Differences in steady-state glutamate levels and variability between ‘non-task-active’ conditions: Evidence from ¹H fMRS of the prefrontal cortex

Abstract

Proton functional magnetic resonance spectroscopy (¹H fMRS) is a noninvasive neuroimaging technique capable of detecting dynamic changes in glutamate related to task-related demands at a temporal resolution under 1 min. Several recent ¹H fMRS studies demonstrated elevated steady-state levels of glutamate of 2% or greater during different ‘task-active’ conditions, relative to a ‘non-task-active’ control condition. However, the ‘control’ condition from these studies does vary with respect to the degree of constraining behavior, which may lead to different glutamate levels or variability between ‘control’ conditions. The purpose of this ¹H fMRS study was to compare the steady-state levels and variability of glutamate in the left dorsolateral prefrontal cortex (dlPFC) of 16 healthy adults across four different putative ‘non-task-active’ conditions: relaxed with eyes closed, passive visual fixation crosshair, visual flashing checkerboard, and finger tapping. Results showed significantly lower glutamate levels during the passive visual fixation crosshair than the visual flashing checkerboard and the finger tapping conditions. Moreover, glutamate was significantly less variable during the passive visual fixation crosshair and the visual flashing checkerboard than the relaxed eyes closed condition. Of the four conditions, the passive visual fixation crosshair condition demonstrated the lowest and least variable glutamate levels potentially reflecting the least dlPFC engagement, but greatest behavioral constraint. These results emphasize the importance of selecting a proper ‘control’ condition to reflect accurately a ‘non-task-active’ steady-state level of glutamate with minimal variability during ¹H MRS investigations.

Introduction

Proton magnetic resonance spectroscopy (¹H MRS) facilitates non-invasive quantification of brain metabolite levels. Improved hardware, acquisition approaches, and higher B₀ field-strength magnets have spurred interest in investigating the temporal dynamics of task-modulated neurochemistry (Duncan et al., 2014a; Stanley and Raz, 2018 - pending revision). Specifically, glutamate, a key neurochemical of interest in ¹H fMRS, is the prominent excitatory neurotransmitter in the brain; ~80% of cortical neurons are glutamatergic (Shepherd and Oxford University Press. 2004). Additionally, glutamate acts as a metabolic substrate in the tricarboxylic acid (TCA) cycle, as a precursor for other metabolites (GABA and glutamine), and is important for ammonia detoxification (Cooper and Jeitner, 2016; Suarez et al., 2002). Recently, several ¹H fMRS studies demonstrated increased steady-state levels of glutamate in response to task-based stimulation related to the hippocampus and in sensory, motor, and prefrontal cortices (Lin et al., 2012; Mangia et al., 2007; Schaller et al., 2013; Schaller et al., 2014; Stanley et al., 2017; Woodcock et al., 2018 - under review). Table 1 provides a summary of ¹H fMRS studies reporting changes in glutamate with task. Task-related increase in glutamate levels relative to a ‘non-task-active’ comparison condition likely reflects increased glucose extraction and metabolism in brain regions associated with stimulation driven by task-related demands (Mangia et al., 2012; Sonnay et al., 2016). The compelling evidence demonstrates high sensitivity in the ability of ¹H fMRS to detect task engagement via glutamate modulation related to the functional specialization of the associated brain area under investigation. However, little attention is given to the selection of the ‘non-task-active’ comparison conditions, which differ greatly between ¹H fMRS studies (Table 1).

Table 1.

Summary of ¹H fMRS studies reporting task-related changes in glutamate.

