Journal of Cerebral Blood Flow & Metabolism (2014) 34, 532–541 & 2014 ISCBFM All rights reserved 0271-678X/14 $32.00 www.jcbfm.com

ORIGINAL ARTICLE

g-Aminobutyric acid (GABA) concentration inversely correlates with basal perfusion in human occipital lobe Manus J Donahue1,2,3,4, Swati Rane1, Erin Hussey3, Emily Mason3,5, Subechhya Pradhan1, Kevin W Waddell1 and Brandon A Ally2,3 Commonly used neuroimaging approaches in humans exploit hemodynamic or metabolic indicators of brain function. However, fundamental gaps remain in our ability to relate such hemo-metabolic reactivity to neurotransmission, with recent reports providing paradoxical information regarding the relationship among basal perfusion, functional imaging contrast, and neurotransmission in awake humans. Here, sequential magnetic resonance spectroscopy (MRS) measurements of the primary inhibitory neurotransmitter, g-aminobutyric acid (GABA þ macromolecules normalized by the complex N-acetyl aspartate-N-acetyl aspartyl glutamic acid: [GABA þ ]/[NAA–NAAG]), and magnetic resonance imaging (MRI) measurements of perfusion, fractional gray-matter volume, and arterial arrival time (AAT) are recorded in human visual cortex from a controlled cohort of young adult male volunteers with neurocognitive battery-confirmed comparable cognitive capacity (3 T; n ¼ 16; age ¼ 23±3 years). Regression analyses reveal an inverse correlation between [GABA þ ]/[NAA–NAAG] and perfusion (R ¼  0.46; P ¼ 0.037), yet no relationship between AAT and [GABA þ ]/[NAA–NAAG] (R ¼  0.12; P ¼ 0.33). Perfusion measurements that do not control for AAT variations reveal reduced correlations between [GABA þ ]/[NAA–NAAG] and perfusion (R ¼  0.13; P ¼ 0.32). These findings largely reconcile contradictory reports between perfusion and inhibitory tone, and underscore the physiologic origins of the growing literature relating functional imaging signals, hemodynamics, and neurotransmission. Journal of Cerebral Blood Flow & Metabolism (2014) 34, 532–541; doi:10.1038/jcbfm.2013.231; published online 8 January 2014 Keywords: arterial spin labeling; fMRI; GABA; inhibition

INTRODUCTION Magnetic resonance imaging (MRI) has been widely applied to obtain a more thorough understanding of the central nervous system. In addition to being an excellent tool for structural tissue characterization, functional MRI (fMRI) can be used to measure regional and global cerebral hemodynamics and to make inferences regarding neuronal activity. Functional MRI, which most commonly exploits blood oxygenation level-dependent (BOLD) contrast,1 does not require exogenous contrast agents and therefore is an invaluable tool for obtaining repeated functional measurements, assessing longitudinal changes in brain function, and for probing pathophysiologic mechanisms in clinical scenarios where contrast agents may be contraindicated. As a result, BOLD fMRI has emerged as the most popular tool for assessing brain function in humans, with a wide variety of clinical, pharmacological, and neuroscience applications to date. However, the BOLD fMRI signal is only an indirect marker of neuronal activity, arising from complex neurochemical, metabolic, and hemodynamic modulations that occur concurrent to both ongoing and stimulus-evoked neuronal activity. Specifically, in stimulus-evoked experiments, BOLD contrast arises owing to a greater increase in cerebral blood flow (CBF; 20% to 50%) relative to cerebral blood volume (CBV; 5% to 30%) and the cerebral metabolic rate of oxygen consumption (0% to 20%), leading to an increase in diamagnetic oxy-hemoglobin relative to paramagnetic deoxy-hemoglobin in capillaries and veins. 1

While much progress has been made in understanding the hemodynamic and metabolic contributions to BOLD contrast, important gaps remain in our ability to relate BOLD signals to underlying neuronal activity and neurotransmission. These gaps significantly hinder BOLD interpretability in many applications, as substantial variability in BOLD responses exists between even healthy individuals, partly accounting for why BOLD data remain largely qualitative in nature. Understanding the physiologic sources of this variability is fundamental to using BOLD as a tool for identifying quantitative differences in brain function between individuals and conditions and for gauging functional response to disease and treatment. The critical barrier to characterizing BOLD signal is that contrast is fueled by a variety of sources, with few direct observables. Alternative fMRI approaches specifically sensitive to individual hemodynamic parameters (e.g., CBF or CBV) can be applied in sequence with BOLD fMRI for more comprehensive investigations.2–6 Yet, the majority of multimodal studies in humans have focused on understanding hemodynamic and metabolic changes, with comparatively less information available on the neurochemical precursors to these changes. However, coordination between excitatory and inhibitory neurotransmission, facilitated by glutamate and g-aminobutyric acid (GABA) respectively, is fundamental to elicit hemodynamic, and therefore BOLD responses. Much elegant work, primarily from the animal literature, has provided important insights into the neuro-energetic mechanisms

Department of Radiology, Vanderbilt University School of Medicine, Nashville, Tennessee, USA; 2Department of Psychiatry, Vanderbilt University School of Medicine, Nashville, Tennessee, USA; 3Department of Neurology, Vanderbilt University School of Medicine, Nashville, Tennessee, USA; 4Department of Physics and Astronomy, Vanderbilt University, Nashville, Tennessee, USA and 5Department of Neuroscience, Vanderbilt University School of Medicine, Nashville, Tennessee, USA. Correspondence: Dr MJ Donahue, Vanderbilt University Institute of Imaging Science, Medical Center North, 1161 21st Avenue South, Nashville, TN 37232, USA. E-mail: [email protected] This work was supported in part by a grant from the National Institute of Neurological Disorders and Stroke (NIH/NINDS 5R01NS078828-02). Received 26 July 2013; revised 14 November 2013; accepted 3 December 2013; published online 8 January 2014

