7
Inuences of negative BOLD responses on positive BOLD responses Carsten M. Klingner a,b, , Kerstin Ebenau a , Caroline Hasler a , Stefan Brodoehl a,b , Yvonne Görlich c , Otto W. Witte a,b a Department of Neurology, University of Jena, Jena, Germany b Brain Imaging Center, University of Jena, Jena, Germany c Department of Occupational Medicine, Georg-August-University of Göttingen, Goettingen, Germany abstract article info Article history: Received 12 October 2010 Revised 7 January 2011 Accepted 11 January 2011 Available online 19 January 2011 Keywords: fMRI BOLD Somatosensory cortex Inhibition Understanding possible interactions between blood oxygenation level-dependent (BOLD) responses is critical for model-based analyses and the interpretation of experiments that deal with stimuli presented close together in time. Such interactions are well documented in the case of successive positive BOLD responses. However, the inuence that a stimulus-induced, negative BOLD response exerts on a subsequent positive BOLD response has yet to be investigated and is the focus of the current study. We performed functional magnetic resonance imaging on 10 healthy subjects during bilateral electrical median nerve stimulation using ve different time intervals between left- and right-sided stimuli. We found an acute interruption of the ongoing negative BOLD response at the onset of the positive BOLD response. Different parameters characterizing the positive BOLD response were estimated. There was no impact of the preceding negative BOLD response on the parameters describing the subsequent positive BOLD response. These ndings indicate that the underlying mechanisms for negative and positive BOLD responses do not engage parallel processes. We hypothesize that the negative BOLD response is caused by a decreased release of the same vasodilatative agents that evoke the positive BOLD response. Additionally, our results demonstrate that there is no need to adjust the model of a positive BOLD response due to a preceding negative BOLD response in the same brain area. © 2011 Elsevier Inc. All rights reserved. Introduction For nearly two decades, functional magnetic resonance imaging (fMRI) has been a powerful tool for studying brain function. Most fMRI techniques are based on measuring blood oxygenation level- dependent (BOLD) contrast using paramagnetic deoxyhemoglobin as an endogenous contrast agent. Following increased neural activity in the brain, local cerebral blood ow (CBF) increases to a greater extent than does consumption of oxygen, resulting in a decrease in deoxyhemoglobin content. The decrease in deoxyhemoglobin content translates into an increased MR signal (positive BOLD response, PBR) due to the paramagnetic nature of deoxyhemoglobin. More recently, negative deections of the BOLD response (negative bold response, NBR) have also been described. These were found, for instance, in the primary motor cortex in response to an ipsilateral motor task (Hamzei et al., 2002; Stefanovic et al., 2004) and in the primary somatosensory cortex (SI) in response to an ipsilateral somatosensory stimulus (Hlushchuk and Hari, 2006; Kastrup et al., 2008; Klingner et al., 2010, 2011). Animal experiments have revealed a tight coupling between NBRs and reduced neuronal activity (Boorman et al., 2010; Shmuel et al., 2006) or enhanced inhibition within sensory systems (Devor et al., 2007). Furthermore, a correlation between the NBR in the ipsilateral SI and changes of the sensory threshold has been shown in humans (Kastrup et al., 2008; Klingner et al., 2010). These studies demonstrate that it is benecial to consider NBRs in addition to PBRs for a more complete understanding of brain functions. Although different methods are available to analyze these signal changes, many neuroscientists use model-based methods, which require prior knowledge about the shape of the BOLD response. These methods analyze the time course of each image voxel to detect a typical BOLD response in correlation with the given stimulus, and they have been shown to be robust and reliable for the analysis of PBRs. However, the quality of the results depends primarily upon the assumptions of the model used, and there are many factors that can inuence the model. For example, nonlinear interactions between events and their evoked responses have been described. These interactions can signicantly change the shape of BOLD responses when they are very proximate in time. Therefore, much effort has been invested in improving existing models of the hemodynamic response, particularly with respect to the possible inuences of consecutive PBRs (for a review, see Buxton et al., 2004). NeuroImage 55 (2011) 17091715 Corresponding author at: Hans Berger Clinic for Neurology, University Hospital Jena, Friedrich Schiller University, Erlanger Allee 101, D 07747 Jena, Germany. Fax: +49 3641 9323402. E-mail address: [email protected] (C.M. Klingner). 1053-8119/$ see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2011.01.028 Contents lists available at ScienceDirect NeuroImage journal homepage: www.elsevier.com/locate/ynimg

Influences of negative BOLD responses on positive BOLD responses

Embed Size (px)

Citation preview

NeuroImage 55 (2011) 1709–1715

Contents lists available at ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r.com/ locate /yn img

Influences of negative BOLD responses on positive BOLD responses

Carsten M. Klingner a,b,⁎, Kerstin Ebenau a, Caroline Hasler a, Stefan Brodoehl a,b,Yvonne Görlich c, Otto W. Witte a,b

a Department of Neurology, University of Jena, Jena, Germanyb Brain Imaging Center, University of Jena, Jena, Germanyc Department of Occupational Medicine, Georg-August-University of Göttingen, Goettingen, Germany

⁎ Corresponding author at: Hans Berger Clinic for NJena, Friedrich Schiller University, Erlanger Allee 101, D 03641 9323402.

