Spatiotemporal features of neurovascular (un)coupling with stimulus-induced activity and hypercapnia challenge in cerebral cortex and olfactory bulb

Abstract

Carbon dioxide (CO₂) is traditionally considered as metabolic waste, yet its regulation is critical for brain function. It is well accepted that hypercapnia initiates vasodilation, but its effect on neuronal activity is less clear. Distinguishing how stimulus- and CO₂-induced vasodilatory responses are (dis)associated with neuronal activity has profound clinical and experimental relevance. We used an optical method in mice to simultaneously image fluorescent calcium (Ca²⁺) transients from neurons and reflectometric hemodynamic signals during brief sensory stimuli (i.e., hindpaw, odor) and CO₂ exposure (i.e., 5%). Stimuli-induced neuronal and hemodynamic responses swiftly increased within locally activated regions exhibiting robust neurovascular coupling. However, hypercapnia produced slower global vasodilation which was temporally uncoupled to neuronal deactivation. With trends consistent across cerebral cortex and olfactory bulb as well as data from GCaMP6f/jRGECO1a mice (i.e., green/red Ca²⁺ fluorescence), these results unequivocally reveal that stimuli and CO₂ generate comparable vasodilatory responses but contrasting neuronal responses. In summary, observations of stimuli-induced regional neurovascular coupling and CO₂-induced global neurovascular uncoupling call for careful appraisal when using CO₂ in gas mixtures to affect vascular tone and/or neuronal excitability, because CO₂ is both a potent vasomodulator and a neuromodulator.

Introduction

The concept of neurovascular coupling refers to chemical signalling in the neuropil that links cellular activity to dynamic hemodynamic alterations, ¹ mechanisms ² which strive to match the high metabolic demands of synaptic activity^3,4 by increasing local cerebral blood flow (CBF) and/or volume (CBV) to deliver more nutrients for oxidative and glucose metabolism (CMR_O2 and CMR_glc).^5,6 Impaired neurovascular coupling indicates onset of pathology as implicated in healthy aging, ischemia, or neurodegeneration.^7,8 Moreover, neurovascular coupling is also critical for techniques that use hemoglobin (Hb) in blood circulation as the image contrast mechanism like functional MRI (fMRI) and near-infrared spectroscopy (fNIRS).^9,10 Each component of Hb, i.e., oxyhemoglobin (oxyHb) and deoxyhemoglobin (deoxyHb), have different magnetic and optical properties, thereby altering ratios of oxyHb to deoxyHb within perfusing blood vessels can affect the blood oxygenation level dependent (BOLD) contrast in fMRI and fNIRS scans.

Carbon dioxide (CO₂) is often thought of as a simple end gas product to be disposed as waste. ¹¹ However, tight regulation of CO₂ within the neuropil is critical for normal brain function ¹² as it impacts respiratory drive, affinity of oxygen binding to Hb, vascular tone to dilate or constrict blood vessels, and interstitial acidity. ¹³ Although CO₂ is internally generated as a localized end-product of nutrient oxidation (as reflected in CMR_O2 and CMR_glc) following an intense change in neuronal activity, its global concentration can be artificially enriched through inhalation, presumably without affecting CMR_O2 and/or CMR_glc.

Deviations from normal CO₂ levels (i.e., 35–45 mmHg; normocapnia) include scenarios of increased CO₂ (i.e., >45 mmHg; hypercapnia) or decreased CO₂ (i.e., <35 mmHg; hypocapnia). Hypercapnia and hypocapnia are widely used in clinical and experimental realms. Since altering the CO₂ in gas mixtures is repeatedly employed in clinical practice (e.g., for neuroprotective measures ¹⁴ ) and to improve neuroimaging contrast (e.g., in calibrated fMRI ¹⁵ ) it is essential to classify the degree to which vascular responses to stimulus-induced activity and hypercapnia are (un)correlated with neuronal activity, both spatially and temporally.

Interstitial CO₂ has direct influence on tissue pH because CO₂ reacts with water to produce carbonic acid (i.e., H₂CO₃) that spontaneously dissociates to form protons (H⁺) and bicarbonate ( ${\text{HCO}}_3^{-}$ ). ¹⁶ Considering interstitial pH affects polarization of cell membranes^17–20 and hypercapnia has been thought to act as an anesthetic ²¹ by decreasing neuronal excitability, ¹⁸ inhaled CO₂ may be a potent neuromodulator. Conversely, hypocapnia increases neuronal excitability and is used clinically not only to induce seizures ²² but also prolong the seizure epochs for therapeutic benefit as in electroconvulsive therapy. ²³

Additionally, acidic pH originating from high CO₂ levels activates voltage-gated potassium (K⁺) channels to hyperpolarize endothelial cells. Thus, interstitial CO₂ acts as a powerful vasomodulator. ²⁴ Short-lived epochs of hypercapnia and hypocapnia have opposite effects on cerebral vasculature promoting vasodilation and vasoconstriction, respectively. ²⁵ Since CO₂ is a potent vasodilator, it affects the radius of the vessels causing fourth order change in CBF and a second order change in CBV, ²⁶ but altering CO₂ concentration is commonly viewed as a safe and easy tool by neuroimaging methods like fMRI and fNIRS to alter contrast.^27,28

We used widefield optical imaging with transgenic mice, expressing genetically encoded calcium (Ca²⁺) indicators (GECIs), ²⁹ to concurrently map fluorescent Ca²⁺ transients from neurons and reflectometric Hb signal to examine neurovascular coupling. Our results show hypercapnia transiently suppresses brain-wide neuronal activity during CO₂-induced vasodilation. The spatiotemporal patterns of neurovascular uncoupling with hypercapnia are starkly different from the classical neurovascular coupling observed with stimuli-induced responses. Thus, we encourage restraint when using CO₂ in gas mixtures to affect neuronal excitability for neuroprotective measures and/or vascular tone for altering neuroimaging contrast. Our results emphasize that CO₂ is potent in both neuromodulation and vasomodulation, and should not be viewed simply as a metabolic waste product.

