Early and late stimulus-evoked cortical hemodynamic responses provide insight into the neurogenic nature of neurovascular coupling

Abstract

Understanding neurovascular coupling is a prerequisite for the interpretation of results obtained from modern neuroimaging techniques. This study investigated the hemodynamic and neural responses in rat somatosensory cortex elicited by 16 seconds electrical whisker stimuli. Hemodynamics were measured by optical imaging spectroscopy and neural activity by multichannel electrophysiology. Previous studies have suggested that the whisker-evoked hemodynamic response contains two mechanisms, a transient ‘backwards' dilation of the middle cerebral artery, followed by an increase in blood volume localized to the site of neural activity. To distinguish between the mechanisms responsible for these aspects of the response, we presented whisker stimuli during normocapnia (‘control'), and during a high level of hypercapnia. Hypercapnia was used to ‘predilate' arteries and thus possibly ‘inhibit' aspects of the response related to the ‘early' mechanism. Indeed, hemodynamic data suggested that the transient stimulus-evoked response was absent under hypercapnia. However, evoked neural responses were also altered during hypercapnia and convolution of the neural responses from both the normocapnic and hypercapnic conditions with a canonical impulse response function, suggested that neurovascular coupling was similar in both conditions. Although data did not clearly dissociate early and late vascular responses, they suggest that the neurovascular coupling relationship is neurogenic in origin.

Introduction

Functional magnetic resonance imaging (fMRI) has revolutionized the field of cognitive neuroscience by allowing insight into the working of the human brain. However, fMRI does not directly measure neuronal activation but rather measures secondary hemodynamic changes closely associated with changes in neuronal activity (Bandettini, 2009; Logothetis and Pfeuffer, 2004). This relationship, termed neurovascular coupling, is still incompletely understood, and a breakdown in neurovascular coupling may be an important contributing factor to several neurodegenerative disease conditions (Girouard and Iadecola, 2006; Resende et al, 2008; Zacchigna et al, 2008). Fortunately, in recent years studies making combined electrophysiological measures of neural activity and optical measures of hemodynamics in animal models have begun to elucidate the stimulus-evoked neurovascular coupling relationships (Berwick et al, 2008; Boorman et al, 2010; Devor et al, 2007; Shmuel et al, 2006). However, important issues remain, such as why does neurovascular coupling exist at all, and how is the relationship controlled (Kleinfeld et al, 2011)? Intuitively, it makes sense that a region of increased neuronal activity requires more energy in the form of glucose and oxygen, and an increase in perfusion to this region would meet this need. An important unresolved question is how the need for increased energy metabolites is relayed from the active neural cells to the vessels that need to be dilated. There are two theories to explain this: the first and most well supported is termed the ‘neurogenic' hypothesis and the second is that neurovascular coupling is driven by direct ‘metabolic' demand. The neurogenic hypothesis proposes that vessels are dilated by mediators released as a product of normal synaptic activity, suggesting neurovascular coupling is just a feed forward mechanism of normal neuronal processing and not related to metabolism per se. The metabolic hypothesis suggests that the need for increased energy substrate sets in process a chain of events that leads to vessel dilation (Gordon et al, 2008; Iadecola and Nedergaard, 2007). In the case of ‘longer' duration stimuli (>8 seconds), there is also the possibility of the presence of different coupling mechanisms, possible neurogenic followed by metabolic, occurring over short and prolonged time scales (Martindale et al, 2005). It is this question that forms the basis of the experiments in the current study. Experiments using optical measures of cortical hemodynamic responses to long duration somatosensory stimuli in rodents within our laboratory and others (Berwick et al, 2008, 2005; Boorman et al, 2010; Bruyns-Haylett et al, 2010; Jones et al, 2005; Kennerley et al, 2005) have shown characteristic peak and plateau hemodynamic response function. We have previously shown that this ‘peak' and ‘plateau' type response could be generated by two separate neurovascular coupling mechanisms, the first being a transient arterial backwards dilation, and the second by in increase in blood volume more closely associated with the site of neuronal activation and potentially of metabolic origin (Martindale et al, 2005). In the present study, we attempt to separate these early and late responses by predilating all the arterial supplying vessels using ‘strong' hypercapnia. We used the whisker barrel cortex rodent model, and stimulated the whisker pad for 16 seconds measuring the hemodynamic response with two-dimensional optical imaging spectroscopy (2D-OIS; Berwick et al, 2008). To investigate whether any putative changes in hemodynamic responses observed during hypercapnia were due to hemodynamic effects alone, cortical neural responses were also measured with multichannel electrophysiology. We also explored whether the recently reported prolonged ‘negative' hemodynamic responses evoked by whisker stimuli in cortical regions adjacent to the whisker barrel somatosensory cortex (Boorman et al, 2010) were altered by hypercapnia. Indeed data suggested that during hypercapnia, stimulus-evoked cortical hemodynamic responses did not contain an initial peak associated with arterial dilation. However, concurrent measurements of neural activity suggested that evoked neural responses were also altered during hypercapnia and subsequent modelling of hemodynamics suggested that this difference in neural response might explain the change in hemodynamic responses seen. Although data did not clearly dissociate early and late vascular responses, they have important implications regarding the neurogenic and metabolic basis of neurovascular coupling. Hypercapnia greatly increased blood volume and saturation within the cortex (and therefore the amount of available oxygen and glucose), but as hemodynamic responses were elicited by stimuli and as neurovascular coupling remained similar, this strongly suggests neurovascular coupling is driven by neurogenic processes, not by metabolic ones.