Study	Sample Size	Acquisition Protocol	Task	Non-Task-Active Control Condition	Results
Mullins et al. (2005)	-12 adults	-4 T -STEAM TE = 20 ms -Bilateral anterior cingulate cortex -2 × 2 × 2 cm3	-Frozen compress (0–4 °C) or sham pain was applied to the base of the left foot -8:32min task epoch preceded by a rest block and followed by two 8:32min rest periods	-Sham condition without the pain stimuli -8:32min rest block length	-Increased glutamate (9 ± 6%) during pain condition vs rest condition
Mangia et al. (2007)	-12 adults	-7 T -STEAM TE = 6 ms -Midline visual cortex -2×2.2 × 2 cm3	-Radial red/black checkerboard covering the entire visual field (8 Hz) -Two protocols: 1) 2 short 5.3min blocks interspersed by rest epochs and 2) 1 long 10.6min block interspersed by rest epochs	-Dark background with small red fixation crosshair -Initial rest block duration 2:40min	-Increased glutamate (3%) during checkerboard vs rest
Gussew et al. (2010)	-6 adults	-3 T -PRESS TE = 30 ms -Left anterior insular cortex -2.5 × 1 × 1cm3	-Heat stimuli were applied to the inner skin area of the left forearm –2 blocks interspersed by rest epochs	-Absence of painful stimuli -Rest block duration 10:40min, which was acquired prior to task paradigm	-Increased glutamate (18 ± 8%) during heat vs rest
Lin et al. (2012)	-10 adults	-7 T -STEAM TE = 15 ms -Midline visual cortex -2×2×2cm3	-Visual stimulation included contrast-defined wedges, moving toward or away from the fixation crosshair and randomized -Two protocols: 1) one 13.2min block interspersed by rest epochs and 2) two 9.9min blocks interspersed by rest epochs	-Median gray background with small fixation crosshair -Either 6:36min or 9:54min rest block durations	-Increased glutamate (2 ± 1%) during single block vs rest -Increased glutamate (3 ± 1%) during the two blocks vs rest
Schaller et al. (2013)	-10 adults	-7 T -SPECIAL TE = 6 ms -Midline visual cortex -2×2.2 × 2cm3	-Reversed black–grey checkerboard (9 Hz) –2 blocks interspersed by rest epochs	-Black background -Each rest block 5:00min in duration	-Increased glutamate (4% ± 1%) during stimulation vs rest
Schaller et al. (2014)	-11 adults	-7 T -SPECIAL TE = 12 ms -Left motor and somatosensory cortices –1.7 × 2 × 1.7 cm3	-Cued finger-to-thumb tapping task with both hands at a frequency of 3 Hz -2 blocks interspersed by rest epochs	-Keep hands in same position and do not move fingers -2:00min initial rest block plus two 5:00min interspersed rest blocks in between two task blocks	-Increased glutamate (2 ± 1%) during finger tapping vs rest
Bednařík et al. (2015)	-12 adults	-7 T -semi-LASER TE = 26 ms -Midline visual cortex -2×2×2cm3	-Red-black checkerboard (7.5 Hz) -2 blocks interspersed by rest epochs	-Black background with small white fixation crosshair -3 rest blocks of 5:18min each	-Increased glutamate (~3%) during checkerboard vs rest
Taylor et al. (2015a)	-7 adults	-7 T -STEAM TE = 10 ms -Dorsal ACC -2 × 2 × 2cm3	-Stroop task with 4 conditions -One block flanked by rest epochs	-Fixation crosshair -One block of 4:00min in duration	-Increased glutamate (2.6 ± 1.0%) during Stroop vs rest
Taylor et al. (2015b)	-16 controls -16 MDD -16 Schizophrenia patients	-7 T -STEAM TE = 10 ms -Dorsal ACC -2×2×2cm3	-Stroop task with 4 conditions -Two blocks interspersed with rest epochs	-Fixation crosshair -One block of 4:00min in duration	-Increased glutamate (3.2%) in controls during 1st Stroop vs rest -Decreased glutamate in MDD during 2nd Stroop vs rest
Huang et al, 2015	-46 trained swimmers (all adults)	-3 T -PRESS TE = 24 ms -ACC (2 × 1.5 × 3cm3) and right primary auditory cortex (1.5 × 2.6 × 2.4 cm3)	-3 blocks: 1) rest #1, 2) mental imagery of swimming and 3) rest #2 -Eyes closed during the entire examination -Each block 12:48min	-Relaxed eyes closed	-Increased glutamate plus glutamine (Glx) in the anterior ACC during mental swim vs rest #1 and rest #2 -Not significant in auditory cortex Glutamate values were not reported -Glutamate values were not reported
Apsvalka et al. (2015)	-19 young adults	-3 T -PRESS TE = 105 ms -Left lateral occipital cortex -2×2×2cm3	-Three different task blocks: novel stimuli and two repeated (6 unique vs 4 unique) stimulus presentations interspersed with rest blocks -Presentation of novel/repeated black-line drawings representing real world objects for 700 ms -4 runs of 8 task blocks/run -each run, 4 novel and 4 repeated blocks -Each block 36sec in duration	-Fixation crosshair -8 rest blocks/run totaling 19:12min in duration	-Increased glutamate (~12%) during novel compared to both rest and repeated presentations
Jahng et al. (2016)	-23 young controls -24 older controls -24 MCI -24 probable Alzheimers disease patients	-3 T -PRESS TE = 35 ms -Precuneus/posterior cingulate -2.5 × 2.5 × 2.5 cm3	-Face-name association task -2 conditions interspersed by rest epochs: 1) novel encoding condition comprised of unfamiliar faces paired with fictional first names and 2) repeat encoding condition comprised of familiar faces paired with fictional first names -Each condition of 60s each ran twice	-Interspersed rest epochs of fixation crosshair (30s in duration for each)	-Independent of group, lower glutamate plus glutamine (Glx) in the novel compared to the repeat condition. -In MCI, increased Glx in the repeat condition compared to rest -In AD, decreased Glx in the novel condition compared to rest -No group differences in glutamate
Woodcock et al. (2017)	-16 healthy adults	-3 T -PRESS TE = 23 ms -Left dlPFC -1.5 × 2.0 × 1.5 cm3	-2-back working memory task -7 task blocks of 64s each interspersed by 32s rest epochs	-Fixation crosshair -One block of 3:28min in duration	-Increased glutamate (2.7%) during n-back vs fixation crosshair
Stanley et al. (2017)	-16 healthy adults	-3 T -PRESS TE = 23 ms -Right anterior hippocampus -1.7 × 3.0 × 1.2 cm3	-Associative learning and memory task -Epochs of encoding (9 unique object-location pairs) and cued-retrieval (of those associated memoranda) and interspersed with rest epochs -8 encoding-retrieval cycles were employed to allow learning to asymptote	-Motor finger tapping in response to a random visual stimulus -One block of 5:24min in duration	-Increased glutamate (5.2% and 4.2%) during both encoding and retrieval, respectively -Applying a median split based on learning proficiency, fast learners showed increased glutamate during the early encoding trials, whereas slow learners showed increased glutamate in the later encoding trials
Lindner et al. (2017)	-19 adults	-3 T -PRESS TE = 32 ms -Right or left border of parietal/occipital cortices -1.5 × 1.5 × 1.5 cm3	-Visuospatial attention task -Button press in response to the tilt orientation of the grating that appeared on the side of the screen cued by an arrow -3 conditions (ipsi, contra and control) randomized -3 blocks interspersed with rest epochs	-Neutral trails when the cued arrow was pointing up and subjects responded by pressing any button -96 interspersed control trials of 5s each totaling 8:00min	-No trial condition effect on glutamate

