Optogenetic investigation of the variable neurovascular coupling along the interhemispheric circuits

Abstract

The interhemispheric circuit connecting the left and the right mammalian brain plays a key role in integration of signals from the left and the right side of the body. The information transfer is carried out by modulation of simultaneous excitation and inhibition. Hemodynamic studies of this circuit are inconsistent since little is known about neurovascular coupling of mixed excitatory and inhibitory signals. We investigated the variability in hemodynamic responses driven by the interhemispheric circuit during optogenetic and somatosensory activation. We observed differences in the neurovascular response based on the stimulation site – cell bodies versus distal projections. In half of the experiments, optogenetic stimulation of the cell bodies evoked a predominant post-synaptic inhibition in the other hemisphere, accompanied by metabolic oxygen consumption without coupled functional hyperemia. When the same transcallosal stimulation resulted in predominant post-synaptic excitation, the hemodynamic response was biphasic, consisting of metabolic dip followed by functional hyperemia. Optogenetic suppression of the postsynaptic excitation abolished the coupled functional hyperemia. In contrast, light stimulation at distal projections evoked consistently a metabolic response. Our findings suggest that functional hyperemia requires signals originating from the cell body and the hemodynamic response variability appears to reflect the balance between the post-synaptic excitation and inhibition.

Introduction

Unlike other organs, the brain has no significant energy storage, and prolonged energy consumption generally increases blood supply to the area of the working brain cells.¹ This phenomenon has given rise to the idea that neuronal activity, oxygen consumption and blood flow are tightly coupled and it has established the blood oxygenated level dependent functional magnetic resonance imaging (BOLD fMRI) as the major non-invasive window into the thinking human brain.² Better understanding of the biological substrates behind neurovascular coupling could answer in what way exactly BOLD fMRI embodies neuronal activity and address a number of disputed topics such as negative BOLD, signal nonlinearity and response variability.^3,4 The recent advances of promoter-driven optogenetics allow for activation of distinct cell types and can cast light on the cell-specific vascular effects.⁵ Increasing evidence suggests that while in general neurovascular coupling is indeed contingent on the oxygen consumption by active neurons, hemodynamic modulation is most likely a combinatorial outcome of cell-specific vasoactive messengers.^6,7 This idea is exemplified by the complex neurovascular events driven by long-distance circuits that actively initiate both excitation and inhibition.

One such mixed circuit is the connection between the left and right mammalian brain. This interhemispheric pathway is crucial for bilateral integration of sensory, motor, and associative functions.⁸ Anatomical studies of interhemispheric connectivity implicate large pyramidal neurons in layers II, III and V projecting fibers along the corpus callosum. These long-distance axons terminate on the dendritic spines of both local excitatory and inhibitory postsynaptic cells in layers II, III and V of the contralateral hemisphere.⁹ Circuit mapping studies show that transcallosal stimulation elicits postsynaptic excitatory currents that are followed by postsynaptic inhibitory currents and the balance between those seem to vary across studies and experimental paradigms.¹⁰

Left-right brain interaction has been a recurring topic in the neurovascular imaging field in the last two decades,^11–18 yet no clear consensus exists on how the mixed effects of interhemispheric excitation and inhibition drive neurovascular and neurometabolic responses. Although interhemispheric functional connectivity is often assumed to reflect cortico-cortical connections, the relay subcortical regions are also shown to modulate the ongoing cortical processing.¹⁹ The neurovascular aspect of this modulation is exposed by the pathway-specific variations in neurovascular and neurometabolic coupling in the same cortical region.^11,14

In the present study, we examined interhemispheric neurovascular and neurometabolic coupling as driven by optogenetic stimulation of callosal projection excitatory neurons. This approach allowed us to isolate the neurovascular effects across a single synapse of cortico-cortical connectivity and avoid the signal modulation from the thalamus and from cells along the multi-synaptic afferent pathway from the periphery.²⁰ We hypothesized that the nature of the vascular response in the areas of mixed excitation and inhibition may be explained by the dominant postsynaptic activity. We also separated the contribution of post-synaptic excitatory cells from the post-synaptic inhibitory cells by suppressing the post-synaptic excitatory cells optogenetically and looked at the vascular contribution of the post-synaptic inhibitory cells. We observed that the same optogenetic stimulus evokes several response categories at the homotopic transcallosal cortical region. The vascular and metabolic characteristics of this response appear to depend on the balance of the evoked post-synaptic inhibition and excitation. In this specific interhemispheric connection between the primary forelimb somatosensory cortices, the post-synaptic inhibition drives primarily vasoconstriction and a largely metabolic response of oxygen consumption without an increase in blood flow, while excitation mainly delivers blood flow to the activated area. The significance of our finding is that while direct stimulation of inhibitory cells was shown to increase blood flow,^7,21,22 as a part of a mixed excitation and inhibition, the inhibitory activity appears to consume oxygen while signaling a decrease in the neurovascular response.