Materials and methods

Animal Preparation and Surgery (n=10)

Female Hooded Lister rats of 230 to 330 g were kept in a 12-hour dark/light cycle at a temperature of 22°C, with food and water supplied ad libitum. Animals were anesthetized with urethane at 1.25 g/kg intraperitoneally and additional doses of 30 mg/kg intraperitoneally urethane were administered if required. Atropine was administered at 0.4 mg/kg subcutaneously to lessen mucous secretions during surgery. Temperature was maintained at 37°C using a homeothermic blanket (Harvard Apparatus, Edenbridge, Kent, UK) through rectal temperature monitoring. The animals were tracheotomized, allowing artificial ventilation with pressurized room air via a medical air compressor (Bambi, Birmingham, UK) and recording of end-tidal CO₂. Blood gas measurements and end-tidal CO₂ measurements were taken to allow correct adjustment of ventilator parameters to keep the animal within physiological limits. The left and right femoral arteries and veins were cannulated to allow the measurement of mean arterial blood pressure and drug infusion. Phenylephrine was infused at 0.13 to 0.26 mg/h to maintain mean arterial blood pressure between 100 and 110 mm Hg. Control values were pO₂=88.4 mm Hg±3.6; pCO₂=30.9 mm Hg ±2.0, blood pressure=103.1 mm Hg±3.9; heart rate=356 bpm±9.1. All procedures were performed in accordance with the 1986 Animal (Scientific Procedures) Act, under approval from the UK Home Office.

Two-Dimensional Optical Imaging Spectroscopy Imaging

Animals were placed in a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA). The skull overlying the somatosensory cortex was thinned with a dental drill. A circular plastic ‘well' was attached over the thinned area of the skull using dental cement. The well was filled with saline to reduce specularities from the skull surface. The 2D-OIS was used to estimate changes in cortical oxyhemoglobin (HbO₂), deoxyhemoglobin (Hbr), and total hemoglobin concentration (Hbt). A Dalsa 1M30P camera (Billerica, MA, USA) operating in 4 × 4 binning mode recorded the images; hence, each image pixel represented 75 × 75 μm of the object. To generate spatial maps of cortical hemodynamic responses, the 2D-OIS technique used a Lambda DG-4 high-speed filter changer (Sutter Instrument Company, Novata, CA, USA). The four wavelengths were specifically chosen as two pairs (495 nm±31 FWHM (full width at half maximum) and 559 nm±16 FWHM; 575 nm±14 FWHM and 587 nm±9 FWHM). The wavelengths in each pair were chosen such that they had a similar total absorption coefficient and thus sample the same tissue volume but have specific absorption coefficients for HbO₂ and Hbr that are as different as is possible to maximize signal-to-noise ratio. The frame rate of the camera was 32 Hz, which was synchronized to the filter switching, thereby giving an 8-Hz effective frame rate for each. To estimate changes in cortical hemodynamics, the data were subject to subsequent spectral analysis was based on the path length scaling algorithm described in detail previously (Berwick et al, 2008, 2005). Briefly, the algorithm uses a modified Beer–Lambert Law with a path length correction factor. We estimated the concentration of hemoglobin in tissue at 104 μmol/L based on the previous measurements (Kennerley et al, 2009) and the saturation was estimated to be 50%. The spectral analysis produced 2D images over time of HbO₂, Hbr, and Hbt. The depth sensitivity of this technique has been investigated with further Monte-Carlo stimulations and been reported previously (Berwick et al, 2005; Kennerley et al, 2005). The technique samples the first 1 mm of tissue and as such is well matched to current source density (CSD) analysis of local field potentials (LFPs).

Stimulus Presentation

Subcutaneous stimulation electrodes insulated to within 2 mm of the tip were inserted in a posterior direction between rows A/B and C/D of the left whisker pad of the rat ensuring the whole whisker pad was stimulated.