E-mail address: [email protected] (C

1053-8119/$ – see front matter © 2011 Elsevier Inc. Aldoi:10.1016/j.neuroimage.2011.01.028

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 October 2010Revised 7 January 2011Accepted 11 January 2011Available online 19 January 2011

Keywords:fMRIBOLDSomatosensory cortexInhibition

Understanding possible interactions between blood oxygenation level-dependent (BOLD) responses is criticalfor model-based analyses and the interpretation of experiments that deal with stimuli presented closetogether in time. Such interactions are well documented in the case of successive positive BOLD responses.However, the influence that a stimulus-induced, negative BOLD response exerts on a subsequent positiveBOLD response has yet to be investigated and is the focus of the current study. We performed functionalmagnetic resonance imaging on 10 healthy subjects during bilateral electrical median nerve stimulation usingfive different time intervals between left- and right-sided stimuli. We found an acute interruption of theongoing negative BOLD response at the onset of the positive BOLD response. Different parameterscharacterizing the positive BOLD response were estimated. There was no impact of the preceding negativeBOLD response on the parameters describing the subsequent positive BOLD response. These findings indicatethat the underlying mechanisms for negative and positive BOLD responses do not engage parallel processes.We hypothesize that the negative BOLD response is caused by a decreased release of the same vasodilatativeagents that evoke the positive BOLD response. Additionally, our results demonstrate that there is no need toadjust the model of a positive BOLD response due to a preceding negative BOLD response in the same brainarea.

eurology, University Hospital7747 Jena, Germany. Fax: +49

.M. Klingner).

l rights reserved.

© 2011 Elsevier Inc. All rights reserved.

Introduction

For nearly two decades, functional magnetic resonance imaging(fMRI) has been a powerful tool for studying brain function. MostfMRI techniques are based on measuring blood oxygenation level-dependent (BOLD) contrast using paramagnetic deoxyhemoglobin asan endogenous contrast agent. Following increased neural activity inthe brain, local cerebral blood flow (CBF) increases to a greater extentthan does consumption of oxygen, resulting in a decrease indeoxyhemoglobin content. The decrease in deoxyhemoglobin contenttranslates into an increased MR signal (positive BOLD response, PBR)due to the paramagnetic nature of deoxyhemoglobin. More recently,negative deflections of the BOLD response (negative bold response,NBR) have also been described. These were found, for instance, in theprimarymotor cortex in response to an ipsilateral motor task (Hamzeiet al., 2002; Stefanovic et al., 2004) and in the primary somatosensorycortex (SI) in response to an ipsilateral somatosensory stimulus

(Hlushchuk and Hari, 2006; Kastrup et al., 2008; Klingner et al., 2010,2011). Animal experiments have revealed a tight coupling betweenNBRs and reduced neuronal activity (Boorman et al., 2010; Shmuelet al., 2006) or enhanced inhibition within sensory systems (Devoret al., 2007). Furthermore, a correlation between the NBR in theipsilateral SI and changes of the sensory threshold has been shown inhumans (Kastrup et al., 2008; Klingner et al., 2010). These studiesdemonstrate that it is beneficial to consider NBRs in addition to PBRsfor a more complete understanding of brain functions.

Although different methods are available to analyze these signalchanges, many neuroscientists use model-based methods, whichrequire prior knowledge about the shape of the BOLD response. Thesemethods analyze the time course of each image voxel to detect atypical BOLD response in correlationwith the given stimulus, and theyhave been shown to be robust and reliable for the analysis of PBRs.However, the quality of the results depends primarily upon theassumptions of the model used, and there are many factors that caninfluence the model. For example, nonlinear interactions betweenevents and their evoked responses have been described. Theseinteractions can significantly change the shape of BOLD responseswhen they are very proximate in time. Therefore, much effort hasbeen invested in improving existing models of the hemodynamicresponse, particularly with respect to the possible influences ofconsecutive PBRs (for a review, see Buxton et al., 2004).

1710 C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

The potential existence of similar effects between NBRs and PBRshas yet to be examined. Such interactions might alter the results ofmodel-based fMRI studies. These interactions might also providesome information about the nature of these processes. Therefore, thecurrent study investigates the influence of an ongoing NBR on theshape of a subsequent PBR. A NBR was elicited by stimulation of theipsilateral median nerve, whereas a PBR was induced by stimulationof the contralateral median nerve. Based on known mechanismsregarding the interactions between PBRs, we have formulated thefollowing three hypotheses that are summarized in Fig. 1:

(i) If the reduced neuronal activity underlying the NBR (Boormanet al., 2010; Shmuel et al., 2006) is caused by inhibition (Devoret al., 2007), then this inhibition might partially shunt orimpede the following PBR. In this case, a reduced amplitude ofthe PBR during the NBR is expected.