GABA and perfusion in visual cortex MJ Donahue et al

533 of excitatory and inhibitory processes.7 For instance, glucose taken up from the blood is converted into glutamate in neurons (glutamate oxidation; CMRglc(ox)), after which glutamate binds to postsynaptic neurons, and subsequently is converted into glutamine in astrocytes, and finally transported back to neurons (rate of glutamate/glutamine cycling: Vcyc).8 Vcyc has been reported using 13C NMR spectroscopy during 13C-labeled glucose in anesthetized9,10 and awake11 rats, as well as in human cortex.12 These studies have shown that CMRglc(ox) and Vcyc are linearly related over a range of anesthesia levels in rats,10,11,13–15 thereby suggesting a close relationship between oxidative energy metabolism and glutamatergic neurotransmission. Additionally, a 13C study combining [1-13C]glucose and [2-13C]acetate infusion in anesthetized rats demonstrated that glutamatergic and GABAergic contributions to glucose oxidation could be separated, and that GABAergic neurons account for a smaller, but still substantial, 23% of total neurotransmitter cycling and 18% of total neuronal tricarboxylic acid cycle flux.16 Therefore, evidence has been provided for oxidative metabolism being very closely linked with changes in both glutamatergic and GABAergic neurotransmission. These findings have been analyzed in the context of energy budgets, which first suggested that the majority of this energy in gray matter of humans may be used to reverse the ion fluxes in excitatory postsynaptic action potentials and currents;17 however, more recently evidence has been provided in human occipital lobe suggesting that coupling between glutamate neurotransmitter cycling and metabolic activity is similar in humans and in rodents, and that most of the metabolic activity is used to support glutamatergic signaling.18 To relate these findings to functional imaging, it is also necessary to understand how CBF varies with excitation and inhibition. While there is general agreement that changes in CBF correlate with oxygen usage during periods of elevated neuronal activity,19 increases in CBF have been reported to extend beyond the locus of oxygen utilization,20 which suggest that CBF is controlled by additional factors beyond signals related to energy production (e.g., CO2, H þ ) alone. Evidence has been provided for glutamate-evoked Ca2 þ influx in postsynaptic neurons activating the production of nitric oxide, arachidonic acid, and adenosine, which in turn produce vasodilation and related increases in CBF.21,22 Largely owing to less available data, more controversy exists with regard to how inhibition contributes to brain energy use, CBF, and thus functional imaging signals. In cortex, exogenous GABA has been shown to increase CBF and CBV,23 yet this observation was not replicated in cerebellum.24 It has also been reported that inhibition does not contribute to BOLD signals, partly because the fraction of inhibitory cells is smaller relative to the fraction of excitatory cells25 and also because mouse astrocytes exposed to glutamate initiate uptake of glucose, whereas no uptake is observed during GABA exposure.26 Therefore, despite elegant work on the neurochemical sequelae of energy production, it remains unclear how some of these findings translate to large-scale changes in CBF, metabolism, and therefore functional imaging signals, in a relatively large (e.g., 3  3  3 mm) fMRI voxel.27,28 An additional complication is that neurotransmitter investigations are frequently conducted under pharmacological manipulation in sedated animals,8,29 or with tissue biopsy or autopsy procedures, with the obvious caveat that in vitro chemical profiles may differ from their in vivo counterparts30 and anesthesia may confound measurements. Furthermore, the greater number of synapses per neuron in primates, relative to rats, lends support for postsynaptic responses playing a more dominant energetic role in humans.28 Therefore, while animal studies have greatly improved our understanding of relationships between metabolism and neurotransmission, animal work has inherent limitations, and thus it is important to additionally understand these relationships in the awake human brain. & 2014 ISCBFM

To gain additional information in awake humans, it is possible to measure in vivo neurotransmitter concentrations using 1H magnetic resonance spectroscopy (MRS) and spectral editing.31–34 Using this approach, it has been shown that basal GABA concentration explains variation in the magnitude of the positive stimulus-evoked hemodynamic responses in human visual cortex,35 and synaptic inhibition, as similarly measured by baseline GABA concentration and MRS, has been shown to correlate inversely with the BOLD fMRI signal change in human visual cortex36 and rat somatosensory cortex,29 and directly with negative BOLD responses in anterior cingulate cortex.37 These findings suggest that BOLD variations are linked to measurable excitation-inhibition balance and cortical network activity. Furthermore, the link between neurotransmission and the BOLD response can be probed noninvasively in vivo using MRI and MRS together. Unexpectedly, a positive trend has been reported between cerebral blood flow-weighted reactivity using pulsed arterial spin labeling (ASL) and baseline GABA concentration in visual cortex.35 In a separate study using a different ASL technique,38 no relationship between CBF and GABA was observed, despite clear trends between temporal features of the evoked BOLD time course and GABA. A lack of relationship between CBF and GABA is unlikely given the reproducible correlations between BOLD and GABA reported. Importantly, a lack of relationship between these parameters casts doubt on the relevance of the BOLD/GABA correlations outlined above (which are frequently reported in relatively small volunteer groups). Alternatively, demonstration of a relationship between GABA and CBF would underscore the relevance of these relationships, and furthermore provide a greater degree of quantitative interpretability. Here, we aim to better characterize these relationships by performing careful measurements of CBF and GABA in human occipital lobe using improved experimental and statistical methods that account for variations in blood velocity and voxel tissue fraction. Importantly, ASL measurements are sensitive to both CBF and arterial arrival time (AAT: time for the labeled blood water to reach the capillary exchange site). Therefore, AAT, not CBF, may positively correlate with GABA and this relationship could explain the ASL–GABA correlation. This would provide support for blood water velocities, likely at the arteriolar level, being associated with GABA. Alternatively, the positive correlation between ASL and GABA may be driven by CBF. Thus, the inverse correlation between BOLD and baseline GABA is likely due to a positive correlation between cerebral metabolic rate of oxygen consumption reactivity and GABA. Here, we systematically investigate these hypotheses by sequentially measuring AAT, CBF, and GABA in the occipital lobe of an age-restrictive cohort of healthy adult male volunteers. We also administered a standard neurocognitive task adapted from an established paradigm,39 as cognitive impairment may confound the relationship between GABA and CBF.

MATERIALS AND METHODS Patient Anonymity, Informed Consent, and Ethics Anonymity of all volunteers was assured by removing the volunteers’ names from data and figures. This study includes experiments on human subjects; procedures were followed in accordance with the ethical standards of the Vanderbilt University Institutional Review Board (IRB Study #101567), the Vanderbilt University Human Research Protection Program, as well as with the Helsinki Declaration of 1975 (and as revised in 1983). Informed, written consent was obtained from all volunteers and all components of this study were in compliance with the Health Insurance Portability and Accountability Act. No animal studies were conducted as part of this work.

Volunteer Demographics Healthy volunteers (n ¼ 16; age ¼ 23±3 years; sex ¼ M) provided informed, written consent in accordance with the Vanderbilt University Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

GABA and perfusion in visual cortex MJ Donahue et al

534 Institutional Review Board. Only adult male volunteers were enrolled owing to the more varied roles of GABA in children and GABA, as well as other metabolite, levels varying regionally between the follicular and luteal phases of the menstrual cycle in females of child-bearing age.40 Additional exclusion criteria included volunteers with prior history of cerebrovascular disease as defined by stroke or transient ischemic attack, and/or prior or current neurologic disorders including dementia, multiple sclerosis, delirium, sleep disorders, or seizure disorders. Volunteers were asked to refrain from consuming caffeine 2 hours before the study and no volunteers were on the following classes of medications known to alter GABAergic tone: benzodiazepines, cholinesterase inhibitors, antipsychotics, opioids, or monoamine oxidase inhibitors. To confirm that all volunteers had similar, healthy cognitive capacities, independent neuropsychological evaluation was performed.

Neurocognitive Task The purpose of this component of the study was simply to ensure that all participants exhibited unimpaired cognitive capacity for their age and demographic. Participants completed a paradigm modeled after Dickerson et al,39 which was modified to be significantly more difficult to identify deficits in otherwise healthy young adults. Four lists of twenty-four unique face/name pairs were counterbalanced so that each face/name had an equal opportunity of being seen as a studied pair, a rearranged pair, or a novel pair. During the study portion, subjects were presented with 96 face/ name pairs and instructed to try to remember as many pairings as possible. Images were presented for 3 seconds followed by a fixation period ( þ ), which was randomly jittered between 3 and 10 seconds in half-second intervals. Subjects were then given a 15-minute rest period, in which they laid in the scanner at rest. During the test portion, subjects were again presented with face/name pairs and asked to indicate whether faces were ‘old’ or ‘new’ via a button box. Subjects were instructed to endorse an item as ‘old’ if it was the same face/name pairing as they had seen in the study portion. They were instructed to endorse an item as ‘new’ if the face/name pairing was completely novel or if the face/name pair had been seen before but was not previously paired together in the study portion (i.e., rearranged pair). Subjects had unlimited time to respond during the test portion. All stimuli were presented on a 53-cm flat screen monitor using the E-Prime 2.0 (Psychology Software Tools, Pittsburgh, PA, USA).