Localization of Whisker Barrel Somatosensory Cortex and Insertion of Electrode

An initial 2D-OIS ‘experiment' was conducted in each animal to localize the whisker barrel cortex region to accurately place a multichannel electrode. The whisker-pad stimulus was 1.2 mA intensity and consisted of a 5-Hz train of pulses for 2 seconds. In each animal, 30 stimulus presentation trials were collected. Each trial was 24 seconds in duration and the stimulus was presented after 8 seconds. From the end of one trial to the start of the next, there was a 2-second ‘rest' period in which no data were collected, resulting in a 26-second interval between stimuli. Trials were averaged to create a ‘mean' trial, which was subjected to the spectral analysis described above. Changes in total hemoglobin concentration were subject to a Statistical Parametric Mapping (SPM) analysis to localize changes associated with presentation of whisker stimuli. The resultant ‘activation maps' were coregistered with camera images of the cortical surface to enable accurate insertion of the electrode normal to the cortical surface, to a depth of 1,600 μm (Figure 1). Insertion of the electrode involved drilling a very small hole in the thinned skull directly above the selected location, piercing the dura, and then inserting the multichannel electrode. The probe was coupled to a preamp and data acquisition device (Medusa Bioamp, TDT, Alachua, FL, USA); using a custom written script in MATLAB (Mathworks, Cambridge, UK).

Figure 1.

Localization of whisker barrel cortex and histological confirmation of electrode placement. (A) Schematic showing the stimulation paradigm used to find the whisker barrel cortex. This trial was presented 30 times and then averaged to create a mean trial response. (B) In-vivo camera image of thinned cranial window preparation showing the surface vasculature overlying the somatosensory cortex of the rat under 575 nm illumination. (C) Statistical Parametric Mapping (SPM) z-score map of the Hbt response to 2 seconds whisker stimulation. The black region represents all pixels within 50% of the peak z-score response. The white regions represent all negative z-score pixels within 50% of the largest negative response. (D) In-vivo image showing 16-channel electrode subsequently placed into the whisker somatosensory area previously localized by the blood volume response. (E) Postmortem histology of a tangential surface section of cortex displaying surface blood vessels filled with photographic emulsion and the location of the electrode lesion. (F) Layer IV tangential histological section of cortex showing the whisker barrel cortex stained for cytochrome oxidase reactivity and the location of the electrode lesion (scale bar=1 mm).

MultiChannel Electrophysiology and Two-Dimensional Optical Imaging Spectroscopy Measurements of 16 Seconds Electrical Whisker-Pad Stimuli

Whisker-pad stimuli (1.2 mA intensity, 5 Hz stimulus pulse train for 16 seconds duration) were presented during 30 stimulus presentation trials for each experimental animal. Each trial was 50 seconds in duration with stimulus presentation occurring after 10 seconds. From the end of one trial to the start of the next, there was a 20-second ‘rest period' in which no data were collected thus resulting in a 70-second interval between successive stimulus presentations. The first 15 of these 30 trials occurred during normocapnia (control). After these first 15 trials, hypercapnia was induced for the remaining 15 trials by raising the amount of CO₂ in the inspired air to 12%, with oxygen (FIO₂) at 20% and nitrogen (FIN) at 68%. The first 15 trials (normocapnia—‘control') and last 13 trials (‘hypercapnia') were averaged to create a ‘mean' trial, which was then subjected to spectral analysis (see above). The ‘baseline' spectral analysis parameters were altered in the hypercapnia condition to 129 μmol/L ‘baseline' hemoglobin concentration and hemoglobin saturation of 0.65. These values were estimated by observing the continuous time series of hemodynamics before subsequent trial averaging (Figure 2). Neural activity was collected simultaneously (in 7 out of the 10 rats) across all 16 channels at a sampling frequency of 6.1 kHz for a period of 26.7 seconds starting 5.2 seconds before stimulation onset within each trial. The animals physiology in the hypercapnic condition measured: pO₂=114.7 mm Hg±3.5; pCO₂=76.0 mm Hg±3.7, blood pressure=116 mm Hg±3.7; heart rate=363 bpm±7.5.

Figure 2.

Baseline neural and hemodynamic changes caused by hypercapnia (FICO₂=12%). (A) Schematic showing the stimulation paradigm used for a single trial. (B) Baseline local field potential (LFP) time series across an entire stimulus presentation experiment (30 individual trials) from electrode channel 8 (∼800 μm below cortical surface from four representative animals). Normocapnic control period is shown in white while the hypercapnic period highlighted in blue. (C) Fast Fourier transform of averaged prestimulus LFP power as a function of frequency from the control and hypercapnia periods. (D) Corresponding baseline hemodynamic time series across an entire stimulus presentation experiment (from four animals) (error bars=s.e.m.). (E) Percentage change in hemoglobin saturation caused by the onset of hypercapnia. 2D-OIS, two-dimensional optical imaging spectroscopy.