(ii) If the mechanisms underlying the NBR and the PBR areindependent (i.e., different mechanisms causing vascularconstriction and dilatation), both effects would sum.

(iii) If the PBR and the NBR are not independent (i.e., they use acommon pathway of neurovascular coupling), the start of thePBR might take over this pathway. In this case, domination ofthe PBR over the preceding NBR is expected.

Regardless of which hypothesis is correct, the resolution of thisissue is important for appropriate interpretation of fMRI experimentsthat deal with different stimuli involving interactions of negative andpositive BOLD responses.

The mathematical equations for the BOLD responses for the threepreviously described hypotheses are as follows:

Hypothesis 1: BR ¼ NBR þ PBR � 1− v1:25 �w

� �

w=absolute value of the maximum amplitude of the NBRv=absolute value of the amplitude of the NBR at the time whenthe PBR starts

Hypothesis 2: BR=NBR+PBRHypothesis 3: BR=NBR|t0

t1+PBR|t1∞

t0=start time of the NBRt1=start time of the PBR

Fig. 1. A schematic illustration of the hypotheses regarding the influence of an ongoing NBR oBOLD response without interactions. Five different colors were used to indicate the time inblack: 5 s).

Materials and methods

Subjects

The study population comprised 10 healthy volunteers (mean age23.6±2.3 years, range 21–29 years, six females, four males) withoutany history of neurological or psychiatric diseases. All subjects wereright-handed according to the Edinburgh Handedness Inventory(Oldfield, 1971). The study was approved by the local ethicscommittee, and all subjects provided their written informed consentaccording to the Declaration of Helsinki.

Stimulation procedures

An electrical median nerve stimulus was used to investigateactivation and deactivation patterns within the SI. The stimulus wasapplied unilaterally and bilaterally at the wrist and consisted of 40 Hzmonophasic square wave pulses (200 μs in duration) generated by aclinical neurostimulator (Digitimer Constant Current Stimulatormodel DS7A). Electrode paste was applied to reduce the electricalresistance between the electrode and the skin. Stimulus intensitieswere determined individually for each subject. First, the motorthreshold for thumb movement was determined. At this currentintensity, no subject reported an unpleasant or painful perception.Wereduced the current by 20% to exclude any motor response. Next, thesubject was asked whether the stimulus sensation was of similarstrength on both hands. If a subject reported an unequal sensation, thestronger stimulus was reduced until the subject reported equality ofsensations between the right and left hands.

MRI measurements were carried out in two experiments. InExperiment 1, stimuli were presented unilaterally to both hands in apseudorandom, event-related regime. The stimulus duration wasfixed at 2.5 s with a jittered stimulus onset asynchrony (interstimulusinterval=12±2 s). A total of 80 stimuli (40 for each hand) werepresented during the acquisition of 491 scans. In Experiment 2, thestimuli were also presented in a pseudorandom, event-related regimewith a fixed stimulus duration of 2.5 s and with a jittered stimulusonset asynchrony (interstimulus interval=10.5±2 s). Here, the lefthand was stimulated first followed by a single stimulation of the righthand with a time delay of 0, 1.25, 2.5, 3.75 or 5 s. This stimulation

n a subsequent PBR. The dashed lines indicate the time course of a positive and negativetervals between the NBR and the PBR (red: 0 s; green: 1.25 s; blue: 2.5 s; gray: 3.75 s;

1711C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

paradigm is depicted in Fig. 2. Each combination of left+right-sidedstimulation (C1–5 in Fig. 2) was repeated 40 times. A total of 200combined stimulations were presented during the acquisition of 1310scans. The timing of stimulus presentations was synchronized withMR image acquisition by the MRI trigger signal.

Functional MRI recordings

A 3.0-Tesla MR scanner (Trio, Siemens, Erlangen, Germany) wasused to obtain 491 echo-planar T2* weighted image volumes (EPI)during Experiment 1 and 1310 EPI volumes during Experiment 2. Thefirst three EPI volumes were discarded due to equilibration effects.Each functional image volume comprised 20 transaxial slices,including the bilateral SI (voxel size=3 mm×3 mm×3 mm, repeti-tion time=2 s, TE=35 ms). After acquisition of functional images,high-resolution, T1-weighted, structural images with a voxel size of1 mm×1 mm×1 mm were recorded to allow precise anatomicallocalization.

