Glutamate Sensitive Imaging and Evaluation of Cognitive Impairment in Multiple Sclerosis

Abstract

Background:

Cognitive impairment profoundly impacts quality of life for patients with multiple sclerosis (MS). Dysfunctional regulation of glutamate in gray matter has been implicated in the pathogenesis of MS by post-mortem pathological studies and in cognitive impairment by in vivo magnetic resonance spectroscopy, yet gray matter pathology is subtle and difficult to detect using conventional T₁- and T₂-weighted MRI. There is a need for high-resolution, clinically-accessible imaging techniques that probe molecular changes in gray matter.

Objective:

To study cortical gray matter pathology related to cognitive impairment in MS using glutamate-sensitive chemical exchange saturation transfer (CEST) MRI at 7T.

Methods:

Twenty patients with relapsing-remitting MS and twenty healthy controls underwent cognitive testing, anatomical imaging, and glutamate-sensitive CEST imaging. Glutamate-sensitive image contrast was quantified for cortical gray matter, compared between cohorts, and correlated with clinical measures of cognitive impairment.

Results and Conclusion:

Glutamate-sensitive contrast was significantly increased in the prefrontal cortex of MS patients with accumulated disability (p<0.05). Additionally, glutamate-sensitive contrast in the prefrontal cortex was significantly correlated with symbol digit modality test (r_S=−0.814) and choice reaction time (r_S=0.772) scores in patients (p<0.05), suggesting that glutamate-sensitive CEST MRI may have utility as a marker for gray matter pathology and cognitive impairment.

Introduction

Cognitive impairment (CI) is a significant symptom of multiple sclerosis (MS) affecting up to ~70% of patients.^{1, 2} CI is the strongest predictor of unemployment^{2, 3} and has profound effects on daily life as it hinders routine household tasks,² self-care,⁴ decision-making,² and the ability to drive safely.⁵ Despite its clinical relevance, the pathological substrate of CI remains unknown, and robust imaging markers for characterizing early biochemical changes and tissue damage underlying CI are lacking.

Magnetic resonance imaging (MRI) studies have investigated associations between CI and structural/functional MRI changes in white matter (WM) and gray matter (GM). While studies focusing on focal and diffuse WM pathologies showed their (partial) role in CI and normal-appearing WM (NAWM) and GM damage have also been implicated in CI,^{1, 6} these studies do not provide insight on molecular mechanisms leading to tissue injury and disability accretion. Early identification of alterations in these molecular pathways is a pressing need, so developing imaging strategies that probe NAWM and GM pathology is critical to further understanding the radiological manifestations of CI in MS.

Dysfunctional regulation of glutamate, the principal excitatory neurotransmitter in the brain, has been implicated in chronic neurodegenerative processes.⁷ In MS, glutamate-mediated excitotoxicity is secondary to increased production by macrophages and microglial cells and leads to neuronal death.^{8, 9} Thus far, magnetic resonance spectroscopy (MRS) has offered a means to non-invasively probe glutamate in MS patients in vivo.^10–12 However, while the importance of glutamate as a target for investigating metabolic changes underlying disease progression in MS is well-established, imaging studies offering large coverage and high resolution combined with CI assessment in MS are needed. These studies will permit investigating glutamate-induced pathology in the cortex, which is an important substrate of cognitive decline in patients.

Chemical exchange saturation transfer (CEST) MRI is a method that holds potential for investigating glutamate abnormalities in the cortex with higher resolution and larger anatomical coverage. CEST harnesses the solute-to-water proton exchange process to enhance detection of endogenous agents (e.g. proteins and neurotransmitters such as glutamate) with protons that resonate at specific offset frequencies from water.¹³ In CEST experiments, labile protons of mobile solutes are selectively saturated using narrow-bandwidth radiofrequency irradiation. Through direct chemical exchange, these saturated protons attenuate the water signal, allowing indirect detection of a low concentration, target solute.^{14, 15} The magnitude of water signal attenuation is related to exchangeable proton concentration of the solute, the exchange rate (k_ex), saturation pulse sequence design, tissue water concentration, and T₁ relaxation time.¹³ Recently, amide proton transfer (APT) CEST MRI sensitive to slowly-exchanging protons associated with amide groups found in protein/peptide backbones was applied to study the molecular composition of the brain and spinal cord in MS at 7.0 Tesla (7T).^{14, 16, 17} This 7T work provided the first CEST data for MS pathology in the brain and demonstrated that APT-CEST has potential as a biomarker of WM pathology.^{14, 17}