Electrophysiological Data Analysis

The 16-channel neural data were analyzed by performing CSD analysis. Recordings were averaged over trials, with stimulus onset ‘jittered' within a 20-ms window to reduce effects of 50 Hz mains noise. The resultant evoked field potential recordings were sampled at 6 kHz with 16-bit resolution. The CSD analysis has been described in detail previously (Martindale et al, 2003). Multilaminar measures of evoked field potentials enabled the use of CSD analysis, which resolves the spatial ambiguities inherent in evoked field potential recordings into a laminar distribution of current sinks and sources. Whisker-pad stimuli result in a large sink-source dipole toward the surface of the cortex. As sinks reflect active excitatory postsynaptic potentials rather than passive (e.g., sources) neural mechanisms, we have typically compared the evoked sinks rather than sources with the accompanying hemodynamics (e.g., Martindale et al, 2003). This ‘primary' current sink has previously been found to colocalize extremely well with layer IV (∼450 μm below the surface of the cortex) as visualized by cytochrome oxidase histology (Jones et al, 2004). An autoradiography study (Gerrits et al, 2000) has shown that maximal cerebral blood flow was also observed in layer IV. For this reason, time series of the layer IV CSD sink was used as the basis of the ‘incoming' neural activity and intracortical processing, from which temporal profiles of the CSD analysis were extracted. Comparing this aspect of the neural response with the accompanying hemodynamics is thus justified from both a signal processing and neurophysiological perspective.

Selection of Regions of Interest for Two-Dimensional Optical Imaging Spectroscopy Data

To select regions of interest (ROIs) to generate subsequent time series, we analyzed the entire image of Hbt changes evoked by whisker stimuli during normocapnia control condition using a general linear model (GLM) SPM approach (Friston et al, 1991). The time series for each pixel was regressed against a design matrix of a representative ‘box car' hemodynamic response function with a ramp and direct current (DC) offset. Subsequent ‘activation' z-scores were calculated on a pixel-by-pixel basis. All pixels within 50% of the maximum z-score were selected and formed the ‘whisker region' chosen for time-series analysis. For 8 out of the 10 animals, regions with negative z-scores were observed adjacent to the whisker region. As such, this ‘negative' ROI was also selected for subsequent time-series analyses. Again, all pixels within 50% of the largest negative z-score were selected for this region. Further, arterial, parenchyma, and vein ‘subregions' were then selected manually from within the ‘whisker region' of interest.

PostMortem Cytochrome Oxidase Histology to Confirm Location of Somatosensory Cortex

After the experiment, some animals were processed to fill cortical vessels with photographic emulsion and stain the barrel cortex for cytochrome oxidase (Wong-Riley and Welt, 1980; Figure 3). This method has been described in detail previously (Berwick et al, 2008).

Figure 3.

Spatial images of trial averaged 16 seconds electrical whisker stimulation-evoked hemodynamic responses from four representative animals during normocapnia (‘control') and hypercapnia. Far left hand column shows in-vivo images of surface vasculature. The highlighted regions on are for whole whisker region (black), and within this area regions for artery (red), parenchyma (green), and vein (blue) were also selected. The surround negative region is shown in white. The red lines depict the main branches of the middle cerebral artery (MCA) for each animal. Middle columns show images of average hemodynamic changes over the entire stimulation period in control and under hypercapnia. Far right hand column shows images of subsequent postmortem histology confirming the location of the whisker barrel cortex in each animal (Hbt, total hemoglobin; HbO₂, oxyhemoglobin; Hbr, deoxyhemoglobin).

Results

Localizing the Whisker Barrel Cortex for Electrode Placement

In all animals, the first experiment localized the whisker barrel cortex using 2D-OIS to a short 2 seconds stimulation of the whisker pad (Figure 1). The averaged Hbt trial data were analyzed using a GLM SPM approach (Friston et al, 1991). This produced a z-score map showing the pixels that corresponded most to a 2-second stimulus design matrix. The black region shows all pixels within 50% of the largest z-score and the white region represents all negative pixels within 50% of the biggest negative z-score (Figures 1B and 1C). This method is used below to select regions for the long duration stimulation. The resultant placement of the 16-channel electrode into the active region is clearly shown (Figure 1D). After in-vivo experimental procedures, some animals underwent histology to fill the surface vessels with photographic emulsion (Figure 1E) and to stain the whisker barrel cortex for cytochrome oxidase reactivity (Figure 1F). In this example (Figures 1E and 1F), the electrode lesion can clearly be seen in both surface and barrel sections.

Baseline Cortical Hemodynamics and Ongoing Neural Activity After Hypercapnic Challenge