Imaging and Spectroscopy Protocol Each volunteer underwent a 3.0T MRI (Philips Medical Systems, Best, The Netherlands) using an 8-channel SENSE coil for reception and quadrature body coil for radiofrequency transmission. The MEGA-PRESS MRS acquisition was performed in the same scan session, except with a birdcage coil for both radiofrequency transmission and reception to increase B1 performance. The MRI protocol consisted of (1) a 3D structural T1-weighted MPRAGE (repetition time/echo time ¼ 5.4/2.5 ms; turbo gradient echo factor ¼ 160; spatial resolution ¼ 1  1  1 mm3), and (2) a gradient-echo single-shot echo planar imaging (factor ¼ 35) multiinversion time (TI) pseudo-continuous arterial spin labeling (pCASL) sequence. The pCASL sequence (repetition time/echo time ¼ 4,350/18 ms; spatial resolution ¼ 3.43  3.43  6 mm3; slices ¼ 15) utilized a Hanning pulse train for blood water labeling with pulse duration ¼ 0.5 ms and total labeling duration ¼ 1,200 ms; postlabeling delay values ¼ 200, 500, 800, 1,100, 1,400, 1,700, 1,900, and 2,300 ms (12 averages per TI), dual adiabatic background suppression for concomitant suppression of gray-matter and white-matter static tissue signal. The pCASL labeling duration was reduced slightly (duration ¼ 1,200 ms) relative to more conventional labeling duration of B1,500 to 2,000 ms41,42 to sensitize the approach to AAT, and the TI range was sampled with a temporal resolution of 300 ms over the approximate range of expected AATs. Finally, an equilibrium magnetization image (M0) was acquired with identical spatial resolution and readout as the pCASL scan, but with repetition time ¼ 20 seconds and all labeling and background suppression pulses removed. For MRS, the 3-p.p.m. GABA peak ([GABA þ ], which reflects GABA þ macromolecules) was measured using a previously published J-difference MEGA-PRESS approach43,44 with repetition time/echo time ¼ 2,000/73 ms, 320 spectra with 2,048 samples each. The editing pulses were toggled between 1.91 and 7.4 p.p.m. on alternate scans and the editing pulses had a bandwidth of 64 Hz. Care was taken to place the spectroscopy voxel (voxel dimensions: 25  30  22 mm) similarly in the occipital cortex between all volunteers and with minimal partial voluming between tissue and sagittal sinus. To provide an additional level of control regarding voxel placement, gray- and white-matter voxel fractions were recorded Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

and included in analysis, as outlined below. Optimization was performed before the spectroscopy acquisition to determine the pulse angle for improved water suppression. Four Regional Saturation Technique slabs using hyperbolic secant pulses targeted to the lipid resonance were placed orthogonally in the anterior–posterior and inferior–superior directions around the voxel to suppress the signal from lipids and to eliminate signal contamination from the scalp lipids. The REgional Saturation Technique slabs had a width of 30 mm, which was chosen to improve the profile relative to much larger REgional Saturation Technique slabs (50 to 100 mm) that are frequently used in other applications. While this approach proved effective to minimize conspicuous lipid contamination, residual lipids upfield of 1.9 p.p.m. were further mitigated by restricting the postprocessing quantification window to 1.9 to 4.0 p.p.m. The orientation of the spectroscopy relative to the pCASL and T1-weighted scans were recorded in all volunteers for subsequent coregistration and tissue volume segmentation.

Analysis: Neurocognitive Assessment For the face/name task, recognition accuracy scores for each subject were calculated using Pr (proportion of hits  proportion of false alarms). Accuracy scores were calculated for hits-novel false alarms (novel face/ name pairs only) and for hits-rearranged false alarms (rearranged face/ name pairs only).

Analysis: Preprocessing First, gray matter, white matter, and cerebrospinal fluid were segmented from the T1-weighted data using a standard hidden Markov random field model and associated expectation-maximization algorithm.45 Second, pCASL data were pair-wise subtracted and a single-compartment kinetic model was applied to simultaneously quantify CBF and AAT:46 0 0otoDt
Dtotot þ Dt 2M0;b fT1;app ae  Dt=T1;b 1  e  ðt  DtÞ=T1;app

2M0;b fT1;app ae  Dt=T1;b e  ðt  t  DtÞ=T1;app 1  e  t=T1;app t þ Dtot

ð1Þ

where f is perfusion (mL/g per second), a ¼ 0.85 is the pCASL labeling efficiency, T1,b ¼ 1.6 seconds is the mean blood water T1, Dt is the AAT of the labeled blood water bolus, t ¼ 1.2 seconds is the labeling duration, M0,b is the equilibrium blood water magnetization, t is the sum of the labeling duration (t) and postlabeling delay time, and 1 1 f ¼ þ T1;app T1 l

ð2Þ

for tissue water T1 ¼ 1.2 seconds and tissue/blood partition coefficient of water l ¼ 0.98 mL/g. M0,b was calculated from the equilibrium image (obtained without spin labeling or background suppression with identical echo planar imaging readout as the pCASL image). f and Dt were simultaneously fit for using standard minimization algorithms (fmincon) in Matlab (Mathworks, Natick, MA, USA). Note, while more complex kinetic models are available, which may account for multiple compartments47 and/or off-resonant magnetization transfer effects,48 at intermediate field strength it has been shown that these more complex models provide similar CBF values within experimental error,49 and therefore to avoid additional assumptions that are required in many of these more complex models, a popular single-compartment model was utilized. Third, for MRS analysis, the edited and unedited spectra were first corrected for phase and frequency variations between the 320 acquisitions using Creatine (Cr) as a reference peak. Since the editing pulse is asymmetrically disposed about the edited resonance, a differential phase results from this procedure. This effect was mitigated by applying a small differential zero-order phase correction to the respectively aligned, odd and even spectra. The corrected mean difference spectrum, obtained then by subtraction of the averaged unedited spectra from the averaged edited spectra, was processed using LCModel50 to obtain a final estimate of the [GABA þ ]/[NAA–NAAG] ratio. To enforce spectral quality control, a CramerRao Lower Bound (CRLB) threshold for [GABA þ ]/[NAA–NAAG] was utilized. Cramer-Rao Lower Bounds compare the known theoretical basis spectra and noise to predict an upper limit on the precision of the output concentrations obtained from LCModel. Larger percentages reflect poorer precision; in this study, CRLBs were restricted to p15% to ensure that data of poor quality were excluded from analysis. Finally, many studies separately normalize [GABA þ ] to [Cr]. Therefore, we additionally investigated the relationship between [GABA þ ]/[NAA–NAAG] and [GABA þ ]/[Cr]. For this purpose, the ratio of [NAA–NAAG]/[Cr] was & 2014 ISCBFM

GABA and perfusion in visual cortex MJ Donahue et al

535

A

2

200 ms

500 ms

800 ms

1100 ms

1400 ms

1700 ms

1900 ms

2300 ms

B

C

0 (a.u.)

1.2

0 (s) 75

0 (ml/100g/min)

Figure 1. Quantification of cerebral blood flow (CBF) and arterial arrival time (AAT) is achieved from acquisition of pseudo-continuous arterial spin labeling (pCASL) data with differing postlabeling delays. (A) pCASL difference images averaged over all volunteers (n ¼ 16; coregistered to Montreal Neurological Institute Atlas) show the inflow of labeled blood water for different postlabeling delays (listed in bottom left), along with the corresponding calculated (B) AAT (units in seconds) and (C) CBF maps (units in mL blood/100 g tissue per minute).

independently estimated from only the even spectra. Multiplication of [GABA þ ]/[NAA–NAAG] by [NAA–NAAG]/[Cr] provided the value for [GABA þ ]/[Cr] ratio.