Materials and methods

Mice preparation

All mice were housed on a 12-hour light/dark cycle. Food and water were available ad libitum. Male and female mice were adults, 6–12 weeks old, 25–30 g, at time of imaging. Three strains of mice were used: Thy1-GCaMP6f (n = 10), OMP-GCaMP6f (n = 15), Thy1-jRGECO1a (n = 4). The cortical mouse strain Thy1-GCaMP6f was made by the Genetically Encoded Neuronal Indicator and Effector (GENIE) Project. It produces green fluorescent calcium (Ca²⁺) indicator, GCaMP6f, in Thy1-expressing pyramidal neurons in the cerebral cortex and is commercially available from Jackson Labs (JAX 024339, GP5.11 line). The olfactory mouse strain OMP-GCaMP6f, where OMP stands for olfactory marker protein, was bred using floxed-GCaMP6f reporter mice (JAX 024105) crossed with OMP-Cre mice (JAX 006668). ³⁰ The resulting OMP-GCaMP6f mice have GCaMP6f expression to study olfactory receptor input neurons in the glomerular layer of the olfactory bulb. Another cortical mouse strain Thy1-jRGECO1a was also made by the GENIE Project and produces red fluorescent Ca²⁺ indicator, jRGECO1a, in Thy1 expressing pyramidal neurons in the cerebral cortex and is also commercially available (JAX # 030536, GP8.31 line).

All surgical procedures were approved by the Yale University Institutional Animal Care and Use Committee (IACUC) and follow the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals. All procedures complied with the regulations of the Animals (scientific procedures) Act 1986 and reporting follows the ARRIVE (animal research: reporting of in vivo experiments) guidelines. We created closed cranial windows above the cerebrum and the olfactory bulb of the mouse head at least 5 days before the first experimental recording. This optical window technique was specially developed with four goals in mind: i) to fix the mouse head with a non-metal head holder; ii) to allow mesoscopic optical measurement from the surface of the brain; iii) to allow for MRI compatibility for recordings of mouse brain without susceptibility effect of the cranial window or head fixation; iv) to keep the cranial window clear and secure for months (Supplementary Methods and Figure S1A).

Multi-modal optical imaging setup

The optical setup was designed to measure wide-field Ca²⁺ fluorescent and reflectometric CBV signals using the same filter and camera configuration (Figure 1), and the distance between the cranial window and the tip of the camera setup was fixed. For Thy1-GCaMP6f and OMP-GCaMP6f mice, data were collected alternating the green fluorescent Ca²⁺ image and the reflectometric CBV image (Figure 1(a)). The GCaMP6f fluorophore was excited with a blue light emitting diode (LED) by Thorlabs M470F3 and bandpass filtered (ET460/30x, Chroma Technology Corporation, Bellows Falls, VT) between 443 and 487 nm (Figure 1(b)). Since we measured the reflectometric signal at 570 ± 5 nm (FB570-10, Thorlabs, Newton, NJ) using a Thorlabs M565F3 LED (Figure 1(b)), we avoided the classic beam splitter/dichroic mirror approach and instead illuminated the surface directly with both light sources as close to perpendicular as the optical setup allowed. Before the experiments we made sure that the illuminating light intensities were as even on the brain surface as possible. The reflected and emitted fluorescent signals were targeted to enter to a built in-house objective, which contained two achromatic doublets (AC050-010-A and AC080-010-A, Thorlabs, Newton, NJ) with a 1 mm radius aperture and a 500 nm longpass filter (FELH0500, Thorlabs, Newton, NJ), and connected to a Rolera-XR camera (QImaging Corporation, Surrey, BC) with 2/3-inch CCD array, 12-bit image collection, and 696 × 520 image resolution. The 500 nm longpass filter allowed both the fluorescent and reflectometric images to be collected by multiplexing. The images were collected with MATLAB Image Acquisition Toolbox (R2019) (The Mathworks Inc., Natick, MA) and the LEDs were switched in synchrony with the camera triggers by CED µ1401 and Spike2 (Cambridge Electronic Design Ltd, Cambridge, UK).

Figure 1.

Dual channel optical recordings of wide-field Ca²⁺ fluorescent and reflectometric CBV signals. (a) The in-house optical assembly includes two achromatic lenses (5- and 8-mm diameter each with 10 mm focal distances), a 1 mm hole (aperture) and a 500 nm long-pass filter. The blue light absorbed by GCaMP6f, producing green Ca²⁺ fluorescence, was recorded with a CCD camera. The alternating green light source at 570 nm (above the absorption spectrum of GCaMP6f) reflects from the tissue and recorded by multiplexing and (b) The theoretical GCaMP6f excitation and emission spectra are shown on the left axis, normalized to their peak value. We recorded above 500 nm in the shaded green area for Ca²⁺ fluorescent signal and in the 570 ± 5 nm range (green-yellow area) for the reflected ΔCBV signal. On the right axis we show the molar extinction coefficient for oxyhemoglobin (HbO₂) and deoxyhemoglobin (Hb). Note that in the shaded areas the absorption of the two chromophores are almost equal (45195 M/cm for Hb and 44729 M/cm for HbO₂), only 1% difference.