Data analysis

Data analysis was performed on a workstation using MATLAB(Mathworks, Natiek, MA) and SPM8 software (Wellcome Departmentof Cognitive Neurology, London, UK, http://www.fil.ion.ucl.ac.uk/spm). For each subject, all images were realigned to the first volumeusing a six-parameter, rigid-body transformation to correct formotion artifacts (Friston et al., 1995). The images were co-registeredwith the subject's corresponding anatomical (T1-weighted) images,re-sliced to correct for acquisition delays (with respect to the tenthslice), normalized to the Montreal Neurological Institute (MNI)standard brain (Evans et al., 1993) to report MNI coordinates andsmoothed using a 6-mm full-width-at-half-maximum Gaussiankernel.

Multiple regression analysis was performed using a general linearmodel to obtain statistical parametric maps for positive and negativeT-contrasts. These maps were calculated for left and right somato-sensory stimulations in Experiment 1. Functional MRI signal timecourses were high-pass filtered at 1/128 Hz to remove low frequencyconfounds. Serial correlations were handled with an autoregressiveAR(1) model. Each experimental condition was modeled using aboxcar input vector convolved with the canonical hemodynamicresponse function. Individual results were projected onto theindividual, co-registered, high-resolution, T1-weighted 3-D data setand were thresholded by the false discovery rate (FDR; Genoveseet al., 2002). The anatomical localization of activations was analyzedwith reference to the standard stereotaxic atlas, and assignment to

Fig. 2. Experimental design. Study setup for somatosensory stimulation

specific brain areas was performed by visual inspection of individualT1-weighted structural data. Individual maps were used to perform arandom effect analysis so that consistent group activation patternscould be obtained. The resulting group statistical maps werethresholded by the FDR. Due to our anatomical a priori hypothesis(deactivations were assumed to occur in the ipsilateral SI), deactiva-tion T-maps were set at a threshold of Pb0.001 (uncorrected).

Hemodynamic response function (HRF) analysis

Wewere interested in the properties of BOLD responses in right SI.We extracted the peristimulus time course of 26 voxels surroundingthe point of maximum activation (single-subject SPM analysis ofExperiment 1). In both experiments, the HRF was estimated from theselected cluster in right SI. In Experiment 1, this yielded twoHRFs: onein response to contralateral stimulation and one in response toipsilateral stimulation. In Experiment 2, we calculated five differentHRFs, one for each different time shift of the left-sided stimulus ascompared to the right-sided stimulus.

For each stimulus condition, averaged BOLD signals within eachregion of interest (ROI) were estimated separately. We performed aleast-squares fit of the experimental signal time courses with a doublegamma-variate function as previously described (Liao and Yen, 2008;Tang et al., 2009).

f tð Þ = td

� �a

e− t−dð Þ=b−ctd0

� �a0

e− t−d0ð Þ=b0 + E:

The following initial values were selected: a=7, b=0.9, d=6.3,a'=14, b'=0.9, d'=12.6, c=0.35, and E=0.05 (Liao and Yen, 2008;Tang et al., 2009). These fitted time courses were used to calculatetime to peak (TTP), amplitude and full width at half maximum(FWHM) of the BOLD response for each subject. A one-way analysis ofvariance for the correlated samples was used to identify differences inthe values induced by the experimental conditions (C1–5). Forcomparisons with significant differences (Pb0.05), a paired t-testwas performed for each combination of conditions. Because one of oura priori hypotheses predicted the absence of any experimentalinfluence on the measured values, we tested specifically forequivalence. To ensure consistency with previously used tests fordifferences, we tested for a β-error less than 5%, which required asignificance threshold of PN0.2 (Bortz and Schuster, 1993).

In addition to the previously described analysis of data fromExperiment 2, we examined whether these experiments coulddistinguish between the three previously stated hypotheses. For thispurpose, we simulated the HRFs for all five experimental conditions

during Experiment 1 (upper row) and Experiment 2 (lower row).

1712 C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

and for all three hypotheses using the HRFs (NBR and PBR) from theright SI obtained in Experiment 1. The resulting BOLD responses werefitted and their characteristics (TTP, maximumamplitude and FWHM)were subjected to the same statistical tests as were the measureddata.

Results

Experiment 1 – localization of positive and negative BOLD responses

Electrical stimulation of both the right and left median nervesevoked a highly significant activation (Pb0.05, FDR corrected) in therandom effect group analysis. These activations were located incontralateral primary somatosensory cortex (BA 3b, 1, 2), in theposterior wall of the precentral gyrus (BA 4) and in the secondarysomatosensory area (Table 1 and Fig. 3). Significant negative BOLDresponses (Pb0.001, uncorrected) were found in the ipsilateral SI(Table 1 and Fig. 3). Activation and deactivation clusters coverednearly homologous areas in both hemispheres (Fig. 3). In single-subject analysis, significant deactivation of the precuneus was seen in7 of 10 subjects. Time courses of the BOLD responses from both SIs areshown in Fig. 3. Visual inspection of the PBR showed a greateramplitude as compared to that of the NBR. Moreover, the PBRdemonstrated a shorter time delay between the stimulus onset andthe increase of the BOLD response (approximately 2 s) as compared tothe delay observed in the NBR (approximately 3 s).

