Materials and Methods - Neurochemical Investigation of pathophysiology

Neurochemical Investigation of pathophysiology

8.3 Materials and Methods

8.3.1 Population Sampling

The study was approved by the local Ethics Committees. All participants gave written informed consent and received monetary compensation for their participation. Forty-three right-handed adult patients with GTS (7 female; 18-65 years) were recruited from the outpatient psychiatry clinic at Hannover Medical School. Demographic and clinical data of all subjects included in the ﬁnal analysis are summarized in Table 8.1. Patients were deemed ineligible if they exhibited severe tics to the head and face, contraindications to Magnetic Resonance (MR) examinations, a history of other signiﬁcant neurological disorders and current abuse of drugs and alcohol. Patients using any psychoactive sub-stances underwent a four-week washout period before participation. After baseline data acquisition, a subset of patients (N=17) received treatment with oral aripiprazole, which was administered using a titration procedure commencing at 2.5mg/day up to maximum daily dosage of 30mg based on treatment response. All patients were diagnosed based on DSM-5 criteria and underwent a thorough clinical assessment battery. The Yale Global Tic Severity Scale (YGTSS) [174] and the modiﬁed Rush Video-based Tic Rating Scale (RVTRS) [175] were used to capture tic severity, while the Premonitory Urge for Tics Scale (PUTS) [176] was used to assess premonitory urges. Clinical information was also collected on comorbid features using various scales that include the Yale-Brown Obses-sive CompulObses-sive Scale (Y-BOCS) [178] and the Revised ObsesObses-sive-CompulObses-sive Inventory (OCI-R) for OCD; the Beck Depression Inventory (BDI-II) [181] for depression; and the Conners’ Adult ADHD Rating Scale (CAARS) [180] for ADHD. Psychiatric comor-bidities were diagnosed as described in Chapter 5. Forty age- and sex-matched healthy control subjects (8 female, 18-65 years) without a history of neurological, psychiatric and tic disorders were also recruited and assessed in a similar manner as the patients.

A subset of the control subjects (N=23) were invited for a second MR exam for test-retest reliability measurements. All subjects were instructed not to drink coﬀee or tea, and to abstain from smoking for at least 2h before the examination. Subjects were also instructed to adhere to a regular sleeping cycle the night before the scan. To minimize the variability that could arise from circadian physiological eﬀects [290], the time of day of the MR exam was matched between patients and controls, with the majority of acquisitions conducted between 10 AM and 4 PM.

Glutamate 80

Table 8.1: Demographic and clinical characteristics of the ¹H-HMRS study sample included in the final analysis

Handedness 36 right 37 right (15 right) 15 right — —

YGTSS- TTS — 21.2±8.2

All the recruited subjects were right handed.

Medication history is reported for drugs taken at least four weeks before data acquisition.

Abbreviations: ADHD = Attention Deficit/Hyperactivity Disorder; BDI-II = Beck Depression Inventory; CAARS = Con-ners’ Adult ADHD Rating Scales; GTS = Gilles de la Tourette Syndrome; OCD = Obsessive-Compulsive Disorder;

OCI-R = Obsessive-Compulsive Inventoryâ ˘A ¸SRevised; PUTS = Premonitory Urge for Tics Scale; QOL= Quality Of Life scale; RVTRS = modified Rush Video-based Tic Rating Scale; Y-BOCS = Yale-Brown Obsessive-Compulsive Scale;

YGTSS-GS = Yale Global Tic Severity Scale Global Score; YGTSS-TTS = YGTSS Total-Tic Score.

Glutamate 81

8.3.2 Magnetic Resonance Data Acquisition

Magnetic resonance measurements were performed on a 3T MAGNETOM Verio (Siemens Healthcare, Erlangen, Germany) equipped with a 32-channel head coil. The patients were instructed to remain still without actively suppressing their tics and thinking of noth-ing in particular. Overall, the scannnoth-ing session lasted approximately 75 minutes and included the acquisition of self-report data on the degree of tic-urges and tic-suppression (see section 8.4.3). A landmark-based pre-scan gradient-echo sequence provided by the vendor (Auto-Align Head, AAH) was applied at the beginning of each scan for automatic detection of the crista galli and the tip of the occipital bone within the mid-sagittal plane [249]. The geometric information was subsequently applied on all imaging pro-tocols and saved for the retest scans. T1-weighted images were acquired in the ﬁrst AAH space using a three-dimensional Magnetization-Prepared 2 Rapid Gradient Echo (MP2RAGE) sequence [271]: repetition time, TR=5s; echo time TE=3.93ms; inversion times, T1=0.7/2.5s; sagittal slab orientation; acquisition matrix 256×256×176, and 1×1×1mm³ nominal resolution [291]. To minimize errors that arise from bulk head-displacement between the anatomical and spectral acquisitions, single-shot ‘dummy’

spectra were localized on the MP2RAGE image immediately after acquisition. The (fast) AAH sequence was applied again before each ¹H-MRS voxel acquisition to co-register the ‘dummy’ scan geometry to the newly deﬁned space. For the retest scans, the same procedure was used to automatically re-localize the ¹H-MRS voxel to the same region.