Since the distance of the dendritic synapse from the cell body appears to influence neuronal integration,²³ we wanted to find out whether independent optogenetic stimulation paradigms either at the cell body or at the distal interhemispheric projections have an effect on the evoked neurovascular response as well. Due to the loss of charge as current flows from distal dendrites to the soma, the inputs distant from the soma have less influence on the action-potential initiation.²³ Therefore, we hypothesized that optogenetic stimulation at the distant transcallosal projections will evoke vascular responses with diminished amplitude. We expected that the nature of the response to have the same categorical variability since we drive the same circuit albeit with less efficiency. Contrary to our expectations, we found that optogenetic stimulation at the distal projections was not sufficient to beckon an increase of cerebral perfusion during any trial despite the local oxygen consumption. We conclude that signals propagating from the cell body are necessary to evoke robust vascular responses in a remotely targeted area.

Our findings agree and expand the scope of previous knowledge^13,15 and contribute to the present-day endeavor to uncover the cellular basis for vascular and metabolic events in the brain.

Methods

Animal preparation

All animal procedures were conducted in accordance with the National Institute of Health guidelines and with the approval of the University of Pittsburgh Institutional Animal Care and Use Committee, and reported in compliance with ARRIVE guidelines.

We used a total of 28 male Sprague Dawley rats (five to seven weeks at the time of the initial procedure, Charles River, Cambridge, MA, USA).

We used viral transduction in the primary forelimb somatosensory cortex (S1FL) to express channelrhodopsin (ChR2) or halorhodopsin (HR) transgene under a promoter for Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) specific to excitatory glutamatergic neurons. Experiments were performed on three different groups: ChR2-only, ChR2 + HR and control. We transduced 15 ChR2-only animals with AAV5-CaMKII.ChR2(H134R)-eYFP, 8 ChR2 + HR animals with AAV5-CaMKII.ChR2(H134R)-eYFP in the right somatosensory cortex and AAV5-CaMKII.HR-mCherry in the left somatosensory cortex. The control group was five animals transduced with control vector AAV5-CaMKII.eYFP or –mCherry. All vectors were obtained from University of North Carolina Vector Core (Chapel Hill, NC, USA). Briefly, animals were anesthetized with 1.5% to 2.2% isoflurane and placed in a stereotaxic frame (Narishige, Tokyo, Japan). A burr hole was drilled into the right hemisphere (for the ChR2-only and the control animals) at stereotaxic coordinates of 0 to 0.5 mm from bregma, 2 mm from the midline. For ChR2 + HR animals, we injected both hemispheres at the above coordinates. The 2-µL virus was infused at 1 mm below the dura by a stereotactic injector (Stoelting Co., Wood Dale, IL, USA) at a rate of 0.2 µL/min.

Animal surgery for experiments

Three to four weeks after the viral injection, the animals were anesthetized with isoflurane (1.5–2.2%); a catheter was placed at the right femoral artery to monitor blood pressure. Two needle electrodes were placed in the forelimb for sensory stimulation. Body temperature was maintained at 37.6℃. Heart rate and blood pressure were continuously monitored (Biopac Systems, Goleta, CA, USA). The skin above the skull was removed and a well of 2-cm diameter was built over both hemispheres with dental cement. Craniotomy was conducted under filtered light (570 ± 10 nm) and 1% agarose was used to maintain intracranial pressure. After surgery, atropine (0.05 to 0.07 mg/kg) was administered intramuscularly to reduce secretions. Then, the anesthesia was either decreased to 0.5–1% isoflurane²⁴ (for direct stimulation of the transcallosal projections) or switched to alpha chloralose (40 mg/kg).

Stimulation

All stimulations were driven by a waveform generator (AMPI, Jerusalem, Israel) and performed in a block design (4 s stimulus delivered every 40 s, with six repetitions). The forelimb stimulation was 1 ms with an amplitude of 1.5 mA at a frequency of 4 Hz with alpha chloralose and 8 Hz with isoflurane. The ChR2 photo-stimulus was delivered via a multi-mode optic fiber with 400-µm core diameter (ThorLabs, Inc., Newton, NJ, USA) coupled to a 473-nm laser diode (CrystaLaser, Inc., Reno, NV, USA). The HR-photostimulus was delivered globally over the brain surface with Mercury-Xenon light source with a power supply unit (Opti-quip model 1600, Highland Mills, NY, USA). Before each experiment, the power of the stimulation light was measured with an optical power meter (Melles Griot, Carlsbad, CA, USA). All power levels are reported for continuous irradiation. For the HR activation, we used continuous yellow light of 40 mW (531 ± 22 nm). For the ChR2, the parameters consisted of three pulse durations (10, 30, and 50 ms) with a fixed power of 5 mW and frequency of 12 Hz.