Experiment 2 – time-dependent interactions of the positive and negativeBOLD responses

To investigate the influence of an ongoing NBR on a PBR, weextracted the peristimulus time course from the right SI (Fig. 4A). InFig. 4B, the same time courses were corrected for different y-values atthe onset of the PBR and were also adjusted for different onset times(shift of the time courses along the x-axis, Fig. 4C). A least-squares fitof the extracted signal was performed with a double gamma-variatefunction for each experimental condition (C1–5; Fig. 4D). For bettervisualization, these fitted signals were plotted on top of each other(Fig. 4E). We estimated TTP, maximum amplitude and FWHM for thefitted response (Table 2). One-way analysis of variance for correlatedsamples revealed no significant differences between any of theexperimental conditions for all three parameters (maximum ampli-tude P=0.86; TTP P=0.73; FWHM P=0.46). In contrast, significantdifferences were found for all three characteristics of the BOLDresponse in the test for equivalence for every combination ofexperimental conditions (C1–5).

Simulation – HRF estimation assuming the three proposed hypothesesNo significant differences but equivalence were found for all

characteristics of the BOLD response. However, this does not provethat the alternative hypotheses proposed can be rejected. For this

Table 1Cortical activation in response to median nerve stimulation.

Brain region Left-sided median nervestimulation

Right-sided median nervestimulation

x y z t-value x y z t-value

SI c 42 −18 66 8.1 −51 −24 60 8.8SI i −48 −24 54 6.1 42 −21 66 6.0SII c 45 −24 21 6.3 −51 −21 12 7.26SII i – – – – 54 −24 21 4.53

MNI coordinates of the activation maxima with corresponding t-values for left- andright-sided median nerve stimulations. Deactivation maximum are highlighted in gray.SI: primary somatosensory cortex, SII: secondary somatosensory cortex, c: contralateral,i: ipsilateral.

purpose, we simulated the HRFs for all five experimental conditionsand for all three hypotheses using the HRFs (NBR and PBR) of the rightSI from Experiment 1. The characteristics of these resulting HRFs werefitted and subjected to the same statistical tests as the measured data.

Hypothesis 1: One-way analysis of variance for correlated samplesrevealed significant differences for all three parameters (maximumamplitude Pb0.0001; TTP P=0.026; FWHM Pb0.0001). Additionally,a paired t-test was performed for each combination of conditions.With respect to the number of resulting tests (10 tests for eachparameter), we corrected for multiple comparisons using theBonferroni correction. For maximum amplitude, 5 out of 10 testsresulted in a significant difference. For FWHM, 4 out of 10 testsresulted in a significant difference. Analysis of TTP did not show anysignificant differences after the Bonferroni correction was applied.

Hypothesis 2: One-way analysis of variance for correlated samplesrevealed significant differences for all three parameters (maximumamplitude Pb0.0001; TTP P=0.00048; FWHM Pb0.0012). Bonferroni-corrected, paired t-tests revealed significant results for maximumamplitude in 9 of 10 tests, for TTP in 2 of 10 tests and for FWHM in 2 of10 tests.

Hypothesis 3: The HRF did not change across different conditions(P=1.0 for all parameters).

Discussion

In the present study, we used fMRI and electrical median nervestimulation to investigate the influence of an ongoingNBR on a PBR. Themain findings of our studywere that the NBRwas acutely interrupted atthe beginning of the PBR and that theNBR did not influence the shape ofthe PBR, irrespective of the time delay between the stimuli.

Electrical median nerve stimulation as used in this experimentproduces robust activation in contralateral SI, which is consistent withprevious studies (Deuchert et al., 2002;Kampeet al., 2000;Nihashi et al.,2005; Rubenet al., 2006). This activation is characterizedbyaPBR. TheSIexhibitedanNBR ipsilateral to the somatosensory stimulus. This result isalso consistent with recent reports on somatosensory processing(Hlushchuk and Hari, 2006; Kastrup et al., 2008; Klingner et al., 2010).The physiological basis of the NBR has not been fully characterized.Several theories have been proposed to explain NBRs, including purelyhemodynamic effects of “blood stealing,” neuronal inhibition, neuronalexcitation or a combination of these processes. It has been suggestedthat the NBR might be associated with an increased level of neuronalactivity that does not lead to an increase in CBF, despite an increase inthe cerebral metabolic rate of oxygen (CMRO2). Recent experiments donot support this hypothesis in studies of healthy subjects (Lin et al., inpress; Stefanovic et al., 2004) or in animal studies (Shmuel et al., 2002).Another possible explanation of a negative BOLD signal is a hemody-namic “stealing” effect (reallocation of blood flow to neighboringactivated regions). Since we did not observe significant activationsaround the deactivated brain area, it seemsunlikely that a stealing effectcould account for the NBRs observed in the current study. Moreover, itwas shown in the visual cortex that a NBR in close spatial relation to aPBR cannot be explained by a stealing effect alone (Smith et al., 2004). Athird explanation assumes that the NBR reflects decreases in neuronalactivity. This assumption is supported by recent animal experimentscombining fMRI with invasive recording techniques (Boorman et al.,2010; Harel et al., 2002; Logothetis, 2002; Shmuel et al., 2006). Thismechanism has also been proposed to underlie NBRs during focal andgeneralized epileptic discharges (Gotman, 2008).

