Abstract

The coupling of neuronal cellular activity to its blood supply is of critical importance to the physiology of the human brain and has been under discussion for more than a century. Linearity in this relationship has been demonstrated in some animal studies, but evidence is lacking in humans. In this study, we compared scalp evoked potentials and the functional magnetic resonance imaging (fMRI) blood oxygen level–dependent (BOLD) signal from healthy human volunteers with changes in the intensity of a somatosensory stimulus. By weighting the fMRI images according to the evoked potential amplitude at corresponding intensities, we tested for positive and negative covariation between these 2 data sets and the extent to which these were linear. Hemodynamic changes in primary somatosensory cortex covaried positively with neuronal activity in a predominantly linear manner, with a small quadratic contribution. Simultaneously, other cortical areas corresponding to the nonstimulated limbs were found to covary negatively and linearly in the hemispheres ipsilateral and contralateral to the stimulus. These concurrent and bilateral cortical dynamics, as well as the intraregional features of this neurovascular coupling, are both more complex than had been considered to date, with considerable implications.

Introduction

The coupling of the brain's neural activity to its blood supply, termed neurovascular coupling, and its mechanisms are a fundamental feature of brain physiology that have been under discussion for more than a century and are still not fully understood. Recent studies have identified a linear relationship between measures of hemodynamic change and neuronal activity in rats (Mathiesen and others 1998; Ngai and others 1999) and primates (Mathiesen and others 1998; Heeger and others 2000; Rees and others 2000; Logothetis and others 2001) and between functional magnetic resonance imaging (fMRI) blood oxygen level–dependent (BOLD) responses in humans and primate neuronal activity (Mathiesen and others 1998; Heeger and others 2000; Rees and others 2000; Logothetis and others 2001). Some nonlinearity in this relationship has also been identified in relation to animal experiments (Ances and others 2000; Jones and others 2004; Hewson-Stoate and others 2005). The relationship is less well characterized in humans, but modern neuroimaging methods combined with more traditional electrophysiological techniques now allow for a definition of the neurovascular relationship in normal human subjects.

A further physiological question relates to whether the brain increases blood flow to functionally active areas at the expense of other nonfunctioning areas. Neuroimaging has thrown some light on this: some studies have identified decreases in blood flow responses in functionally related (but not adjacent) cortical areas in both sensory (Drevets and others 1995; Peyron and others 1999) and motor areas (Allison and others 2000) during activation. However, whether these findings are related to decreased neuronal activity causing reductions in hemodynamic change (i.e., an underlying negative neurovascular coupling mechanism), or are a purely vascular phenomenon, remains unresolved.

The experiments presented here investigate the direction and linearity of the neurovascular coupling relationship in normal human subjects. We compared changes in cerebral blood flow (CBF) (using blocked design fMRI BOLD) and scalp electrophysiology (using somatosensory evoked potentials [SEPs]) in parallel experiments with changes in the intensity of a median nerve electrical stimulus. By weighting the fMRI images according to the evoked potential amplitudes at corresponding intensities, we sought to test the hypothesis that these 2 data sets covaried with each other, either positively or negatively, in a linear or nonlinear manner. We thus identified whether these relationships might vary within the somatosensory cortex itself or between hemispheres.

Methods

Stimulation

Six healthy adults participated (4 males; mean age 24.33 years, range 22–29 years), recruited from local university members. All studies were performed under Local Ethics Committee Approval guidelines, with full informed consent obtained. Stimuli were 0.2-ms square-wave electrical pulses delivered to the median nerve at the wrist for 30-s blocks. Stimulation intensity values were chosen to span a range from just above sensory threshold to the highest level bearable for 30 s but did not exceed 30 mA or individual pain thresholds. Values were normalized to individual motor thresholds to enable comparisons between data sets and across the group. During fMRI scanning, stimuli were delivered at 100 Hz to ensure that a detectable BOLD response could be recorded (Kampe and others 2000). During SEPs recording, stimuli were delivered at 20 Hz to allow accurate identification of cortical SEP components in transient mode. Current limiting resistors were placed in the stimulating cables during fMRI as a safety precaution (Lemieux and others 1997). The null hypothesis in this experiment was that there would be no significant covariance between fMRI BOLD responses in somatosensory cortex and SEP amplitudes during changes in stimulus intensity.

Somatosensory Evoked Potentials

SEPs were recorded using Ag/AgCl 10-mm disc electrodes from contralateral parietal cortex, 3 cm posterior and 7 cm lateral to the vertex (Cz) referenced to Fz, and from the mixed nerve at the elbow of the stimulated arm. Electrode impedances were maintained at less than 8 kΩ. Over 450 averages were made of 50 ms bin width and stored for subsequent off-line analysis. Scalp potentials were amplified using a band-pass filter of 3–3000 Hz. An automatic artifact rejection system excluded from the averages all runs containing transients exceeding ± 50 μV at any recording channel, commonly due to muscular or pulsatile artifacts. Stimulation intensity was pseudorandomized between successive recordings. The analysis was concentrated on the initial cortical component of the short-latency SEP, the N20–P25 component (termed “SEP amplitude”). The N20–P25 waveform of the human SEP is generated by a well-circumscribed population in the bank of the postcentral sulcus, that is, primary somatosensory cortex (Broughton and others 1969; Goff and others 1977; Allison and others 1980; Grimm Schreiber and others 1998). Pearson's rank test was used for correlation analysis (e.g., between SEP amplitudes and intensity level) using the Statistical Package for the Social Sciences. Statistical significance was set at P < 0.05. Statistical significance threshold was set at P < 0.05 throughout.