Hemoglobin-based optical intrinsic signal

The illumination source consisted of oblique light guides connected to a halogen light source (Thermo-Oriel, Stratford, CT, USA) to transmit filtered light of 600 ± 50 nm. Reflected light from the cortex was split in two spectrally distinct images (Dual-Cam, Photometrics, Tucson, AZ, USA), which were filtered at 572 ± 7 nm and 620 ± 7 nm (for the ChR2 only animals) and 540 ± 7 nm (for the ChR2 + HR animals). Based on the light absorption of oxyhemoglobin and deoxyhemoglobin, the reflectance at the 540 nm and the 570 nm isosbestic point is indicative of the total hemoglobin (HbT-weighted), whereas the reflectance at 620 nm is primarily deoxyhemoglobin (dHb-weighted). The optical intrinsic signal (OIS) at 540 nm and 570 nm is therefore representative of cerebral blood volume (CBV), while the OIS at 620 nm is more sensitive to the blood oxygenation level and comparable to the BOLD fMRI signal. The OIS data were recorded at 30 fps with two analog CCD cameras (Sony XT-75, Tokyo, Japan) and an analog-to-digital frame-grabbing board (Matrox, Inc., Dorval, Quebec, Canada). The field of view was 5 × 4 mm² (for unilateral) or 8.8 × 7.4 mm² (for bilateral) depending on the magnification setting of the macroscope (MVX-10, Olympus, Tokyo, Japan).

Collected data were analyzed using MATLAB software (Mathworks, Natick, MA, USA). OIS images were averaged across repetitions within the same trial and smoothed by a 2-pixel-wide Gaussian filter. Activation maps were computed by the intensity difference between each data point and a 2-s baseline period preceding the stimulus onset.

Electrophysiology

When light stimulation was at the bodies of the ChR2-expressing cells, we recorded electrical activity at the opposite hemisphere using 16 channel linear electrode (A1x16-3 mm-100-703-CM16LP, Neuronexus Technologies. Ann Arbor, MI), implanted into the homotopic S1FL. Electrophysiological recordings were sampled at 24,414 Hz with a recording system (RX7, Tucker-Davis Technologies, Alachua FL). The raw data stream was filtered to produce LFP (1-300 Hz) and spike (300 – 5000 Hz) data streams using a second-order Butterworth filter. The LFP analysis was carried out through the quantification of LFP power spectrum with a custom MATLAB script based on a previously published method.²⁵ Current source density (CSD) analysis was performed by computing the average evoked (stimulus-locked) LFP at each site, and then calculating the second spatial derivative.

When light stimulation was at the distal projections, a glass-coated electrode with a tip diameter of 5 µm and nominal impedance of about 1 MΩ (Carbostar, Kation Scientific, Minneapolis, MN, USA) was placed adjacent to the optic fiber below the core of the illuminated area at a depth of 200 to 600 µm. Electrophysiological activity was recorded at a frequency of 20 kHz (Plexon, Inc., Dallas, TX, USA).

Statistical analysis

All statistical analyses were performed in MATLAB, using standard statistical functions. Group data were obtained from the percent changes in amplitude during the time course of the deoxyhemoglobin-based (dHb) and hemoglobin (Hb) intrinsic optical signal in 10 animals. Briefly, a region of interest (ROI) was drawn over the exposed brain of each animal to obtain the time series of change in reflectance. The time series were temporally averaged to 1-s bins and for dHb datasets, they were separated visually into negative, biphasic or no-response group. The analysis on the corresponding HbT datasets was performed in the same response categories. One-way ANOVA was used to determine whether there is a difference between the means of the baseline, dip and overshoot in each group for dHb. The baseline was the signal average for 2 s before the stimulus onset; the dip was the local minima during the first 10 s and the overshoot was the local maximum 25 s after the stimulus onset. The ANOVA was followed by a post hoc Tukey’s test for multiple comparison of the group means with alpha = 0.05 to determine which group was significantly different from the rest. For the corresponding HbT datasets, we used one-way t-test comparing the mean of 2 s baseline to the local minima 25 s after the stimulus onset.

For the multichannel electrophysiology, the power frequency spectrum at the top, middle and bottom four recording sites on the linear electrode was averaged across animals (N = 8) and the standard error of the distribution was used to compute the significance between the power spectrum before and after the stimulation. For the glass-coated electrode recording at the axonal terminals, the frequency time series were normalized by their baseline during 5 s before the stimulation and binned into low (1–25 Hz) and high (25–170 Hz) spectral bins. Each resulting time series of changes in the power of the high and low spectral bins were then averaged across animals (N = 8) and the standard error was computed from the distribution.

For the correlation analysis between the stimulation parameters and the amplitude of response, as well as between hemodynamic imaging and electrophysiology, we used linear regression followed by an F-statistics to test the significance of the linear model.