active’ comparison condition is expected to be inert with respect to the functional specialization of that region leading to relatively lower

The intent of the ‘task-active’ condition is to drive metabolic demand, and thus, modulate glutamate levels in the brain region of interest, based on the region's putative functional specialization. In contrast, the ‘non-task-steady-state and less variable glutamate levels. In sensory and motor cortices, the ‘non-task-active’ condition may appear obvious. For example, an optimal comparison condition for the investigation of visual stimulation in the occipital cortex is the absence of visual stimuli, such as a blank screen or blank screen with a crosshair (Table 1). However, the complexity of conducting ¹H fMRS in executive and associative cortices, such as the dorsolateral prefrontal cortex (dlPFC), is magnified. The dlPFC is responsible for numerous cognitive functions and is susceptible to modulation by intrinsic neural activity, such as intrusive thoughts or environmental distractions, that encompass the broader umbrella of executive functions (Miller and Cohen, 2001). Thus, the dlPFC is a particularly challenging region to establish a proper ‘non-task-active’ comparison condition for task-based ¹H fMRS studies. In a ¹H fMRS study, Woodcock et al. (2018 - under review) used a visual crosshair condition as the ‘non-task-active’ condition which preceded the N-back working memory task, because the interspersed rest epochs between the N-back blocks demonstrated greater variability in the dlPFC glutamate compared to the glutamate during the N-back blocks. Similarly, Stanley et al. (2017) used a finger tapping condition to contrast a hippocampal learning and memory task because the variability of glutamate in the hippocampus during rest epochs using a crosshair was much greater than the task epochs (unpublished results). Additionally, several ¹H fMRS studies have noted greater variability in glutamate during the ‘rest’ comparison condition than the ‘task-active’ condition (Ip et al., 2017; Lin et al., 2012; Mangia et al., 2007; Schaller et al., 2013; Schaller et al., 2014). Collectively, these observations suggest that the level of constraint on behavior imposed during these different ’non-task-active’ control conditions (e.g., relaxed eyes closed vs fixation on a crosshair) may impact the steady-state level of glutamate and its variability during acquisition. Thus, the focus of the present ¹H fMRS study is to assess the degree to which different control conditions truly reflect a ‘non-task-active’ state.