The 570 nm reflectometric signal was recorded at an isosbestic point of Hb (Figure 1(b)), i.e., the oxygen saturation of Hb does not influence the CBV signal intensity. We assumed that we measured the changes of the total Hb concentration (i.e., [oxyHb] + [deoxyHb]) which is the negative equivalent of the CBV change. This reflected signal was also used for hemodynamic correction of the fluorescent Ca²⁺ signal. The fluorescent Ca²⁺ signal change was equivalent to the variation of Ca²⁺ ion intracellular concentration in the observed area. In Thy1-GCaMP6f and Thy1-jRGECO1a mice the Ca²⁺ signal reflected the change of the excitatory neuronal activity, whereas for OMP-GCaMP6f mice the Ca²⁺ signal described the olfactory input to glomeruli. For Thy1-jRGECO1a mice, we used a similar optical setup, but the Ca²⁺ excitation wavelength was 560 ± 40 nm and recorded above a 600 nm cutoff, whereas the CBV was determined by measuring Hb absorption at 685 nm and 800 nm isosbestic wavelengths.

Experimental setup for multi-modal imaging and physiological monitoring

On the day of the experiment the surgically prepared and spontaneously breathing mice were anesthetized with 0.25–0.5% isoflurane and subcutaneous infusion of dexmedetomidine, an α2-adrenergic receptor agonist (Dexdomitor, Covetrus, Elizabethtown, PA), with 0.25 mg/kg/hour dosage. The mice were laid in supine position on a heated platform (ATC1000, World Precision Instruments, Saratoga, FL) with the headpost fixed to a stand. LED lights illuminated the optical window evenly through fiber optics as close to perpendicular to the cranial window as possible. The camera was positioned exactly above the head with a micromanipulator. Rectal temperature and respiration were continuously recorded during the experiment. The respiration rate was measured with a laser Doppler probe (Oxyflow 2000; Oxford Optronics, Oxford, UK) positioned to the chest wall allowing spontaneous breathing movement recorded in the laser Doppler signal.

We obtained cortical data from Thy1-GCaMP6f and Thy1-jRGECO1a mice, and bulbar data from OMP-GCaMP6f mice. In Thy1-GCaMP6f and Thy1-jRGECO1a mice two stimulating electrodes were attached to the hindpaw, between toes 1–2 and 4–5 with removable microclips. The conductivity between skin and electrodes was increased with electrode gel. Hindpaw stimulation was provided by a stimulus/isolator unit (A365, WPI, Inc, Saratoga, FL), which was controlled by the Spike2 software. Stimulation parameters were 1–1.5 mA, 10 Hz, with 1% duty cycle for 30 s duration. The 10 Hz stimulation was selected for maximal response under the selected anesthesia,^31–33 and either 1 or 1.5 mA were used to evoke a response (without muscular contractions) that differentiated from baseline noise. In OMP-GCaMP6f mice the odor stimulation was 10% of ethyl-butyrate in mineral oil perfused with the breathing air using Spike2 controlled solenoids for 30 s.^34,35 Hypercapnic challenges were performed by adding 5% CO₂ (inducing ∼50 mmHg PaCO₂ ³⁶ ) to the breathing gas mixture for 30 s by the Spike2 controlled solenoids.^37,38 During the hypercapnic episode the percentage of other gases decreased, but we avoided hypoxia because the breathing gas mixture (N₂: 72%, O₂: 28%) has slightly higher O₂ concentration, which provided normoxic O₂ concentration even during the hypercapnic challenge. The mice and the whole experimental setup during recording were positioned in a black box to avoid ambient light contamination from the room. We first measured stimulation-induced CBV and Ca²⁺ responses in the cerebral cortex or the olfactory bulb, and then we applied hypercapnic challenges. Thereafter, stimulation and hypercapnic challenges were repeated several times in random order.

Data processing for Ca²⁺ fluorescent and CBV reflectometric signals

Multi-modal data were collected with 10 fps temporal resolution using 60 ms exposure time (i.e., 60% duty cycle). Ca²⁺ and CBV data were collected alternatively (multiplexed recording), where the switch on/off cycles of the two different light sources were not overlapped to minimize cross-contamination. Data were cropped to the embedding rectangular area of cortex/bulbar to decrease file size. Then images were separated into raw Ca²⁺ and CBV signals. Since imperfections created during optical window preparation can locally alter the recorded signal, such areas were identified in the vessel/anatomical image and excluded from further analysis for both Ca²⁺ and CBV signals. Both Ca²⁺ and CBV data were adjusted by the dark image (collected only with ambient light). Since the origins of the two signals were different, further analysis required different preprocessing steps. For example, for GCaMP6f the green emitted Ca²⁺ fluorescence signal is obtained with blue excitation to calculate fluorescence change (ΔF/F₀). While the emitted light intensity depends on the Ca²⁺ ion flux (measured ΔF/F₀ signal), the surrounding Hb can absorb the emitted green light (attenuated ΔF/F₀ signal). Since both Ca²⁺ and CBV signals change were measured concurrently, we needed local information of Hb in the vicinity of the emitted green light (i.e., same pixel) to correct the Ca²⁺ intensities. We achieved this by measuring the attenuation of the reflectometric 570 nm light. The 570 nm is one of the isosbestic points of the oxyHb and deoxyHb absorption curves, so the oxygenation change of the Hb did not confound this reflectometric signal to represent CBV changes in the region of interest (ROI). Based on the modified Beer-Lambert equation, we used the raw CBV and raw Ca²⁺ signals to estimate the corrected Ca²⁺ signal (Figure S1B),^39–41

$$\text{corrected}\hspace{0.17em}{\text{Ca}}^{2+}=\text{raw}\hspace{0.17em}{\text{Ca}}^{2+}\times \text{exp}[-\text{ln}\left(\text{raw}\hspace{0.17em}\text{CBV}\right)\times \text{d}\left(\text{GFP}\right)$$

$$\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\times \text{ε}\left(\text{GFP}\right)/\text{d}\left(\text{CBV}\right)/\text{ε}\left(\text{CBV}\right)]$$