fMRI Acquisition

Imaging was performed on a Bruker Medical S300 scanner (Bruker Medical, Ettlingen, Germany), acquiring gradient echo-planar imaging (EPI) fMRI BOLD images of 25 contiguous 5-mm oblique axial slices at 3 T. The matrix size was 128 × 64 (an in-plane resolution of 2 × 4 mm), repetition time of 4000 ms, echo time of 27 ms, and 90° flip angle. Subjects lay supine in the scanner and wore earplugs and ear defenders for noise attenuation throughout, according to our local protocols. A functional imaging series comprised 72 sequential images of each slice during stimulus presentation in a blocked design, alternating 8 scans (32 s) on, 8 scans off, that is, a total of 4 min and 48 s. The first 8 scans were discarded to allow for T1 saturation effects. Stimulation intensities were unchanged during a 72-scan sequence but were pseudorandomized between sequences. Scanning also included the once-only acquisition of a fast gradient echo T1-weighted anatomical reference image of the whole brain. No neurological abnormalities were identified in any of the subjects studied.

fMRI Preprocessing

All image preprocessing and statistical analysis were done using statistical parametric mapping (SPM99; Wellcome Department of Cognitive Neurology, http://www.fil.ion.ucl.ac.uk/spm/) on Matlab (Mathworks Inc., MA) under Linux. Each image volume was reoriented, adjusted for acquisition slice timing, and realigned to the first of each sequence. The images were further spatially normalized into a standardized stereotactical space (Montreal Neurological Institute [MNI] space; the EPI template provided in SPM99) before being smoothed using a Gaussian filter of 4 mm full-width half maximum. Coordinates are therefore given in MNI space; areas were anatomically defined by transforming these coordinates into “Talairach space” (Talairach and Tournoux 1988).

fMRI BOLD Analysis

Blocks of stimuli were modeled using a boxcar function, incorporating a delay appropriate to the hemodynamic response. The size of the hemodynamic response was measured by calculating signal minus average baseline response across all scans. The modulation of hemodynamic responses by task-related activation was further characterized using both the (mean corrected) intensity and SEP amplitudes as linear regressors. This identified brain regions in which the size of task versus baseline hemodynamic response covaried linearly with these measures on a voxel-by-voxel basis (either positively or negatively).

We used SEP amplitudes as independent regressors in the fMRI analysis in order to show which voxels would best covary with SEP amplitudes. This approach does not require an a priori hypothesis regarding fMRI activation. This approach therefore eliminates the inherent assumption that the fMRI locus that correlates best with stimulus intensity will be that which correlates best with SEP amplitude. Nonlinear covariances of the fMRI with the SEP data were similarly modeled using a second-order (quadratic) derivative of SEP amplitudes (Robson 1958). Results were averaged across the group in a fixed-effects analysis to form a group-mean image, as it is widely recognized that at least 12 subjects are required for reasonable population effects to be seen in a random-effects model, due to intersubject variability (Holmes and Friston 1998). The images are shown in the standard radiological convention. All statistical maps were thresholded at P < 0.05, after correcting for multiple comparisons. Where clusters of activations were identified as significant, the coordinates of the voxel with highest t-score are given.

Results

Effects of Stimulus Intensity

Effects of Stimulus Intensity on SEP Amplitude

N20–P25 amplitudes of the cortical SEP correlated linearly with stimulus intensity in all subjects examined (P < 0.01 for each subject, group data P < 0.001, Fig. 1A, an example from one subject is given in Fig. 1C). Largest amplitudes were reached at 125% of motor threshold, which showed marked individual variability in absolute amplitude and intensity.

Figure 1.

The correlation of SEP N20–P25 amplitudes with stimulus intensity (A) is linear across subjects (mean ± standard error [SE] of mean; P < 0.001). An example from one subject is given in (C). Largest amplitudes were reached at 125% of motor threshold. The correlation of fMRI BOLD percent signal intensity change with stimulus intensity (B) is linear across subjects (mean ± SE mean; P < 0.05). Group data of n = 6 are shown.

Figure 1.

The correlation of SEP N20–P25 amplitudes with stimulus intensity (A) is linear across subjects (mean ± standard error [SE] of mean; P < 0.001). An example from one subject is given in (C). Largest amplitudes were reached at 125% of motor threshold. The correlation of fMRI BOLD percent signal intensity change with stimulus intensity (B) is linear across subjects (mean ± SE mean; P < 0.05). Group data of n = 6 are shown.

fMRI BOLD Activity

Irrespective of intensity, the fMRI BOLD voxel of maximal stimulus-induced activation was found in contralateral somatosensory cortex in each subject. Gradient echo fMRI BOLD voxel Z score and percent signal change at this peak voxel increased significantly with increasing stimulus intensity (P < 0.05, P < 0.05, respectively; Fig. 1B).

Testing for Linear Covariation with Stimulus Intensity

Positive Linear fMRI BOLD Covariation with Stimulus Intensity

The fMRI BOLD areas in contralateral somatosensory cortex (peak voxel coordinates: 38, −24, 62; Z score 10.23; cluster size 626; P < 0.001), contralateral thalamus (peak voxel coordinates: 16, −22, 02; Z score 6.81; cluster size 30; P < 0.001), and ipsilateral cerebellum (peak voxel coordinates: −20, −52, −34; Z score 5.38; cluster size 17; P < 0.001) showed significant linear covariation with stimulus intensity (Table 1; Fig. 2A).

Table 1.

fMRI BOLD areas that covary linearly with stimulus intensity

Figure Covariation Brain region Coordinates x, y, z Z score Cluster size 
2A Positive Contralateral SI 38, −24, 62 10.23** 636** 
50, −22, 56 10.16** 
44, −16, 54 10.04** 
 Positive Contralateral thalamus 16, −22, 2 6.81** 30** 
2B Negative Contralateral SI 2, −26, 58 4.80** 1* 
 Negative Ipsilateral SI −40, −34, 64 5.09* 5* 
Figure Covariation Brain region Coordinates x, y, z Z score Cluster size 
2A Positive Contralateral SI 38, −24, 62 10.23** 636** 
50, −22, 56 10.16** 
44, −16, 54 10.04** 
 Positive Contralateral thalamus 16, −22, 2 6.81** 30** 
2B Negative Contralateral SI 2, −26, 58 4.80** 1* 
 Negative Ipsilateral SI −40, −34, 64 5.09* 5* 

Note: Fixed-effects group analysis of n = 6 shown. **P < 0.001, *P < 0.05; SI = primary somatosensory cortex.