To address this issue, we conducted a series of ¹H fMRS experiments in the left dlPFC of healthy young adults using four different putative ‘non-task-active’ conditions: relaxed eyes closed, passive visual fixation crosshair, visual flashing checkerboard, and a finger tapping task. Steady-state glutamate levels and their variability during these conditions were quantified and compared. The dlPFC is a region of broad interest across psychiatric disorders and cognitive neuroscience. Additionally, these ‘non-task-active’ conditions were selected due to their lack of primary neural engagement associated with the functional specialization of the dlPFC. However, because of the variability in the level of behavioral engagement between conditions, we are hypothesizing significant differences in the steady-state level of glutamate and variability between conditions. We hypothesize the visual flashing checkerboard will exhibit lower and less variable steady-state levels of glutamate.

Materials and methods

Participants

The Wayne State University Institutional Review Board approved all study procedures. Sixteen right hand dominant adult volunteers narrowly selected for age (9 males; mean age ± SD: 24.3 ± 3.5 years old) were recruited. Participants reported no psychiatric or neurological disorders, MRI contraindications, or current use of psychoactive medications. Participants provided written informed consent and were compensated for their time.

Experimental conditions

All scans were completed in the morning (between 9:30 a.m. and 11:30 a.m.). The ¹H fMRS data were acquired during each of the four ‘non-task-active’ conditions. Subjects were given an explicit verbal description of each one of the four conditions prior to the scans. The four conditions were as follows: 1) relaxed eyes closed – subjects were instructed to relax and keep eyes closed; 2) passive visual fixation crosshair - subjects focused on a static, white crosshair projected in the center of the screen; 3) visual flashing checkerboard - subjects focused on a flashing grayscale checkerboard pattern (3 Hz) projected in the center of the screen; and 4) finger tapping - subjects pressed a button with their right index finger each time a white box (1 Hz) was projected in the center of the screen. The relaxed eyes closed and visual flashing checkerboard conditions were always the first and last to be administered, respectively; passive visual fixation crosshair and finger tapping were administered randomly (Fig. 1).

Fig. 1.

The four ¹H fMRS ‘control’ conditions acquired as separate scans from left to right include: relaxed with eyes closed, passive visual fixation crosshair, visual flashing checkerboard and motor finger tapping.

The acquisition order of the imaging protocol included acquiring T₁-weighted MRI images followed by setting up the ¹H fMRS (voxel placement and shimming) followed by acquiring the ¹H fMRS of the four conditions. The rationale for always starting with the “relaxed eyes closed condition”, which was the least constrained condition, was to ensure subjects were indeed relaxed without being confounded by thoughts from previous conditions as may be the case if the order was random. The visual checkerboard condition was always presented last because it was the first part of a longer two-part N-back working memory task (Woodcock et al., 2018 - under review). Brief breaks were given, and verbal instructions provided, before the start of the next condition.

¹H fMRS acquisition and processing

All scans were completed on a 3T Siemens Verio system with a 32-channel volume head-coil. T₁-weighted images were collected using the Magnetization Prepared Rapid Gradient Echo (MPRAGE) sequence (TR = 2.2s, TE = 3 ms, TI = 799 ms, flip angle= 13°, field-of-view (FOV) = 256 × 256 × 160 mm³, matrix= 256 × 256 × 160, 1 mm³ pixel resolution). Shimming of the B₀-field was initially conducted using FASTESTMAP (Tkac and Gruetter, 2005) and by specifying a slightly larger voxel covering the left dlPFC (25 × 25 × 25 mm³) because FASTESTMAP is ineffective when prescribing a voxel that is rotated or angulated. Finally, the T₁-weighted images were used to guide voxel placement (15 × 20 × 15 mm³; 45 mm³) in the left dlPFC (Brodmann Areas 45 and 46) using the automated voxel placement (AVP) approach (Woodcock et al., 2018 - pending revision) in 15 of 16 subjects (the one exception due to experimenter error).