where d(GFP) and d(CBV) are the mean optical pathlength of emission of green fluorescent protein (GFP) at 500–530 nm and the mean optical pathlength of CBV signal at 570 nm, 270 and 280 µm, respectively,^39–41 and ε(GFP) and ε(CBV) are the molar extinction coefficients of averaged Hb at the same wavelengths, 20897.4 cm⁻¹ M⁻¹ and 44784 cm⁻¹ M⁻¹, respectively, as previously described by Scott Prahl (https://omlc.org/spectra/hemoglobin). Thus, all presented data will show the hemodynamically corrected Ca²⁺ signal. We also simulated the corrected Ca²⁺ signals by modifying the second part of the equation from 20% to 240% of the original value to estimate the potential error of hemodynamic under-correction of the hypercapnic Ca²⁺ signal (Figure S1C). We found that not even 180% change can completely nullify the Ca²⁺ signal change.

Because of fast, burst-like Ca²⁺ fluorescence activity in the cortex, we generated Student’s t-test maps by comparing the resting period to consecutive 1 s bins for the entire 30 s stimulation epoch (using ttest2 function in Matlab), where the aggregate was Gaussian filtered with a kernel size of 3 to generate (de)activated t-map ROIs. We extracted the average time series from those ROI voxels where the p-value was smaller than 0.05. The same process was applied for the CBV data. Trial based time-courses were averaged pixel-wise and shown as their mean and standard deviation (SD), unless otherwise specified. Overlap of the Ca²⁺ and CBV t-maps were calculated as the Sørensen-Dice similarity coefficient (using dice function in MATLAB). We also calculated time-to-peak (TTP) and return-to-rest time (RRT).

Results

We measured the raw fluorescent Ca²⁺ signal from the cerebral cortex (hindpaw (HP) stimulus and CO₂ exposure) and olfactory bulb (odor stimulus and CO₂ exposure) along with the reflectometric Hb absorption change at the isosbestic wavelength. The reflectometric signal was used to calculate the CBV change from the same spatial location as the Ca²⁺ signal and for hemodynamic correction of the fluorescent Ca²⁺ signal.^39–41 The sensory stimulation data were needed to confirm that the hypercapnia data are physiologically relevant. Moreover, data from the bulb and cortex were used to confirm universality of responses to sensory and CO₂ stimuli. Respiratory rate significantly changed from 138 ± 11 breaths/min at baseline to 213 ± 33 breaths/min during CO₂ exposure and returned to 134 ± 10 breaths/min one minute after cessation of hypercapnic challenge.

Cortical recordings

During HP stimulation bilaterally localized cortical activations were observed in both Ca²⁺ and CBV t-maps, as shown from one representative Thy1-GCaMP6f mouse (Figure 2(a)) and across other Thy1-GCaMP6f mice (Figure S2A), where the spatial overlap between Ca²⁺ and CBV t-maps was ∼75% (Table S1).

Figure 2.

Representative single trial data from one Thy1-GCaMP6f mouse during hindpaw (HP) stimulation and hypercapnia challenge (or CO₂ exposure). During HP stimulation cortical (a) t-maps of Ca²⁺ (left) and CBV (right) signals, and (b) temporal responses of Ca²⁺ (red) and CBV (green) signals. With HP stimulation there is marked increase in Ca²⁺ signal during increase in CBV signal. During CO₂ exposure cortical (c) t-maps of Ca²⁺ (left) and CBV (right) signals, and (d) temporal responses of Ca²⁺ (purple) and CBV (blue) signals. With CO₂ exposure there is marked decrease in Ca²⁺ signal during increase in CBV signal. (e–f) Spatiotemporal neurovascular coupling and uncoupling in cerebral cortex during stimulation and hypercapnia from the same Thy1-GCaMP6f mouse. Ca²⁺ (top) and CBV (bottom) activity patterns are shown relative to the start of the of hypercapnic challenge (30 s in absolute time) with 1 s binned maps.

Cortical Ca²⁺ and CBV responses with HP stimulation exhibited rapid signal increase/decrease coinciding well with stimulus onset/offset timing, indicating strong neurovascular coupling (Figure 2(b) and (e)). In Thy1-GCaMP6f mice the TTP with HP stimulation for cortical Ca²⁺ response (∼7 s) was faster than the cortical CBV response (∼14 s), whereas the RRT were quite similar (∼22 s) for cortical Ca²⁺ and CBV responses (Table S2). Average magnitudes of cortical Ca²⁺ and CBV responses were +0.3% and +1.3%, respectively.

Upon CO₂ exposure bilateral global increases were detected in cortical CBV t-map but with bilateral global deactivation in cortical Ca²⁺ t-map, as shown in the same representative Thy1-GCaMP6f mouse (Figure 2(c)) and across other Thy1-GCaMP6f mice (Figure S2B). Despite neurovascular uncoupling with CO₂ exposure, the spatial overlap between cortical Ca²⁺ and cortical CBV t-maps was ∼96% (Table S1).

CO₂-induced cortical Ca²⁺ and CBV responses exhibited very slow, but global, signal increase/decrease coinciding poorly with onset/offset timing of CO₂ exposure (Figure 2(d)), indicating distinct spatiotemporal patterns of neurovascular uncoupling (Figure 2(f)). In Thy1-GCaMP6f mice, the TTP for CO₂-induced cortical Ca²⁺ response (∼30 s) was similar to CO₂-induced cortical CBV response (∼36 s), whereas the RRT were quite similar (∼50 s) for the CO₂-induced Ca²⁺ and CBV responses (Table S2). Average magnitudes of cortical Ca²⁺ and CBV responses were, respectively, −0.5% and +3.0% for the entire CO₂ exposure epoch, respectively.