The experimental paradigm in each animal consisted of 30 stimulus presentation trials in which 16 seconds duration electrical whisker-pad stimuli were presented (Figure 2A). During the first 15 trials, electrical whisker-pad stimuli were presented during ‘baseline' normocapnic conditions whereas the last 15 trials involved presentation of electrical whisker-pad stimuli during hypercapnia. Hypercapnia causes an increase in blood flow and volume in the cerebral cortex through vasodilatation. Thus, to accurately assess this expected baseline (rather than stimulus-evoked) increase and any putative changes in baseline neural activity, in four experimental animals, we extended the data collection period of each stimulus presentation trial from 50 to 68 seconds. This effectively decreased the ‘rest period' in which no data were collected to 2 seconds between trials. Therefore, in these four animals the 70-second interstimulus interval was identical to other experimental subjects but allowed collection of almost continuous data to enable greater examination of hemodynamic changes elicited by hypercapnia. An ROI was selected from the cortical area activated by whisker stimuli, using the GLM analysis described in Materials and methods. We then collated the time series for each individual trial together such that an effectively continuous time series could be produced. We also collected continuous electrophysiological data from a single channel (channel 8–800 μm below the cortical surface) to assess any change in ongoing ‘baseline' LFPs (Figure 2B). During normocapnia, the LFP reflected the slow wave sleep typically observed after standard doses of urethane anesthetic in rat (Angel and Harris, 1998; Nunez, 1996) characterized by high-amplitude low-frequency oscillations (1 to 14 Hz, Steriade et al, 1990). A reduction in mean amplitude of the continuous LFP time series after the onset of hypercapnia (marked by the red arrow in Figure 2B) was observed. The power spectra of the prestimulus LFP (Figure 2C) in control and hypercapnia showed that the lower frequency oscillations were dramatically reduced in the hypercapnic condition, an observation similar to that seen in recent mild hypercapnia studies in primate and human (Thesen et al, 2011; Zappe et al, 2008). Hypercapnia was also accompanied by a large increase in Hbt, HbO₂, and saturation, with a decrease in the level of Hbr (Figures 2D and 2E), thus suggesting a large increase in the baseline level of oxygenated hemoglobin in cortex.

Image Montage and Time-Series Responses Across Animals

Data from four example animals are presented (Figure 3) to display the hemodynamic response to 16 seconds whisker stimulation during normocapnic and hypercapnic conditions both spatially and in magnitude. The left hand column shows the in-vivo charge coupled device camera images of the surface cortex under 575 nm illumination. The activated whisker somatosensory cortex and ‘surround' ROIs selected from the GLM analysis are overlaid. The middle columns show images of stimulus-evoked hemodynamic changes during normocapnia and hypercapnia for each animal. Each image represents an average over the entire 16-second period of whisker stimulus presentation. Thus, the color of the image represents an average micromolar change from baseline produced by whisker-pad stimulation. In the normocapnic ‘control' condition, stimulus-evoked changes in Hbt were largest in the surface branches of the middle cerebral artery. For Hbr, there was a large decrease, which was localized to the main ‘draining' veins. The HbO₂ increase is more uniform across all cortical vascular compartments than are the changes in Hbt or Hbr. There was a marked difference in the spatial distribution of the stimulus-evoked hemodynamic response during the hypercapnia condition compared with that elicited during normocapnia (Figure 3, left side). The evoked increases in Hbt and HbO₂ were no longer biased to the large arteries and were more localized to the tissue between the major branches of the middle cerebral artery. Importantly, during hypercapnia there was little decrease in Hbr in the draining veins. Postmortem histology is shown in the right hand column (Figure 3).

Time Series of Stimulus-Evoked Cortical Hemodynamic Responses in Different Vascular Compartments During Normocapnia and Hypercapnia

To assess how the hemodynamic responses changed under the normocapnic and hypercapnic conditions, data were taken from ROIs and averaged across animals (Figure 4). The whole whisker region and surrounding negative region were selected based on a GLM analysis described above. Arterial, parenchyma, and vein regions were then manually selected from within the ‘whisker' region. It should be stated that although the responses from each region will be biased toward that particular vascular compartment, the 2D-OIS is effectively sampling data from the first millimeter of tissue (see Berwick et al, 2005), so these time series should be seen as having a biased weighting toward a particular vascular compartment rather than being solely from a vascular compartment per se (see Bruyns-Haylett et al, 2010). Regardless of the vascular compartment, the time courses of stimulus-evoked hemodynamic responses were different during hypercapnia compared with those elicited during normocapnia. In the normocapnic ‘control' condition (Figure 4A), the stimulus-evoked changes in Hbt and HbO₂ were characterized by a ‘peak' and subsequent plateau, which then displayed a return to baseline after the end of the stimulation period. The time to the Hbt peak in the parenchyma region across all animals was 3.14 seconds±0.3 Hbr showed a similar ‘peak' and ‘plateau' response to that observed for Hbt and HbO₂ but was inverted. The largest increase in Hbt was observed in the arterial ROIs and was smallest in the venous ROIs. The time course of the stimulus-evoked hemodynamic response in all vascular compartments changed considerably during hypercapnia (Figure 4B). For Hbt and HbO₂, there was an absence of the initial peak observed during normocapnic ‘control' conditions, with a slow increase to the peak that did not occur until the end of the stimulus presentation period and as such the time to peak across the animals for the Hbt parenchyma response was 15.9±1.95 seconds. The stimulus-evoked decreases in Hbr across all vascular compartments evident during normocapnia were absent during hypercapnia.

Figure 4.

Time series of cortical hemodynamic responses from regions of interest selected from different ‘vascular' compartments during normocapnia and hypercapnia (n=10). (A) Cortical hemodynamic responses elicited during normocapnia (‘control'). (B) Cortical hemodynamic responses elicited during hypercapnia condition. (C) Predicted blood oxygen level-dependent (BOLD) functional magnetic resonance imaging (fMRI) response when using the hemodynamic parenchyma time-series response as input to a biophysical model (Hbt, total hemoglobin; HbO₂, oxyhemoglobin; Hbr, deoxyhemoglobin; error bars=s.e.m.).