Figure 2.

The fMRI BOLD areas that covary linearly, (A) positively, and (B) negatively with stimulus intensity. These are contralateral somatosensory cortex, contralateral thalamus, and ipsilateral cerebellum (covary positively; A) and contralateral somatosensory cortex near the midline and ipsilateral somatosensory cortex (covary negatively; B). A fixed-effects group analysis of n = 6 is shown. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 1.

Figure 2.

The fMRI BOLD areas that covary linearly, (A) positively, and (B) negatively with stimulus intensity. These are contralateral somatosensory cortex, contralateral thalamus, and ipsilateral cerebellum (covary positively; A) and contralateral somatosensory cortex near the midline and ipsilateral somatosensory cortex (covary negatively; B). A fixed-effects group analysis of n = 6 is shown. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 1.

Negative Linear fMRI BOLD Covariation with Stimulus Intensity

Areas that covaried negatively with increasing stimulus intensity were contralateral somatosensory cortex near the midline (peak voxel coordinates: 2, −26, 58; Z score 4.80; cluster size 1; P < 0.05) and ipsilateral somatosensory cortex (peak voxel coordinates: −40, −34, 64; Z score 5.09; cluster size 5; P < 0.05; Table 1; Fig. 2B).

Testing for Linear Covariation with SEP Amplitude

Positive Linear fMRI BOLD Covariation with SEP Amplitude

Contralateral somatosensory cortex and ipsilateral cerebellum covaried significantly in a positive linear fashion with SEP amplitudes in each subject (Table 2, Fig. 3). The group analysis also showed most significant linear covariation with SEP amplitudes in contralateral somatosensory cortex (peak voxel coordinates: 40, −26, 64; Z score 10.17; cluster size 458; P < 0.001), contralateral thalamus (peak voxel coordinates: 16, −22, −02; Z score 6.08; cluster size 15; P < 0.05), and ipsilateral cerebellum (peak voxel coordinates: −12, −56, −32; Z score 5.21; cluster size 15; P < 0.05), across subjects (Table 3; Fig. 4A).

Table 2.

fMRI BOLD areas that covary positively and linearly with SEP amplitudes in individual subjects

Figure Subject Brain region Coordinates x, y, z Z score Cluster size 
3A Contralateral SI 50, −22, 52 10.88** 466** 
 Ipsilateral Cerebellum −16, −62, −30 6.34** 27** 
3B Contralateral SI 40, −22, 62 9.22** 150** 
3C Contralateral SI 38, −24, 58 7.34** 106** 
 Ipsilateral Cerebellum −20, −50, −38 5.75** 10** 
3D Contralateral SI 44, −28, 50 8.14** 113** 
3E Contralateral SI 52, −20, 52 11.74** 237** 
 Ipsilateral Cerebellum −24, −62, −30 7.49** 102** 
3F Contralateral SI 46, −36, 58 5.85** 29** 
Figure Subject Brain region Coordinates x, y, z Z score Cluster size 
3A Contralateral SI 50, −22, 52 10.88** 466** 
 Ipsilateral Cerebellum −16, −62, −30 6.34** 27** 
3B Contralateral SI 40, −22, 62 9.22** 150** 
3C Contralateral SI 38, −24, 58 7.34** 106** 
 Ipsilateral Cerebellum −20, −50, −38 5.75** 10** 
3D Contralateral SI 44, −28, 50 8.14** 113** 
3E Contralateral SI 52, −20, 52 11.74** 237** 
 Ipsilateral Cerebellum −24, −62, −30 7.49** 102** 
3F Contralateral SI 46, −36, 58 5.85** 29** 

Note: Fixed-effects group analysis of n = 6 shown. **P < 0.001, *P < 0.05; SI = primary somatosensory cortex.

Figure 3.

The fMRI BOLD areas that covary linearly and positively with SEP N20–P25 amplitudes in individual subjects. These are contralateral somatosensory cortex and ipsilateral cerebellum. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 2.

Figure 3.

The fMRI BOLD areas that covary linearly and positively with SEP N20–P25 amplitudes in individual subjects. These are contralateral somatosensory cortex and ipsilateral cerebellum. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 2.

Table 3.

fMRI BOLD areas that covary with SEP amplitude

Figure Covariation Brain region Coordinates x, y, z Z score Cluster size 
4A Positive, linear Contralateral SI 40, −26, 64 10.17** 458** 
50, −20, 56 10.14** 
 Positive, linear Contralateral thalamus 16, −22, −2 6.08** 15* 
4C Positive, nonlinear Contralateral SI 34, −34, 58 4.85* 2* 
4B Negative, linear Contralateral SI 4, −44, 58 6.28** 127** 
 Negative, linear Ipsilateral SI −42, −32, 62 6.04** 15* 
— Negative, nonlinear None    
Figure Covariation Brain region Coordinates x, y, z Z score Cluster size 
4A Positive, linear Contralateral SI 40, −26, 64 10.17** 458** 
50, −20, 56 10.14** 
 Positive, linear Contralateral thalamus 16, −22, −2 6.08** 15* 
4C Positive, nonlinear Contralateral SI 34, −34, 58 4.85* 2* 
4B Negative, linear Contralateral SI 4, −44, 58 6.28** 127** 
 Negative, linear Ipsilateral SI −42, −32, 62 6.04** 15* 
— Negative, nonlinear None    

Note: Fixed-effects group analysis of n = 6 shown. **P < 0.001, *P < 0.05; SI = primary somatosensory cortex.

Figure 4.

The fMRI BOLD areas that covary positively and linearly with SEP N20–P25 amplitudes (A), negatively and linearly with SEP N20–P25 amplitudes (B), and positively and nonlinearly with SEP N20–P25 amplitudes (C). These are contralateral somatosensory cortex and contralateral thalamus (covary positively and linearly, A), contralateral somatosensory cortex near the midline and ipsilateral somatosensory cortex (covary negatively and linearly, B), and contralateral somatosensory cortex (covary positively and nonlinearly, C). A fixed-effects group analysis of n = 6 is shown. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 3.