We also recorded from Thy1-jRGECO1a mice which emit red Ca²⁺ fluorescence (Figures 3 and S3). In Thy1-jRGECO1a mice HP-induced and CO₂-induced changes were comparable to observations in Thy1-GCaMP6f mice. In Thy1-jRGECO1a mice the spatial overlap between cortical Ca²⁺ and CBV t-maps was also high, with HP stimulation (Figure 3(a)) and CO₂ exposure (Figure 3(b)). In Thy1-jRGECO1a mice the TTP for CO₂-induced cortical Ca²⁺ response (∼32 s) was similar to the CO₂-induced cortical CBV response (∼33 s), whereas the RRT were quite similar (∼50 s) for the CO₂-induced Ca²⁺ and CBV responses (Figure 3(c)).

Figure 3.

Representative data from one Thy1-jRGECO1a mouse during hindpaw (HP) stimulation and hypercapnia challenge (or CO₂ exposure). (a) During HP stimulation cortical t-maps of Ca²⁺ (left) and CBV (right) signals. (b) During CO₂ exposure cortical t-maps of Ca²⁺ (left) and CBV (right) signals. Both (a) and (b) show single trial data. (c) Averaged (mean ± SD) dynamics for all responses of cortical Ca²⁺ (purple) and CBV (blue) signals during CO₂ exposure all Thy1-jRGECO1a mice (n = 4) shown in fractional changes.

Bulbar recordings

During odor stimulation bilaterally localized bulbar activations were observed in both Ca²⁺ and CBV t-maps, as shown from one representative OMP-GCaMP6f mouse (Figure 4(a)) and across other OMP-GCaMP6f mice (Figure S4A), where the spatial overlap between the bulbar Ca²⁺ and CBV t-maps was ∼87% (Table S1).

Figure 4.

Representative single trial data from one OMP-GCaMP6f mouse during odor stimulation and hypercapnia challenge (or CO₂ exposure). During odor stimulation bulbar (a) t-maps of Ca²⁺ (left) and CBV (right) signals, and (b) temporal responses of Ca²⁺ (red) and CBV (green) signals. With odor stimulation there is marked increase in Ca²⁺ signal during increase in CBV signal. In contrast to the cortical response during hindpaw stimulation, in the bulb there is a sharp increase in Ca²⁺ signal immediately after onset of odor stimulus. During CO₂ exposure bulbar (c) t-maps of Ca²⁺ (left) and CBV (right) signals, and (d) temporal responses of Ca²⁺ (purple) and CBV (blue) signals. With CO₂ exposure there is marked decrease in Ca²⁺ signal during increase in CBV signal. In contrast to the cortical response during CO₂ exposure, in the bulb there is a sharp increase in Ca²⁺ signal immediately after offset of hypercapnia challenge. (e-f) Spatiotemporal neurovascular coupling and uncoupling in olfactory bulb during stimulation and hypercapnia from the same Thy1-GCaMP6f mouse. Ca²⁺ (top) and CBV (bottom) activity patterns are shown relative to the start of the of hypercapnic challenge (30 s in absolute time) with 1 s binned maps.

Odor-induced bulbar Ca²⁺ and CBV responses exhibited rapid signal increase/decrease coinciding well with stimuli onset/offset timing (Figure 4(b)), indicating well-accepted neurovascular coupling. In OMP-GCaMP6f mice, the TTP for odor-induced bulbar Ca²⁺ response (∼1 s) was much faster than the odor-induced bulbar CBV response (∼5 s), whereas the RRT were quite similar (∼27 s) for the odor-induced bulbar Ca²⁺ and CBV responses (Table S2). Magnitudes of bulbar Ca²⁺ and CBV responses were, respectively, +3.3% and +1.1% for the entire odor stimulation duration.

During CO₂ exposure bilateral global increase was measured in the bulbar CBV t-map and bilateral global deactivation was observed in the bulbar Ca²⁺ t-map, as shown in the same representative OMP-GCaMP6f mouse (Figure 4(c)) and across other OMP-GCaMP6f mice (Figure S4B), where the spatial overlap between bulbar Ca²⁺ and CBV t-maps was ∼82% (Table S1).

CO₂-induced bulbar Ca²⁺ and CBV responses exhibited very slow, but global, signal increases/decreases delayed from the onset/offset timing of CO₂ exposure (Figure 4(d)), indicating noticeable spatiotemporal features of neurovascular uncoupling (Figure 4(f)). In OMP-GCaMP6f mice the TTP for CO₂-induced bulbar Ca²⁺ response (∼27 s) was similar to the CO₂-induced bulbar CBV response (∼32 s), whereas the RRT were quite similar (∼32 s) for the CO₂-induced Ca²⁺ and CBV responses (Table S2). Magnitudes of bulbar Ca²⁺ and CBV responses were, respectively, −1.0% and +1.0% for the entire CO₂ exposure epoch.

Comparison of dynamic Ca ^{2 +} and CBV responses with HP/odor stimuli and CO₂ exposure

Overall dynamic patterns of Ca²⁺ and CBV responses obtained from all Thy1-GCaMP6f (n = 10) and OMP-GCaMP6f (n = 15) mice (Figure 5; Tables S1, S2) showed similar trends as the individual GCaMP6f mice (Figures 2, S2, 4, S4). While Ca²⁺ responses to hypercapnia in the cortex and bulb were quite similar in amplitude and dynamics, the magnitude of the Ca²⁺ response to HP stimulation was significantly lower than with odor stimulation (Figures 5(a) and (b)). However, the Ca²⁺ response to hypercapnia in the bulb had a peculiar signal reversal immediately following the CO₂ offset, a pattern that was not observed in the cortex (Figures 5(a) and (b)). While the CBV responses in the cortex and bulb were generally similar in dynamics, the CBV response to hypercapnia in the cortex was almost four times greater than in the bulb, and moreover, the CBV response in the bulb peaked much earlier than in the cortex (Figures 5(c) and (d)).