Data Analysis and Statistical Considerations The primary objective is to assess the correlation between the MRI measurements (i.e., baseline CBF and AAT) and the MRS measurement (i.e., [GABA þ ]/[NAA–NAAG]). To achieve this, the GABA voxel was superimposed onto the quantified CBF and AAT (Dt) maps and CBF and AAT were recorded only for common cortical regions that overlapped between the CBF, AAT and MRS regions. A measure of error was calculated as the standard deviation (STD) of the CBF and AAT over the common region-ofinterest or as the value of the CRLB of the [GABA þ ] measurement in the J-edited spectra. This analysis pipeline ensured that the same regions were & 2014 ISCBFM

being compared for the MRS and MRI comparisons. Next, normal probability plots and Pearson’s R values were calculated separately between the dependent variable (CBF) and the independent variables ([GABA þ ]/[NAA–NAAG] and AAT). Additional secondary analyses were performed as well. Importantly, gray-matter fraction within the voxel may vary between volunteers, and therefore we additionally performed multiple regression analysis whereby the gray-matter voxel fraction, calculated from the T1-weighted segmentation procedure described above, was added as an additional parameter in the model. R2, adjusted R2, and P values were recorded. Finally, it is important to note that the AAT may contribute to final interpretation in two ways. First, AAT may correlate with [GABA þ ]/ [NAA–NAAG], which was tested here using the multi-TI pCASL approach and sequential measurement of [GABA þ ]/[NAA–NAAG]. Alternatively, there may be no correlation between these two parameters, however, Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

GABA and perfusion in visual cortex MJ Donahue et al

536

A

B

C

D

Figure 2. Region-of-interest procedure. The location of the g-aminobutyric acid (GABA) voxel in occipital cortex as recorded from the scanner in a representative subject. Orthogonal slices of (A) the T1-weighted image, (B) the voxel location overlaid on the T1-weighted image, (C) the gray matter within the voxel, and (D) the white matter within the voxel. The quantified cerebral blood flow (CBF) map was similarly coregistered to the T1-weighted scan. Care was taken to place the voxel similarly in all volunteers, and to avoid regions of cerebrospinal fluid (CSF) and the sagittal sinus. CBF quantified from single-TI pCASL data may show greater variability when uncorrected for AAT (e.g., a constant AAT assumption simply adds more uncertainty to the CBF measurement). Therefore, we repeated the above analyses when only the postlabeling delay ¼ 1,700 ms pCASL data were used for CBF quantification and the AAT was fixed at 0.8 second (calculated mean AAT over visual cortex).

RESULTS For the neurocognitive task, all subjects performed within a normal range on this task with an average Pr of 66.7±14.7 for hitsnovel false alarms and a Pr of 22.9±16.5 for hits-rearranged false alarms. Neurocognitive scores were interpreted by a clinical neuropsychologist (coauthor: BAA) and it was determined that the range of scores was typical for cognitively normal volunteers of this age group. Figure 1 shows the subject-averaged (A) pCASL difference images for different postlabeling delay times, whereby the changes in contrast between these postlabeling delay times are due to inflow of labeled blood water into the imaging slice. By Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

applying the general kinetic model (equation (1)), it is possible to calculate (B) AAT and (C) CBF. All AAT and CBF calculations were performed in the native space of the pCASL images. Figure 2 shows a typical location for the MRS voxel. This region was used for analysis of both [GABA þ ]/[NAA–NAAG] (from the MRS data) and CBF (from the pCASL MRI data). The fractional gray matter and white matter within the voxel are shown as well (B–D). Gray-matter fraction within the voxel was found to be 36.3±2.0%. Therefore, the analysis region contains substantial contributions from white matter, and, to a lesser extent, cerebrospinal fluid. This gray-matter fraction is smaller than is typically reported in fMRI voxels (70% to 90%) due to the much larger size of the MRS voxel relative to typical fMRI voxel dimensions (e.g., 3  3  3 mm). However, the range of variability in gray-matter concentration was small (±2.0%), suggesting that bias from partial volume effects was similar in all volunteers. Figure 3 shows representative J-edited spectra from two subjects (A), along with the LCModel fit of these two spectra (B). In both cases, the GABA doublet at B3.0 p.p.m. is clearly visible; all & 2014 ISCBFM

GABA and perfusion in visual cortex MJ Donahue et al

537

A

Subject 7 Glx

4.0

Subject 8 GABA

2.0

3.0 Chemical Shift (ppm)

B

4.0

Glx

GABA

2.0

3.0 Chemical Shift (ppm)

C

Subject 7

70

Subject 8

0 CBF (ml/100g/min)

Figure 3. g-Aminobutyric acid (GABA) and arterial spin labeling (ASL). (A) Spectra (A) and LCModel fit (B) results from two representative subjects, showing the GABA doublet at B3.0 p.p.m. (C) Orthogonal slice representations of the cerebral blood flow (CBF) maps for the same two subjects, which show lower CBF in the occipital cortex of the subject with higher GABA (Subject 8) and higher CBF in the occipital cortex of the subject with lower GABA (Subject 7). The units of the color bar are mL blood/100 g tissue per minute.

volunteers were required to have CRLBsp15% and therefore the quality of spectra from other subjects is similar to the quality of these presented cases. It is clear that Subject 7 shows a reduced GABA peak relative to Subject 8, and this volunteer was found also to have higher occipital CBF (C). Figure 4 shows scatter plots from all subjects and data for (A) CBF (multi-TI) versus [GABA þ ]/[NAA–NAAG] (R ¼  0.46; P ¼ 0.037), (B) CBF (single-TI) versus [GABA þ ]/[NAA–NAAG] (R ¼  0.12; P ¼ 0.33), and (C) AAT versus [GABA þ ]/[NAA–NAAG] (R ¼  0.13; P ¼ 0.32). It is observed that when accounting for blood transit time variations in the multi-TI pCASL fitting procedure, a significant albeit weak inverse relationship is found between CBF and [GABA þ ]/[NAA–NAAG]. This relationship disappears when the AAT is assumed to be constant in all & 2014 ISCBFM

volunteers, and finally the AAT itself does not appear to correlate with [GABA þ ]/[NAA–NAAG]. Figure 5 shows the results of the multiple regression analyses whereby multi-TI CBF (dependent variable) is investigated relative to [GABA þ ]/[NAA–NAAG], gray-matter voxel fraction, and AAT for the multi-TI fitting procedure. The adjusted R2 is found to be 0.61 (F ¼ 6.2), with most of the variability explained by [GABA þ ]/[NAA– NAAG] (P ¼ 0.096) and AAT (P ¼ 0.005), but not gray-matter voxel fraction (P ¼ 0.361). We also performed a supplementary multiple regression analysis whereby [GABA þ ]/[NAA–NAAG] was the dependent variable. In this case, adjusted R2 is found to be 0.27 (F ¼ 1.5), with most of the variability explained by CBF (P ¼ 0.096), but very little by AAT (P ¼ 0.825) and gray-matter voxel fraction (P ¼ 0.770). These collective findings lend support for, over the Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

GABA and perfusion in visual cortex MJ Donahue et al

538

A

whether [GABA þ ]/[NAA–NAAG] values were similar to [GABA þ ]/ [Cr] values. It was found that [GABA þ ]/[NAA–NAAG] and [GABA þ ]/[Cr] correlated extremely tightly (Po0.01) and therefore choice of the normalizing metabolite is unlikely to drive the observed correlations. Table 1 provides CBF, AAT, and gray-matter concentration for all volunteers.