At 7T, glutamate-sensitive CEST (“GluCEST”) MRI has been applied in healthy human brains, in the GM of patients with psychosis, and the hippocampi of patients with epilepsy.^18–20 At 7T, Δω and spectral resolution are increased, improving the ability to selectively saturate metabolites of interest even at faster k_ex.²¹ Increases in B₀ also offer increased SNR, which can be utilized for higher resolution or decreased imaging time. These benefits of 7T are important for observing glutamate, which contains amine protons with intermediate/fast exchange rates.

On the premise that glutamate dysfunction is an important precursor of neurodegeneration and that cortical degeneration has a significant impact on CI, we applied GluCEST imaging at 7T to investigate cortical GM of patients with MS and examine the relationship between GluCEST MRI and clinical measures of CI.

Methods

Twenty patients with relapsing-remitting MS and 20 group age- and sex-matched healthy controls (HCs) were consecutively enrolled (Table 1). Patients underwent a clinical examination to rate physical disability using the Expanded Disability Status Scale (EDSS) score.²² All subjects underwent neuropsychological assessment (Table 2), and a brain MRI. An additional group of seven HCs was enrolled and scanned twice for Bland Altman repeatability analysis. All data were acquired under a protocol approved by the local institutional review board and signed, informed consent was obtained prior to the study.

Table 1:

Demographic information

	All MS Patients (n=20)	MS with EDSS > 0 (n=11)	Healthy Controls (n=20)
Females	15	8	15
Age, mean ± SD	37.8 ± 3.8 years	37.7 ± 4.4 years	34.5 ± 8.4 years
Years of Education, mean ± SD	15.6 ± 2.7 years	15.6 ± 2.4 years	17.3 ± 1.7 years
EDSS, median (range)	1.25 (0 – 6)	2.0 (1 – 6)	-
Disease Duration, mean ± SD	8.8 ± 6.6 years	10.5 ± 7.7 years	-

MS = multiple sclerosis; EDSS = Expanded Disability Status Scale

Table 2:

Cognitive tests administered to study participants.

Cognitive Test Name	Description
Symbol Digit Modalities Test1	Written format; assesses information processing speed
Paced Auditory Serial Addition Test2	3 second interval version; assesses auditory information processing speed, working memory, and calculation ability
Simple Reaction Time3	One stimulus; assesses information processing speed and motor speed
Choice Reaction Time3	Four stimuli; assesses information processing speed and motor speed
Brief Visuospatial Memory Test-Revised4	Evaluated for total learning and delayed recall; assesses visuospatial learning and memory
Buschke Selective Reminding Test5	Evaluated for total recall across 8 trials; assesses verbal learning and memory
Trail Making Tests A and B6	Paper format; assesses visual attention, cognitive flexibility, and executive functions
Global Deterioration Scale7	Rating scale (1-7) assesses the level of cognitive decline
Brief Cognitive Rating Scale8	Assesses cognitive abilities with an interview covering 8 axes: concentration and calculation, recent memory, remote memory, orientation, functioning and self-care, language, motoric functioning, mood and behavior

¹Smith A. Symbol digit modalities test: Manual. Los Angeles: Western Psychological Services. Manual, 1982.

²Forn C, Belenguer A, Parcet-Ibars MA, et al. Information-processing speed is the primary deficit underlying the poor performance of multiple sclerosis patients in the Paced Auditory Serial Addition Test (PASAT). J Clin Exp Neuropsychol 2008; 30: 789-796. DOI: 10.1080/13803390701779560.