There are many conflicting opinions regarding which aspects ofinformation processing are the most energy consuming (Buzsaki et al.,2007; Logothetis et al., 2001; Shulman et al., 2004). It is neverthelessassumed that increases or decreases in neuronal activity also increase ordecrease energy demand (Ames, 2000; Paulson et al., 2010). Reducedenergy consumption reduces the extraction fraction of oxyhemoglobin,causing an initial undershoot of paramagnetic deoxyhemoglobin (Devor

Fig. 3. Random effect group (n=10) analysis. Activations (Pb0.05, FDR corrected) and deactivations (Pb0.001, uncorrected) in response to event-related (2.5 s) left median nervestimulation (A) and right median nerve stimulation (B) are shown superimposed on a slightly inflated individual brain. Yellow–red encodes positive BOLD signals whereas blueencodes negative BOLD signals.

1713C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

et al., 2007) with a corresponding small overshoot of the NBR. This isfollowed by a reduction in local CBF and in local cerebral blood volume(CBV; Boorman et al., 2010; Devor et al., 2007). This decreased CBVovercompensates for the reduction in energy consumption and

Fig. 4.Dependence of the HRF of the right SI on the time delay between right-hand and left-haData were also averaged across all subjects. The original BOLD time courses for all five experand x-axis (C) to correct for signal differences at the stimulus onset due to the ongoing NBR.were plotted together to ensure visual comparability.

increases the deoxyhemoglobin concentration, resulting in the obser-vance of a NBR (Boorman et al., 2010; Devor et al., 2007; Harel et al.,2002). This mechanism has been demonstrated to occur in areassurrounding a PBR (Devor et al., 2007; Harel et al., 2002) as well as in

nd stimulation. All data were averaged across 26 voxels surrounding the peak response.imental conditions are shown (A). These time courses were shifted along the y-axis (B)Panel D shows the fitted functions of the BOLD responses. In Panel E, all fitted functions

Table 2Fitting results for BOLD responses in right SI.

Stimulus condition Time to peak, in s Amplitude, in %signal change

FWHM, in s

C1 5.45±0.28 1.73±0.67 4.11±0.57C2 5.38±0.27 1.67±0.66 3.97±0.51C3 5.38±0.23 1.74±0.84 4.17±0.56C4 5.42±0.34 1.65±0.70 4.16±0.62C5 5.30±0.24 1.71±0.63 4.07±0.66MEAN 5.39 1.70 4.10

Fitting results of BOLD responses and the corresponding characteristics in the right SI.BOLD responses were fitted with a double gamma-variate model for each subject andeach experimental condition separately. These curves were used to derive threeparameters, including TTP, BOLD amplitude and FWHM. The mean and standarddeviation for the individual fitting results are shown.

1714 C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

ipsilateral barrel cortex of rats following electrical whisker stimulation(Boorman et al., 2010).

Intuitively, one might assume that the inhibition underlying theNBR partially shunts the excitation that causes the PBR. There areseveral explanations why this result was not observed in the presentexperiments. On the one hand, the electrical signals are much shorterthan the resulting hemodynamic responses, and the major inhibitionmay have ended long before the superimposed excitative responseceased. On the other hand, we do not know whether the dominatingdeactivation is due to inhibition or to disfacilitation; the latter wouldnot occur alongside a shunting of excitation.

From these contradictions, one could derive the second proposedhypothesis; that is, the PBR and NBR are independent and shouldsuperimpose on one another. This hypothesis is further strengthenedby the observation of different time courses in these two processes.Thus, the NBR starts with a typical delay of 1–2 s. However, the dataobserved in the present study do not support the model of twosuperimposed, independent processes. Instead, the NBR seems to beinterrupted with the onset of the PBR and does not influence theamplitude or time course of the subsequent PBR. This finding indicatesthat the processes that mediate the positive and negative BOLDresponses are not entirely parallel. We hypothesize that the PBR andNBR engage the same pathway of neurovascular coupling or that thepathways at least interact to some extent.