Multi-TI CBF (ml/100g/min)

120.0 100.0 80.0 60.0 40.0 20.0

N = 16 R = -0.46 P = 0.037

0.0 -0.1

B

0.1 0.3 [GABA+]/[NAA-NAAG]

Single-TI CBF (ml/100g/min)

120.0 100.0 80.0

0.5

N = 16 R = -0.13 P = 0.32

60.0 40.0 20.0 0.0

-0.1

C

0.1 0.3 [GABA+]/[NAA-NAAG]

0.5

Arterial Arrival Time (sec)

1.20 1.00 0.80 0.60 0.40 0.20

N = 16 R = -0.12 P = 0.33

0.00 -0.1

0.1 0.3 [GABA+]/[NAA-NAAG]

0.5

Figure 4. Scatter plots describing the relationships between [GABA þ ]/[NAA–NAAG] and the hemodynamic parameters. (A) A significant (P ¼ 0.037) inverse relationship is observed between the occipital [GABA þ ]/[NAA–NAAG] and CBF as calculated from the multiinversion time (multi-TI) fitting procedure. (B) When only a singleTI point is used (postlabeling delay ¼ 1,700 ms) with an assumed, constant arterial arrival time (AAT) ¼ 0.8 second, this trend disappears. (C) No significant relationship is found between [GABA þ ]/[NAA– NAAG] and AAT. These data demonstrate that inaccurate assumptions regarding the AAT can distort the detection of relationships between CBF and [GABA þ ]/[NAA–NAAG]. CBF, cerebral blood flow; GABA, gaminobutyric acid; NAA, N-acetyl aspartate; NAAG, N-acetyl aspartyl glutamic acid.

small range of gray-matter fractional variability, most variability in CBF is explained by (of our measured parameters) variability in [GABA þ ]/[NAA–NAAG]. Finally, we performed multiple regression analyses between [GABA þ ]/[NAA–NAAG] and CBF and gray-matter fraction, separately for the CBF calculated from the multi-TI versus single-TI fitting procedure. Previous correlative studies between GABA and CBF have only utilized single-TI procedures, which may be suboptimal when transit time varies between volunteers. In this regression, AAT was not included as a regressor as this metric is only available in the multi-TI fitting procedure (see Figure 5). Here as well, the multi-TI CBF described the [GABA þ ]/[NAA–NAAG] best (R ¼ 0.43), with only a very weak correlation (R ¼ 0.32) between the single-TI CBF and [GABA þ ]/[NAA–NAAG], and in neither approach did gray-matter fraction contribute significantly to the outcome (P40.05). In this study we chose to normalize [GABA þ ] by [NAA–NAAG], however a popular choice is also [Cr]. Therefore, we also evaluated Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

DISCUSSION The primary finding of this work is that measurements of occipital GABA ([GABA þ ]/[NAA–NAAG]) using J-edited MEGA-PRESS spectroscopy inversely correlate with CBF in the same region. This finding provides some physiologic basis for the recently reported inverse relationships between evoked BOLD responses and GABA. A secondary finding, which is important for explaining the prior inconsistencies between ASL-measured CBF and GABA, is that while blood arrival times do not appear to correlate strongly with [GABA þ ]/[NAA–NAAG], failure to account for arrival time in measurements of CBF can reduce or even eliminate the detectability of relationships between CBF and GABA. It is important to consider these findings in the context of the AAT measurement. Recent work has performed multiecho ASL measurements and quantified T2, which is dependent on vascular compartment, to show that following spin labeling in feeding cervical vessels, blood water leaves the vasculature and enters tissue in less than 2 seconds.51 Importantly, in ASL measurements where only a single postlabeling delay time is sampled (which is common given time constraints in many applications), it is not possible to measure the time that the blood water bolus reaches the imaging slice and rather a common value for this parameter is assumed. This metric has been used as a marker in steno-occlusive cerebrovascular disease52 and has been implicated in memory disorders,41 where blood velocity is reduced secondary to vascular stenosis, or alternatively, arrival time is increased owing to increased collateralization. Here, we observe that by not accounting for intersubject AAT variability, the relationship between ASL-measured CBF and [GABA þ ]/[NAA–NAAG] is considerably weakened. It is also important to note that the AAT measure is similar, yet fundamentally different from more common measures of blood arrival time using dynamic susceptibility contrast MRI with exogenous paramagnetic contrast agents. Here, either time-to-peak (time for the tracer to maximally attenuate signal in a voxel) or mean transit time (ratio of CBV/CBF or approximate time for tracer to traverse the capillary bed) is reported. Recent reports53,54 have reviewed quantitative comparisons of ASL and dynamic susceptibility contrast in the context of different patient and healthy populations. g-Aminobutyric acid has recently been found to correlate with a range of behavioral and neuroimaging metrics,35–37,55,56 which has greatly increased interest in measuring GABA to better understand both group level differences in performance and functional imaging signals. To accurately measure the relationship between CBF and GABA in this study, we therefore controlled for sex, age (within one decade), and cognitive performance in our volunteers, and additionally implemented multi-TI pCASL with increased sensitivity to blood AAT to understand how arrival time may contribute to GABA in a voxel. Finally, we performed multiple regression analyses to understand the extent to which different hemodynamic metrics and quantification procedures correlate with GABA. Therefore, we believe that this study provides important information about the relationship between basal GABA and CBF in vivo. It was unexpected to find that the [GABA þ ]/[NAA–NAAG] measurement correlated only very weakly with the amount of gray matter in the voxel, as it is well known that GABA levels are higher in cortex than in white matter. A similar finding has been reported previously by other groups as well.36 The likely reason for & 2014 ISCBFM

GABA and perfusion in visual cortex MJ Donahue et al

539 100.0

100.0 Mean CBF (ml/100g/min)

Mean CBF (ml/100g/min)

A

80.0 60.0 40.0 20.0 0.0 0.1

0.2 0.3 [GABA+]/[NAA-NAAG]

80.0 60.0 40.0 20.0

0.50

0.70 Mean AAT (s)

0.90

1.10

80 60 40 20 0

0.4

20

40 60 Sample Percentile

80

100

0.4 [GABA+]/ [NAA-NAAG]

[GABA+]/ [NAA-NAAG]

20.0

0

0.3 0.2 0.1 0 0.0

20.0

40.0 60.0 Mean CBF (ml/100g/min)

80.0

0.2 0.1 0 0.320 0.330 0.340 0.350 0.360 0.370 0.380 0.390 0.400 Gray Matter Fraction

0.2 0.1 0.50

0.70 Mean AAT (s)

0.90

1.10

0.4 [GABA+]/ [NAA-NAAG]

0.3

0.3

0 0.30

100.0

0.4 [GABA+]/ [NAA-NAAG]

40.0

100

100.0

0.0 0.320 0.330 0.340 0.350 0.360 0.370 0.380 0.390 0.400 Gray Matter Fraction

B

60.0

0.0 0.30

0.4

Mean CBF (ml/100g/min)

Mean CBF (ml/100g/min)