³Reicker LI, Tombaugh TN, Walker L, et al. Reaction time: An alternative method for assessing the effects of multiple sclerosis on information processing speed. Arch Clin Neuropsychol 2007; 22: 655-664. DOI: 10.1016/j.acn.2007.04.008.

⁴Benedict R. Brief visuospatial memory test-revised: professional manual. Odessa, Florida: Psychological Assessment Resources. Inc, 1997.

⁵Buschke H and Fuld PA. Evaluating storage, retention, and retrieval in disordered memory and learning. Neurology 1974; 24: 1019-1025.

⁶Lezak MD. Neuropsychological Assessment. Oxford University Press, USA, 1995.

⁷Reisberg B, Ferris SH, de Leon MJ, et al. The Global Deterioration Scale for assessment of primary degenerative dementia. The American journal of psychiatry 1982; 139: 1136-1139. DOI: 10.1176/ajp.139.9.1136.

⁸Tsolaki M, Drevelegas A, Karachristianou S, et al. Correlation of dementia, neuropsychological and MRI findings in multiple sclerosis. Dementia 1994; 5: 48-52.

MRI experiments

All MR imaging was performed using a Philips Achieva 7.0T MR Scanner (Philips Healthcare, The Netherlands) with 2 channel volume transmit and 32-channel receive head coil (Nova Medical, Wilmington, MA). The scan protocol included a 3D Magnetization Prepared Rapid Acquisition Gradient Recalled Echo (MPRAGE) with a field of view covering the whole brain for segmentation, a 2D axial CEST sequence, and a B₁ map corresponding to the CEST. See Supplementary Methods and Appendix for acquisition parameters.

Image processing

For CEST acquisitions, data from each offset frequency were registered to the target reference acquisition (S₀) using intensity-based affine registration in Matlab (Mathworks, Natick, MA). The registered CEST data were normalized to S₀ and CEST z-spectra were corrected voxel-wise for B₁ and B₀ inhomogeneities (cf. Supplementary Methods). GluCEST contrast was calculated from corrected spectra as:

$$GluCEST=\frac{S\left(-\mathrm{\Delta }\omega \right)-S\left(\mathrm{\Delta }\omega \right)}{S(-\mathrm{\Delta }\omega )}\times 100,$$ [1]

where Δω (offset frequency between water and solute frequencies) is 3.0 ppm for sensitivity to glutamate amine protons.¹⁸

Image analysis

GM, WM, and cerebrospinal fluid (CSF) regions were segmented in SPM12 from the MPRAGE using the “Segment” tool. GM was divided into distinct cortical regions (i.e. prefrontal, parietal, motor, somatosensory, and occipital cortices) using multi-atlas labeling applied to the MPRAGE volume.²³ Details on segmentation are provided in the Supplementary Methods. GluCEST contrast (Eq. 1) was studied in the cortical GM using tissue masks consisting of all cortical GM voxels, prefrontal cortex voxels, and parietal cortex voxels. Motor cortex voxels were also obtained to compare a region that is not anticipated to be related to cognitive deficits. Histograms of GluCEST contrast as well as mean and median GluCEST values were computed for each mask.

Statistical analysis

A Bland Altman repeatability analysis²⁴ was explored for the seven HCs scanned twice. A two-sample t-test was performed to test differences in age and cognitive test scores between HCs and MS patients. Linear regression models were used to compare GluCEST contrast for the entire cortical ribbon and for specific cortical regions between HC and MS groups with age included as a covariate. Although the age difference between groups was only 3.3 years (p=0.12), we included the age covariate because there is evidence that age influences glutamate levels in the brain.²⁵ In patients with MS, associations between GluCEST and measures of cognitive function were investigated using a Spearman partial correlation with correction for age. Although age was similar between the patient and HC groups, age has significant effects on several cognitive tests; thus, partial correlations were preferred to correct for age as a confounding factor in these analyses. Years of education were also similar between cohorts (15.6±2.7 and 17.3±1.7 years for MS and HC, respectively), and did not correlate with cognitive scores after adjusting for age, so this variable was not included as a covariate. Associations between GluCEST and cognitive test scores were not examined in the HC group, because variance within physiologic levels of glutamate is expected to be highly regulated and not expected to be related to cognition in an otherwise healthy individual.

