bottom row, right column].

Statistics were done on each parameter separately. First, an Anderson-darling test was done to determine if the data conformed to a normal distribution. If this was not the case, a non-parametric Friedman test was performed. Otherwise, an analysis of variance (ANOVA) was performed, and the sphericity of the assumptions was checked. If the sphericity assumption was not met, a Greenhouse–Geisser correction was performed. Because of the high number of statistical tests (especially relevant for statistics on the correlation matrices later in this article), there is a large chance of a false positive in the results. To correct this, a false discovery rate correction (FDR, Benjamini–Hochberg) was performed, providing us with adjusted $p$-values, hereafter referred to as $q$-values. Pairwise comparisons (Tukey–Kramer) were calculated on $p$-values that retained their significance after the FDR ($q$-values). Table 1 in Appendix D shows which tests were done for which parameter (A), and the $p$-values per seed pair of the post-hoc testing (B). Figure 10(a) in Appendix E shows NVC curves per mouse, per brain area.

2.3.3.

Ultrasound

The assumption of sampling from a normal distribution was checked with an Anderson–Darling test, and homogeneity of variance was tested with a two-sample $F$ test. As both assumptions were met, CaCl and Sham groups were compared using a two-sample $t$-test [Fig. 9(b) in Appendix C].

2.3.4.

NVC with rigidified artery

The same parameters as NVC in Sham were calculated (Fig. 11 in Appendix F). A similar approach for determining correct tests as in NVC Sham was also used. However, the sample sizes between the groups differed (CaCl $n=14$, Sham $n=17$), and thus, a Mack–Skillings⁴⁷ test (code from Pingel⁴⁸) was used instead of a Friedman test in non-parametric instances. Which tests were used eventually can be found in Table 2 in Appendix G. None of the parameters showed a significant difference among groups, so no post-hoc tests were performed.

2.3.5.

Correlation Matrices

The correlation value of all seed pair combinations was calculated (Fig. 12 in Appendix H). The use of global signal regression (GSR) has been a topic of much debate in resting state studies.⁴⁹ In this paper, the presented correlation matrices and statistics are without the use of GSR, to prevent the accidental removal of signal of interest. However, correlation matrices with GSR can be found in Fig. 13 in Appendix I. A Fisher’s $z$-transform was executed to calculate statistics.⁵⁰ The data of the correlation matrices were not normally distributed (tested with the Kolmogorov–Smirnov test), and therefore, a non-parametric Mann–Whitney $U$ test was performed for each ROI pair for comparisons between CaCl and Sham groups, and a Friedman test was chosen for tests over the three time points. An FDR was run on all seedpairs. To keep the possibility of seeing trends in the data, both $p$-values and $q$-values of the correlation matrixes can be found in Table 3 in Appendix J.

4.1.1.

General NVC Observations

Indications of neurovascular coupling in resting state have been found in previous studies^{10,12,41,54,55} as well as in the current one. Ma et al.¹² showed a clear neurovascular coupling in the resting state, and our results coincide very well with theirs, in both curve-shape and timing. In addition, Bruyns-Haylett et al.¹⁰ found hemodynamic curves as a response toward detected spontaneous activations that resemble our results very closely, albeit with a slower response. This difference might be explained by the presence of anesthesia in their study, which can alter the response.^52,56 We found a delay of roughly 1 s between the HbO and GCaMP peaks, similar to other resting state studies and ultrashort stimuli findings.^12,57,58 Studies with stimulations often show a later peak in HbO.^59–61 This is likely due to the stronger or prolonged neuronal response that is evoked in comparison to the resting state, requiring a greater influx of HbO that in turn takes longer to reach.

In the GCaMP-detected spontaneous activation data, we can see oscillations in both GCaMP and hemodynamic data. In general, GCaMP activation fluctuates around a baseline. Because of this, a spontaneous activation in a resting state recording is likely preceded by a dip in activity, and vice versa. Although a stimulus can cut into the timecourse and “demand” an activation at any point, the spontaneous activation will rely on the fluctuations around the baseline and build upon the natural increase to reach the threshold. When the spontaneous fluctuation frequency of neurons differs by even a small amount, eventually, the timecourses will fall out of sync and cancel each other out, which will result in an average that is back at baseline.

This oscillation is seen not only in GCaMP but also in hemodynamic data. The dip before the spontaneous neuronal activity is occasionally seen in other works. In our current results, the initial dip is clear, and the results resemble the preliminary results of Bruyns–Haylett very closely.¹⁰ We can clearly distinguish it in Ma et al.¹² as well, although to a lesser extent. In Wright et al.,⁶² we can spot it in the example of a single timecourse. On the other hand, the undershoot after the HbO influx is much more often observed.^{10,12,54–58,61} This dip only seems to occur in detected spontaneous activations^10,12,52,55 and very short and/or weak stimuli^{52,54–58,61} and disappears if the stimuli get stronger.^10,56,61

In the HbO-detected spontaneous activations, the HbO increases several seconds before crossing the threshold. This might be due to slow-wave, background fluctuations of HbO, that have a slower frequency than the response curve toward GCaMP. The threshold of HbO detection might only be reached if both the slow wave and higher-frequency fluctuations are at their peak.

4.1.2.

Left versus right brain area differences

When comparing the strength of the GCaMP activations above the detection threshold, we found significantly different responses between brain areas between all homotopic areas, with the right area showing significantly higher peaks than the left. A possible explanation for this could be the imaging setup. Mice were always positioned in the same manner. The location of the researcher, the angle of the LEDs to activate the GCaMP and measure the hemodynamics (acquisitions were otherwise performed in a dark room), and other factors could have had an effect on the acquisition. Interestingly, the number of times that GCaMP timecourses cross the threshold in the different areas does not show this pattern with the same intensity (Fig. 8 in Appendix B).

Although the GCaMP peaks are generally higher on the right side of the brain, the hemodynamics do not show this same pattern, with the homotopic regions differing significantly in neither HbO/HbR peak, nor increase/decrease between left and right regions. With a difference in GCaMP activity, but none in hemodynamics, we logically end up with a different strength in neurovascular coupling between right and left areas. This is more visible when looking at the strength as calculated with the increases, with all areas except for the retrosplenial showing a significantly stronger NVC in the left areas compared with the right.

4.1.3.

Medial versus lateral brain area differences

As regards the timing differences between brain areas, we found the retrosplenial cortex to have a significantly slower response than most other brain areas. The retrosplenial areas also have the highest HbR peak and the biggest HbR decrease. Last, we can generally distinguish two groups in the HbO data: the motor and retrosplenial areas with a lesser response, and the sensory and visual areas with a greater response, although these differences are not always significant. The difference in timing, HbO influx, and HbR peak might all stem from the same phenomenon. The vascular density is not homogeneous over the brain.⁶³ The cerebral arteries feeding large parts of the more superficial parts of the brain wrap around the brain and come in laterally. This means that the middle cerebral artery comes across the sensory region, and the posterior cerebral artery feeds the visual area. The motor region and the retrosplenial area are located more medially, which might mean that the blood supply to those areas is less efficient, and thus, the delay is longer, and the peak in HbO values is lower. This would also explain the higher HbR peak; as we can see in the delay, the sensory and visual areas receive the needed HbO increase significantly quicker than the retrosplenial areas, whereas the motor areas stay in between those values.