Figure 5.

Average (mean ± SD) responses from all Thy1-GCaMP6f (cortical; n = 10) and OMP-GCaMP6f (bulbar; n = 15) mice. (a) Dynamic cortical Ca²⁺ responses during hindpaw (HP) stimulation (red) and CO₂ exposure (purple) from Thy1-GCaMP6f mice. (b) Dynamic bulbar Ca²⁺ responses during odor stimulation (red) and CO₂ exposure (purple) from OMP-GCaMP6f mice. (c) Dynamic cortical CBV responses during hindpaw (HP) stimulation (green) and CO₂ exposure (blue) from Thy1-GCaMP6f mice and (d) Dynamic bulbar CBV responses during odor (green) and CO₂ exposure (blue) from OMP-GCaMP6f mice.

Discussion

Regulating CO₂ in the brain is critical ¹² due to its impact on factors such as: respiration, oxygen-hemoglobin binding, vascular tone, and interstitial acidity. ¹³ Although stimulus-induced activity and hypercapnia trigger vasodilation, the mechanisms by which interstitial CO₂ levels are raised differ (by metabolism and inhalation, respectively). Thus, establishing how these vasodilatory responses are (dis)associated with neuronal activity has clinical^17–23 and experimental^24–28 relevance.

We simultaneously measured reflectometric CBV signal and fluorescent Ca²⁺ signal (hemodynamic corrected) from mouse brain (Figures 1 and S1). For cortical recordings we used green fluorescent Ca²⁺ in Thy1-GCaMP6f mice (Figures 2 and S2) and red fluorescent Ca²⁺ in Thy1-jRGECO1a mice (Figures 3 and S3), whereas for bulbar recordings we used green fluorescent Ca²⁺ in OMP-GCaMP6f mice (Figures 4 and S4). These multi-modal recordings were used to examine the neurovascular (un)coupling during stimuli and CO₂ exposure (Figure 5; Tables S1, S2). Our results reveal contrasting effects of hypercapnia on neuronal response (deactivation or decrease) and vasodilatory effects (activation or increase), which are spatiotemporally disassociated in a complex manner when compared to conventional neurovascular coupling, with vasodilation and increased neuronal activity.

Technology development to measure brain-wide neurovascular (un)coupling

In conjunction with green and red fluorescent Ca²⁺ signal, the reflectometric CBV signal was measured at two isosbestic points of Hb absorption: at 570 nm to record the green Ca²⁺ signal (in Thy1-GCaMP6f and OMP-GCaMP6f mice) simultaneously with CBV, and at 800 nm to measure the red Ca²⁺ signal (Thy1-jRGECO1a mice) simultaneously with CBV.

Using the GCaMP6f as an example (Figure 1), we recorded the CBV and Ca²⁺ signals in two different spectral bands, one for the green fluorescent signal of neuronal activity and another for the hemodynamic signal of Hb concentration changes. Ideally, the blue absorption spectral band below 490 nm excited the green fluorescent chromophore emitting light within the 500–530 nm band. Applying an appropriate dichroic beamsplitter, the two spectral regimes were well separated to avoid cross-talk. The clear advantage of this widely used method is that the excitation and emitted light go through the same optical system ensuring that only the same areas are illuminated and recorded. ⁴² The intensity of emitted fluorescent light depends on both the concentration of the fluorophores and the concentration of absorptive chromophores in the tissue. Since we calculate the ΔF/F₀ signal, its value remains stable until there is no time dependent change in the absorptive chromophore concentration. The highest concentration chromophores are the Hb molecules in the cerebral vessels and because of neurovascular coupling the stimulation itself changes not only the neuronal activity but also CBV. Therefore, we had to measure the CBV independently, not only for the hemodynamic response but for the correction of fluorescence signal intensity with the changing Hb concentration.^39–41 This can be achieved either below the absorption spectrum of the fluorophore with ultraviolet^43,44 or above it with visible light. Since ultraviolet light requires specific lenses and sensitive camera sensor, we used the isosbestic wavelength of Hb absorption at 570 nm. The parallel application of the reflectometric measurement forced us to give up the classic GFP dichroic mirror and emission filter to illuminate the surface separately from the objective. This approach is still viable because Hb absorbance declines rapidly above 550 nm, the higher end emissions can minimally influence the fluorescent signal and the 570 nm is above the absorption spectrum of the green fluorescent Ca²⁺ indicator.

Mechanisms of CO₂-induced neuronal depression and vasodilation

Traditionally CO₂ is simply thought of as a metabolic waste product. ¹¹ However, recent work proposes CO₂ as a mediator of neurovascular coupling, ⁴⁵ and our results strongly suggest CO₂ is a vasomodulator and a neuromodulator. It is well documented that CO₂ is a potent vasodilator, ²⁴ through a process which involves activating voltage-gated K⁺ channels to hyperpolarize endothelial cells. However, the mechanisms of CO₂-induced neuromodulation are more complex, as the CO₂-mediated neuronal inhibition may be due to a combination of direct and indirect mechanisms initiated after hypercapnic onset.