Results

Reproducibility of CEST and feasibility in MS patients

Anatomical and CEST imaging were performed on all participants without adverse events. Figure 1 shows representative anatomical images (Figure 1A, E), B₀ shift maps (Figure 1B, F), and GluCEST maps (Figure 1C-D, G-H) for a HC (27-year-old female) (A-D) and a patient with MS (37-year-old female, EDSS=2.5, duration=6 years) (E-H). Across all subjects, the mean B₀ frequency shift ranged from −0.55 ± 0.19 ppm to 0.25 ± 0.26 ppm. GluCEST maps calculated from the CEST data (Δω=3.0ppm) (Figure 1C, G) show that GluCEST contrast in WM is lower than that of GM (segmented in Figure 1D, H), as expected.¹²

Figure 1: Anatomical images, B₀ shift maps, GluCEST contrast maps, and segmented GM GluCEST contrast maps for 7T MRI.

A healthy control (27-year-old female) is shown in (A-D) and a patient with MS (37-year-old female, EDSS = 2.5, duration = 6 years) is shown in (D-F). GM = gray matter.

In seven HCs, two scans were completed in separate visits (1-9 weeks apart) to evaluate repeatability of CEST imaging and determine the sensitivity to clinically-relevant changes in GluCEST contrast. Bland-Altman analysis was performed on mean cortical GM GluCEST and resulted in a repeatability coefficient of 1.57: the absolute difference between any two future measurements is estimated to be no greater than 1.57 on 95% of occasions (Supplementary Figure 1). The difference between the two measurements for each subject fell within the limits of agreement, and the 95% confidence interval contains 0 indicating no bias between repeat visits.

Group differences in CEST-derived measures

In Figure 2A, the average B₁- and B₀-corrected z-spectra for cortical GM voxels in HCs and MS patients are compared and show an upward shift in the baseline values (outer edges) of the z-spectra and greater inter-subject variability for MS patients relative to HCs. Figure 2B shows the distribution of cortical GM GluCEST across both cohorts, and mean GluCEST (Figure 2E) shows a trend toward increased GluCEST in patients. In the subset of patients with some degree of accumulated disability (EDSS>0, n=11) (Figure 2C-E), the trend toward greater cortical GM GluCEST compared to HCs remains as evidenced by linear regression with age included as a covariate (Figure 2F, p=0.081 for HCs vs. MS with EDSS>0).

Figure 2:

GluCEST in cortical GM. CEST z-spectra (A, C) and cortical GM GluCEST (B, D) differ between healthy controls (solid blue line) and patients with MS (dashed red line) for the entire patient cohort (A, B) and for the subset of MS patients with EDSS scores greater than 0 (C, D). There is a trend toward greater cortical GM GluCEST in MS as shown by the raw GluCEST data (E) and age-corrected disease effects plots (F). Shaded areas represent standard deviation for each cohort. For (A, B), n=20 healthy controls and 20 patients with MS. For (C, D), n = 20 healthy controls and 11 patients with MS. EDSS = Expanded Disability Status Scale, GM = gray matter.

Regional analyses of the cortical GM showed that the prefrontal and parietal cortices were consistently visualized in the CEST slice for all subjects, while the tissue masks for the other three regions only contained few voxels (Supplementary Figure 2). Thus, we focused further analysis on the prefrontal and parietal cortices. In the prefrontal cortex of MS patients, an upward shift in the z-spectrum baseline (Figure 3A) and greater inter-subject variability in GluCEST contrast (Figure 3B, E) relative to HCs remained. In patients with EDSS>0 (Figure 3C-E), the differences in prefrontal cortex GluCEST between HCs and patients were more apparent, and the age-corrected effect of disease on GluCEST was significant (Figure 3F, p<0.05 for HCs vs. MS with EDSS>0). While the z-spectra differed between patients and controls in the parietal region (Supplementary Figure 3A, C) and the distribution of GluCEST contrast was wider in patients (Supplementary Figure 3B, D), the average GluCEST contrast values were similar (Supplementary Figure 3E) and there were no significant age-corrected disease effects for the parietal cortex (Supplementary Figure 3F).