Figure 4.

The fMRI BOLD areas that covary positively and linearly with SEP N20–P25 amplitudes (A), negatively and linearly with SEP N20–P25 amplitudes (B), and positively and nonlinearly with SEP N20–P25 amplitudes (C). These are contralateral somatosensory cortex and contralateral thalamus (covary positively and linearly, A), contralateral somatosensory cortex near the midline and ipsilateral somatosensory cortex (covary negatively and linearly, B), and contralateral somatosensory cortex (covary positively and nonlinearly, C). A fixed-effects group analysis of n = 6 is shown. These images are thresholded at P < 0.05 corrected for multiple comparisons. The corresponding data are given in Table 3.

Negative Linear fMRI BOLD Covariation with SEP Amplitude

Two areas which covaried negatively with SEP amplitudes were found: the first in ipsilateral primary somatosensory cortex, in a similar location to that which covaried positively with SEP amplitudes above but in the opposite hemisphere (upper limb hand areas; peak voxel coordinates: −42, −32, 62; Z score 6.04; cluster size 15; P < 0.05; Table 3; Fig. 4B); and the second in contralateral primary somatosensory cortex areas, close to the midline (leg and/or foot areas; peak voxel coordinates: 4, −44, 58; Z score 6.28; cluster size 127; P < 0.001; Table 3; Fig. 4B).

Testing for Nonlinear Covariation with SEP Amplitude

Positive Quadratic fMRI BOLD Covariation with SEP Amplitude

Nonlinear contributions were also modeled, and a small area (peak voxel coordinates: 34, −34, 58; Z score 4.85; cluster size 2; P < 0.05; Table 3; Fig. 4C) of fMRI BOLD activity in contralateral primary somatosensory cortex was found to covary significantly with SEP amplitudes. This cluster fell within the boundaries of the larger cluster of 458 voxels that covaried linearly with SEP amplitudes (Fig. 4A).

Negative Quadratic fMRI BOLD Covariation with SEP Amplitude

No clusters reached significance thresholding for negative, nonlinear covariations with SEP amplitudes (results not shown).

No other types of nonlinearity reached significance in preliminary data analysis, neither positively nor negatively.

Discussion

SEP amplitudes and fMRI BOLD responses correlated linearly with stimulus intensity in all subjects, and across subjects. Both modalities therefore exhibited the same qualitative pattern of experimental effects in primary somatosensory cortex when measured in parallel. Furthermore, the covariation of fMRI BOLD responses with SEP amplitudes (indicative of neurovascular coupling) was found to be strongly linear in this area. Short-latency SEPs are attributed mainly to synchronized extracellular currents from summated postsynaptic potentials of pyramidal cells in primary somatosensory cortex (Eccles 1951; Creutzfeldt and others 1966; Lopes da Silva and Storm van Leeuwan 1978; Nunez 1981; Lopes da Silva 1991). fMRI BOLD responses predominantly measure the CBF response from cortical vessels due to the inherent magnetic changes in hemoglobin during activation (a transient drop in the deoxy:oxy-hemoglobin ratio [Fox and Raichle 1986; Ogawa and others 1990; Kwong and others 1992; Malonek and others 1997]). Although the correlation found in these experiments does not prove causation, these findings together imply that the synaptic activity of a population of somatosensory cortical neurons play a major role in signaling the needs of the neuron to the vasculature. This is consistent with previous findings in primates (Logothetis, Pauls and others 2001) and the empirical evidence that cortical action potentials may contribute relatively little to the metabolic demand of the brain (Creutzfeldt 1995).

However, electrophysiological responses have inherent differences in signal to noise between different techniques, such that electrical signals that can be measured by extracellular electrodes are preferentially biased toward the slow oscillations of membrane potential (local field potential [LFP]) rather than action potentials or spikes. SEP measurements are too coarse to detect spiking activity, and therefore their contribution to the fMRI BOLD signals observed cannot be meaningfully analyzed using these techniques. Given evidence that human fMRI BOLD activity may be proportional to primate single-cell firing activity, this suggests that action potentials may be important in the neurovascular coupling relationship (Heeger and others 2000; Rees and others 2000). The hemodynamic response may therefore be better related to the underlying ensemble electrical activity, including both LFP and spiking activity. The importance of this possibility and several other factors that may affect the lack of a perfect relation between BOLD and SEPs have been discussed elsewhere at length (Arthurs and Boniface 2002; Heeger and Rees 2002; Rees and others 2002; Logothetis and Pfeuffer 2004). Our results suggest that the cortical activity measured using the early component of the SEP makes a strong linear contribution to the vascular changes that dominate the human BOLD response.

The exact metabolic nature of the neurovascular signal currently remains unknown, although there are many possibilities, including the astrocytic recycling of glutamate (the “astrocyte-neuron lactate shuttle” hypothesis) (Pellerin and others 1998), increased potassium levels causing vasodilatation (Paulson and Newman 1987), and/or increases in nitric oxide and adenosine (Dirnagl and others 1994). All these metabolic candidates fail to demonstrate the necessary temporal and precise spatial relationship between accumulations and flow increase and have been previously discussed in detail with other possibilities (Lou and others 1987; Villringer and Dirnagl 1995; Kuschinsky 1997).

Nonlinearity

In this study, a much smaller area of fMRI BOLD activity in primary somatosensory cortex was also found to covary with SEP amplitudes in a quadratic fashion (Fig. 4C). This area fell within the larger cluster in somatosensory cortex that correlated linearly (Fig. 4A). Given the small nature of this response, any inference made must be speculative. However, a small nonlinear component to the overwhelming linear neurovascular coupling relationship (2 nonlinear/458 linearly covarying voxels, i.e., 0.44%) is consistent with most (Mathiesen and others 1998; Brinker and others 1999; Heeger and others 2000; Ogawa and others 2000; Rees and others 2000; Logothetis and others 2001) but not all recent studies (Ances and others 2000). Previous studies have found nonlinearities at the extreme ends of this relationship, such that the relationship is well approximated by a linear function over the midrange of stimuli (Hewson-Stoate and others 2005), as seen here.