¹H fMRS spectra were continuously acquired every 16s during each condition (PRESS with OVS and VAPOR, TE = 23 ms, TR =4.0s, 4 averages per measurement, 13 measurements per condition, bandwidth =2 kHz, 2048 complex data points; scan duration per condition= 3:28min). The scan duration for each condition was chosen to mimic the typical block length of control condition measurements observed in ¹H fMRS studies (Table 1) without being too lengthy to avoid fatigue and/or loss of task engagement. Prior to quantification, raw spectra, without water suppression, were phase-, shift-, and eddy current-corrected. After removing the first ¹H fMRS measurement for each condition to avoid data not meeting the steady-state condition, consecutive spectra were averaged to increase the signal-noise ratio, resulting in a 32s temporal resolution or 6 quantitated measurements per condition (representative spectrum shown in Fig. 2). LC Model (version 6.3) with a simulated basis set for a priori knowledge was used to quantify metabolite levels (Provencher, 1993). Additional details on the fully automated post-processing and quantification are provided by Stanley et al. (2017). Briefly, B₁-field correction and voxel tissue segmentation were done on the T₁-weighted structural images via FreeSurfer and FSL tools (Dale et al., 1999; Smith et al., 2004). Voxel tissue composition and unsuppressed water levels were used to calculate absolute glutamate concentration (Gasparovic et al., 2006).

Fig. 2.

Depiction of a typical ¹H MRS spectrum acquired with eight averages yielding a 32s temporal resolution. The acquired spectrum in black is superimposed on the LC Model fitted spectrum in red. Below includes the glutamate signal (Glu) in blue and the residual of the fitting in grey. The chemical shift axis in ppm is depicted at the bottom.

Statistical analyses

The two primary questions were to test whether glutamate levels and/or glutamate variability differed across conditions. The outcome variables of interest were glutamate levels and the percent coefficient of variation (% CV) from the 6 measurements for each condition. Other metabolites, N-acetyl-aspartate (NAA), phosphocreatine plus creatine (PCr + Cr), trimethylamines [glycerophosphocholine plus phosphocholine (GPC + PC)], and myo-inositol, were assessed as well. The main effect of condition was examined using repeated measures generalized estimating equations and post-hoc differences were evaluated using least square means (SAS GENMOD; SAS Institute Inc.). Given the exploratory nature of this pilot study, post-hoc pairwise comparisons were carried out even if the main condition effect was non-significant. Descriptive statistics are presented mean ± one standard deviation, unless otherwise noted. The threshold for significance was: p ≤ .05.

Results

Voxel overlap

Geometric voxel overlap indicated highly accurate (mean % overlap of each subjects’ voxel and the template voxel in template space = 92.3 ± 4.7%) and reliable (voxel overlap across all subjects = 88.6%) voxel placement across subjects (Fig. 3). Mean (±1 SD) voxel tissue composition was 36.8 ± 3.8% gray matter and 60.8 ± 4.5% white matter.

Fig. 3.

Voxel overlap across all subjects depicted in the template brain. Red-to-white color gradient represents percentage of voxel overlap across subjects (white = 100% voxel overlap; red = 0% voxel overlap).

LC model fit characteristics

The overall quality of the ¹H fMRS data representing a 32s temporal resolution was reasonable and consistent across conditions, with an overall mean S/N of 11.7 ± 2.0, FWHM of 4.8 ± 1.0 Hz, and CRLB for glutamate of 7 ± 1%, which did not significantly differ between conditions (Table 2).

Table 2.

LCModel fitting characteristics (values ± SEM).