Figure 3:

GluCEST in prefrontal cortex. CEST z-spectra (A, C) and GluCEST for the prefrontal region of cortical GM (B, D) differ between healthy controls (solid blue line) and patients with MS (dashed red line) for the entire patient cohort (A-B) and for a subset of patients with EDSS scores greater than 0 (C-D). Raw GluCEST data show a trend toward greater prefrontal cortex GluCEST in MS (E), and age-corrected comparisons between healthy controls and patients with MS show a significant increase in prefrontal cortex GluCEST for the subset of patients with EDSS > 0 (G, *p<0.05). Shaded areas represent standard deviation for each cohort. For (A, B), n=20 healthy controls and 20 patients with MS. For (C, D), n = 20 healthy controls and 11 patients with MS. EDSS = Expanded Disability Status Scale.

Associations between CEST-derived measures and CI

Group differences in cognitive performance are reported in Table 3. Significant differences were observed for the Global Deterioration Scale, Brief Cognitive Rating Scale, Symbol Digit Modalities Test, and Paced Auditory Serial Addition Test (p<0.001) and the Brief Visuospatial Memory Test-Revised (total recall), Buschke Selective Reminding Test (total recall, 8 trials), Trail Making Test B, Simple Reaction Time, and Choice Reaction Time (p<0.05). When the subset of patients with EDSS>0 (n=11) was compared to HCs, test scores differed significantly between groups for all cognitive assessments.

Table 3:

Measures of cognitive function in healthy controls and patients with multiple sclerosis.

Cognitive Assessment, mean ± SD	All MS Patients (n=20)	MS with EDSS > 0 (n=11)	Healthy Controls (n=20)	Higher Score = “Better”
Global Deterioration Scale	2.3 ± 0.8a	2.5 ± 0.7b	1.0 ± 0
BCRS	14.5 ± 4.3a	16.0 ± 2.9b	8.0 ± 0
SDMT	49.4 ± 11.3a	44.5 ± 11.1b	62.4 ± 9.4	✓
BVMT-R (total recall)	24.5 ± 5.0a	22.5 ± 4.8b	28.6 ± 5.2	✓
BVMT-R (delayed recall)	8.9 ± 2.9	8.3 ± 2.5b	10.3 ± 2.3	✓
Buschke (total recall)	82.3 ± 15.5a	81.2 ± 18.7b	93.9 ± 14.4	✓
PASAT	65.4% ± 21.0%a	53.0% ± 17.4%b	86.9% ± 10.9%	✓
Trail Making Test A	27.5 ± 13.7 s	33.9 ± 14.4 sb	21.5 ± 7.3 s
Trail Making Test B	55.1 ± 20.0 sa	66.2 ± 17.7 sb	41.0 ± 13.9 s
Simple Reaction Time	306.1 ± 36.6 msa	319.3 ± 30.7 msb	272.6 ± 49.1 ms
Choice Reaction Time	506.9 ± 79.0 msa	526.0 ± 83.6 msb	439.0 ± 67.2 ms

MS = multiple sclerosis; EDSS = Expanded Disability Status Scale; BCRS = Brief Cognitive Rating Scale; SDMT = Symbol Digit Modalities Test; BVMT-R = Brief Visuospatial Memory Test-Revised; Buschke = Buschke Selective Reminding Test; PASAT = Paced Auditory Serial Addition Test.

^aScores of patients with MS (all) differ significantly from those of healthy controls (p< 0.05).

^bScores of patients with MS (EDSS > 0) differ significantly from those of healthy controls (p< 0.05).

Table 4 shows the associations between cognitive test scores and CEST-derived measures in patients. Over all MS patients, Global Deterioration Scale scores were significantly correlated with mean GluCEST for cortical GM and with mean and median GluCEST in the prefrontal cortex. Brief Cognitive Rating Scale scores were also significantly correlated with mean GluCEST in the entire cortical ribbon. Scores on the Buschke Selective Reminding Test (total recall, 8 trials) were significantly correlated with median GluCEST in the prefrontal cortex. For the subset of patients with EDSS>0, the Symbol Digit Modalities Test and Choice Reaction Time scores correlated significantly with mean and median GluCEST in the cortical GM and in the prefrontal cortex (Figure 4). GluCEST in the prefrontal cortex was also strongly associated with Global Deterioration Scale and Brief Cognitive Rating Scale scores. In contrast, these tests did not correlate significantly with GluCEST in the parietal cortex. However, Paced Auditory Serial Addition Test scores were strongly associated with GluCEST in the parietal cortex for the group with EDSS>0, as were Brief Visuospatial Memory Test-Revised (total recall) scores for the group of all patients (Table 4).

Table 4:

Spearman partial correlations between cognitive test scores and GluCEST values.

	All MS Patients		MS with EDSS > 0
	GluCEST Mean	GluCEST Median	GluCEST Mean	GluCEST Median
Global Deterioration Scale
All Cortical GM	r = 0.519b	0.430c	0.453	0.439
Prefrontal Cortex	0.484b	0.485b	0.398	0.555c
Parietal Cortex	0.196	0.149	0.130	−0.003
Brief Cognitive Rating Scale
All Cortical GM	0.462b	0.326	0.391	0.374
Prefrontal Cortex	0.407c	0.351	0.583c	0.583c
Parietal Cortex	0.288	0.254	0.362	0.201
Symbol Digit Modalities Test
All Cortical GM	−0.284	−0.208	−0.818a	−0.816a
Prefrontal Cortex	−0.167	−0.131	−0.777a	−0.814a
Parietal Cortex	−0.097	−0.086	−0.368	−0.199
BVMT-R (total recall)
All Cortical GM	0.311	0.351	−0.023	0.075
Prefrontal Cortex	0.229	0.215	0.026	−0.070
Parietal Cortex	0.359	0.453c	0.234	0.448
BVMT-R (delayed recall)
All Cortical GM	(Cognitive test score did not differ significantly between groups)		0.000	−0.010
Prefrontal Cortex			0.286	0.130
Parietal Cortex			0.013	0.152
Buschke (total recall)
All Cortical GM	−0.352	−0.368	−0.405	−0.403
Prefrontal Cortex	−0.434c	−0.495b	−0.340	−0.492
Parietal Cortex	−0.061	−0.032	−0.014	−0.023
PASAT
All Cortical GM	0.121	0.210	0.189	0.221
Prefrontal Cortex	−0.052	−0.011	−0.133	−0.234
Parietal Cortex	0.220	0.261	0.471	0.571c
Trail Making Test A
All Cortical GM	(Cognitive test score did not differ significantly between groups)		0.480	0.577c
Prefrontal Cortex			0.117	0.229
Parietal Cortex			0.500	0.402
Trail Making Test B
All Cortical GM	0.016	−0.019	0.204	0.236
Prefrontal Cortex	0.003	0.011	0.042	0.176
Parietal Cortex	−0.162	−0.193	−0.014	−0.254
Simple Reaction Time
All Cortical GM	−0.342	−0.360	−0.009	−0.003
Prefrontal Cortex	−0.199	−0.177	0.149	0.267
Parietal Cortex	−0.353	−0.389	−0.057	−0.201
Choice Reaction Time
All Cortical GM	0.320	0.267	0.646b	0.720b
Prefrontal Cortex	0.342	0.344	0.750b	0.772a
Parietal Cortex	−0.061	−0.085	0.255	0.181

MS = multiple sclerosis; EDSS = Expanded Disability Status Scale; GM = gray matter; BVMT-R = Brief Visuospatial Memory Test-Revised; Buschke = Buschke Selective Reminding Test; PASAT = Paced Auditory Serial Addition Test.