0

80.0

0.3 0.2 0.1 0 0

20

40 60 Sample Percentile

80

100

Figure 5. Multiple regression performed on the multiinversion time (multi-TI) imaging and spectroscopy data. Data points are shown along with the multiple-regression predicted-model fit (dashed line). (A) Results of the multiple regression analyses whereby multi-TI cerebral blood flow (CBF) (dependent variable) is investigated relative to [GABA þ ]/[NAA–NAAG], gray-matter voxel fraction, and arterial arrival time (AAT) for the multi-TI fitting procedure. The adjusted R2 is found to be 0.61 (F ¼ 6.2), with most of the variability explained by [GABA þ ]/[NAA–NAAG] (P ¼ 0.096) and AAT (P ¼ 0.005), but not gray-matter voxel fraction (P ¼ 0.361). (B) We also performed a supplementary multiple regression analysis whereby [GABA þ ]/[NAA–NAAG] was the dependent variable. In this case, adjusted R2 is found to be 0.27 (F ¼ 1.5), with most of the variability explained by CBF (P ¼ 0.096), but very little by AAT (P ¼ 0.825) and gray-matter voxel fraction (P ¼ 0.770). These collective findings lend support for, over the small range of gray matter fractional variability, most variability in CBF is explained by (of our measured parameters) variability in [GABA þ ]/[NAA–NAAG]. GABA, g-aminobutyric acid; NAA, N-acetyl aspartate; NAAG, N-acetyl aspartyl glutamic acid.

this pertains to the small range of variability in gray-matter fractions across volunteers, which simply does not supply sufficient range to observe this finding. It is likely that studies with larger voxel volumes, placed less consistently, may reveal this trend. Regarding the phase correction procedure implemented in processing the MEGA-PRESS spectra, a recently published study has confirmed that alignment reduces variance and demonstrates that the quantification procedure should include additional corrections for optimal accuracy.57 In this work, a differential phase between odd and even spectra is used to approximate the differential phase; however, an optimal method would be to use prior knowledge for the correction. Optimization of such correction procedures in future work is warranted. These findings should also be considered in light of their potential biologic and clinical relevance to studies of behavioral control disorders, as well as poststroke plasticity mechanisms. First, it is well known that excitation-inhibition imbalance is implicated in a range of neuropsychological disorders. In practice, measuring CBF, with magnetic resonance, single photon emission computed tomography, or positron emission tomography, is considerably simpler than measuring GABA, and therefore it & 2014 ISCBFM

would be convenient if CBF represented a surrogate marker of GABA. While our results lend support for an inverse relationship between GABA and CBF, even in our strict volunteer cohort, this trend was not of sufficient statistical strength (P ¼ 0.037) to justify CBF being a direct indicator of GABA concentration. This is not unexpected, and furthermore suggests that multiple additional mechanisms likely reflect the influence of GABA on hemodynamics in a voxel. One likely contributor is the cerebral metabolic rate of oxygen metabolism, which was not measured here and in practice is extremely difficult to measure reliably using MRI. It is likely that it is not the GABAergic voxels specifically that drive the CBF relationships, but rather the ratio of GABAergic to glutamatergic activity in a voxel, and these relationships will vary regionally throughout the brain. In this study, we additionally performed analysis of [GABA þ ]/[Glx], whereby Glx represents the measured glutamate and glutamine signal in a voxel; however, we found no different trends relative to what was presented here ([GABA þ ]/[NAA–NAAG] as an independent variable). The likely reason for this pertains to the varied role of glutamate, as a neurotransmitter but also as a protein amino acid, its varied metabolic roles, and its role as a precursor to GABA. Therefore, Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

GABA and perfusion in visual cortex MJ Donahue et al

540 Table 1.

Measurements in all volunteers

Subject

1 (106 776) 2 (106 794) 3 (106 993) 4 (107 092) 5 (107 640) 6 (107 710) 7 (108 052) 8 (108 095) 9 (108 311) 10 (108 281) 11 (108 369) 12 (108 997) 13 (109 259) 14 (109 313) 15 (109 343) 16 (107 244) Mean STD

Age (years)

23 23 23 21 25 24 23 29 19 26 21 23 24 24 24 19 23 3

Graymatter fraction

0.394 0.355 0.369 0.385 0.355 0.379 0.332 0.341 0.334 0.354 0.363 0.335 0.377 0.381 0.383 0.379 0.363 0.020

[GABA þ ]/ [NAA– NAAG]

0.196 0.127 0.159 0.164 0.193 0.211 0.176 0.334 0.271 0.255 0.333 0.204 0.219 0.267 0.248 0.065 0.214 0.071

[GABA þ ]/ [NAA– NAAG] CRLB

0.04 0.05 0.06 0.04 0.04 0.06 0.05 0.03 0.03 0.06 0.03 0.06 0.04 0.04 0.03 0.15 0.051 0.029

Multi-TI pCASL

Multi-TI pCASL

Multi-TI pCASL

Multi-TI pCASL

Single-TI pCASL

Single-TI pCASL

Mean CBF (mL/100 g per minute)

STD CBF (mL/100 g per minute)

Mean AAT (s)

STD AAT (s)

Mean CBF (mL/100 g per minute)

STD CBF (mL/100 g per minute)

60.5 73.9 52.3 78.9 40.3 94.8 75.5 34.4 33.4 72.4 51.3 65.3 78.2 65.8 15.3 72.7 60.3 21.0

10.0 17.3 11.5 15.7 11.4 20.6 14.7 10.0 8.2 18.1 15.3 12.2 14.3 10.8 6.5 11.0 13.0 3.8

0.81 0.72 0.77 0.86 0.74 0.87 0.91 0.76 0.86 0.91 0.60 0.81 1.03 0.95 0.42 0.77 0.80 0.14

0.17 0.22 0.16 0.16 0.21 0.23 0.19 0.26 0.19 0.27 0.28 0.14 0.20 0.15 0.27 0.11 0.20 0.05

36.0 63.3 56.3 80.2 47.0 80.1 73.3 44.2 46.4 54.6 66.1 53.1 62.7 85.2 14.2 55.4 57.4 18.2

6.6 19.4 17.1 17.8 12.4 16.2 14.0 10.8 10.0 13.6 23.3 12.9 12.9 10.8 7.4 11.8 13.6 4.4

TI, inversion time; pCASL, pseudo-continuous arterial spin labeling; AAT, arterial arrival time; CBF, cerebral blood flow; STD, standard deviation; CRLB, CramerRao Lower Bound; GABA, g-aminobutyric acid; NAA, N-acetyl aspartate; NAAG, N-acetyl aspartyl glutamic acid.