Motivated by previous imaging, genetic and post-mortem work [30,31,124,288],¹H spec-tra were acquired from three cuboid Regions of Interest (ROIs), with Point-RESolved Spectroscopy (PRESS) [292] and TE=30ms; TR=3s; 1024 time-domain data points; 80 (ROIs 1 and 2; see below) or 128 (ROI 3) water-suppressed and 16 water-unsuppressed averages. A ﬁrst 25×16×16mm³ voxel was prescribed on the anterior Mid-Cingulate Cortex (aMCC) with the center on the level of the genu of the corpus callosum and an orientation parallel to the hippocampal axis (Figure 8.1A). A second28×16×16mm³ voxel was centered on the bilateral thalamus while maximizing the amount of grey mat-ter (GM) (Figure 8.1B). A third 20×15×20mm³ voxel was localized within the left corpus striatum as anterior as possible to include maximal portions of the caudate; as inferior as possible to include the most inferior portion of the putamen; and as medially as possible without including any portions of the lateral ventricles (Figure 8.1C). On average, the voxel contained GM portions from the putamen (63.2%), pallidum (22.3%), caudate (12.8%) and nucleus accumbens (1.7%). ROIs were shimmed automatically us-ing FASTESTMAP with 4-5 iterations [293, 294]. Comprehensive imaging parameters are provided in Appendix D.

Glutamate 82

8.3.3 Absolute Metabolite Quantitation

As MRS measurements involve the summation of multiple averages to build the Signal-to-Noise Ratio (SNR), subject motion and drifts in the magnetic ﬁeld during acquisition cause frequency and phase errors [295]. Such drifts give rise to incoherent signal averag-ing, line broadenaverag-ing, lineshape distortion and reduced SNR, which may ultimately aﬀect metabolite quantitation and group comparisons. As such, non-averaged time-domain raw data were exported from the scanner and a non-linear least-squares minimization operation was employed to ﬁt each signal average, (wheret is time) to a reference scan, Sref(t) (here taken to be the ﬁrst average), by the adjustment of the frequency, ω, and phase, φ, of the signal according to [295,296]:

minimize

f,ϕ∈R kR(t)−G(t, f, ϕ)k (8.1) where

G(t, f, ϕ) =S(t)·e^{2π(f t+}³⁶⁰^φ ⁾ (8.2)

To enable ﬁtting of complex data while avoiding non-physical parameter estimates, the vectors and Sref(t) were modiﬁed prior to minimization by concatenating their real and imaginary parts into a single real-valued vector. Motion corrupted outlier signals were additionally removed prior to spectral averaging. This was accomplished by calculating the root-mean-square of the diﬀerence spectrum between each individual acquisition and the average and discarding acquisitions deviating by more than three standard deviations (SD) from the mean [296]. The performance of the algorithm is demonstrated in Figure 8.1 (D-E).

The water signal from the non-suppressed spectra was used as a concentration reference.

However, spectroscopic voxels in diﬀerent subjects contain varying proportions of GM, white matter (WM) and cerebro-spinal ﬂuid (CSF); the GM and WM compartments have diﬀerent water concentrations with diﬀerent T1 and T2 relaxation times for water;

and negligible signals of most metabolites arise from the CSF compartment. To con-sider tissue compartmentation, the MRS voxel was ﬁrst registered to anatomical space by calculating the transformation matrix from the raw ﬁle header, and a binary mask representing the voxel limits was then constructed to map the voxel onto the anatomical image [250,251]. SPM12 New-Segment (http://www.ﬁl.ion.ucl.ac.uk/spm) was used to automatically segment the brain into probabilistic GM, WM and CSF maps, which were binned at 0.5 to make the three tissue classes add up to 100%. The GM masks were op-timized to include subcortical nuclei generated via FSL-FIRST’s Bayesian model-based

Glutamate 83

Figure 8.1: Voxel localization and spectral data pre-processing. Console screenshots illustrating the prescription of the voxels in the(a)anterior mid-cingulate cortex (aMCC),(b)bi-lateral thalamus and(c) corpus striatum on MP2RAGE images in single subject. (d) The effect of frequency- and phase-drift (FPD) correction on data acquired from a striatal voxel of a GTS patient (top-panel) is clearly visible in the corrected data (bottom-panel). The performance of the non-linear least-squares minimization operation is illustrated on a single scan (the 25th average) superimposed on the reference signal (inset plots). For the aMCC spectra, which exhibited a high SNR, spectral registration was conducted on an approximate range between 1.8-4.2 ppm (red shaded area). For spectra with a lower SNR, the strength of the water signal was utilized for spectral registration on an approximate range between 4.2-7.5ppm (green shaded area). (e) The benefit of the correction on the linewidth and SNR is clearly visible on the red coloured spectrum. As some of the spectra exhibited a spurious residual signal in the ppm range between 3.6-4.0 ppm (blue shaded area), possibly due to insufficient spoiling, the LCModel fitting range was adjusted to 0.2-3.67ppm (orange

shaded area).