^aSignificant correlation (p<0.01).

^bSignificant correlation (p<0.05).

^cMeasures were strongly associated (p<0.10).

Figure 4:

Associations between cognitive measures and GluCEST. Performance on the Symbol Digit Modalities Test (A-B) and Choice Reaction Time (C-D) correlates significantly with GluCEST in the overall cortical GM (A, C) and in the prefrontal cortex (B, D) in the subset of patients with EDSS scores greater than 0. GM = gray matter, SDMT = symbol digit modalities test, CRT = choice reaction time.

Associations between GluCEST of the motor cortex and CI

In addition to the prefrontal and parietal cortices, GluCEST in the motor cortex was evaluated, and there were no significant differences in GM GluCEST contrast between patients and HCs. However, GluCEST in the motor cortex correlated significantly with EDSS (r_S=−0.481 for median GluCEST, p<0.05) and Simple Reaction Time (r_S=−0.543 for mean, r_S=−0.500 for median, p<0.05) in the group of all 20 patients. Cognitive tests that correlated significantly with prefrontal and parietal GluCEST did not correlate with motor cortex GluCEST.

Discussion

In this work, we applied glutamate-sensitive CEST MRI to study cortical GM changes related to pathology in MS. Our work delivers two main findings which support the potential role of glutamate-sensitive CEST MRI as biomarker of disease for patients with MS.

1. Glutamate-sensitive CEST MRI detects pathological changes in the cortex of MS patients

We found a trend toward increased GluCEST contrast in the cortical GM of patients with MS relative to HCs, and in the prefrontal cortex this increase was significant (Figure 3). While GluCEST is not a direct measurement of glutamate concentration, we interpret an increase in GluCEST contrast to correspond with increased glutamate levels. Inter-subject variability increased in the patient group, and the z-spectrum baseline shifted upward, indicating a decrease in the magnetization transfer (MT) effects.²⁶ This corresponds with MT differences in GM of relapsing-remitting MS that have been observed in previous studies.²⁷ Negative CEST asymmetry contrast results from MT asymmetry, which peaks ~2.3 ppm upfield (Δω=−2.3 ppm) from water.²⁸

Previous MRS studies in MS have shown mixed results regarding glutamate,^{10–12, 29} and are likely related to differences in field strength, imaging protocols, image resolution (1.87×1.87×10 mm³ for GluCEST versus 1 cm³ (whole brain) to 15.6 cm³ (single voxel) for MRS), patient cohort characteristics, and the region of interest studied. A whole-brain 3T MRS study observed decreases in glutamate + glutamine in several regions, including prefrontal and deep GM,¹² while another study showed increased glutamate in active lesions and NAWM and normal levels in chronic lesions.¹⁰ Others have detected decreases in glutamate in the hippocampus, thalamus, and cingulate associated with worse visuospatial memory in MS, and a decrease in parietal cortex glutamate that did not correlate with memory.¹¹

CEST offers the ability to investigate molecular and biochemical tissue pathology through changes in proteins, neurotransmitters, and other metabolites.^{13–15, 18} Glutamate-sensitive CEST at 7T has been applied in the healthy human brain¹⁸ as well as in the brain of patients with epilepsy¹⁹ and psychosis,²⁰ yet has not been studied in cortical GM pathology or MS. The current work differs from previous studies by evaluating CEST in multiple eloquent regions of the brain simultaneously with higher spatial resolution relative to MRS,^{10, 11} and by targeting glutamate rather than amide protons.^{14, 17} Additionally, we focused on cortical GM pathology rather than WM.¹⁴ We observed increased GluCEST contrast in the prefrontal cortex, which is consistent with post-mortem studies of MS that revealed alterations in glutamate reuptake mechanisms that correlated with excitotoxicity-induced neuronal and synaptic damage.⁹ In contrast to inflammatory WM lesions, cortical inflammation is marked by activated microglia which impair glutamate reuptake as well as release excess glutamate. Furthermore, the cortex is especially sensitive to glutamate-mediated excitotoxicity due to its high number of glutamatergic synapses and higher glutamate compared to WM. Although we did not detect a significant increase in GluCEST contrast in the parietal cortex, there was a trend toward increased GluCEST in patients with EDSS>0, which agrees with Geurts et al. using MRS in the parietal cortex.²⁹