Recent findings suggest a key role for astrocytes in the vascularreaction. Astrocytes sense perisynaptic activity and release Ca2+

intracellularly in astrocytic endfeet. Ca2+ in the endfeet thenmediatesthe release of vasodilatative agents, causing a slow dilatation of localcerebral arterioles (Iadecola and Nedergaard, 2007; Koehler et al.,2009; Carmignoto and Gomez-Gonzalo, 2010). Short-term vasodila-tative agents (e.g., nitric oxide) released by postsynaptic neurons mayalso play a role in this process (Attwell et al., 2010; Koehler et al.,2009; Toda et al., 2009). We speculate that decreased synaptic activitywill decrease the amount of Ca2+ in astrocytic endfeet, resulting in apassive vasoconstriction and a NBR. A subsequent episode of neuralexcitation would then cause a largely undisturbed release of Ca2+.Unfortunately, the mechanisms underlying the NBR are still incom-pletely understood, and it is well conceivable that our observationsare due to other mechanisms (e.g., an impact at the level of restingCBF; Attwell et al., 2010; Shen et al., 2008).

The present investigation also leads us to conclude that studiesusing complex stimulus designs do not need to readjust the PBRmodel with respect to preceding stimuli that might possibly cause aNBR in the target area.

Conclusion

The current study investigated the influence of a NBR on asuccessive PBR. The NBR was found to end acutely at the beginning ofthe PBR without further convolution of the PBR. This finding suggests

that the negative and positive BOLD responses are not mediated byentirely parallel processes, and we hypothesize that they rely on thesame mechanisms of neurovascular coupling.

Acknowledgments

The authors thank the reviewers for their helpful comments andinsights.

References

Ames, A., 2000. CNS energymetabolism as related to function. Brain Res. Rev. 34, 42–68.Attwell, D., Buchan, A.M., Charpak, S., Lauritzen, M., Macvicar, B.A., Newman, E.A., 2010.

Glial and neuronal control of brain blood flow. Nature 468, 232–243.Boorman, L., Kennerley, A.J., Johnston, D., Jones, M., Zheng, Y., Redgrave, P., Berwick, J.,

2010. Negative blood oxygen level dependence in the rat: a model for investigatingthe role of suppression in neurovascular coupling. J. Neurosci. 30, 4285–4294.

Bortz, J., Schuster, C., 1993. Statistik für Human- und Sozialwissenschaftler.Buxton, R.B., Uludag, K., Dubowitz, D.J., Liu, T.T., 2004. Modeling the hemodynamic

response to brain activation. Neuroimage 23 (Suppl 1), S220–S233.Buzsaki, G., Kaila, K., Raichle, M., 2007. Inhibition and brain work. Neuron 56, 771–783.Carmignoto, G., Gomez-Gonzalo, M., 2010. The contribution of astrocyte signalling to

neurovascular coupling. Brain Res. Rev. 63, 138–148.Deuchert, M., Ruben, J., Schwiemann, J., Meyer, R., Thees, S., Krause, T., Blankenburg, F.,

Villringer, K., Kurth, R., Curio, G., Villringer, A., 2002. Event-related fMRI of thesomatosensory system using electrical finger stimulation. NeuroReport 13, 365–369.

Devor, A., Tian, P., Nishimura, N., Teng, I.C., Hillman, E.M., Narayanan, S.N., Ulbert, I.,Boas, D.A., Kleinfeld, D., Dale, A.M., 2007. Suppressed neuronal activity andconcurrent arteriolar vasoconstriction may explain negative blood oxygenationlevel-dependent signal. J. Neurosci. 27, 4452–4459.

Evans, A.C., Collins, D.L., Mills, S.R., Brown, E.D., Kelly, R.L., Peters, T.M., 1993. 3Dstatistical neuroanatomical models from 305 MRI volumes. IEEE Nuclear ScienceSymposium and Medical Imaging Conference, pp. 1813–1817.

Friston, K.J., Frackowiak, R.S.J., Heather, J.D., Frith, C.D., Poline, J.B., Ashburner, J., 1995.Hum. Brain Mapp. 2.

Genovese, C.R., Lazar, N.A., Nichols, T., 2002. Thresholding of statistical maps infunctional neuroimaging using the false discovery rate. Neuroimage 15, 870–878.

Gotman, J., 2008. Epileptic networks studied with EEG-fMRI. Epilepsia 49 (Suppl 3),42–51.

Hamzei, F., Dettmers, C., Rzanny, R., Liepert, J., Buchel, C., Weiller, C., 2002. Reduction ofexcitability (“inhibition”) in the ipsilateral primary motor cortex is mirrored byfMRI signal decreases. Neuroimage 17, 490–496.

Harel, N., Lee, S.P., Nagaoka, T., Kim, D.S., Kim, S.G., 2002. Origin of negative bloodoxygenation level-dependent fMRI signals. J. Cereb. Blood Flow Metab. 22,908–917.

Hlushchuk, Y., Hari, R., 2006. Transient suppression of ipsilateral primary somatosen-sory cortex during tactile finger stimulation. J. Neurosci. 26, 5819–5824.

Iadecola, C., Nedergaard, M., 2007. Glial regulation of the cerebral microvasculature.Nat. Neurosci. 10, 1369–1376.