As the positive blood oxygen level-dependent (BOLD) fMRI response commonly used as ‘mapping' signal in functional neuroimaging is predominantly due to stimulus-evoked decreases in Hbr it would be anticipated that the stimulus-evoked BOLD signal changes would be reduced during the level of hypercapnia in the present investigation. Thus, to explore the relevance of our optical data to functional neuroimaging, we estimated BOLD fMRI responses in the control and hypercapnia conditions. Time series of hemodynamics from a parenchymal ROI served as input for a Monte-Carlo simulation of magnetic resonance (MR) signal attenuation (Martindale et al, 2008) to estimate the changes in fMRI, gradient-echo, BOLD signal expected at 7 T with echo time=15 ms. For the purpose of the Monte-Carlo simulation, we assumed an extravascular compartment with a mean vessel radius of 10 μm. We considered only the extravascular compartment of the MR signal at 7 T only a small fraction of signal originates from the intravascular compartment (Martindale et al, 2008). Predictions show an expected 3% positive BOLD signal in response to stimulation under normocapnia. The estimated BOLD response appears absent under hypercapnic conditions (Figure 4C) despite a clear hemodynamic response evident in the total hemoglobin and oxyhemoglobin time series.

Current Source Density Analysis of Electrical Whisker-Pad Stimulus-Evoked Local Field Potentials During Normocapnia (‘Control') and Hypercapnia

Displaying evoked hemodynamic responses elicited during normocapnia and hypercapnia on the same axes further facilitates their comparison (Figure 5A). The question arises as to whether this difference in stimulus-evoked hemodynamics is due to possible differences in evoked neural activity during hypercapnia compared with that observed during normocapnia. To investigate this possibility, CSD analysis was performed on the multilaminar measure of LFP provided by concurrent multichannel electrophysiology in the majority of animals. A major current sink in layer IV was localized after CSD analysis, and then spatially averaged to produce a CSD sink time series as described previously (Boorman et al, 2010; Martindale et al, 2003). This CSD sink is commonly attributed to the afferent, sensory-evoked excitatory postsynaptic potential activity in the cortex and represents synaptic input and cortical processing that is most correlated with the associated stimulus-evoked cortical hemodynamic response. The CSD responses in each animal were normalized to the first impulse of the neural profile in the control condition (Figure 5B). The response elicited by whisker-pad stimuli during normocapnia displayed the typical neural profile that we have shown a number of times after the same stimulus presentation paradigm (Berwick et al, 2008; Boorman et al, 2010). The response to the first impulse was the largest followed by a sharp reduction of response magnitude that gradually settled to a ‘plateau' level for the remainder of the stimulation train. Under hypercapnia (Figure 5C), the first neural impulse was reduced to that observed during normocapnia with subsequent neural responses to individual pulses of the stimulus train occurring at a similar height to normocapnia.

Figure 5.

Time series of cortical hemodynamic and electrophysiological responses during normocapnia and hypercapnia. (A) Average total hemoglobin (Hbt) time-series response from a parenchymal region of interest during normocapnia (‘control') and hypercapnia. (B) Layer IV current source density sink (CSD sink) response to electrical whisker-pad stimulation (16 seconds, 5 Hz, 1.2 mA) in the normocapnic control condition. (C) CSD sink response during hypercapnia.

Convolution of Neural Response with Canonical Impulse Response Function

Thus, it appeared that like stimulus-evoked hemodynamic responses, neural responses also differed during hypercapnia compared with those elicited during normocapnia. However, it is difficult to infer by observation alone, whether the two phenomena were related. Thus, to assess whether the difference in hemoglobin response was related to the altered neural response, we convolved the normalized inverted average neural activation profile across animals (Figures 6A and 6B) with a canonical hemodynamic impulse response function that was obtained in a previous study (Martindale et al, 2003). Indeed, the time series estimated by convolution appears to qualitatively resemble the average Hbt time series of responses elicited during normocapnia and hypercapnia, thus suggesting that the differences in total hemoglobin response were most likely due to differences in evoked activity. To further quantify the fits to the actual data of these predictions, coefficients of determination (R²) were calculated for stimulus duration (0 to 16 seconds) and suggested a reasonable fit to the data: normocapnic control (r²=0.87) and hypercapnic (r²=0.98) (Figures 6D and 6E). The fit for the return to baseline after stimulation (16 to 34 seconds) for the control condition (r²=0.88) was better predicted than for hypercapnia (r²=0.78). The overall coefficient of determination reflected this for the entire time-series control was r²=0.89 and hypercapnia r²=0.79.

Figure 6.

Predicting whisker pad-evoked cortical hemodynamic responses using linear convolution of a canonical hemodynamic impulse response function. (A) ‘Inverted' averaged normalized neural response profile from data obtained during normocapnia. (B) Inverted averaged normalized neural response profile during hypercapnia condition. (C) Canonical hemodynamic impulse response function. (D) Averaged measured and predicted whisker pad-evoked cortical Hbt time-series responses in the control condition. (E) Average measured and predicted Hbt time-series responses in the hypercapnia condition. Hbt, total hemoglobin.