unlike GABA whose role is largely that of an inhibitory neurotransmitter, glutamate is much less specific and until the neurotransmitter pool of glutamate can be isolated specifically (or manipulated pharmacologically), MRS measurements of glutamate for purposes of increasing BOLD interpretability may not be particularly telling. Animal studies have suggested direct roles of neurotransmitters after postischemic injury, and the relevance of poststroke GABA levels in humans has been suggested using pharmacological manipulation of GABA.58 After stroke, there is a decrease in inhibitory activity, and some data have shown that in wellrecovered patients increasing inhibition through GABAA agonists causes prior stroke symptoms reemerge.58 Reduced GABA in M1 after ischemic stroke, and a correlation between GABA reduction and functional recovery has recently been reported.59 Therefore, regional inhibition and corresponding changes in CBF could be a biomarker of functional outcomes. As additional information becomes available regarding the quantitative pathways in which these parameters are related, along with improved imaging approaches for measuring these parameters quickly and reliably, it may be possible to improve patient selection for neurorehabilitative therapies based on the coordinated neurochemical and hemodynamic phenomena. Furthermore, many stroke studies of inhibition have measured cortical excitability using transcranial magnetic stimulation, yet transcranial magnetic stimulation does not always elicit a motor evoked potential in stroke patients. Therefore, MRS offers a new perspective on understanding interhemispheric inhibition after stroke. In terms of functional imaging, it is important to note that BOLD signals are only indirect markers of neuronal activity. Recent correlative findings among neurotransmission, behavior, and BOLD responses are being presented more frequently, yet few efforts have been made to our knowledge to describe these relationships as more than qualitative correlations. Therefore, as more information regarding correlative relationships between in vivo neurotransmitter levels and hemodynamic metrics become available, it will be important to incorporate these observations to propose a quantitative framework that can be used to interrogate neurochemical and hemodynamic reactivity relationships. Recent reports have outlined quantitative frameworks relating neurochemical balance within a voxel to hemodynamic and BOLD fMRI Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541

reactivity,60 however, detailed experimental validation of these models have not yet been conducted. In conclusion, we measured basal GABA and CBF in vivo in human visual cortex in a controlled cohort of young adult males using MEGA-PRESS MRS and multi-TI pCASL in sequence. Findings indicate CBF, but not AAT, inversely correlates with GABA in occipital cortex. This finding reinforces the credibility of recently measured inverse correlations between evoked BOLD functional imaging signals and inhibition, and additionally reconciles recent work whereby no or paradoxical correlations were observed between measurements of CBF-weighted ASL contrast and baseline GABA. DISCLOSURE/CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGMENTS The authors are grateful to David Pennell, Leslie McIntosh, Kristen George-Durrett, Donna Butler, Kevin Wilson, and Chuck Nockowski for experimental support.

REFERENCES 1 Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci USA 1990; 87: 9868–9872. 2 Donahue MJ, Blicher JU, Ostergaard L, Feinberg DA, MacIntosh BJ, Miller KL et al. Cerebral blood flow, blood volume, and oxygen metabolism dynamics in human visual and motor cortex as measured by whole-brain multi-modal magnetic resonance imaging. J Cereb Blood Flow Metab 2009; 29: 1856–1866. 3 Lu H, Xu F, Grgac K, Liu P, Qin Q, van Zijl P. Calibration and validation of TRUST MRI for the estimation of cerebral blood oxygenation. Magn Reson Med 2012; 67: 42–49. 4 Lu H, Golay X, Pekar JJ, Van Zijl PC. Sustained poststimulus elevation in cerebral oxygen utilization after vascular recovery. J Cereb Blood Flow Metab 2004; 24: 764–770. 5 Gu H, Lu H, Ye FQ, Stein EA, Yang Y. Noninvasive quantification of cerebral blood volume in humans during functional activation. Neuroimage 2006; 30: 377–387. 6 Hua J, Qin Q, Donahue MJ, Zhou J, Pekar JJ, van Zijl PC. Inflow-based vascularspace-occupancy (iVASO) MRI. Magn Reson Med 2011; 66: 40–56. 7 Buzsaki G, Kaila K, Raichle M. Inhibition and brain work. Neuron 2007; 56: 771–783. 8 Hyder F, Patel AB, Gjedde A, Rothman DL, Behar KL, Shulman RG. Neuronal-glial glucose oxidation and glutamatergic-GABAergic function. J Cereb Blood Flow Metab 2006; 26: 865–877.

& 2014 ISCBFM

GABA and perfusion in visual cortex MJ Donahue et al

541 9 Sibson NR, Dhankhar A, Mason GF, Behar KL, Rothman DL, Shulman RG. In vivo 13C NMR measurements of cerebral glutamine synthesis as evidence for glutamate-glutamine cycling. Proc Natl Acad Sci USA 1997; 94: 2699–2704. 10 Patel AB, de Graaf RA, Mason GF, Kanamatsu T, Rothman DL, Shulman RG et al. Glutamatergic neurotransmission and neuronal glucose oxidation are coupled during intense neuronal activation. J Cereb Blood Flow Metab 2004; 24: 972–985. 11 Oz G, Berkich DA, Henry PG, Xu Y, LaNoue K, Hutson SM et al. Neuroglial metabolism in the awake rat brain: CO2 fixation increases with brain activity. J Neurosci 2004; 24: 11273–11279. 12 Shen J, Petersen KF, Behar KL, Brown P, Nixon TW, Mason GF et al. Determination of the rate of the glutamate/glutamine cycle in the human brain by in vivo 13C NMR. Proc Natl Acad Sci U S A 1999; 96: 8235–8240. 13 Sibson NR, Dhankhar A, Mason GF, Rothman DL, Behar KL, Shulman RG. Stoichiometric coupling of brain glucose metabolism and glutamatergic neuronal activity. Proc Natl Acad Sci USA 1998; 95: 316–321. 14 Choi IY, Lei H, Gruetter R. Effect of deep pentobarbital anesthesia on neurotransmitter metabolism in vivo: on the correlation of total glucose consumption with glutamatergic action. J Cereb Blood Flow Metab 2002; 22: 1343–1351. 15 de Graaf RA, Mason GF, Patel AB, Rothman DL, Behar KL. Regional glucose metabolism and glutamatergic neurotransmission in rat brain in vivo. Proc Natl Acad Sci USA 2004; 101: 12700–12705. 16 Patel AB, de Graaf RA, Mason GF, Rothman DL, Shulman RG, Behar KL. The contribution of GABA to glutamate/glutamine cycling and energy metabolism in the rat cortex in vivo. Proc Natl Acad Sci USA 2005; 102: 5588–5593. 17 Attwell D, Laughlin SB. An energy budget for signaling in the grey matter of the brain. J Cereb Blood Flow Metab 2001; 21: 1133–1145. 18 Hyder F, Rothman DL, Bennett MR. Cortical energy demands of signaling and nonsignaling components in brain are conserved across mammalian species and activity levels. Proc Natl Acad Sci USA 2013; 110: 3549–3554. 19 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; 96: 9403–9408. 20 Malonek D, Grinvald A. Interactions between electrical activity and cortical microcirculation revealed by imaging spectroscopy: implications for functional brain mapping. Science 1996; 272: 551–554. 21 Niwa K, Araki E, Morham SG, Ross ME, Iadecola C. Cyclooxygenase-2 contributes to functional hyperemia in whisker-barrel cortex. J Neurosci 2000; 20: 763–770. 22 Kaufmann WE, Worley PF, Pegg J, Bremer M, Isakson P. COX-2, a synaptically induced enzyme, is expressed by excitatory neurons at postsynaptic sites in rat cerebral cortex. Proc Natl Acad Sci USA 1996; 93: 2317–2321. 23 Fergus A, Lee KS. GABAergic regulation of cerebral microvascular tone in the rat. J Cereb Blood Flow Metab 1997; 17: 992–1003. 24 Li J, Iadecola C. Nitric oxide and adenosine mediate vasodilation during functional activation in cerebellar cortex. Neuropharmacology 1994; 33: 1453–1461. 25 Waldvogel D, van Gelderen P, Muellbacher W, Ziemann U, Immisch I, Hallett M. The relative metabolic demand of inhibition and excitation. Nature 2000; 406: 995–998. 26 Chatton JY, Pellerin L, Magistretti PJ. GABA uptake into astrocytes is not associated with significant metabolic cost: implications for brain imaging of inhibitory transmission. Proc Natl Acad Sci USA 2003; 100: 12456–12461. 27 Magistretti PJ, Pellerin L. Cellular mechanisms of brain energy metabolism and their relevance to functional brain imaging. Philos Trans R Soc Lond B Biol Sci 1999; 354: 1155–1163. 28 Attwell D, Iadecola C. The neural basis of functional brain imaging signals. Trends Neurosci 2002; 25: 621–625. 29 Chen Z, Silva AC, Yang J, Shen J. Elevated endogenous GABA level correlates with decreased fMRI signals in the rat brain during acute inhibition of GABA transaminase. J Neurosci Res 2005; 79: 383–391. 30 Perry TL, Hansen S, Gandham SS. Postmortem changes of amino compounds in human and rat brain. J Neurochem 1981; 36: 406–410. 31 Mescher M, Merkle H, Kirsch J, Garwood M, Gruetter R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 1998; 11: 266–272. 32 Edden RA, Barker PB. Spatial effects in the detection of gamma-aminobutyric acid: improved sensitivity at high fields using inner volume saturation. Magn Reson Med 2007; 58: 1276–1282. 33 Near J, Evans CJ, Puts NA, Barker PB, Edden RA. J-difference editing of gammaaminobutyric acid (GABA): Simulated and experimental multiplet patterns. Magn Reson Med 2012; 70: 1183–1191. 34 Zhu H, Edden RA, Ouwerkerk R, Barker PB. High resolution spectroscopic imaging of GABA at 3 Tesla. Magn Reson Med 2011; 65: 603–609. 35 Donahue MJ, Near J, Blicher JU, Jezzard P. Baseline GABA concentration and fMRI response. Neuroimage 2010; 53: 392–398.