Kampe, K.K., Jones, R.A., Auer, D.P., 2000. Frequency dependence of the functional MRIresponse after electrical median nerve stimulation. Hum. Brain Mapp. 9, 106–114.

Kastrup, A., Baudewig, J., Schnaudigel, S., Huonker, R., Becker, L., Sohns, J.M., Dechent, P.,Klingner, C., Witte, O.W., 2008. Behavioral correlates of negative BOLD signalchanges in the primary somatosensory cortex. Neuroimage 41, 1364–1371.

Klingner, C.M., Hasler, C., Brodoehl, S., Witte, O.W., 2010. Dependence of the negativeBOLD response on somatosensory stimulus intensity. Neuroimage 53, 189–195.

Klingner, C.M., Huonker, R., Flemming, S., Hasler, C., Brodoehl, S., Preul, C., Burmeister, H.,Kastrup, A., Witte, O.W., 2011. Functional deactivations: multiple ipsilateral brainareas engaged in the processing of somatosensory information. Hum. Brain Map. 32,127–140.

Koehler, R.C., Roman, R.J., Harder, D.R., 2009. Astrocytes and the regulation of cerebralblood flow. Trends Neurosci. 32, 160–169.

Liao, C.C., Yen, C.T., 2008. Functional connectivity of the secondary somatosensorycortex of the rat. Anat Rec (Hoboken) 291, 960–973.

Lin, P., Hasson, U., Jovicich, J., Robinson, S., in press. A Neuronal Basis for Task-NegativeResponses in the Human Brain. Cereb. Cortex (2010 Aug 30. [Epub ahead of print]).

Logothetis, N.K., 2002. The neural basis of the blood-oxygen-level-dependentfunctional magnetic resonance imaging signal. Philos. Trans. R. Soc. Lond. B Biol.Sci. 357, 1003–1037.

Logothetis, N.K., Pauls, J., Augath, M., Trinath, T., Oeltermann, A., 2001. Neurophysio-logical investigation of the basis of the fMRI signal. Nature 412, 150–157.

Nihashi, T., Naganawa, S., Sato, C., Kawai, H., Nakamura, T., Fukatsu, H., Ishigaki, T., Aoki,I., 2005. Contralateral and ipsilateral responses in primary somatosensory cortexfollowing electrical median nerve stimulation–an fMRI study. Clin. Neurophysiol.116, 842–848.

Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburghinventory. Neuropsychologia 9, 97–113.

Paulson, O.B., Hasselbalch, S.G., Rostrup, E., Knudsen, G.M., Pelligrino, D., 2010. Cerebralblood flow response to functional activation. J. Cereb. Blood Flow Metab. 30, 2–14.

Ruben, J., Krause, T., Taskin, B., Blankenburg, F., Moosmann, M., Villringer, A., 2006. Sub-area-specific Suppressive Interaction in the BOLD responses to simultaneous fingerstimulation in human primary somatosensory cortex: evidence for increasingrostral-to-caudal convergence. Cereb. Cortex 16, 819–826.

1715C.M. Klingner et al. / NeuroImage 55 (2011) 1709–1715

Shen, Q., Ren, H., Duong, T.Q., 2008. CBF, BOLD, CBV, and CMRO(2) fMRI signal temporaldynamics at 500-msec resolution. J. Magn. Reson. Imaging 27, 599–606.

Shmuel, A., Yacoub, E., Pfeuffer, J., Van de Moortele, P.F., Adriany, G., Hu, X., Ugurbil, K.,2002. Sustained negative BOLD, blood flow and oxygen consumption response andits coupling to the positive response in the human brain. Neuron 36, 1195–1210.

Shmuel, A., Augath, M., Oeltermann, A., Logothetis, N.K., 2006. Negative functional MRIresponse correlates with decreases in neuronal activity in monkey visual area V1.Nat. Neurosci. 9, 569–577.

Shulman, R.G., Rothman, D.L., Behar, K.L., Hyder, F., 2004. Energetic basis of brainactivity: implications for neuroimaging. Trends Neurosci. 27, 489–495.

Smith, A.T., Williams, A.L., Singh, K.D., 2004. Negative BOLD in the visual cortex:evidence against blood stealing. Hum. Brain Mapp. 21, 213–220.

Stefanovic, B., Warnking, J.M., Pike, G.B., 2004. Hemodynamic and metabolic responsesto neuronal inhibition. Neuroimage 22, 771–778.

Tang, L., Avison, M.J., Gore, J.C., 2009. Nonlinear blood oxygen level-dependentresponses for transient activations and deactivations in V1 - insights into thehemodynamic response function with the balloonmodel. Magn. Reson. Imaging 27,449–459.

Toda, N., Ayajiki, K., Okamura, T., 2009. Cerebral blood flow regulation by nitric oxide inneurological disorders. Can. J. Physiol. Pharmacol. 87, 581–594.