Histology

After imaging, animals were perfused transcardially with phosphate-buffered saline, then with 4% paraformaldehyde. The brains were frozen in optimal cutting temperature compound (EMS, Hatfield, PA, USA) and cryosectioned using a Leica CM1850 cryostat (Leica, Wetzlar, Germany) in 25 -µm thick slices, and mounted on glass slides. The sections were imaged using a Zeiss AxioImager (Oberkochen, Germany).

Results

To examine the spatiotemporal dynamics of neurovascular coupling, we shined blue light at the bodies of ChR2-positive excitatory neurons – the virus injection site – and recorded the response from the opposite hemisphere across the full cortical thickness. We also simultaneously evaluated deoxyhemoglobin-weighted (dHb) and total-hemoglobin-weighted (HbT) signals by optical imaging in the same hemisphere (N = 5). In some animals, both hemispheres were imaged (N = 5). Graphical representation of the experimental set up is shown in Figure 1(a). Initially, we verified the vector-mediated expression of ChR2 at the site of the viral injection (Figure 1(b), arrow). In the majority of animals (10 out of 13), we observed ChR2-positive regions on the opposite hemisphere dispersed across the contralateral homotopic cortex. We placed the recording electrode in the opposite hemisphere at the brightest region of the long-distance projections (Figure 1(b), asterisk). We also verified using histology that those bright regions are indeed long-distance projections originating from the corpus callosum (Figure 1(c), arrow). At the level of the somatosensory cortex, the corpus callosum had a large number of ChR2-positive axonal fibers, traversing from one hemisphere to the other (Figure 1(d)). At the site of the viral injection, there was a robust expression across all layers and a substantial number of cell projections were entering the corpus callosum (Figure 1(e)).

Figure 1.

Viral expression of ChR in the rat somatosensory cortex follows the interhemispheric circuit allowing light-driven functional investigation. (a) Experimental set up. (b) Top view of the cortex showing ChR-YFP expression at the injection site (arrow) and axonal projections reaching the contralateral hemisphere (asterisk). Composite image of bright field and yellow fluorescence. (c) Fluorescent image of histological section showing ChR-positive axonal projections exiting the transcallosal tract (arrow) and synapsing on the local cells. (d) ChR-positive axons crossing from one hemisphere to the other. (e) ChR-positive neurons at the AAV-injected site enter the corpus callosum to travel towards the opposite hemisphere; scale bar (c–d) = 250 µm.

We used several light pulse durations (10, 30 and 50 ms) to test whether the interhemispheric response scaled with the size of the optogenetic stimulation. While the primary site of excitation at the cell bodies exhibited clear correlation with increasing stimulus strength as we previously reported,²⁶ the response on the opposite hemisphere was not correlated to the strength of the stimulus (F-statistic of the linear regression model, p > 0.05). Hence, all of the conditions were pooled together in the response assessment. We also tested weaker stimuli (5 ms pulse duration and power of 1 mW in four animals); however, these lower levels of light stimulation did not evoke detectable response on the contralateral side, either optically or electrophysiologically, and those stimulation parameters were not tested further.

Figure 2 shows dHb-weighted images and the corresponding time series of the typical transcallosal responses after optogenetic stimulation at the cell bodies. The blue lightning indicates the stimulus side and the red arrow points at the side of electrical recording on the opposite hemisphere. The black and red curves are the time series generated over the respective elliptical regions. The image brightening indicates an increase in oxygenated blood to the region due to blood flow and the darkening on the dHb-weighted signal suggests an increase in oxidative metabolism.^27–30 While the dHb-weighted signal at the optical stimulation site was always accompanied by large oxygenation (black eclipse region), the dHb-weighted signal evoked in the opposite side varied greatly. These variable interhemispheric responses were categorized into three types – negative (Figure 2(a) and (b)), biphasic (Figure 2(c) and (d)) or no-response (Figure 2(e) and (f)).

Figure 2.

Optogenetic stimulation in the right somatosensory cortex (black ellipse) evokes variable deoxyhemoglobin-based signal (dHb) on the left (contralateral) side (red ellipse). (a) In half of the trials, the contralateral side exhibited prolonged dip indicating oxygen consumption without evoking blood flow. (b) Black curve shows the time series over the black ellipse and the red curve over the red ellipse on the contralateral side. (c) On occasion, exactly the same stimulation also evoked a biphasic response in the contralateral side with a dip followed by a rise in oxygenation. (d) Time series over the two hemispheres show that the oxygen rise arrives faster at the stimulus side (black curve) comparing to the transcallosal side (red curve). (e) An example of transcallosal response absence to the same light stimulation. (f) Time series showing hemodynamic response at the site of the stimulus (black curve) and no response at the transcallosal side (red curve).