Hemodynamic Response Cortical Region Adjacent the Whisker Barrel Region

Our previous research has showed that electrical stimulation of the whisker pad produces reliable reductions in blood volume and saturation in regions surrounding the whisker barrel cortex, which in turn are responsible for a prolonged negative BOLD signal observed in fMRI studies (Boorman et al, 2010). In 8 out of the 10 animals, there was a ‘negative' hemodynamic response observed in regions adjacent to the whisker barrel cortex suggested by SPM analysis (Figure 7A). This time series displayed the expected ‘inverted' hemodynamic response commensurate in direction with a prolonged negative BOLD fMRI response, namely that Hbt and HbO₂ decreased and Hbr increased above baseline after stimulus presentation. In the hypercapnic condition, all aspects of the hemodynamic response from the surround region were altered (Figure 7B) in that Hbt and HbO₂ and to a lesser extent Hbr monotonically increased with extremely long latency reaching a peak only ∼20 seconds after stimulation onset.

Figure 7.

Cortical hemodynamic responses from a surrounding negative region to the whisker barrel somatosensory cortex (n=8, two animals had no negative response in the control). (A) During normocapnia (‘control'). (B) During hypercapnia (Hbt, total hemoglobin; HbO₂, oxyhemoglobin; Hbr, deoxyhemoglobin; error bars=s.e.m.).

Discussion

The present study investigated whether early and late neurovascular responses could be dissociated after presentations of a prolonged 16 seconds stimulus to contralateral whisker pad. We attempted to inhibit the ‘early' aspect of the stimulus-evoked response by presenting stimuli during hypercapnia. The main findings were (1) the blood volume response under control normocapnic conditions contained the typical peak and plateau profile seen previously (Berwick et al, 2008; Boorman et al, 2010). The early peak response was absent under the hypercapnic condition, leaving only the long latency response, suggesting we had revealed a putative arterial mechanism for the initial hemodynamic response component. (2) However, hypercapnia induced a change in both baseline neural and hemodynamic activity. Under hypercapnia, changes in the ongoing LFP were observed in addition to large increases in blood volume and saturation. (3) These ongoing LFP changes were also accompanied by substantial changes in the temporal profile of evoked neural activity compared with control. Convolving these temporal profiles with a hemodynamic impulse response function produced qualitatively similar time series to the hemodynamic responses observed in both the normocapnic and hypercapnic conditions, therefore suggesting that the differences in evoked activity were likely to have been the main factor in producing the differences in the time series of hemoglobin. (4) Intriguingly, a ‘negative hemodynamic' response found in a ‘surround region' adjacent to the whisker barrel somatosensory cortex observed in the normocapnic control condition was absent in the hypercapnic condition, and instead this area showed the presence of a long latency response.

Early and Late Neurovascular Responses

A hypothesis of this study was that the time course of neurovascular coupling to a longer duration stimulus (16 seconds) was the product of two distinct mechanisms, the first, an initial large dilation of the arterial ‘tree' which created the ‘peak' aspect of the hemodynamic response, followed by the second, a prolonged ‘plateau' response more localized to the site of neural activation. This hypothesis was based on the results from our laboratory and others, which suggested these separate mechanisms could exist (Kasischke et al, 2004; Martindale et al, 2005). We attempted to remove the peak response by predilating all the arteries with hypercapnia. Examination of the hemodynamics alone supported the hypothesis, as during hypercapnia the ‘peak' response was absent and the evoked hemodynamics were more localized to the tissue overlying the whisker barrels (Figure 3). Without any measure of the neural response (an approach common in many fMRI studies) we would have concluded that separate early and late mechanisms existed. However, concurrent multichannel electrophysiology suggested hypercapnia caused large changes in the baseline cortical state, which were associated with an altered profile of stimulus-evoked neural activity. Compared with that observed during normocapnia the neural response elicited by whisker stimuli displayed a reduction in the initial pulse of the whisker-pad stimulus train followed by a reduced degree of sensory adaptation. This difference in somatosensory-evoked neural response is similar to that reported by Castro-Alamancos when comparing cortical responses elicited by whisker stimuli during differences in ongoing LFP (Castro-Alamancos, 2004, 2009; Jones et al, 2008). To investigate whether this difference in evoked activity had altered the temporal profile of the hemodynamic response, we convolved the temporal profile of the neural responses with a standard hemodynamic impulse response function and showed that the change in the temporal profile of evoked neural activity was likely to be responsible for the altered hemodynamic response. We, therefore, concluded that the stimulus-evoked neurovascular coupling relationship was similar during both normocapnia and hypercapnia conditions. However, our findings do not preclude different mechanisms for producing the early and late stages of hemodynamic responses to long duration stimuli, but rather suggest that employing a multimodal approach such as we have used here, is essential for further exploring the subtle nature of this neurovascular coupling relationship (Berwick et al, 2008; Boorman et al, 2010).