There is currently a debate as to whether this nonlinearity may itself be attributable to a nonlinear neuronal response, including the contribution of transiently high neuronal-spiking activity to a hemodynamic response that is primarily synaptically driven (Bandettini and Ungerleider 2001; Logothetis and others 2001; Birn and Bandettini 2005), or alternatively due to the inherent nonlinear relationship between metabolic demand and the BOLD signal. The BOLD response is linearly related neither to CBF nor to the cerebral metabolic rate of oxygen consumption (CMRO2) (Rees and others 1997; Hoge and others 1999; Mandeville and others 1999). Changes in cerebral oxygen consumption have been shown to increase linearly with synaptic activity but demonstrate a threshold effect, also contributing to nonlinearities (Sheth and others 2004). Alternatively, the nonlinearity may arise from slight differences in sensitivity of our acquisition techniques, as discussed later in this section.

Negative Coupling

Two areas of negative linear neurovascular coupling were also found in this study: in ipsilateral somatosensory cortex corresponding to the upper limb including the hand area (mirroring the area of positive linear coupling in the opposite hemisphere) and in contralateral somatosensory cortex corresponding to the sensory representation of the leg and foot (Fig. 4B). This suggests an efficient suppression of blood flow to relatively “inactive” limb cortical areas. These findings were accessible only by virtue of this type of voxel-by-voxel covariance analysis, and there are a number of possible interpretations.

We use the term “negative” BOLD to mean a reduction in BOLD activity relative to baseline which corresponds to the timing of the stimulus, also referred to as deactivation. There are 2 such types of negative BOLD signal identified in the current literature. The first is a transient initial negative dip in BOLD signal to stimulus activation characterizing the hemodynamic response function. This is thought either to be a hemodynamic steal phenomenon or to be caused by oxygen consumption in the absence of a hemodynamic response (Rother and others 2002). The other, demonstrated in this experiment, is a sustained negative BOLD response usually seen at distances of centimeters away from stimulated areas but in physiologically correlated areas, such as opposite motor (Hamzei and others 2002), sensory (Drevets and others 1995), and visual areas (Smith and others 2004). Initially thought to be a hemodynamic steal phenomenon caused by a redistribution of blood flow to adjacent areas of cortex (Harel and others 2002), it now appears much more likely to represent a neuronally driven inhibitory phenomenon, when observed at sites distant to, but functionally related to, active brain areas. No areas of blood flow decreases were observed in the penumbral region of the activated area in somatosensory cortex in our study.

The fMRI BOLD signal decreases seen in ipsilateral motor cortex during unilateral hand movements are proportional to the task-related increases in contralateral M1 (in parallel with duration of movement) (Newton and others 2005). Studies have also shown that the reductions in BOLD signal in this area during contralateral activations are linearly related to the metabolic down-regulation, that is, CBF and cerebral metabolic rate of oxygen consumption (CMRO2) changes, suggesting an inhibitory neuronal signal underlying this negative BOLD response (Stefanovic and others 2004). Negative BOLD has also been reported during electroencephalography (EEG) spiking activity and found to occur at sites that are distant from anatomical areas related to spikes (Kobayashi and others 2005), suggesting neuronal inhibition.

Decreases in activation have also previously been attributed to higher level cortical function changes, such as the “anticipation” of expected stimuli elsewhere: CBF decreases in hand and face zones of ipsilateral somatosensory cortex (while attending to toe stimulation) have been correlated with anxiety levels during anticipation of stimuli (Drevets and others 1995). This suggests the suppression of ipsilateral responses in order to “focus on,” or “attend to,” contralateral responses where stimuli are expected in direct proportion to the anxiety level. fMRI BOLD responses in our experiment might therefore covary negatively with SEP amplitudes because they, in turn, covary with increasing stimulus intensity. However, although decreases in fMRI BOLD signal from ipsilateral somatosensory cortex do correlate with stimulus intensities, P values were lower and voxel t-statistics less significant than the equivalent analyses with SEP amplitudes (Tables 1 and 3, respectively). Ipsilateral cortical responses may therefore be more closely (albeit negatively) related to contralateral responses rather than to the stimulus.

An alternative consideration is the changes in ongoing event-related desynchronization or synchronization at particular frequency bands (Pfurtscheller and Lopes da Silva 1999; Pfurtscheller 2001), which may represent increased cortical activation and deactivation, respectively, where “activation” represents increased resonance-like behavior of connected subnetworks. These types of changes are time locked to the event but not phase locked and therefore cannot be extracted using conventional linear methods such as averaging (as in SEP recordings), but require frequency analysis. They have been demonstrated in somatosensory and visual cortices (Neuper and Pfurtscheller 2001; Pfurtscheller and others 2001; Singh and others 2002; Moosmann and others 2003). The amplitude of negative fMRI BOLD responses to acoustic stimulation has been shown to correlate positively with measures of EEG synchronization during sleep (Czisch and others 2004), suggesting a relationship between cortical deactivation and negative BOLD signals. Further, analysis of these network synchronization changes may give a greater understanding of the underlying signal causing negative BOLD changes.

What we have shown is that the negative fMRI BOLD activity seen in contralateral somatosensory cortex correlates with neuronal activity (as indexed by the SEP) and fMRI BOLD changes in the “active” cortical area, that is, they are directly related to markers of neuronal activity elsewhere. If these findings reflect a neuronally mediated corticocortical inhibition, such that ipsilateral cortical activity is inhibited in proportion to increases in contralateral cortical activity, then it is possible that lesions of the corpus callosum might disrupt these neuronal connections.