The nature of the response varied within the same animal and its distribution across the three response categories was not significantly different between animals. We separated visually all the transcallosal responses into three major categories: negative, biphasic or no-response and analyzed the variance of the time series averaged across animals (Figure 3) (see ‘Methods’). A significant metabolic dip was observed for the negative and biphasic response categories (Figure 3(a) and (b)), a significant overshot was observed only for the biphasic response category (Figure 3(b)). There were no differences between baseline and later parts of the time series in the no-response category (Figure 3(c)). The pie charts in Figure 3(d) summarize the response distribution across categories for total of 50 optogenetic and 27 forelimb experiments in 10 animals. Most commonly we observed only negative response (53% for optogenetic and 45% for forelimb stimulation). In the remaining cases, we observed either the previously described biphasic response¹³ (33% for optogenetic and 28% for forelimb stimulation) or no-response. The peak of the negative response (negative dHb signal) varied greatly between experiments (4–8 s after the onset of the 4-s stimulus). For the biphasic responses, the positive amplitude of the response differed by an order of magnitude between the two hemispheres (0.2% vs. 2%).

Figure 3.

Group data summarizing the variable transcallosal deoxyhemoglobin-based signal (dHb) described in Figure 3. (N = 10 animals). The mean amplitudes of the baseline, the dip and the overshoot of the time-series were compared using one-way ANOVA, followed by the Tukey’s method for multiple comparison of the group means with alpha=0.05. The means of the groups that were significantly different are marked with asterisks. (a) Negative-only response indicating oxygen consumption without increase in blood flow (no significant overshoot). (b) Biphasic response with significant metabolic dip followed by an increase in oxygenation in the area. (c) No detectable dHb signal. (d) Group data showing the frequency of each type of transcallosal response (negative, biphasic or none) for both ChR2 and forelimb stimulation.

The corresponding HbT-weighted images and the time series over the drawn regions of interest are shown in Figure 4. Using statistical analysis of the HbT time-series, we compared the mean value at baseline to the local minima in the 25 s after the stimulus onset. Overall, the changes in HbT in the hemisphere opposite to the optogenetic stimulus paralleled the changes in dHb described in Figures 2 and 3. We observed no significant changes in HbT (Figure 4(a) and (b)) when dHb was purely metabolic indicating that these responses likely indicate an increase in oxidative metabolism without accompanied functional hyperemia. The only statistically significant change in the blood volume HbT (one way t-test, p < 0.05, Figure 4(c) and (d)) was during the biphasic dHb response indicating an increase in blood volume during the dHb overshoot. As expected, there were no significant HbT signal changes (Figure 4(e) and (f)) when the simultaneously recorded dHb also showed no response.

Figure 4.

Bilateral blood volume changes based on total hemoglobin signal (HbT) after unilateral optogenetic stimulation (30 ms, 5 mW, 8 Hz) in the somatosensory cortex at the black ellipse. (a, c and e) show the HbT response acquired simultaneously with the dHb panels on Figure 2. (b, d and f) plots represent the time series over the light-stimulated (black) side and the transcallosal hemisphere (red).

The primary site of optogenetic stimulation (Figures 2 and 4), as well as the primary cortex for the forelimb response exhibited dHb and HbT changes congruent with our previously published reports.^26,31

To examine neural basis of these metabolic and vascular responses, we analyzed the types of neural activity that correspond to the three intrinsic signal categories described (Figure 5). For the negative response case, without rise in blood oxygenation, we observed upsurge in the power of gamma frequency band in layers II-III and layer IV, consistent with synchronized inhibitory activity³² (Figure 5(a)). These were also the cortical layers that we observed histologically to have the highest synaptic density of the terminating ChR2 projections coming from the contralateral side (Figure 5(b)). We also quantified the changes in the power across the frequency spectrum at the top, middle and bottom four recording sites on the linear electrode, averaged across animals (N = 8) and compared those changes before and after the light stimulus (Figure 5(c)). We found significant differences in the gamma band for the top and bottom layers (top and bottom panels on Figure 5(c)), that likely indicate inhibitory activity in layers II-III and layer V correspondingly. This increase in gamma power however did not correlate significantly with the maxima of the accompanied dHb dip reported above (F-statistic of the linear regression model, p > 0.05). These data are consistent with previously reported anatomical and functional interhemispheric connectivity and zones of interhemispheric inhibition.^9,10

Figure 5.

Electrical activity at the site transcallosal to the blue light stimulus. (a) Power across the frequency spectrum at different cortical depths shows increase in gamma band frequency in layers II – III and V (ellipses) associated with inhibitory activity during the rise of oxygen consumption in the absence of blood flow. (b) Histological section of the somatosensory cortex showing the ChR2-YFP expression at the distal axonal projections coming from the corpus callosum (arrow). Layers II-III and V show brighter regions with higher synaptic density (ellipses). (c) Average changes of the power across all frequencies before (black) and right after the stimulation (blue) demonstrating increase in gamma band in the upper (top panel) and lower (bottom panel) cortical layers. There was no significant changes in the middle layers (middle panel). Shaded regions represent the standard error of the group of 8 animals. (d) Current source density diagrams across the cortical thickness indicate excitatory activity in the presence of increased blood volume during interhemispheric ChR2 activation (left), ipsilateral forelimb (FL) (middle) and contralateral FL (right). (e) Average trace of the local field potential across all channels of the same data sets as in (d).

Conversely, the biphasic response case showed a rise in the local field potential (LFP) across the cortical thickness as presented by the CSD profiles in Figure 5(d). The CSD during optogenetic stimulation (Figure 5(d), left panel) was similar to the ipsilateral forelimb stimulation CSD (Figure 5(d), middle panel). Generally, the signal appeared to arrive first in the upper and lower layers consistent with earlier activation of layers II, III and V (Figure 5(d)). We also averaged the LFP across the cortical thickness to distinguish the comparative amplitude and temporal evolution (Figure 5(e)) of the neural response. Due to the difference in pulse duration between the light stimulus and the forelimb electrical stimulus, we could not compare directly the temporal profiles, although the optogenetically evoked LFP and the corresponding CSD appeared to last longer, consistent with the longer light pulses (up to 50 ms, compared with the electrical forelimb stimulation pulses, 1 ms). The average LFP and the CSD produced by contralateral forelimb stimulation was the largest in amplitude and arrived the soonest (Figure 5(d) and (e)) consistent with previously published data from our laboratory.^26,31 The amplitude of the LFP averaged across layers did not correlate significantly with the overshoot of the dHb biphasic signal and the size of the HbT response measured simultaneously (F-statistic of the linear regression model, p > 0.05).

We dissected further the separate contributions of the postsynaptic cells by selectively inhibiting the post-synaptic excitatory cells with HR under excitatory-specific promoter (N = 8 animals). The experimental set up of this approach is shown in Figure 6(a). We confirmed the expression of HR conjugated to mCherry (HR-mCherry) in the opposite hemisphere from the ChR2-YFP expression (Figure 6(c)). Due to the light emission of the two fluorescent proteins and the separate excitation bands of the two opsins, we could only record OIS at hemoglobin isosbestic point and get information about HbT changes. Shining blue light at the bodies of the ChR2-positive cells evoked an increase in LFP in 30% of all trials in N = 8 animals, revealing the expected CSD profile (Figure 6(b), left panel), concurrent with the increase in blood volume (Figure 6(d), top panel). When we stimulated ChR2 in the presence of yellow light, we abolished this response and the CSD profile was preserved only at the lower cortical layers (Figure 6(b), right panel) congruent with limited light penetrance in the deeper tissue areas. Consistent with this electrophysiological event, there was an absence of HbT response when we stimulated ChR2 in the presence of yellow light (Figure 6(d), bottom panel). The mean amplitude of the blood volume increase across all datasets in the biphasic group differed significantly from the mean when post-synaptic activity in excitatory neurons was abolished optogenetically in the same animals (one way t-test, p < 0.05, N = 8 animals, two to three trials per animal).

Figure 6.

Inhibition of the post-synaptic excitatory cells abolishes the stimulus-evoked rise in blood volume. (a) Experimental set up. (b) Current source density (CSD) maps of neural activity during blue light stimulation (ChR2, left) and during the same stimulus with yellow light on (ChR2 + HR) to inhibit the postsynaptic excitatory cells (right). (c) Composite image of bright field and fluoresce showing ChR2-YFP expression on the right (top) hemisphere and HR-mCherry expression on the left hemisphere (bottom). (d) Total hemoglobin weighted (HbT) intrinsic signal during blue light stimulation (ChR2, top) and during the same stimulus with yellow light on (ChR2 + HR, bottom). Gray scale, % signal change from baseline.

Finally, we investigated whether the nature of the hemodynamic response differs depending on the location of the light stimulus – cell bodies versus distal cell projections. We wanted to ascertain whether activating directly the distal projections of the long-distance excitatory neurons can evoke the same response variability. We targeted the blue light at the brightest region of the ChR2-positive interhemispheric projections. In order to avoid the light artifact on the large multichannel recording electrode, we used a single channel recording instead. The experimental set up of this approach is presented in Figure 7(a). The stimulation of the ChR2 transcallosal projections failed to evoke any activation in approximately half of the experiments – OIS and neuronal activity. As in the earlier set of experiments, we observed an increase in the gamma band power in the rest of the experiments (56% of the time). Average of the temporal evolution of the high (blue curve) and low (black curve) frequencies across all animals (N = 8) is shown in Figure 7(c). We saw no change in the lower frequencies typical for the firing rates of excitatory neurons (Figure 7(b), black curve). This electrophysiological signature was reliably accompanied by significant metabolic response as indicated by the group data in Figure 7(c) and illustrated by the dip in the dHb-weighted images (Figure 7(d), top panel). We found no significant statistical correlation between the amplitude of the gamma frequencies and the size of the metabolic dip (F-statistic of the linear regression model, p > 0.05). We used a smaller field of view centered around the stimulation site in order to focus on the local change of the intrinsic signal. At this higher resolution, we were able to see a considerably more detailed field of view with better signal-to-noise and detected a significant brightening on the HbT images (Figure 7(d), bottom panel, one-way t-test, p > 0.05, comparing the baseline with local maxima in the time series, N = 8 animals, two to three trials per animal), consistent with vasoconstriction and reduction of the total blood volume in the area, similar to previous reports.^7,13 This response was also independent from the size of the stimulus, and was absent in control animals.