During CO₂ exposure carbonic anhydrase, a ubiquitously expressed enzyme in all cell types, increases extracellular H⁺ through the spontaneous CO₂ + H₂O → H⁺ +  ${\text{HCO}}_3^{-}$ reaction. The subsequent pH reduction is known to blunt neural excitability and disrupt membrane potential regulation by inhibiting voltage-dependent Ca²⁺ channels and hyperpolarization-gated and cyclic nucleotide-activated cation (HCN) channels.^46,47 Additionally, increased arterial CO₂ can give rise to elevated levels of adenosine, a potent neuromodulator, acting through G-protein coupled adenosine A1 receptors to inhibit neuronal activity by inhibiting excitatory Ca²⁺ channels, stimulating inhibitory K⁺ channels, or diminishing the excitatory synaptic release of glutamate.^46,48 The rise in interstitial adenosine presumably originates/derives from CO₂-mediated inhibition of adenosine kinase, the enzyme responsible for metabolizing adenosine for the production of ATP, or CO₂/pH modulation of purinergic receptors on local astrocytes or microglia. ⁴⁹

While our results show consistent responses to CO₂ exposure across the surface of the murine brain, we have also identified a novel post-stimulus effect in the olfactory bulb. Immediately following the end of the CO₂ exposure and diminished Ca²⁺ fluctuations, we report a post-stimulus “positive” peak in Ca²⁺ activity (Figure 5(b)). This complex response may be a resulting from activation of acid-sensing ion channels (ASICs) known to be densely expressed in the olfactory bulb ⁵⁰ and other deeper regions of the brain. ⁵¹ Following this mechanism of excitation, the excess H⁺ from CO₂ conversion would bind and activate ASICs facilitating cellular influx of sodium (Na⁺), increasing neuronal excitability. ⁵² This may be a protective mechanism against hypercapnic acidosis clearing excess CO₂/H⁺, which seems preferential to the bulb.

Features of spatiotemporal neurovascular (un)coupling during induced activity and hypercapnia

The vast difference in the spatial extent of stimuli-induced local neurovascular coupling vs. CO₂-induced global neurovascular uncoupling (Figures 2 and 4) can be explained by specific wiring connections of each stimulus to localized brain regions vs. non-specific distribution of inhaled gas to the entire body. Stimuli-induced CBV increases were temporally shorter than the CO₂-induced in both the cortex and the bulb (Figures 5(c) and (d)). Additionally, the CO₂-induced global neurovascular uncoupling in the cortex produced a much larger peak CBV amplitude compared to the stimuli-induced, while the responses to both challenges were generally similar in amplitude for bulb (Figures 5(c) and (d)). The observed difference in cortical CBV amplitude could be due to a domination of CBV (and BOLD) response from the larger pial vessels which are located mostly in the superficial lamina of the cortex.^53,54 Interestingly, while HP stimulation and CO₂ exposure elicit an increase in CBV, the comparison of Ca²⁺ responses between the two challenges highlights the CO₂-induced neurovascular uncoupling (Figures 5(a) and (b)). The differences in fluorophore distribution in the cortex (diffuse) and the bulb (condensed) further highlight the breadth and magnitude of the neurovascular uncoupling, respectively.

In Thy1-GCaMP6f the source of the cortical Ca²⁺ signal is likely dendrites within the neuropil. A previous calibrated fMRI and electrophysiology study in rat cerebral cortex showed laminar distribution of sensory-evoked electrical activity, where the highest magnitude was located in the middle and deeper layers. ⁵⁴ In the mouse cerebral cortex ⁵⁵ we estimate the distance from the surface of the cranial window to the location of highest neuropil density in middle to deeper cortical layers to be 2–3 mm, which is outside the range of fluorescent optical recordings, even with the 1.2–2.5 mm thick transparent region of the cranial window. Similarly, in OMP-GCaMP6f the origin of the bulbar Ca²⁺ signal is olfactory receptor input neuron synapsed in the glomerular layer. A recent fMRI and optical imaging study in the rat olfactory bulb demonstrated highest odor-induced responses in the glomerular layer. ⁴² In the mouse olfactory bulb ⁵⁵ we expect the gap from the top of the cranial window to the glomerular layer to be around 1.5 mm. These suggest a factor two higher optical signal for the bulbar data.

As discussed above, the neuronal inhibition effects of CO₂ are multifaceted involving activities of carbonic anhydrase, voltage-dependent Ca²⁺ channels, hyperpolarization-gated and cyclic HCN channels, and adenosine A1 receptors. It is likely these cellular machineries are widely distributed in cerebral cortex and olfactory bulb. However, the presence of a post-stimulus “positive” peak in the bulb (Figure 5(b)) is most likely due to the activity of ASICs, which are more prominent in specific brain regions including the bulb.^50,51

Impact of current results on using CO₂ gas mixtures in experimental practice

Calibrated fMRI requires measurements of BOLD signal along with CBF and/or CBV to estimate CMR_O2. ⁵⁶ Calibration is often performed using brief exposures to slightly elevated CO₂ with inhaled air to obtain maximal CBF (or CBV) response. ⁵⁷ This standard method for calibrated fMRI is presumed to purely produce vascular changes in BOLD signal via CBF/CBV, with minimal or no CMR_O2 changes. ⁵⁸ However, this iso-metabolic presumption remains under debate due to variations in the duration of the hypercapnic challenge, the concentration of CO₂ inhaled or degree of hypercapnia induced, methods of exposure, techniques used to identify increases in arterial CO₂, and brain regions assessed – all of which contribute to substantial outcome variability in the literature. ¹⁵ Some studies demonstrate no statistically significant change in CMR_O2 with CO₂ concentrations of 5–6%, ⁵⁹ whereas others have identified either a significant decrease^60–62 or increase ⁴⁹ in estimated CMR_O2. Although these variations in CMR_O2 estimation may arise due to experimental differences, any CMR_O2 change with hypercapnia depends on the correlative accuracy of these methods in the absence of actual changes in neuronal activity.