Experimental Confounds

We must acknowledge some differences in the implementation of the fMRI BOLD and SEP protocols used in this experiment, although these are unlikely to create substantial experimental confounds. First, electrical stimulation of the medial nerve was applied at 100 Hz during fMRI recording and at 20 Hz during SEP recording, as in a previous study (Arthurs and others 2000). The lower frequency allows reliable identification of early components in the SEP recording, whereas the higher stimulation frequency is more efficient for determining the fMRI BOLD responses (Kampe and others 2000). SEPs at 100 Hz are inherently difficult to record due to the stimulus artifact, and the standing waveform generated is difficult to interpret at this frequency in the absence of more sophisticated analysis methodology. Short-latency SEP intensity–dependent stimulus response curves have not been found to vary significantly between 20 and 100 Hz (O. Arthurs and S. Boniface, unpublished data). However, 2 different frequencies are required to optimize each signal (fMRI and SEP), and this suggests that the different frequencies may generate subtly different responses. The use of multichannel EEG or magnetoencephalography recording may be required to accurately identify SEP component generators at high stimulation frequencies to resolve these issues.

Second, we also only modeled quadratic, second-order nonlinearities in the data and did not further investigate third order or other types of nonlinearity. However, no other types of nonlinearity reached significance in preliminary data analysis. Further, detailed modeling of this relationship may reveal more subtle nonlinearities.

Third, we used block design fMRI recording and event-related SEP recordings, although both SEPs and the fMRI BOLD response were summarized over a 30-s block, in order to eliminate and minimize any initial adaptive responses. Given the long periods over which these data sets were recorded in parallel, this is unlikely to make a significant difference. However, we acknowledge that these practical constraints and differences in techniques could account for small changes, such as the nonlinear covariations observed.

Summary

In conclusion, the simultaneous activation and suppression of functionally related cortical areas, as well as the intraregional features of the neurovascular coupling response, appear considerably more complex than has been considered to date and require further investigation.

This project was supported by Technology Foresight, UK and by Oxford Instruments, UK. Merck Sharp and Dohme supported OJA on an MB/PhD program. Conflict of Interest: None declared.

References

Allison
JD
Meador
KJ
Loring
DW
Figueroa
RE
Wright
JC
Functional MRI cerebral activation and deactivation during finger movement
Neurology
 , 
2000
, vol. 
54
 
1
(pg. 
135
-
142
)
Allison
T
Goff
W
Williamson
PD
Van Gilder
JC
Desmedt
JE
On the neural origin of early components of the human somatosensory evoked potential
Progress in clinical neurophysiology: clincal uses of cerebral, brainstem and spinal somatosensory evoked potentials
 , 
1980
, vol. 
7
 
Basel, Switzerland
Karger
(pg. 
51
-
68
)
Ances
BM
Zarahn
E
Greenberg
JH
Detre
JA
Coupling of neural activation to blood flow in the somatosensory cortex of rats is time-intensity separable, but not linear
J Cereb Blood Flow Metab
 , 
2000
, vol. 
20
 
6
(pg. 
921
-
930
)
Arthurs
OJ
Boniface
SJ
How well do we understand the neural origins of the fMRI BOLD signal?
Trends Neurosci
 , 
2002
, vol. 
25
 
1
(pg. 
27
-
31
)
Arthurs
OJ
Williams
EJ
Carpenter
TA
Pickard
JD
Boniface
SJ
Linear coupling between functional magnetic resonance imaging and evoked potential amplitude in human somatosensory cortex
Neuroscience
 , 
2000
, vol. 
101
 
4
(pg. 
803
-
806
)
Bandettini
PA
Ungerleider
LG
From neuron to BOLD: new connections
Nat Neurosci
 , 
2001
, vol. 
4
 
9
(pg. 
864
-
866
)
Birn
RM
Bandettini
PA
The effect of stimulus duty cycle and “off” duration on BOLD response linearity
Neuroimage
 , 
2005
, vol. 
27
 
1
(pg. 
70
-
82
)
Brinker
G
Bock
C
Busch
E
Kerp
H
Hossmann
KA
Hoehn-Berlage
M
Simultaneous recording of evoked potentials and T2*-weighted MR images during somatosensory stimulation of rat
Magn Reson Med
 , 
1999
, vol. 
41
 
3
(pg. 
469
-
473
)
Broughton
R
Meier-Ewert
KH
Ebe
M
Evoked visual, somato-sensory and retinal potentials in photosensitive epilepsy
Electroencephalogr Clin Neurophysiol
 , 
1969
, vol. 
27
 
4
(pg. 
373
-
386
)
Creutzfeldt
OD
Cortex cerebri: performance, structural and functional organisation of the cortex
1995
New York
Oxford University Press
Creutzfeldt
OD
Watanabe
S
Lux
HD
Relations between EEG phenomena and potentials of single cortical cells. I. Evoked responses after thalamic and erpicortical stimulation
Electroencephalogr Clin Neurophysiol
 , 
1966
, vol. 
20
 
1
(pg. 
1
-
18
)
Czisch
M
Wehrle
R
Kaufmann
C
Wetter
TC
Holsboer
F
Pollmacher
T
Auer
DP
Functional MRI during sleep: BOLD signal decreases and their electrophysiological correlates
Eur J Neurosci
 , 
2004
, vol. 
20
 
2
(pg. 
566
-
574
)
Dirnagl
U
Niwa
K
Lindauer
U
Villringer
A
Coupling of cerebral blood flow to neuronal activation: role of adenosine and nitric oxide
Am J Physiol
 , 
1994
, vol. 
267
 
1 Pt 2
(pg. 
H296
-
H301
)
Drevets
WC
Burton
H
Videen
TO
Snyder
AZ
Simpson
JR
Jr
Raichle
ME
Blood flow changes in human somatosensory cortex during anticipated stimulation
Nature
 , 
1995
, vol. 
373
 
6511
(pg. 
249
-
252
)
Eccles
JC
Interpretation of action potentials evoked in the cerebral cortex
Electroencephalogr and Clin Neurophysiol
 , 
1951
, vol. 
3
 (pg. 
449
-
464
)
Fox
PT
Raichle
ME
Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects
Proc Natl Acad Sci USA
 , 
1986
, vol. 
83
 
4
(pg. 
1140
-
1144
)
Goff
GD
Matsumiya
Y
Allison
T
Goff
WR
The scalp topography of human somatosensory and auditory evoked potentials
Electroencephalogr Clin Neurophysiol
 , 
1977
, vol. 
42
 
1
(pg. 
57
-
76
)
Grimm
C
Schreiber
A
Kristeva-Feige
R
Mergner
T
Hennig
J
Lucking
CH
A comparison between electric source localisation and fMRI during somatosensory stimulation
Electroencephalogr Clin Neurophysiol
 , 
1998
, vol. 
106
 
1
(pg. 
22
-
29
)
Hamzei
F
Dettmers
C
Rzanny
R
Liepert
J
Buchel
C
Weiller
C
Reduction of excitability (“inhibition”) in the ipsilateral primary motor cortex is mirrored by fMRI signal decreases
Neuroimage
 , 
2002
, vol. 
17
 
1
(pg. 
490
-
496
)
Harel
N
Lee
SP
Nagaoka
T
Kim
DS
Kim
SG
Origin of negative blood oxygenation level-dependent fMRI signals
J Cereb Blood Flow Metab
 , 
2002
, vol. 
22
 
8
(pg. 
908
-
917
)
Heeger
DJ
Huk
AC
Geisler
WS
Albrecht
DG
Spikes versus BOLD: what does neuroimaging tell us about neuronal activity?
Nat Neurosci
 , 
2000
, vol. 
3
 
7
(pg. 
631
-
633
)
Heeger
DJ
Rees
D
What does fMRI tell us about neuronal activity?
Nat Rev Neurosci
 , 
2002
, vol. 
3
 
2
(pg. 
142
-
151
)
Hewson-Stoate
N
Jones
M
Martindale
J
Berwick
J
Mayhew
J
Further nonlinearities in neurovascular coupling in rodent barrel cortex
Neuroimage
 , 
2005
, vol. 
24
 
2
(pg. 
565
-
574
)
Hoge
RD
Atkinson
J
Gill
B
Crelier
GR
Marrett
S
Pike
GB
Linear coupling between cerebral blood flow and oxygen consumption in activated human cortex
Proc Natl Acad Sci USA
 , 
1999
, vol. 
96
 
16
(pg. 
9403
-
9408
)
Holmes
AP
Friston
KJ
Generalisability, random effects and population inference
Neuroimage
 , 
1998
, vol. 
7
 pg. 
S754
 
Jones
M
Hewson-Stoate
N
Martindale
J
Redgrave
P
Mayhew
J
Nonlinear coupling of neural activity and CBF in rodent barrel cortex
Neuroimage
 , 
2004
, vol. 
22
 
2
(pg. 
956
-
965
)
Kampe
KK
Jones
RA
Auer
DP
Frequency dependence of the functional MRI response after electrical median nerve stimulation
Hum Brain Mapp
 , 
2000
, vol. 
9
 
2
(pg. 
106
-
114
)
Kobayashi
E
Bagshaw
AP
Grova
C
Dubeau
F
Gotman
J
Negative BOLD responses to epileptic spikes
Hum Brain Mapp
 , 
2005
 
[Epub ahead of print]
Kuschinsky
W
Neuronal-vascular coupling: a unifying hypothesis
Adv Exp Med Biol
 , 
1997
, vol. 
413
 (pg. 
167
-
176
)
Kwong
KK
Belliveau
JW
Chesler
DA
Goldberg
IE
Weisskoff
RM
Poncelet
BP
Kennedy
DN
Hoppel
BE
Cohen
MS
Turner
R
Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation
Proc Natl Acad Sci USA
 , 
1992
, vol. 
89
 
12
(pg. 
5675
-
5679
)
Lemieux
L
Allen
PJ
Franconi
F
Symms
MR
Fish
DR
Recording of EEG during fMRI experiments: patient safety
Magn Reson Med
 , 
1997
, vol. 
38
 
6
(pg. 
943
-
952
)
Logothetis
NK
Pauls
J
Augath
M
Trinath
T
Oeltermann
A
Neurophysiological investigation of the basis of the fMRI signal
Nature
 , 
2001
, vol. 
412
 
6843
(pg. 
150
-
157
)
Logothetis
NK
Pfeuffer
J
On the nature of the BOLD fMRI contrast mechanism
Magn Reson Imaging
 , 
2004
, vol. 
22
 
10
(pg. 
1517
-
1531
)
Lopes da Silva
F
Neural mechanisms underlying brain waves: from neural membranes to networks
Electroencephalogr Clin Neurophysiol
 , 
1991
, vol. 
79
 (pg. 
81
-
93
)
Lopes da Silva
F
Storm van Leeuwan
W
Brazier
MAB
Petsche
H
The cortical alpha rhythm in dog: the depth and surface profile of phase
Architectonics of the cerebral cortex
 , 
1978
New York
Raven Press
(pg. 
319
-
333
)
Lou
HC
Edvinsson
L
MacKenzie
ET
The concept of coupling blood flow to brain function: revision required?
Ann Neurol
 , 
1987
, vol. 
22
 
3
(pg. 
289
-
297
)
Malonek
D
Dirnagl
U
Lindauer
U
Yamada
K
Kanno
I
Grinvald
A
Vascular imprints of neuronal activity: relationships between the dynamics of cortical blood flow, oxygenation, and volume changes following sensory stimulation
Proc Natl Acad Sci USA
 , 
1997
, vol. 
94
 
26
(pg. 
14826
-
14831
)
Mandeville
JB
Marota
JJ
Ayata
C
Moskowitz
MA
Weisskoff
RM
Rosen
BR
MRI measurement of the temporal evolution of relative CMRO(2) during rat forepaw stimulation
Magn Reson Med
 , 
1999
, vol. 
42
 
5
(pg. 
944
-
951
)
Mathiesen
C
Caesar
K
Akgoren
N
Lauritzen
M
Modification of activity-dependent increases of cerebral blood flow by excitatory synaptic activity and spikes in rat cerebellar cortex
J Physiol
 , 
1998
, vol. 
512
 