Figure 7.

Stimulation of the distal projections of ChR2-expressing neurons evokes metabolic response without an increase in blood volume (a) Experimental set up. (b) Average power levels over time of low (black) and high (blue) frequency bands showing increase of high frequency power during the stimulus duration (0–4 s).The shaded regions represent the standard error of the group N = 8 animals. (c) Summary of the simultaneously recorded intrinsic signal demonstrating that the increases in the high frequency in (b) are accompanied by metabolic dip and no significant increase of oxygenation. Error bars represent the standard error of the same group of N = 8 animals. The mean amplitudes of the baseline, the dip and the overshoot of the time-series were compared using one-way ANOVA, followed by the Tukey’s method for multiple comparison of the group means with alpha=0.05. (d) Example of deoxyhemoglobin (dHb, top) and total hemoglobin (HbT, bottom) intrinsic signal in response to local blue light stimulation illustrating the group data in (c). Gray scale, % signal change from baseline.

Discussion

The emerging evidence of cell specificity in the brain energy metabolism implies the existence of rich and coordinated vascular signaling mechanisms.^3,6,7 In the background of continuous cortical inhibition and excitation, it appears unlikely that blood flow in the brain can be governed solely by oxygen consumption and modeled by membrane depolarization and hyperpolarization alone. Our data assert that cellular inhibition and excitation are rarely sovereign processes in terms of vascular control, at least in the context of long-range interhemispheric neurovascular coupling. Our results agree with prior studies showing dissociation of blood flow and metabolism in somatosensory cortex ipsilateral to forelimb stimulation.^12,13 The observed lasting negative changes in blood oxygenation and the apparent vasoconstriction are hypothesized to be a result of predominant inhibition and we confirm that finding by showing a concomitant increase in gamma band consistent with inhibitory activity. Previous studies have demonstrated that fast-spiking inhibitory neurons are the major generator of gamma rhythm³² and this oscillation is thought to emerge from the coordinated interaction of excitation and inhibition.³³ Functionally gamma oscillation shapes the coherence of sensory signals and controls the excitation spread.³⁴ Both of these processes are particularly important in the somatosensory left-right brain interaction as they mange the feedback during manual coordination.⁸

Fitting the summation of LFPs to the amplitude of the metabolic response (dHb dip) was not a feasible option in our survey of transcallosal coupling since cell bodies and projections of inhibitory neurons are not aligned in space to produce synchronous changes in LFP.³⁵ Whenever we observed the typical layer-specific LFP, we ascribed it to the activity of large pyramidal neurons spanning the full cortical thickness with processes parallel to each other. It is worth noting that while the intrinsic optical signal originates from the superficial layers, the multichannel recording spanned the full cortical thickness. Comparing findings across these data sets has to be done with awareness that they are collected in different space and time continuum. Nonetheless, they do have a causative relationship and the superficial regions in the brain are shown to propagate the vascular events from the deeper layers.^3,13,36 Similar to our results, a prior study observed excitatory LFP arriving transcallosally in only portion of the trials, the rest was a combination of downward deflection of LFP or no response.¹² We did not use the extracellular multiunit activity (MUA) as a correlate to vascular or metabolic events since we are investigating firing of mixed cell populations, while the extracellular recording by nature is heavily weighted towards action potentials from large pyramidal neurons. In contrast, the firing of small inhibitory cells randomly dispersed across layers is difficult to detect with in vivo extracellular recording and traditionally the presence of inhibitory activity is inferred by the reduction of ongoing excitatory activity.³⁵ In the cases of simultaneous inhibition and excitation, the overall change in neuronal firing rate is difficult to discern. Instead of looking at the conventional LFP and MUA, we chose to evaluate the power changes across the frequency spectrum since different frequency bands originate from the different spiking rate of excitatory and inhibitory neurons.^32,35 Since inhibition in this transcallosal circuit appears to be modulatory and not a direct driver of vascular response, not surprisingly we did not find any significant correlation between amplitudes of gamma power and size of Hb-based signal beyond the described observations. We did not detect any vascular or neural response in the opposite hemisphere during low-intensity light stimulation (power of 1 mW, light pulse duration of 5 ms). This outcome suggests an existence of activity threshold necessary to evoke blood flow across long-distance akin to previous observations.^37,38 We detected a transcallosal response only at higher powers (5 mW, light pulse duration of 10 ms, 30 ms and 50 ms, and frequency of 12 Hz), and its amplitude did not scale appreciably with the amplitude of the stimulus parameters, at least in the ranges that we used. Our effort to modulate the transcallosal response with graduation of stimuli was limited to the light power that does not evoke response in naïve animals.²⁴ Since the amplitude of the transcallosal response was relatively small, a reliable correlation between the response and the stimulus strength may be difficult to capture with our methods. Similar outcome was observed at the ipsilateral cortex during forelimb activation.¹³ Previous studies using direct electrical stimulation of the interhemispheric circuit indicated frequency dependence, particularly in the high frequency range.^12,14 One explanation for this disparity may be the non-specific nature of direct electrical stimulation in the cortex. Without the cell specificity conferred by the ChR2 promoter, the projections from cells originating from the opposite hemisphere are also activated as well as local inhibitory cells. Antidromic propagation of the direct electrical stimulus could also affect the characteristics of the response.^11,39