Baseline State Changes and Implications for Functional Magnetic Resonance Imaging Studies

The study suggests that an important consideration for fMRI studies is that changes in baseline hemodynamic state disproportionately affects the washout of Hbr compared with other aspects of the hemodynamic response. As the BOLD signal is predominantly due to the stimulus-evoked decreases in Hbr, then this commonly used mapping signal may also be affected (Jones et al, 2005; Jones et al, 2008). Indeed, using an appropriate biophysical model (Martindale et al, 2008), the BOLD response in the hypercapnia condition was predicted to be nonexistent (Figure 4C). From the current investigation, it appears that changes in total blood volume are remarkably well preserved in spite of underlying changes in baseline hemodynamics. Therefore, for future MRI studies in which baseline state changes could be a complicating factor, cerebral blood volume weighted or arterial spin labelling (assuming Grub relationship holds) methodologies may produce data more reliably aligned to the underlying neural activity than measuring the BOLD response alone.

Hypercapnia also produced changes in baseline neural activity that were similar to those observed in two recent studies conducted in awake human and anaesthetized primate (Thesen et al, 2011; Zappe et al, 2008). There is still intense debate about the appropriateness of using hypercapnia as a calibration procedure for fMRI as this assumes hypercapnia has little effect on cortical activity or metabolism (Yablonskiy, 2011). However, although the current investigations suggested hypercapnia-induced changes in activity, the degree of hypercapnia was far greater that that typically used in calibrated fMRI studies.

Metabolic or Neurogenic? Recent Controversies Regarding Neurovascular Coupling

Although investigating the origin of neurovascular coupling was not the main research question addressed in this study, our data suggest stimulus-evoked neurovascular coupling to be ‘neurogenic' in origin, rather than being driven by absolute levels of metabolic demand. If neurovascular coupling was metabolic in origin then it would be expected that hemodynamics evoked would change under the condition of hypercapnia, because during hypercapnia saturation and blood volume in the cortex are increased by ∼20%, therefore making glucose and oxygen more freely available to the stimulated region of cortex. This interpretation, however, in the current investigation is confounded by the fact that baseline neural activity also changed during this level of hypercapnia and thus future research measuring the tissue oxygen concentration may assess whether the change in cortical state is associated with changes in metabolic demand that alter the baseline tissue oxygenation. Although, it has been previously reported that excessive glucose or oxygen supply does not alter stimulus-evoked CBF responses (e.g., Wolf et al, 1997), the novel finding here is that despite alterations in the temporal profile of both stimulus-evoked neural and hemodynamics response produced by severe hypercapnia, neurovascular coupling relationships remain similar to that observed during normocapnia.

There have been several recent publications that have called into question the stability of the neurovascular coupling relationship (Devor et al, 2008; Rossier, 2009; Sirotin and Das, 2009; Vanzetta and Slovin, 2010). The results of this study suggest that neurovascular coupling is tightly controlled even when measured during large increases in baseline saturation and blood volume in the cortex. However, it does appear that the washout of Hbr is most susceptible to changes in baseline and thus studies attempting to measure brain function with the BOLD fMRI response alone should be mindful of putative differences in baseline which could affect subsequent evoked responses. The current study also suggests that there are some aspects of the neurovascular response that may need to be investigated in greater detail (Figure 7). For instance, there was an extremely long latency response seen in the region ‘surrounding' the whisker barrel cortex during hypercapnia. Furthermore, if this response type, which may be related to the cortical state occurs on a wider spatial scale it may help explain the recent controversial finding of nonneuronal ‘predictive' hemodynamic responses shown in the awake primate by Sirotin and Das (2009). Rather than being predictive they could be showing long latency nonlocalized hemodynamic effects of previous stimuli. This might be even more prominent for short interstimulus intervals (as used in the Sirotin and Das study), and we are currently exploring if these long latency nonlocalized hemodynamic responses persist for very short single impulse stimuli under varying interstimulus intervals. However, it is important to note that other plausible explanations for the effect reported by Sirotin and Das (2009) have also been postulated (Kleinschmidt and Muller, 2010).

Conclusions

This study provides evidence that neurovascular coupling is remarkably robust even in the face of large baseline changes in saturation and blood volume in cortex. These changes strongly suggest that neurovascular coupling is primarily neurogenic in origin. The Hbr ‘washout' signal, which is the aspect of the hemodynamics response most closely related to the BOLD fMRI signal, appears to be the hemodynamic measure of neurovascular coupling most vulnerable to these changes, thus having implications for any BOLD fMRI studies that may involve baseline state changes. This highlights the importance of a multimodal approach for investigating neural function when using modern neuroimaging techniques, as relying solely on the BOLD fMRI signal could lead to false negatives purely due to changes in baseline state. This study has also revealed stimulus driven extremely long latency nonlocalized hemodynamic responses linked to cortical state which may require further investigation to interpret neuroimaging more accurately.

The authors declare no conflict of interest.