& 2014 ISCBFM

36 Muthukumaraswamy SD, Edden RA, Jones DK, Swettenham JB, Singh KD. Resting GABA concentration predicts peak gamma frequency and fMRI amplitude in response to visual stimulation in humans. Proc Natl Acad Sci USA 2009; 106: 8356–8361. 37 Northoff G, Walter M, Schulte RF, Beck J, Dydak U, Henning A et al. GABA concentrations in the human anterior cingulate cortex predict negative BOLD responses in fMRI. Nat Neurosci 2007; 10: 1515–1517. 38 Muthukumaraswamy SD, Evans CJ, Edden RA, Wise RG, Singh KD. Individual variability in the shape and amplitude of the BOLD-HRF correlates with endogenous GABAergic inhibition. Hum Brain Mapp 2012; 33: 455–465. 39 Dickerson BC, Salat DH, Greve DN, Chua EF, Rand-Giovannetti E, Rentz DM et al. Increased hippocampal activation in mild cognitive impairment compared to normal aging and AD. Neurology 2005; 65: 404–411. 40 Harada M, Kubo H, Nose A, Nishitani H, Matsuda T. Measurement of variation in the human cerebral GABA level by in vivo MEGA-editing proton MR spectroscopy using a clinical 3 T instrument and its dependence on brain region and the female menstrual cycle. Hum Brain Mapp 2011; 32: 828–833. 41 Donahue MJ, Hussey E, Rane S, Wilson T, van Osch M, Hartkamp N et al. Vessel-encoded arterial spin labeling (VE-ASL) reveals elevated flow territory asymmetry in older adults with substandard verbal memory performance. J Magn Reson Imaging advance online publication, 30 April 2013; doi:10.1002/jmri.24150; e-pub ahead of print. 42 Rane S, Ally BA, Hussey E, Wilson T, Thornton-Wells T, Gore JC et al. Inverse correspondence between hippocampal perfusion and verbal memory performance in older adults. Hippocampus 2013; 23: 213–220. 43 Waddell KW, Zanjanipour P, Pradhan S, Xu L, Welch EB, Joers JM et al. Anterior cingulate and cerebellar GABA and Glu correlations measured by (1)H J-difference spectroscopy. Magn Reson Imaging 2011; 29: 19–24. 44 Waddell KW, Avison MJ, Joers JM, Gore JC. A practical guide to robust detection of GABA in human brain by J-difference spectroscopy at 3 T using a standard volume coil. Magn Reson Imaging 2007; 25: 1032–1038. 45 Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging 2001; 20: 45–57. 46 Buxton RB, Frank LR, Wong EC, Siewert B, Warach S, Edelman RR. A general kinetic model for quantitative perfusion imaging with arterial spin labeling. Magn Reson Med 1998; 40: 383–396. 47 Zhou J, Wilson DA, Ulatowski JA, Traystman RJ, van Zijl PC. Two-compartment exchange model for perfusion quantification using arterial spin tagging. J Cereb Blood Flow Metab 2001; 21: 440–455. 48 Wang J, Zhang Y, Wolf RL, Roc AC, Alsop DC, Detre JA. Amplitude-modulated continuous arterial spin-labeling 3.0-T perfusion MR imaging with a single coil: feasibility study. Radiology 2005; 235: 218–228. 49 Donahue MJ, Lu H, Jones CK, Pekar JJ, van Zijl PC. An account of the discrepancy between MRI and PET cerebral blood flow measures. A high-field MRI investigation. NMR Biomed 2006; 19: 1043–1054. 50 Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993; 30: 672–679. 51 Liu P, Uh J, Lu H. Determination of spin compartment in arterial spin labeling MRI. Magn Reson Med 2011; 65: 120–127. 52 MacIntosh BJ, Lindsay AC, Kylintireas I, Kuker W, Gunther M, Robson MD et al. Multiple inflow pulsed arterial spin-labeling reveals delays in the arterial arrival time in minor stroke and transient ischemic attack. AJNR Am J Neuroradiol 2010; 31: 1892–1894. 53 Donahue MJ, Strother MK, Hendrikse J, Novel MRI. approaches for assessing cerebral hemodynamics in ischemic cerebrovascular disease. Stroke 2012; 43: 903–915. 54 Wintermark M, Sesay M, Barbier E, Borbely K, Dillon WP, Eastwood JD et al. Comparative overview of brain perfusion imaging techniques. Stroke 2005; 36: e83–e99. 55 Sumner P, Edden RA, Bompas A, Evans CJ, Singh KD. More GABA, less distraction: a neurochemical predictor of motor decision speed. Nat Neurosci 2010; 13: 825–827. 56 Edden RA, Crocetti D, Zhu H, Gilbert DL, Mostofsky SH. Reduced GABA concentration in attention-deficit/hyperactivity disorder. Arch Gen Psychiatry 2012; 69: 750–753. 57 Evans CJ, Puts NA, Robson SE, Boy F, McGonigle DJ, Sumner P et al. Subtraction artifacts and frequency (Mis-)alignment in J-difference GABA editing. J Magn Reson Imaging 2012; 38: 970–975. 58 Lazar RM, Berman MF, Festa JR, Geller AE, Matejovsky TG, Marshall RS. GABAergic but not anti-cholinergic agents re-induce clinical deficits after stroke. J Neurol Sci 2010; 292: 72–76. 59 O’Shea J, Boudrias MH, Stagg CJ, Bachtiar V, Kischka U, Blicher JU et al. Predicting behavioural response to TDCS in chronic motor stroke. Neuroimage 2014; 85(Pt 3): 924–933. 60 Sotero RC, Trujillo-Barreto NJ. Modelling the role of excitatory and inhibitory neuronal activity in the generation of the BOLD signal. Neuroimage 2007; 35: 149–165.

Journal of Cerebral Blood Flow & Metabolism (2014), 532 – 541