Glutamate 84

segmentation of subcortical nuclei [297]. Within-voxel tissue proportions were then cal-culated based on the optimized tissue segmentation masks. Averaged frequency- and phase-drift-corrected spectra were ﬁt with LCModel [199]. As some spectra contained artifacts in the frequency range above 3.7 ppm, presumably due to insuﬃcient spoiling, spectral ﬁtting was performed in a 0.2-3.67 ppm range. Results for the default 0.2-4.0 ppm range exhibited correspondence and are not presented here. For ﬁnal inclusion into the statistical models, a step-wise semi-automated quality-control method was utilized.

Detectable metabolites were ﬁrst identiﬁed for each voxel in the control sample if the relative Cramér-Rao Lower Bounds (CRLBs) of a given metabolite were below 100%

in at least 50% of the subjects [298]. Cut-oﬀ values of absolute CRLBs (CRLBlim), deﬁned as 50% of the mean metabolite concentration in the control sample) were then calculated for each metabolite to threshold cases with excessive ﬁtting errors [299]. Final inclusion criteria were: (a)correct voxel prescription,(b) SNR>10 (LCModel output), (c) full width at half maximum (FWHM) < 11Hz (LCModel output) for good quality spectra and (d) absolute CRLB < CRLBlim for individual metabolites. All surviving spectra were visually inspected to ensure the quality of included data. Within-voxel compartmentation was considered by applying Eq. 2 [254]:

Cm= I_m

where cm is the metabolite concentration in tissue; c^w_o =55.6mol/L is concentration of bulk water [245]; Im and Iw are the amplitudes of the metabolite and the water signal, respectively; andNm is the number of protons within the molecule contributing to the metabolite signal. Partial-voluming and the presence of non-water substances are accounted for by the scaling factorξ, where f_ε and α_ε are, respectively, the volume fraction and relative water content of tissueε= GM, WM, CSF), and Rεconsiders water relaxation. Relaxation eﬀects of metabolites were ignored since they have similarT₁ and T1 in GM and WM [246]. Relaxation time and relative water tissue content values are outlined in Table8.2.

Glutamate 85

Table 8.2: WaterT1 andT2 relaxation times and relative water content (α) in GM, WM and CSF

GM WM CSF

T1 1820¹ 1084¹ 4163²

T2 99¹ 69¹ 503³

α 0.81^4,5 0.71^4,5 1^4,5

1(Stanisz et al., 2005) [197];²(Lin et al., 2001) [300]; ³(Piechnik et al., 2009) [196]; ⁴(Choi et al., 2006) [301];⁵(Norton et al., 1966) [302]

8.3.4 Statistical Analysis

Statistical analysis was performed in the Python programming language (Scipy v.0.15.1 and Statsmodels v.0.6.1) [303, 304]. All data exhibited a Gaussian distribution as as-sessed via the Kolmogorov-Smirnov test. Between- and within- group diﬀerences for all metabolites were assessed using two-way independent sample t-tests and paired-sample t-tests, respectively, with a signiﬁcance level set at <0.05 (uncorrected). A multiple-linear-regression model accounting for age and gender was employed to examine the re-lationship between Gln/Glu concentrations and clinical measures (YGTSS total-tic score, RVTRS, PUTS, Y-BOCS and CAARS). Metabolite test-retest reliability measures were assessed in the control sample by calculating the coeﬃcient of variation (COV), percent-age diﬀerence and paired-sample t-tests. The reliability of spatial re-localization was assessed using the Sørensen–Dice metric [252,253].

Glutamate 86

8.4 Results

8.4.1 Demographic and clinical characteristics

Complete datasets were acquired from 37 of 43 recruited patients and 36 of 40 recruited controls. MR data was not collected from(a)three patients and three controls due to se-vere head motion during acquisition;(b)two patients and one control due to claustropho-bia; and(c) one patient due to previously unreported MR contraindications. Seventeen of the 37 patients underwent a four-week aripiprazole therapy, and complete datasets were collected from 15 patients. Control and patient subjects were comparable in terms of age (t71=0.045, P=0.96), gender (Fischer’s odds ratio 0.80, P=0.76) and handedness (Table8.1). Control subjects signiﬁcantly diﬀered from the patients on obsessions/com-pulsions (OCI-R, Y-BOCS), ADHD (CAARS), and depression (BDI-II). In the subset of patients that received aripiprazole treatment, signiