Study of the Spatial Correlation Between Neuronal Activityand BOLD fMRI Responses Evoked by Sensoryand Channelrhodopsin-2 Stimulation in the RatSomatosensory Cortex

Nan Li & Peter van Zijl & Nitish Thakor & Galit Pelled

Received: 12 September 2013 /Accepted: 22 December 2013# Springer Science+Business Media New York 2014

Abstract In this work, we combined optogenetic tools withhigh-resolution blood oxygenation level-dependent functionalMRI (BOLD fMRI), electrophysiology, and optical imagingof cerebral blood flow (CBF), to study the spatial correlationbetween the hemodynamic responses and neuronal activity.We first investigated the spatial and temporal characteristics ofBOLD fMRI and the underlying neuronal responses evokedby sensory stimulations at different frequencies. The resultsdemonstrated that under dexmedetomidine anesthesia, BOLDfMRI and neuronal activity in the rat primary somatosensorycortex (S1) have different frequency–dependency and distinctlaminar activation profiles. We then found that localized acti-vation of channelrhodopsin-2 (ChR2) expressed in neuronsthroughout the cortex induced neuronal responses that wereconfined to the light stimulation S1 region (<500 μm) withdistinct laminar activation profile. However, the spatial extentof the hemodynamic responses measured by CBF and BOLDfMRI induced by both ChR2 and sensory stimulation wasgreater than 3 mm. These results suggest that due to thecomplex neurovascular coupling, it is challenging to deter-mine specific characteristics of the underlying neuronal activ-ity exclusively from the BOLD fMRI signals.

Keywords Hemodynamic . fMRI . Optical imaging .

Optogenetics . Somatosensory cortex

Introduction

Functional MRI (fMRI) techniques have rapidly grown in im-portance to enable characterization of the neuronal activity thatoccurs in the healthy brain and during pathology at the systemlevel. Similar to other hemodynamics-based functional brainimaging techniques, the blood oxygenation level-dependent(BOLD) fMRI signal is an indirect measurement of the neuronalactivity. The exact relationship between the neuronal responsesand hemodynamic responses remains unclear and under debates(Attwell and Iadecola 2002; Logothetis 2008; Vanzetta andGrinvald 2008; Enager et al. 2009; van Eijsden et al. 2009;Ekstrom 2010; Yen et al. 2011). Indeed, in recent years therehave been great efforts to resolve this controversial topic. Thevast majority of themicroscopic research is focused on exploringthe molecular factors implicated in the underlying neurovascularcoupling, such as vasoactive ions, vasoactive factors related toenergymetabolism, vasoactive factors/neurotransmitters releasedby neuronal activation, and the role astrocytes play inneurovascular coupling (Lauritzen 2005; Iadecola andNedergaard 2007; Devor et al. 2008; Koehler et al. 2009;Lecrux and Hamel 2011). The majority of the macroscopicresearch is focused on investigating the temporal correlationbetween the magnitude of the neuronal responses and the hemo-dynamic responses. Studies have demonstrated that the BOLDfMRI responses are tightly correlated to increases in local fieldpotential (LFP) and spiking activity (Logothetis et al. 2001;Mukamel et al. 2005; Shmuel et al. 2006; Viswanathan andFreeman 2007; Kim et al. 2010); on the other hand, dissociation

N. Li : P. van Zijl :G. Pelled (*)F. M. Kirby Center for Functional Brain Imaging, Kennedy KriegerInstitute, Baltimore, MD, USAe-mail: Pelled@kennedykrieger.org

N. Li : P. van Zijl :G. PelledThe Russell H. Morgan Department of Radiology and RadiologicalScience, Division ofMRResearch, Johns Hopkins University Schoolof Medicine, 707 N Broadway, Baltimore, MD, USA

N. ThakorThe Department of Biomedical Engineering, Johns HopkinsUniversity School of Medicine, Baltimore, MD, USA

J Mol NeurosciDOI 10.1007/s12031-013-0221-3

between LFP and spiking activity to hemodynamic responseshas been reported (Caesar et al. 2003; Devor et al. 2007; Pelledet al. 2009; Lauritzen et al. 2012). In addition, the question of co-localization between BOLD responses and the correspondingneuronal activity remains difficult to address due to mechanism-based limitations in the MRI methodology.

The focus of this work was to investigate whether BOLDfMRI signal can be used to deduce about the underlyingneuronal activity and to what degree it co-localized withneuronal activity. Until recently, electrical stimulation viaelectrodes was used to map and modulate the stimulus re-sponse in the brain (Sultan et al. 2007). Optogenetic tools nowenable neuronal manipulations that are precise, reversible, andcell specific (Boyden et al. 2005; Zhang et al. 2010). Thesenew tools provide the ability to elucidate detailed neuronalmechanisms associated with brain function. For example,optogenetic tools are being utilized to address questions re-garding the basis of the neuronal activity that underlie theBOLD fMRI signals that were challenging to measure before(Lee et al. 2010; Desai et al. 2011; Kahn et al. 2011; Scott andMurphy 2012; Airan et al. 2013; Vazquez et al. 2014).

Conventional electrical forepaw stimulations were conductedto evoke neuronal firing rate in response to sensory afferentinputs through the thalamocortical circuits.We used optogeneticsas amean to produce localized stimulation of cortical neurons viaoptic fibers that have been demonstrated to not cause significantsusceptibility effect that can interfere with MRI acquisition (Leeet al. 2010; Desai et al. 2011; Kahn et al. 2011; Li et al. 2011;Airan et al. 2013). A combination of multi-channel and micro-electrodes for electrophysiology recordings, optical imaging ofcerebral blood flow (CBF), andBOLD fMRIwere used to assesstemporal and spatial characteristics of neuronal and hemodynam-ic responses. The results demonstrate that the BOLD fMRI,CBF, and neuronal responses in S1 evoked by sensory stimula-tion are frequency-dependent, have a distinct laminar profile, andextend 3 mm along S1. The BOLD fMRI and CBF responses inS1 evoked by activation of channelrhodopsin-2 (ChR2) resultedin responses that were similar in their spatial extent to the oneinduced by conventional limb stimulation. However, the neuro-nal responses evoked by ChR2 stimulation were confined to thelight stimulation region which was below 0.5 mm. Thus, it ismanifested that it is plausible that under certain conditions, theBOLD fMRI responses is only partially co-localized with theneuronal responses.