	Non-Task-Active ‘Control’ Conditions
	Relaxed Eyes Closed	Fixation Crosshair	Flashing Checkerboard	Motor Finger Tapping	X2	p-value
FWHM in Hz	4.7 (0.8)	4.9 (1.1)	4.9 (0.9)	4.9 (1.0)	0.02	0.10
S/N	11.9 (1.9)	11.8 (2.0)	11.6 (1.9)	11.4 (2.1)	0.15	0.98
% CRLB of Glutamate	7 (1)	7 (1)	7 (1)	7 (1)	0.06	0.10

Glutamate levels across conditions

The overall mean (±1 SD) glutamate level for the four conditions; relaxed with eyes closed, passive visual fixation, finger tapping, and visual checkerboard, were 11.9 ± 1.1, 11.6 ± 0.9, 12.0 ± 0.9, and 12.2 ± 0.9 (mmol/kg wet weight), respectively. When analyzed for mean concentration differences across the four conditions, the condition term for glutamate failed to reach significance (X² = 5.09; p = 0.17). However, significance was revealed by the post-hoc comparisons; the glutamate level was significantly lower during the passive visual fixation crosshair (11.6 ± 0.9 mmol/kg wet weight) compared to both the visual flashing checkerboard (12.2 ± 0.9 mmol/kg wet weight; p = 0.0080) and the finger tapping (12.0 ± 0.9 mmol/kg wet weight; p = 0.023) conditions (Fig. 4). The condition term was not significantly different for the other metabolites, NAA, PCr + Cr, GPC + PC and myo-inositol.

Fig. 4.

Mean (±SEM) levels of glutamate based on the six 32s acquisitions per subjects for each condition and expressed in institutional units (IU). Significant differences from the post-hoc analyses between conditions are noted with ‘*’ (p = 0.023) and ‘**’ (p = 0.0080).

Glutamate variability across conditions

As with the analysis of the mean glutamate levels, fluctuation of the glutamate signal at a temporal resolution of 32s and expressed as a % CV failed to reach significance across conditions (X² = 6.60, p = 0.086) (Fig. 5). However, post-hoc analysis demonstrated significantly lower % CV for glutamate during the visual flashing checkerboard compared to both the relaxed eyes closed (p = 0.0037) and finger tapping (p = 0.035) conditions. In addition, glutamate levels during the passive visual fixation crosshair were significantly less variable than the relaxed eyes closed (p = 0.0087) (Fig. 6) condition.

Fig. 5.

Variability of glutamate across the six 32s acquisitions for the four ‘control’ conditions: relaxed eyes closed (A), passive visual fixation crosshair (b), visual flashing checkerboard (C) and motor finger tapping (D). Values are expressed as the mean % change (±SEM) relative to the mean glutamate from the six 32s acquisitions of each subjects.

Fig. 6.

The mean (±SEM) variability of glutamate expressed a % CV based on the six 32s acquisitions per subject for each condition. Significant differences from the post-hoc analyses between conditions are noted with ‘*’ (p = 0.035), ‘**’ (p = 0.0037) and ‘***’ (p = 0.0087).

Passive visual fixation crosshair and visual flashing checkerboard exhibited the least glutamate signal variability (lowest % CV) among the four conditions examined. Glutamate was the only metabolite to show significant differences in the % CV across conditions.

Discussion

The motivation of this study was to investigate whether differences in glutamate can be detected across different putative ‘non-task-active’ conditions, knowing that ¹H fMRS is sensitive to detecting glutamate modulation induced by neural engagement. Specifically, the steady-state level and variability of glutamate in the left dlPFC were assessed during different conditions of varying degree of behavioral/cognitive constraints (e.g., visual vs. no visual, motor vs. visual, task engaged vs. rest). These conditions were chosen specifically because the left dlPFC is not the dominant brain region engaged. Unexpectedly, steady-state glutamate level was the lowest during the passive visual fixation crosshair condition, while the visual flashing checkerboard and motor finger tapping conditions had significantly higher glutamate (by 4.7% and 3.2%, respectively) (Fig. 4). Moreover, the passive visual fixation crosshair and visual flashing checkerboard conditions produced the least variability in glutamate with CV values under 5%, which were both significantly lower than the eyes closed condition (mean CV = 6.7%; Figs. 5 and 6).