The sensitivity of calibrated fMRI to CMR_O2 estimation requires an accurate measurement of hemodynamic and neuronal, or lack thereof, responses to hypercapnia. While studies have demonstrated increases, decreases, and no change in CMR_O2 during hypercapnia, ¹⁵ in calibrated fMRI studies hypercapnic challenge is still the most popular method to get the maximal BOLD response while assuming no CMR_O2 changes. However, gas-free calibrated fMRI methods are available and have been applied in murine and human brain.^63–67 We measured the influence of brief (30 s) and moderate (5% CO₂) hypercapnic exposure on neuronal and hemodynamic activity, and our results clearly show that widespread neuronal activity is indeed influenced by CO₂ with global neurovascular uncoupling.

Avoiding the confounding neurological response to CO₂ can be difficult.^21–24 There are two other methods for calibrated fMRI using gas challenges, hyperoxia and carbogen. A hyperoxic challenge (breathing enriched O₂ typically 50–100% as compared with 21% in normal air) completely avoids the neurological impact of CO₂. However, a pitfall of using hyperoxia is the necessity to assume a constant value of resting oxygen extraction fraction (OEF₀), and hyperoxia has been shown to also impact neural activity. ⁶⁸ Moreover, carbogen (typically 10% CO₂ balanced with 90% O₂) is favorably viewed as an ideal gas mixture for calibrated fMRI, ⁶⁹ but based on our findings and those of prior studies ¹⁵ we presume that carbogen combines the confounding factors of both hypercapnia and hyperoxia. While gas-free calibrated fMRI using machine learning methods ⁷⁰ may be the best method to avoid confounding neurological effects, these methods often use a hypercapnic or hyperoxic challenge as a foundation for calibrated fMRI modeling. ⁷¹

Limitations of current study with considerations for future directions

Most human neuroimaging studies which use CO₂ exposure with specific assumptions ¹⁵ are conducted in the awake state, whereas most animal studies (like the present one) are executed under different types of anesthetics to achieve different levels of global brain activity. ⁷² While this may be considered a weakness, we predict that future experiments in the awake mouse paradigm will generate similar results. We selected the anesthetic dexmedetomidine to establish robust neural and hemodynamic responses, since dexmedetomidine anesthesia is generally considered to a level close to the awake state. ³³ However, dexmedetomidine alone can cause vasoconstriction and sometimes epileptic responses, therefore we balanced these effects with low dose of isoflurane. ³³ Isoflurane alone is not an ideal anesthetic for rodent neuroscience studies because it is known to cause burst-suppression neural activities ⁷³ and cerebral vasodilation. ⁷⁴ In awake human brain sustained increases in CO₂ suppressed neuronal activity^57,62,75 as measured by magnetoencephalography^62,75 and electroencephalography. ⁵⁷ However, high CO₂ also depressed neuronal activity in anesthetized non-human primate brain as measured by electrophysiology. ⁷⁶ These studies, did not measure neuronal and hemodynamic responses to CO₂ concurrently to demonstrate spatiotemporal patterns of uncoupling.

Neuroscience has benefited tremendously from the evolution of GECIs enabling cell-specific recordings of Ca²⁺ activity in brains of transgenic rodents. ²⁹ Two-photon imaging and/or histology can validate location/activity of these cells (excitatory/inhibitory neurons; astrocytes), whereas single-photon mesoscopic Ca²⁺ fluorescence imaging can measure from entire cortical (or bulbar) surfaces. Each imaging modality has its strengths and weaknesses. One weakness in mesoscopic Ca²⁺ fluorescence imaging that we faced was observed in the variations in signal strength in our results. These variations may be attributed to weaker region-specific responses to stimuli or variations in wavelength penetration through surface tissue. Due to these weaknesses, GECI models are constantly being remodeled and refined to produce stronger fluorophore emissions.

A weakness in our present study is that we did not measure astrocytic responses, nor distinguish between glutamatergic and GABAergic neurons. Most GECIs specifically target neurons because there is little (or no) commercially available GCaMP6f or jRGECO1a mouse lines to target astrocytes. For this reason researchers are required to create their own using ‘floxed’ GECI lines cross-bred with astrocyte-specific expression of Cre-recombinase, or by viral delivery using recombinant AAV vectors. ⁷⁷

While this study is similar to prior studies that have combined neuronal signal using either green or red fluorescence with CBV,^39–41 future studies can be designed where the green and red channels are used together to measure from neurons and astrocytes simultaneously. Alternatively, the green or red could be used for either neurons or astrocytes, leaving the other channel to probe cerebral microvessels by using plasma-borne fluorescent probes. ⁷⁸ This type of multi-modal fluorescent imaging would require careful design of optical hardware to measure the CBV signal concurrently.

Conclusion

We utilized novel widefield Ca²⁺ and hemodynamic imaging of the murine cerebral cortex and olfactory bulb to demonstrate complex neurovascular responses to various stimuli including CO₂. Our current investigation shows that exposure to 5% CO₂ induces vasodilation and neural inhibition simultaneously across large swathes of the brain. These findings challenge the experimental validity of using CO₂ inhalation in calibrated fMRI experiments, specifically when these epochs of gas exposure are not brief. Other clinical applications to modify vascular tone should also consider the neuromodulating capacity of CO₂.

Supplemental Material

sj-pdf-1-jcb-10.1177_0271678X231183887 - Supplemental material for Spatiotemporal features of neurovascular (un)coupling with stimulus-induced activity and hypercapnia challenge in cerebral cortex and olfactory bulb

Supplemental material, sj-pdf-1-jcb-10.1177_0271678X231183887 for Spatiotemporal features of neurovascular (un)coupling with stimulus-induced activity and hypercapnia challenge in cerebral cortex and olfactory bulb by Shaun James, Simon Sanggaard, Adil Akif, Sandeep K Mishra, Basavaraju G Sanganahalli, Hal Blumenfeld, Justus V Verhagen, Fahmeed Hyder and Peter Herman in Journal of Cerebral Blood Flow & Metabolism