Materials and Methods

All animal procedures were conducted in accordance with theNIH Guide for the Care and Use of Laboratory Animals andapproved by the Johns Hopkins University Animal Care andUse Committee.

Animal Preparation

Lentivirus production and transduction were performed ac-cording to Gradinaru et al. (2009). The lentivirus (5 μl) con-taining Lenti-CaMKIIα-ChR2-EYFP-WPRE was deliveredinto the lateral ventricle to P3 rat pups as described previously(Li et al. 2011). Twenty-two rats expressing ChR2 werestudied. Electrophysiology, optical imaging, and fMRI mea-surements were performed on 10–11-week-old rats expressingChR2 (220±50 g). For electrophysiology and fMRI measure-ment, a craniotomy (∼1-mm square) was performed above theright S1 (centered at AP=−0.5 mm, ML=3.7 mm, relative toBregma). For optical imaging, the skull above right S1 (4×5 mm) was thinned to ∼100 μm until the underlying cerebralvessels were visualized. After surgery, isoflurane wasdiscontinued and a bolus of dexmedetomidine (0.05 mg/kg,S.C.) was given. The anesthesia was then maintained by acontinuous infusion (0.1 mg/kg/h) of dexmedetomidine (Zhaoet al. 2008; Pawela et al. 2010). Respiration rate, oxygensaturation, and heart rate were continuously monitoredthroughout all measurements by an oximeter (MouseOx,STARR Life Sciences Corp, PA, USA).

Forepaw Stimulation

Electrical stimulation (World Precision Instruments, Sarasota,FL, USA) of the left forepaw was induced by two shortelectrodes that were inserted between the digits. The stimula-tion current was consisted of 3 mA, 300 μs pulses. Sixdifferent frequencies (1, 3, 6, 9, and 20 Hz) were applied for20 s, respectively. After the frequency–dependency study, themaximum BOLD response frequency 9 Hz, as in agreementwith previous studies (Zhao et al. 2008), was selected for theforepaw stimulation in the rest of experiments of this work.

ChR2 Stimulation

A 3-ft-long, 400-μm-diameter optic fiber (0.39NA,FT400EMT, Thorlabs, Newton, NJ, USA) was coupled to a473-nm laser source (AL-473-60TA, Aixiz, Huston, TX,USA) through a customized collimation setup. The otherend of the optic fiber was mounted on top of the center ofthe right S1 craniotomy (AP=−0.5 mm, ML=3.7 mm) with-out touching the brain tissue. The power of the light at the endof the optical fiber was measured with a digital optical powermeter (PM100, Thorlabs, Newton, NJ, USA). Various lightintensities at the fiber output were tested in control rats notexpressing ChR2, and 63.66 mW mm−2 was selected underwhich no significant cerebral blood flow and BOLD fMRIchanges were observed in response to the light illumination.Light stimulation consisted of 20ms of light pulses repeated at20 Hz controlled by a shutter controller (SC10, Thorlabs,Newton, NJ, USA) triggered via TTL signals which was

J Mol Neurosci

generated and programmed using Spike 2 software(Cambridge Electronic Design, Cambridge, UK).

Electrophysiology Recording

Extracellular neuronal activity recording was performed todetermine the laminar S1 responses. A 12 channel axial arraymicroelectrode with contacts spaced 150 μm apart (200 μm indiameter, 1 MΩ, FHC, Bowdoin, ME, USA) was used. Theaxial electrode was placed in the center of S1 (AP=−0.5 mm,ML=3.7 mm), proximate to the optic fiber (<100 μm), andslowly lowered to the target depth using a micromanipulator(FHC).

In order to determine the spatial S1 responses, a customizedgrid of nine tungsten electrodes (125μm in diameter with a tipsize of a few micrometers, 350 KΩ, FHC, Bowdoin, ME,USA) was used. The tungsten electrodes were placed in across pattern and were spaced 500 μm apart. The whole gridcovered a 2×2-mm area. The electrodes grid was lowered tothe target depth (lamia IV; 700–800 μm below the pial sur-face) using a micromanipulator (FHC). The optic fiber waspositioned proximate (<100 μm) to the center electrode in thegrid (AP=−0.5 mm, ML=3.7 mm).

Local field potential (LFP) and multi-unit activity (MUA)were sampled at 1 and 11 k Hz and band-pass filtered between0.1–500 Hz and 500–5 k Hz, respectively. Discriminatedsignals were collected from a CED interface and Spike2software (Cambridge Electronic Design, Cambridge, UK).Three sets of data of 20 s forepaw or ChR2 stimulation werecollected. To identify spiking events, the standard deviation(SD) of the MUA signals without stimulations was calculatedfor 5 s. MUA signals with amplitude greater than four times ofthe SD were defined as spiking activity. Post-stimulus timehistogram analysis was performed to define stimulus evokedspiking activity. MUA that showed increased activity in one 5-ms bin during the first 40 ms following the onset of thestimulation were considered to show stimulus-evoked re-sponse. The number of stimulus-evoked spiking activity with-in the first 40 ms was calculated. LFP waveforms were aver-agedwith respect to the stimulation triggers after removing theDC offset. The mean amplitude of the negative deflection wascalculated for each train of stimuli. For LFP maps, data fromthe 12 contacts was resampled to 120 rows by using cubicspline interpolation.

Functional MRI

MRI images were acquired on a 9.4-T horizontalMRI scanner(Bruker, Ettlingen, Germany). Rats were placed in a custom-built MRI rat cradle equipped with a dedicated holder forpositioning the light source coupled MRI compatible opticfiber. A custom-built 1.1-cm-diameter surface coil was used totransmit and receive MR signals (Airan et al. 2013). A

gradient-echo, echo-planar imaging (GE EPI) sequence wasused: echo time (TE)=21 ms, repetition time (TR)=1,000 ms,bandwidth=250 kHz, field of view (FOV)=1.92×1.92 cm,and matrix size=128×128, resulting in an in-plane resolutionof 150×150×1,000 μm. T2-weighted rapid acquisition withrelaxation enhancement (RARE) sequence was used to ac-quire high-resolution anatomical images with the followingparameters: TE=11 ms, TR=2,000 ms, number of averages=2, bandwidth=250 KHz, FOV=1.92×1.92 cm, and matrixsize=256×256. Three coronal slices of images centered atBregma 0 were acquired. Two epochs of 20 baseline and 20scans during forepaw or ChR2 stimulation were collected.The FMRIB Software Library (FSL 4.1.9) software was usedfor analysis (Smith et al. 2004). Activation maps were obtain-ed using the general linear model (GLM). The following pre-statistics processing was applied: motion correction usingFMRIB’s Linear Image Registration Tool, grand-mean inten-sity normalization of the entire 4D dataset by a single multi-plicative factor, and high-pass temporal filtering (Gaussian-weighted least-squares straight line fitting, with sigma=30 s).The GLM method used in FMRI Expert Analysis Tool withlocal autocorrelation correction was used. Double-gammahemodynamic response function convolution kernel was ap-plied to the basic event block to generate the design matrix. Z-score statistics were cluster-size thresholded for effective sig-nificance of p<0.05. The activation threshold was set at 2.3.Group and laminar analysis were performed by custom-written software running in MATLAB R2011b (MathWorksInc., Natick, MA, USA). In order to characterize the laminarBOLD fMRI signals, the surface of S1 forepaw representation(defined by a brain atlas) was defined by the user based onhigh-resolution (75×75×1,000 μm) RARE images. Three re-gions of interest (ROI) were defined: laminae II+III (150 μm–600 μm), lamina IV (600 μm–900 μm), and laminae V+VI(900 μm–1,800 μm). The number of activated voxels wascalculated for each ROI. For each animal, laminar-specificBOLD fMRI response time courses were obtained by averag-ing all the active pixels from each ROI. We determined thespatial extent of the BOLD fMRI responses (within the medial–lateral axis) by calculating the number of active pixels alongthe ROI that represents lamia IV.

Optical Imaging

Laser speckle contrast imaging (LSCI), a minimally invasiveimaging modality, was used to collect CBF images with highspatiotemporal resolution (Dunn et al. 2001; Li et al. 2009).All raw laser speckle images were acquired using BaslerpiA640-210gm GigEvision camera (Basler VisionTechnologies, Germany). A 60-mm, f/2.8 macro-lens with amaximum reproduction ratio of 1:1 (Nikon Inc., Melville, NY,USA) was mounted on the camera using a Cmount to a NikonF mount adapter. A 632-nm laser (0.5 mW, JDSU, Milpitas,

J Mol Neurosci

CA, USA) provided coherent illumination. A red light filterD620/30 m (Chroma, Rockingham, VT, USA) was insertedbetween the camera and the lens to prevent 475 nm light fromentering the CCD sensor. A thinned-skull cranial window of4×5 mm area centered at the right S1 was made for imaging.Before LSCI imaging, white light reflectance images wereobtained to adjust the camera to focus on the cranial windowarea of interest under the illumination of a cold light sourcewith fiber optics Luminator HGY3. Once the magnifi-cation of the camera was set, the aperture was set to theoptimal condition for photographing speckles (Boas andDunn 2010). The shutter speed was set to get an opti-mal exposure as 5 ms. Images were taken continuouslyat a rate of 25 frames per second. Software was writtento control the image acquisition in LabView 2009(National Instruments, Austin, TX, USA). Five sets ofdata from 20 s of baseline activity followed by 20 s offorepaw or ChR2 stimulation were collected. Rawspeckle images were co-registered, and the contrast was cal-culated according to Li et al. (2009). The intensity of LSCIcontrast image is inversely proportional to the blood flow.Activation maps were obtained using two-tailed Z-test(p<0.05) with an activation threshold set to 3.5 and a clustersize of >10 pixels.

Immunohistochemistry

All rats were sacrificed and perfused at the end of the exper-iments to confirm ChR2 expression in pyramidal excitatoryneurons. Free floating 50-μm-thick coronal brain sectionswere immunostained with nuclear staining 4′,6-diamidino-2-phenylindole (DAPI; a nuclear marker), anti-enhanced yellowfluorescent protein (EYFP; indicative of ChR2 expression),and anti-calcium/calmodulin-dependent protein kinase IIα(CaMKIIα; marker of excitatory pyramidal neurons) antibod-ies as described previously (Gradinaru et al. 2009). Imageswere acquired using a Zeiss microscope (Axio ImagerM1 Stand Mot, Carl Zeiss Microscopy, LLC, Thornwood,NY, USA). In order to assess the number of neurons express-ing ChR2, across three brain sections positioned at Bregma 0,1, and 2, we counted the total number of DAPI stained cells(N_c) in S1. Since neurons consist of ∼63 % of cortical cells(Miller and Potempa 1990), we calculated the total number ofneurons (N_n). From the GFP staining results in these brainsections, we counted the number of neurons that expressedChR2 (N_ChR2). The percentage of neurons in S1 expressingChR2 was calculated by 100×N_ChR2/N_n.

Statistics

Two-tailed Student’s t test was performed between the differ-ent stimulation conditions using Matlab (Mathworks). Resultsand figures show the average±standard error of mean (SEM).

Results

Frequency–Dependency of Forepaw Stimulus-Evoked BOLDfMRI and Neuronal Responses

Electrical stimulation of the forepaw was applied to 11 rats.Six of them were tested in 9.4 T MRI scanner to evaluate theforepaw stimulus-evoked BOLD fMRI responses. The activa-tion Z-map of the BOLD fMRI responses was calculated by Z-test analysis with the effective significance of p<0.05.Electrophysiological recordings to measure the forepawstimulus-evoked neuronal responses were tested on the otherfive rats. LFP and MUAwere obtained throughout the depthof S1 by using the 12-channel axial array microelectrodepositioned at the center of S1 forepaw representation(Bregma 0, ML 3.7 mm). Figure 1a shows activation Z-maps of the BOLD responses evoked by contralateral forepawstimulations at 1, 3, 6, 9, and 20 Hz. The highest amplitude ofthe BOLD fMRI percentage change averaged across S1 wasobserved at 9 Hz (6.45±0.66 %, n=6), while the lowest valueat 1 Hz (4.29±0.59 %, n=6). The highest spatial extent ofBOLD fMRI response, calculated by the number of the acti-vated pixels in the functional Z-maps of BOLD signal, wasobserved at 6 Hz (327.17±37.23, n=6). On the other hand, thehighest amplitude of the stimulus-evoked LFP and MUAwasobserved at 3 Hz (1.87±0.28 μV and 421.07±58.13, respec-tively; n=5). These results demonstrate the different frequen-cy–dependency of BOLD fMRI and neuronal responses ofrats under dexmedetomidine anesthesia.

Laminar Characteristics of Forepaw Stimulus-Evoked BOLDfMRI and Neuronal Responses

Based on the above frequency–dependency results, the spatialcomparison of signals were performed at 9 Hz, which was thefrequency of where the greatest BOLD fMRI response wasobserved. The activated pixels, calculated by Z-test analysis ofBOLD fMRI signals with effective significance of p<0.05,were found across the cortical depth in S1 contralateral to theforepaw stimulation. Three regions of interest (ROI) weredefined: laminae II+III (150–600 μm), lamina IV (600–900 μm), and laminae V+IV (900–1,800 μm). The timecourse and the amplitude of BOLD fMRI responses werecalculated for each ROI. The increases of the BOLD fMRIsignal occurred immediately following the stimulation onsetand persisted throughout the duration of the stimulation(Fig. 2a). The highest amplitude of the BOLD fMRI percent-age change was observed in laminae II+III (7.72±0.48 %),followed by lamina IV (7.66±0.44 %) and laminae V+VI(4.31±0.14 %; n=6). The highest amplitude of the stimulus-evoked LFP and MUA were found in laminar IV (LFP:laminae II+III 0.26±0.04 μV, lamina IV 1.49±0.13 μV, lam-inae V+VI 1.04±0.11 μV; MUA: laminae II+III 230.22±

J Mol Neurosci

28.53, lamina IV 454.68±26.91, laminae V+VI 287.01±30.69; n=5). Figure 2b, c shows a representative exampleand the group averages of the laminar spatial distribution ofthe electrophysiology responses, respectively. In agreementwith previous reports (Silva and Koretsky 2002), we alsoobserved differences in the spatial distribution of the re-sponses between imaging and electrophysiology.

Lateral Spatial Profiles of Forepaw Stimulus-Evoked BOLDfMRI and Neuronal Responses

After comparing the laminar characteristics of forepawstimulus-evoked BOLD fMRI and neuronal responses, we theninvestigated how these neuronal responses are co-localizedwith the accompanying hemodynamic responses in the lateralspatial extent. Both BOLD fMRI and CBF responses were

measured to evaluate the hemodynamic changes in S1 evokedby forepaw stimulation. Laser speckle contrast imaging (LSCI),a minimally invasive optical imaging method (Dunn et al.2001; Li et al. 2009), was performed to detect CBF responsesin S1 representation (FOV=3.75×5.1 mm). This technique issensitive to CBF changes of the upstream pial arterioles anddraining venules within a cortical depth of approximately 0.6–0.8 mm (Dunn et al. 2005). The lateral spatial extent of the CBFZ-maps responses revealed by LSCI (Z-test analysis with effec-tive significance of p<0.05) to forepaw stimulation was ob-served across FOV that covered the right S1 forepaw represen-tation (n=5) (Fig. 3a). The percent change of the BOLD fMRIresponses across the activated pixels within lamina IV wascalculated for each individual animal for each condition andaveraged across subjects (n=6). Consistent with the directoptical CBF measurements, the spatial extents of the BOLD

Fig. 1 BOLD fMRI and neuronal activity evoked by sensory stimulationhave different frequency–dependency in S1. aExamples of BOLD fMRIactivation maps in response to contralateral forepaw stimulation at 1, 3, 6,9, and 20 Hz. b Averaged time courses of BOLD fMRI percentagechanges in response to forepaw stimulation at different frequencies(mean, n=6). The gray color bar depicts the length of stimulation. c

Averaged time courses and amplitudes of LFP in response to forepawstimulation at different frequencies (mean, n=5). d Normalized groupresults (mean±SEM) showed that the maximal magnitude of LFP and thespiking rate of MUA (n=5) was observed when the stimulus was pre-sented at 3 Hz, while the peak of the BOLD fMRI percentage change andits extent was observed at 9 and 6 Hz, respectively (n=6)

Fig. 2 BOLD fMRI and neuronal activity evoked by sensory stimulationat 9 Hz have different laminar profiles. aTime courses and amplitudes ofBOLD fMRI responses to forepaw stimulation across the cortical depth(mean±SEM; n=6) demonstrate that hemodynamic responses initiatedand had the highest amplitude in laminae II+III. b Example of evoked

LFP responses to forepaw stimulation demonstrates that neuronal re-sponses initiated and had the highest amplitude in lamina IV and thenspread throughout the cortical depth. cNormalized group results (mean±SEM) of averaged LFP amplitude (n=5), spiking rates of MUA (n=5),and magnitude of BOLD fMRI (n=6) responses across the cortical depth

J Mol Neurosci

fMRI Z-maps responses to forepaw covered a region in medi-al–lateral distance of 3,600±198 μm (Fig. 3b).

The lateral spatial characteristics of neuronal responsesevoked by forepaw stimulation were tested by a grid of ninetungsten electrodes. This grid covered a 2×2-mm area withinthe S1 forepaw representation (Fig. 3c). LFP and MUA re-sponses were recorded simultaneously from all nine elec-trodes that were lowered into lamina IV. Forepaw stimulationresulted in LFP and MUA responses that were detected in allthe electrodes positioned across the right S1 representation(Fig. 3d, e).

Spatial Activation Profiles of Optogenetics-Induced BOLDfMRI and Neuronal Responses

To further investigate the co-localization between the BOLDfMRI and neuronal responses, we induced localized neuronalactivation by optogenetic approaches. Fifteen animals weregenetically engineered to express ChR2 in excitatory corticalneurons. Immunohistochemistry was performed on brainslices obtained from all the rats that participated in this study(Fig. 4a). Cells that were positively stained for DAPI (nuclearstaining), EYFP (indicative of ChR2 expression), andCaMKIIα (marker for excitatory neurons) were found acrossthe cortical depth at laminae II to VI in S1 granular zone andconsisted about 35 % of all neurons in that area. An opticalfiber was used for light delivery, and laser pulses of 20 Hz for20 ms were applied to produce reproducible neuronal re-sponses (Gradinaru et al. 2009; Lee et al. 2010).

Figure 4b shows the laminar activation profiles of the LFPresponses in S1 that were evoked by the ChR2 stimulation.

The neuronal responses were confined to lamina IV (with thepeak LFP amplitude of 0.12±0.02 μV) and the supragranularlaminae with response amplitudes ten times smaller comparedto forepaw stimulation (Student t test, p<0.005, n=5).Figure 4c, d demonstrates the lateral spatial profiles of theLFP and MUA responses across lamina IV of S1 that weremeasured by nine-channel electrodes grid. Both evoked LFPand MUA responses by ChR2 stimulation were detected onlyin the electrode proximate (<100 μm) to the optic fiber thatdelivered the light stimulation, with minimal evoked re-sponses in the other eight electrodes (Fig. 4c, d). Thus, theChR2 stimulation resulted in neuronal responses that werelimited to laminar IV and II+III with a spatial extent of lessthan 500 μm.

We then characterized the spatial characteristics of thehemodynamic responses evoked by ChR2 stimulations acrossthe forepaw representation in S1. Although the changes inneuronal firing rate induced by ChR2 stimulation, as demon-strated above, were localized to the stimulated cortical area,the hemodynamic responses were found to cover a muchbroader area and were only partially co-localized with theunderlying neuronal responses (Fig. 4e, f). The extents ofthe CBF Z-maps responses to ChR2 stimulations in S1 wereobserved across the FOV that covered the S1 forepaw repre-sentation (3,700±233 vs. 3,750±180 μm, n=5; Fig. 4g). Theaverage amplitude of the CBF responses across S1 was abouttwo times less in response to ChR2 stimulation compared toforepaw stimulation (15.07±2.12 vs. 28.40±1.75 %; Student ttest, p<0.005, n=5). Consistent with the direct optical CBFmeasurements, the spatial extents of the BOLD fMRI Z-mapsresponses to ChR2 stimulation were similar to the activation

Fig. 3 Hemodynamic responses and neuronal activity evoked by sensorystimulation have similar spatial extent across S1. a Example of CBFactivation Z-map overlaid on the original CBF contrast image shows thatforepaw stimulation resulted in extensive hemodynamic response in theright S1. The tip of the optic fiber used for ChR2 wasmasked off from theimages. b Examples of BOLD fMRI activation Z-map overlaid on the

anatomical images. cAn illustration of the position of the nine recordingelectrodes (red dots) and the optic fiber (blue star) overlaid on a LSCIimage of the rat S1. dForepaw stimulation resulted in LFP responses thatwere observed in all nine electrodes that were positioned at lamia IVandcovered 2×2mmof S1. eForepaw stimulation resulted inMUA that wereobserved in all nine electrodes across S1

J Mol Neurosci

maps obtained following forepaw stimulation (3,450±244 vs.3,600±198 μm, n=5; Fig. 4h). The average amplitude of theBOLD responses across the activated pixels within S1 wastwo times less in response to ChR2 compared to forepawstimulation (2.49±0.54 vs. 5.51±0.26 %; Student t test,p<0.05, n=5). We also characterized the laminar profiles ofBOLD fMRI responses to the ChR2 stimulation (Fig. 4i). Thehighest BOLD fMRI increases were detected in laminae II+III, followed by lamina IV and lamina V+VI, which weresimilar to the laminar activation profiles observed as a re-sponse to forepaw stimulation.

Discussion

We studied the relationship between the neuronal and BOLDfMRI responses in the somatosensory cortex by comparingneuronal and hemodynamic responses evoked by sensory andlocalized ChR2 stimulations. Multi-model measurements in-cluding multi-channel electrophysiology recordings, opticalimaging of CBF, and BOLD fMRI were applied. In agreementwith previous studies (Sanganahalli et al. 2008; Kim et al.2010), the evoked neuronal and its accompanying BOLDfMRI responses to tactile stimulation demonstrated differ-ences in their stimulus frequency–dependency and laminarcharacteristics. We then sought to further investigate the spa-tial co-localization of the neuronal and hemodynamic

responses by using optogenetics as a mean to activate alocalized subset of S1 neurons. The results demonstrated thatstimulation of ChR2-expressing excitatory neurons in S1 in-duced neuronal activity that was mainly confined to laminaIV. However, the extent of the CBF and BOLD fMRI hemo-dynamic responses due to ChR2 stimulation was greater thanthe neuronal activity measured by electrophysiology and sim-ilar to the hemodynamic responses evoked by tactilestimulation.

The neuronal pathway by which peripheral tactile stimula-tion activates cortical neuronal assemblies and its activationpatterns across the cortical laminae and the somatosensoryrepresentation is well described. Recently, several studiesusing transgenic mice expressing ChR2 in excitatory neuronsin lamina V of sensory and motor cortices focused on thecorrelation between the neuronal and the hemodynamic re-sponses (Lee et al. 2010; Desai et al. 2011; Kahn et al. 2011;Scott and Murphy 2012; Kahn et al. 2013; Vazquez et al.2014). These studies have demonstrated high temporal line-arity and correlation between the BOLD fMRI responses dueto the optogenetic stimulation and the neurophysiologicalmeasured responses. However, unlike neuronal responsesevoked by tactile sensory stimulation that are known to initiatein lamina IV, both Khan et al. (2013) and Vazquez et al. (2014)reported that the maximal neuronal activity observed in thetransgenic mice that express ChR2 specifically in lamina Vwas detected in lamina V.

Fig. 4 Hemodynamic responses and neuronal activity evoked by ChR2stimulation are partially co-localized in S1. aExamples of neurons withinS1 immunostained with antibodies targeted towards EYFP (green, indic-ative of ChR2 expression) and CaMKIIα (red, marker of excitatorypyramidal neurons). A large field of view fluorescence image shows thatChR2 expressing neurons are located throughout the cortex. bExample ofevoked LFP responses to ChR2 stimulation shows that responses wereconfined to laminae II+III and IV. cChR2 stimulation resulted in evokedLFP responses that were observed only in electrodes proximate to theoptic fiber with a distance less than 500 μm. dChR2 stimulation resultedin increased MUA spiking activities only in electrodes proximate to theoptic fiber. e Example of CBF activation Z-map shows that ChR2

stimulations resulted in extensive CBF increases in S1. f Example ofBOLD fMRI activation Z-map shows that ChR2 stimulations resulted inextensive BOLD fMRI responses in S1. g Group average of the CBFresponses across S1 representation (in the medial–lateral axis) demon-strates similar spatial extent of the hemodynamic responses induced byforepaw and ChR2 stimulations (mean±SEM, n=5). hGroup average ofthe BOLD fMRI responses across lamina IV S1 representation (in themedial–lateral axis) demonstrates similar extent and localization of thehemodynamic responses induced by forepaw and ChR2 stimulations(mean±SEM, n=5). i Time courses of BOLD fMRI responses acrossthe cortical laminae evoked by ChR2 (mean±SEM, n=5)

J Mol Neurosci

The principal difference between our animal model and theabove studies is that in our study, the ChR2 was expressedthroughout the cortical depth of the rat’s brain and was notconfined to deep cortical laminae. Consequently, we couldinduce ChR2 activity in supragranular laminae and elicit local-ized neuronal activity in lamina IV, which is also the primarylamina receiving thalamocortical projections and responding tothalamocortical input. This has led to cortical responses thatexhibited different neuronal characteristics compared to con-ventional tactile stimulation in terms of the neuronal activitydistribution across the cortical depth and the somatosensoryrepresentation. Nevertheless, it produced hemodynamic re-sponses that had similar characteristics to the ones detectedfollowing tactile stimulation. Interestingly, a recent study dem-onstrated that administration of glutamate receptor antagonistssignificantly decreased sensory evoked hemodynamic re-sponses, but had little effect on ChR2 evoked hemodynamicresponses in transgenic mice expressing ChR2 in lamina V(Scott and Murphy 2012). These findings further support thathemodynamic responses may be independent of local excitato-ry synaptic transmission. Together, these findings suggest afundamental difference in the spatial extent between the neuro-nal and the hemodynamic responses and that under certainconditions the BOLD fMRI responses may only partially co-localized with the neuronal responses.

Simultaneous electrophysiology recordings and fMRI mea-surements in the cat’s visual cortex (Kim et al. 2004) demon-strated that the BOLD fMRI signals that were averaged acrossmillimeters scale were tightly correlated to the electrophysiolog-ical responses. However, BOLD fMRI signals that were aver-aged for voxel sizes smaller than 2.6×2.6 mm2 showed highvariance and correlation of less than 50 % with the electrophys-iological responses. It emerges that the spatial correspondencebetween BOLD fMRI, optical imaging, and neuronal activity isdetermined by intrinsic properties of both the neurovascularcoupling mechanism and the imaging technologies. BOLDfMRI signals originate from partially deoxygenated blood ves-sels, such as capillaries, venules, and downstream veins (Daviset al. 1998; van Zijl et al. 1998), while optical CBF imaginggenerally detects blood flow in arterioles and upstream pialarteries (Li et al. 2009; Boas and Dunn 2010). Gradient echo(GE) pulse sequence, as was used in this study, is the commonBOLD fMRI method. However, the GE-BOLD fMRI signalsare highly sensitive for large draining veins which reduce thecapability of the technology to detect neuronal responses insubmillimeter resolution (Zhao et al. 2006). Other methods suchas spin echo BOLD fMRI, cerebral blood volume, and cerebralblood flow MRI measurements may provide superior spatiallocalization of neuronal activities (Ugurbil et al. 2003; Zhaoet al. 2004; Moon et al. 2007). However, they offer lowertemporal resolution and signal-to-noise ratios compare to theGE-BOLD fMRI method and are therefore less applied in func-tional brain studies (Kim and Ogawa 2012).

In our study, even though the neuronal responses as a resultof ChR2 stimulation were found to be localized in electrodesadjacent to the light stimulation area, the accompanying he-modynamic changes were detected more than 3 mm further inS1. Capitalizing on previous studies, the results suggest thatdue to the complex hemodynamic mechanism that is the basisof the BOLD fMRI contrast, there may be differences in thespatial extent between the neuronal and the hemodynamicresponses.

Acknowledgments The authors would like to thank Dr. KarlDeisseroth for kindly providing the ChR2 plasmids, Drs. Assaf Giladand Piotr Walczak for helping with virus preparation, and John Downeyand Yang Han for assisting with the animal imaging. This work wassupported by NIH/NINDS R01NS072171 (G.P.).

References

Airan RD, Li N, Gilad AA, Pelled G (2013) Genetic tools to manipulateMRI contrast. NMR Biomed 26:803–809

Attwell D, Iadecola C (2002) The neural basis of functional brain imagingsignals. Trends Neurosci 25:621–625

Boas DA, Dunn AK (2010) Laser speckle contrast imaging in biomedicaloptics. J Biomed Opt 15:011109

Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005)Millisecond-timescale, genetically targeted optical control of neuralactivity. Nat Neurosci 8:1263–1268

Caesar K, Thomsen K, Lauritzen M (2003) Dissociation of spikes,synaptic activity, and activity-dependent increments in rat cerebellarblood flow by tonic synaptic inhibition. Proc Natl Acad Sci U S A100:16000–16005

Davis TL, Kwong KK, Weisskoff RM, Rosen BR (1998) Calibratedfunctional MRI: mapping the dynamics of oxidative metabolism.Proc Natl Acad Sci U S A 95:1834–1839

Desai M, Kahn I, Knoblich U, Bernstein J, Atallah H, Yang A et al (2011)Mapping brain networks in awake mice using combined opticalneural control and fMRI. J Neurophysiol 105:1393–1405

Devor A, Tian P, Nishimura N, Teng IC, Hillman EM,Narayanan SN et al(2007) Suppressed neuronal activity and concurrent arteriolar vaso-constriction may explain negative blood oxygenation level-dependent signal. J Neurosci 27:4452–4459

Devor A, Hillman EM, Tian P, Waeber C, Teng IC, Ruvinskaya L et al(2008) Stimulus-induced changes in blood flow and 2-deoxyglucoseuptake dissociate in ipsilateral somatosensory cortex. J Neurosci 28:14347–14357

Dunn AK, Bolay H, Moskowitz MA, Boas DA (2001) Dynamic imagingof cerebral blood flow using laser speckle. J Cereb Blood FlowMetab 21:195–201

Dunn AK, Devor A, Dale AM, Boas DA (2005) Spatial extent of oxygenmetabolism and hemodynamic changes during functional activationof the rat somatosensory cortex. Neuroimage 27:279–290

Ekstrom A (2010) How and when the fMRI BOLD signal relates tounderlying neural activity: the danger in dissociation. Brain ResRev 62:233–244

Enager P, Piilgaard H, Offenhauser N, Kocharyan A, Fernandes P, HamelE et al (2009) Pathway-specific variations in neurovascular andneurometabolic coupling in rat primary somatosensory cortex. JCereb Blood Flow Metab 29:976–986

Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K(2009) Optical deconstruction of parkinsonian neural circuitry.Science 324:354–359

J Mol Neurosci

Iadecola C, Nedergaard M (2007) Glial regulation of the cerebral micro-vasculature. Nat Neurosci 10:1369–1376

Kahn I, Desai M, Knoblich U, Bernstein J, Henninger M, Graybiel AMet al (2011) Characterization of the functional MRI response tem-poral linearity via optical control of neocortical pyramidal neurons. JNeurosci 31:15086–15091

Kahn I, Knoblich U, DesaiM, Bernstein J, Graybiel AM, Boyden ES et al(2013) Optogenetic drive of neocortical pyramidal neurons gener-ates fMRI signals that are correlated with spiking activity. Brain Res1511:33–45

Kim SG, Ogawa S (2012) Biophysical and physiological origins of bloodoxygenation level-dependent fMRI signals. J Cereb Blood FlowMetab 32:1188–1206

Kim DS, Ronen I, Olman C, Kim SG, Ugurbil K, Toth LJ (2004) Spatialrelationship between neuronal activity and BOLD functional MRI.Neuroimage 21:876–885

Kim T, Masamoto K, FukudaM, Vazquez A, Kim SG (2010) Frequency-dependent neural activity, CBF, and BOLD fMRI to somatosensorystimuli in isoflurane-anesthetized rats. Neuroimage 52:224–233

Koehler RC, Roman RJ, Harder DR (2009) Astrocytes and the regulationof cerebral blood flow. Trends Neurosci 32:160–169

Lauritzen M (2005) Reading vascular changes in brain imaging: isdendritic calcium the key? Nat Rev Neurosci 6:77–85

Lauritzen M, Mathiesen C, Schaefer K, Thomsen KJ (2012) Neuronalinhibition and excitation, and the dichotomic control of brain hemo-dynamic and oxygen responses. Neuroimage 62:1040–1050

Lecrux C, Hamel E (2011) The neurovascular unit in brain function anddisease. Acta Physiol (Oxf) 203:47–59

Lee JH, Durand R, Gradinaru V, Zhang F, Goshen I, Kim DS et al (2010)Global and local fMRI signals driven by neurons definedoptogenetically by type and wiring. Nature 465:788–792

Li N, Jia X, Murari K, Parlapalli R, Rege A, Thakor NV (2009) Highspatiotemporal resolution imaging of the neurovascular response toelectrical stimulation of rat peripheral trigeminal nerve as revealedby in vivo temporal laser speckle contrast. J Neurosci Methods 176:230–236

Li N, Downey JE, Bar-Shir A, Gilad AA, Walczak P, Kim H et al (2011)Optogenetic-guided cortical plasticity after nerve injury. Proc NatlAcad Sci U S A 108:8838–8843

Logothetis NK (2008) What we can do and what we cannot do withfMRI. Nature 453:869–878

Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A (2001)Neurophysiological investigation of the basis of the fMRI signal.Nature 412:150–157

Miller MW, Potempa G (1990) Numbers of neurons and glia in mature ratsomatosensory cortex: effects of prenatal exposure to ethanol. JComp Neurol 293:92–102

Moon CH, Fukuda M, Park SH, Kim SG (2007) Neural interpretation ofblood oxygenation level-dependent fMRI maps at submillimetercolumnar resolution. J Neurosci 27:6892–6902

Mukamel R, Gelbard H, Arieli A, Hasson U, Fried I, Malach R (2005)Coupling between neuronal firing, field potentials, and FMRI inhuman auditory cortex. Science 309:951–954

Pawela CP, Biswal BB, Hudetz AG, Li R, Jones SR, Cho YR et al (2010)Interhemispheric neuroplasticity following limb deafferentation de-tected by resting-state functional connectivity magnetic resonanceimaging (fcMRI) and functional magnetic resonance imaging(fMRI). Neuroimage 49:2467–2478

Pelled G, BergstromDA, Tierney PL, ConroyRS, ChuangKH, YuD et al(2009) Ipsilateral cortical fMRI responses after peripheral nervedamage in rats reflect increased interneuron activity. Proc NatlAcad Sci U S A 106:14114–14119

Sanganahalli BG, Herman P, Hyder F (2008) Frequency-dependent tactileresponses in rat brain measured by functional MRI. NMR Biomed21:410–416

Scott NA, Murphy TH (2012) Hemodynamic responses evoked by neuro-nal stimulation via channelrhodopsin-2 can be independent ofintracortical glutamatergic synaptic transmission. PLoSOne 7:e29859

Shmuel A, Augath M, Oeltermann A, Logothetis NK (2006) Negativefunctional MRI response correlates with decreases in neuronal ac-tivity in monkey visual area V1. Nat Neurosci 9:569–577

Silva AC, Koretsky AP (2002) Laminar specificity of functional MRIonset times during somatosensory stimulation in rat. Proc Natl AcadSci U S A 99:15182–15187

Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE,Johansen-Berg H et al (2004) Advances in functional and structuralMR image analysis and implementation as FSL. Neuroimage23(Suppl 1):S208–S219

Sultan F, Augath M, Logothetis N (2007) BOLD sensitivity to corticalactivation induced by microstimulation: comparison to visual stim-ulation. Magn Reson Imaging 25:754–759

Ugurbil K, Toth L, Kim DS (2003) How accurate is magnetic resonanceimaging of brain function? Trends Neurosci 26:108–114

van Eijsden P, Hyder F, Rothman DL, Shulman RG (2009)Neurophysiology of functional imaging. Neuroimage 45:1047–1054

van Zijl PC, Eleff SM, Ulatowski JA, Oja JM, Ulug AM, Traystman RJet al (1998) Quantitative assessment of blood flow, blood volumeand blood oxygenation effects in functional magnetic resonanceimaging. Nat Med 4:159–167

Vanzetta I, Grinvald A (2008) Coupling between neuronal activity andmicrocirculation: implications for functional brain imaging. HFSP J2:79–98

Vazquez AL, Fukuda M, Crowley JC, Kim SG (2013) Neural andhemodynamic responses elicited by forelimb- and photo-stimulation in channelrhodopsin-2 mice: insights into the hemody-namic point spread function. Cereb Cortex (in press)

Viswanathan A, Freeman RD (2007) Neurometabolic coupling in cere-bral cortex reflects synaptic more than spiking activity. Nat Neurosci10:1308–1312

Yen CC, Fukuda M, Kim SG (2011) BOLD responses to differenttemporal frequency stimuli in the lateral geniculate nucleus andvisual cortex: insights into the neural basis of fMRI. Neuroimage58:82–90

Zhang F, Gradinaru V, Adamantidis AR, Durand R, Airan RD, de LeceaL et al (2010) Optogenetic interrogation of neural circuits: technol-ogy for probing mammalian brain structures. Nat Protoc 5:439–456

Zhao F, Wang P, Kim SG (2004) Cortical depth-dependent gradient-echoand spin-echo BOLD fMRI at 9.4 T. Magn Reson Med 51:518–524

Zhao F, Wang P, Hendrich K, Ugurbil K, Kim SG (2006) Cortical layer-dependent BOLD and CBV responses measured by spin-echo andgradient-echo fMRI: insights into hemodynamic regulation.Neuroimage 30:1149–1160

Zhao F, Zhao T, Zhou L,WuQ, HuX (2008) BOLD study of stimulation-induced neural activity and resting-state connectivity inmedetomidine-sedated rat. Neuroimage 39:248–260

J Mol Neurosci