Increased glutamate modulation related to task-demand is consistent with new metabolic steady-states from the neural output that has been postulated to be driven by shifting the net glutamatergic excitatory and GABAergic inhibitory (E/I) balance towards greater excitability (Duncan et al., 2014a; Isaacson and Scanziani, 2011; Lauritzen et al., 2012; Maffei, 2017; Stanley and Raz, 2018 - pending revision; Tatti et al., 2017). Likewise, a relative decrease in the steady-state glutamate level is consistent with a net neural output reflecting a shift in the E/I balance towards a state of less excitability (or a greater inhibitory drive). This implies that the four ‘control’ conditions differed in their level of engagement of the left dlPFC by shifting the level of excitability (i.e., E/I drive). One may speculate that the level of directed attention or executive control exercised during these different conditions, which would modulate dlPFC engagement, was greatest for the visual flashing checkerboard, and to a lesser degree the finger tapping condition, relative to the passive visual fixation crosshair condition. Interestingly, the relatively most engaged condition, visual flashing checkerboard, produced the least glutamate variability, which is consistent within the framework. Counterintuitive to our hypothesis was the passive visual fixation crosshair condition, which revealed the lowest level of steady-state glutamate, as well as the second to least variable glutamate levels. These findings indicated the passive visual fixation condition was the most robust ‘non-task-active’ condition of the four tested herein. Lastly, the relaxed eyes closed condition, which is viewed as the least behaviorally constrained task (i.e., with the most ambiguous set of instructions), produced the largest variability.

This series of experiments exhibits the criticality of developing an appropriate ‘non-task-active’ comparison condition for task-based ¹H fMRS and provides evidence supporting a dependency on the level of constraint of behavior for the brain area of interest. For some brain regions, the ‘non-task-active’ state can be presumed to be the absence of any neural engagement. In sensory and motor cortices, this is the absence of task-related stimulation (e.g., blank screen for the visual cortex) (Lin et al., 2012; Mangia et al., 2007c; Schaller et al., 2013; Schaller et al., 2014). However, executive and associative brain regions, such as the dlPFC, which is a central node of the executive network, receiving, and processing a variety of sensory inputs for multiple cognitive functions, pose a greater challenge (Miller and Cohen, 2001).

The observation that the passive visual fixation crosshair condition induced the lowest steady-state glutamate level, in addition to low glutamate level variability, has important implications for both functional and non-functional ¹H MRS studies. Though, our findings are limited to the dlPFC, the conceptual framework of constraining behavior is generalizable across brain areas accordingly to their functional specialization. As noted above, several ¹H fMRS studies investigating different brain areas have reported greater variability in glutamate during the control condition relative to the ‘task-active’ condition, which illustrates the effect observed herein is not limited to the dlPFC (Ip et al., 2017; Lin et al., 2012; Mangia et al., 2007; Schaller et al., 2013; Schaller et al., 2014). Intuitively, minimizing glutamate variability, especially for ¹H fMRS experiments using block designs, will lead to greater statistical power in detecting significant differences between alternating periods of task-active and non-task-active. However, the majority of non-functional ¹H MRS studies are acquired without any specific instructions or behavioral constraints, aside from asking the participants to relax and keep their head still during acquisition. This is potentially problematic knowing that the steady-state glutamate levels may differ depending on whether behavior is being constrained (or not) during data collection, including the level of vigilance, anxiety or agitation, which can vary between participants or patient groups being studied. Also, the impact of participants sleeping during scans has shown differences in glutamine levels (Bartha et al., 1999). Collectively, this new knowledge illustrates the importance of constraining behavior during ¹H MRS to ensure that the steady-state level of glutamate best reflects the ‘non-task-active’ condition with minimal variability. Ultimately, further studies are warranted to confirm its impact on research investigating different disorders or conditions.

Not surprising were the negative results in the other tested metabolites: NAA, PCr + Cr, GPC + PC, and myo-inositol, which displayed no significant differences in steady-state levels nor % CV between the four conditions. This is consistent with the growing evidence from ¹H fMRS studies primarily showing task-related modulation is specifically limited to glutamate, GABA, lactate, aspartate and glucose (Duncan et al., 2014b; Stanley and Raz, 2018 - pending revision).

We do acknowledge that our explanation of different steady-state glutamate levels and variability between conditions reflecting differences in the level of engagement in the left dlPFC is broadly defined and lacks empirical support. Other limitations include the small sample size, which limited statistical power.

Conclusions

In all, this study demonstrated that putative ‘non-task-active’ control conditions require careful consideration for ¹H fMRS investigations. We found that steady-state glutamate levels, and the variability therein, differed across ‘control’ conditions. Our findings indicated that the passive visual fixation crosshair condition was the most optimal ‘non-task-active’ condition, of those tested herein, for investigations of glutamate levels in the dlPFC. This finding provides compelling evidence that glutamate levels are highly susceptible to the behavioral constraints of ‘non-task-active’ conditions.