The push and pull of the simultaneous inhibition and excitation can explain the three scenarios we observed – negative, biphasic and no-response. The concurrently released dilatory and constrictive agents from different cell types can have an opposing effect and a certain threshold of coordinated synaptic release must be reached to trigger a robust hemodynamic response.^7,13,37,38 The dissociation between metabolic and vascular responses we observed can be further explained by the different origins of metabolic rise and vascular control. The consumption of oxygen and the resulting metabolic dip in deoxyhemoglobin intrinsic signal is directed by ATP turnover,^13,36 while stimulation-induced blood flow appears to be initiated by ionic gradients.⁴⁰ Further evidence comes from the observed threshold effect between synaptic activity, oxygen consumption and vascular response consistent with energy source buffer.^36–38 This is an important point to emphasize since at first look our results show that negative BOLD appears to be caused by inhibition, while excitatory activity is necessary to evoke robust flow; however, based on the negative or positive BOLD signal alone, we cannot ascertain that the underlying activity is excitatory or inhibitory.⁴¹ For example in our data, the initial dip during the biphasic response can be either inhibition that is overtaken by excitation, or early sub-threshold excitation that eventually drives the blood flow to the energy-consuming area.

When we stimulated the distal endings of the ChR2-positive projections located on the opposite hemisphere from the cell body, we observed only metabolic response with concurrent increase in the gamma band (Figure 7). One possible explanation is that focal stimulation at a subset of distal projections has a strength below the threshold of vascular signaling since it recruits a smaller number of cells comparing to the stimulation of many cell bodies on the other side.³⁹ Alternatively, there may be a difference in the vascular control when the same cell is stimulated at different regions. Pyramidal neurons integrate differentially proximal and distal synaptic inputs²³ and this may also provide opportunity for differential vascular control by the same circuit based on the input location. In support of this hypothesis, previous work showed that the activation of two excitatory afferent systems at different locations of the same target cell elicited cerebral blood flow by different signaling mechanisms.⁴²

It is also worth noting that energy consumption in the brain is not constrained to oxidative phosphorylation. There is a mounting evidence that cells can shift between oxidative phosphorylation and glycolysis and a significant degree of decoupling may exist between cerebral oxygen and glucose metabolic rate.⁶ In the same transcallosal circuit, it was demonstrated that upon forelimb activation the glucose consumption surpassed the oxygen consumption on the ipsilateral side.¹³ An important question to be addressed by future research is whether such metabolic shifts and the participation in the astrocyte-neuronal lactate shuttle⁶ differ between excitatory and inhibitory cells.

Finally, the balance between inhibition and excitation and the evoked vascular response may not depend solely on the stimulus attributes but also on the energetic state of the system when the stimulus arrives.^43,44 Traditionally, the amplitude of OIS is computed by subtracting background signal; however, the ongoing activity may be the decisive factor for the response to an incoming stimulus, especially in threshold events such as interhemispheric interactions. This phenomenon could also account for our observation that the same stimulus parameters evoked variable interhemispheric response. Further work is needed to characterize how the neurovascular signaling can be propagated, amplified or diminished depending on the ongoing brain activity.

Conclusions

This work aims to bridge the gap between cellular approaches and whole brain functional imaging. The end goal is to understand both the fundamental principles of brain function as well as the biological basis of functional signals. At a first approximation, the conventional view of BOLD as a surrogate of neuronal activity holds true. However, upon closer inspection, our work supports the accumulating evidence that vascular changes are not ruled simply by the depleted oxygen, but are result of continuously retuning concert of events that engages diverse cellular players.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by NIH grants NINDS T32 NS086749 (BI); R01NS094404 (AV), R01 NS094396 (TK), NIBIB EB003324 and EB018903 (MF); and IBS-R015-D1 (SGK)

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

BI designed and performed all experiments, analyzed the data and wrote the manuscript, AV and TDYK participated in the design, data collection and analysis, MF and SGK participated in the design and edited the manuscript.