2. Cognitive impairment correlates with GM GluCEST in patients with disability

Our patient cohort presented with CI affecting most commonly impaired domains, with differences seen in processing speed, executive function, and new learning. These differences were seen in the entire cohort as well as the subgroup with more advanced disability. Two types of analyses were performed. First we assessed associations between CI and GluCEST in all patients. Since about half of our patients had EDSS=0 and minimal CI, we applied an arbitrary cutoff of EDSS>0 to consider a subgroup with more advanced disease to more adequately capture CI and its associated metabolic substrates. In both groups, Brief Cognitive Rating Scale scores were associated with GluCEST of prefrontal cortex. The Brief Cognitive Rating Scale (used also in patients with Alzheimer’s disease) assesses both concentration and memory.³⁰ In patients with MS, this likely captures the difficulty in recalling information due to attention deficits which impede attentiveness to new data, and it is not surprisingly associated with disease in the prefrontal lobe.³¹ In the subgroup with EDSS>0, associations were seen between prefrontal cortex GluCEST and Symbol Digit Modalities Test and Choice Reaction Time with robust correlation coefficients (r_S=−0.814 and 0.772, respectively) (Figure 4). These measures of cognitive function assess information processing speed, the most common cognitive domain affected by MS.¹ Higher-order association areas in prefrontal cortex have pivotal importance in cognitive function as previously demonstrated in diseases affecting corticostriatal loops between the prefrontal cortex and the basal ganglia, such as MS and advanced Parkinson’s disease.^{32, 33}

In contrast, topographically non-specific correlations such as those involving the motor cortex were not significant. Rather, motor cortex GluCEST correlated with Simple Reaction Time (relevant to motor speed) and EDSS, thus confirming the topographic specificity of our findings.

Study limitations

This study had some limitations. First, CEST effects in vivo may be influenced by competing MT and nuclear Overhauser enhancement (NOE) effects, and relaxation time changes (T₁ and T₂).³⁴ Methods for correcting for these effects³⁴ and the impact on GluCEST quantification should be evaluated in future studies. GM GluCEST contrast may also be influenced by cortical atrophy.³⁵ We sought to minimize partial volume effects from enlarged CSF spaces by excluding voxels with a narrow z-spectrum FWHM (characteristic of CSF) from the GM mask. GluCEST results were treated with a histogram-style analysis rather than normalized to GM voxel count, because 1) we did not want to weight some subjects more heavily in group comparisons, and 2) atrophy is a global process in the brain in MS while our CEST data is currently limited to a smaller volume. It is also possible that cortical lesions contributed to the observed cortical GM GluCEST contrast, and the contributions of cortical lesions and normal-appearing GM might be separated by acquiring additional sequences tailored to macroscopic lesions. In the current study, a 2D slice was acquired, as whole-brain CEST is time-prohibitive. In future studies, CEST acquisitions could be optimized to achieve greater anatomical coverage (enabling comparisons with global atrophy effects) and minimize competing effects. Finally, when calculating the statistical significance of the univariate correlations, we did not apply a correction for multiple comparisons due to the exploratory nature of the study and to minimize the risk of type II errors. However, all correlations reported to have a significant p-value (p<0.05) had at least a moderate strength (i.e. r_S>0.40).

Conclusions

We showed that CEST contrast has the potential for adding biochemical specificity to GM pathology in vivo. Detecting biochemical changes that precede macroscopic GM lesions may aid in earlier detection of definite cognitive impairment and open the horizons to new experimental therapeutics targeting the molecular mechanisms of neurodegeneration.