Pt 2
(pg. 
555
-
566
)
Moosmann
M
Ritter
P
Krastel
I
Brink
A
Thees
S
Blankenburg
F
Taskin
B
Obrig
H
Villringer
A
Correlates of alpha rhythm in functional magnetic resonance imaging and near infrared spectroscopy
Neuroimage
 , 
2003
, vol. 
20
 
1
(pg. 
145
-
158
)
Neuper
C
Pfurtscheller
G
Evidence for distinct beta resonance frequencies in human EEG related to specific sensorimotor cortical areas
Clin Neurophysiol
 , 
2001
, vol. 
112
 
11
(pg. 
2084
-
2097
)
Newton
JM
Sunderland
A
Gowland
PA
fMRI signal decreases in ipsilateral primary motor cortex during unilateral hand movements are related to duration and side of movement
Neuroimage
 , 
2005
, vol. 
24
 
2
(pg. 
1080
-
1087
)
Ngai
AC
Jolley
MA
D'Ambrosio
R
Meno
JR
Winn
HR
Frequency-dependent changes in cerebral blood flow and evoked potentials during somatosensory stimulation in the rat
Brain Res
 , 
1999
, vol. 
837
 
1–2
(pg. 
221
-
228
)
Nunez
PL
Electric fields of the brain
1981
New York
Oxford University Press
Ogawa
S
Lee
TM
Kay
AR
Tank
DW
Brain magnetic resonance imaging with contrast dependent on blood oxygenation
Proc Natl Acad Sci USA
 , 
1990
, vol. 
87
 
24
(pg. 
9868
-
9872
)
Ogawa
S
Lee
TM
Stepnoski
R
Chen
W
Zhu
XH
Ugurbil
K
An approach to probe some neural systems interaction by functional MRI at neural time scale down to milliseconds
Proc Natl Acad Sci USA
 , 
2000
, vol. 
97
 
20
(pg. 
11026
-
11031
)
Paulson
OB
Newman
EA
Does the release of potassium from astrocyte endfeet regulate cerebral blood flow?
Science
 , 
1987
, vol. 
237
 
4817
(pg. 
896
-
898
)
Pellerin
L
Pellegri
G
Bittar
PG
Charnay
Y
Bouras
C
Martin
JL
Stella
N
Magistretti
PJ
Evidence supporting the existence of an activity-dependent astrocyte-neuron lactate shuttle
Dev Neurosci
 , 
1998
, vol. 
20
 
4–5
(pg. 
291
-
299
)
Peyron
R
Garcia-Larrea
L
Gregoire
MC
Costes
N
Convers
P
Lavenne
F
Mauguiere
F
Michel
D
Laurent
B
Haemodynamic brain responses to acute pain in humans: sensory and attentional networks
Brain
 , 
1999
, vol. 
122
 
Pt 9
(pg. 
1765
-
1780
)
Pfurtscheller
G
Functional brain imaging based on ERD/ERS
Vision Res
 , 
2001
, vol. 
41
 
10–11
(pg. 
1257
-
1260
)
Pfurtscheller
G
Krausz
G
Neuper
C
Mechanical stimulation of the fingertip can induce bursts of beta oscillations in sensorimotor areas
J Clin Neurophysiol
 , 
2001
, vol. 
18
 
6
(pg. 
559
-
564
)
Pfurtscheller
G
Lopes da Silva
FH
Event-related EEG/MEG synchronization and desynchronization: basic principles
Clin Neurophysiol
 , 
1999
, vol. 
110
 
11
(pg. 
1842
-
1857
)
Rees
G
Friston
K
Koch
C
A direct quantitative relationship between the functional properties of human and macaque V5
Nat Neurosci
 , 
2000
, vol. 
3
 
7
(pg. 
716
-
723
)
Rees
G
Howseman
A
Josephs
O
Frith
CD
Friston
KJ
Frackowiak
RS
Turner
R
Characterizing the relationship between BOLD contrast and regional cerebral blood flow measurements by varying the stimulus presentation rate
Neuroimage
 , 
1997
, vol. 
6
 
4
(pg. 
270
-
278
)
Rees
G
Kreiman
G
Koch
C
Neural correlates of consciousness in humans
Nat Rev Neurosci
 , 
2002
, vol. 
3
 
4
(pg. 
261
-
270
)
Robson
DS
A simple method for constructing orthogonal polynomials when the independent variable is unequally spaced
Biometrics
 , 
1958
, vol. 
15
 (pg. 
187
-
191
)
Rother
J
Knab
R
Hamzei
F
Fiehler
J
Reichenbach
JR
Buchel
C
Weiller
C
Negative dip in BOLD fMRI is caused by blood flow–oxygen consumption uncoupling in humans
Neuroimage
 , 
2002
, vol. 
15
 
1
(pg. 
98
-
102
)
Sheth
SA
Nemoto
M
Guiou
M
Walker
M
Pouratian
N
Toga
AW
Linear and nonlinear relationships between neuronal activity, oxygen metabolism, and hemodynamic responses
Neuron
 , 
2004
, vol. 
42
 
2
(pg. 
347
-
355
)
Singh
KD
Barnes
GR
Hillebrand
A
Forde
EM
Williams
AL
Task-related changes in cortical synchronization are spatially coincident with the hemodynamic response
Neuroimage
 , 
2002
, vol. 
16
 
1
(pg. 
103
-
114
)
Smith
AT
Williams
AL
Singh
KD
Negative BOLD in the visual cortex: evidence against blood stealing
Hum Brain Mapp
 , 
2004
, vol. 
21
 
4
(pg. 
213
-
220
)
Stefanovic
B
Warnking
JM
Pike
GB
Hemodynamic and metabolic responses to neuronal inhibition
Neuroimage
 , 
2004
, vol. 
22
 (pg. 
2
(pg. 
771
-
778
)
Talairach
T
Tournoux
P
Co-planar stereotaxic atlas of the human brain
 , 
1988
Stuttgart, Germany
Thieme
Villringer
A
Dirnagl
U
Coupling of brain activity and cerebral blood flow: basis of functional neuroimaging
Cerebrovasc Brain Metab Rev
 , 
1995
, vol. 
7
 
3
(pg. 
240
-
276
)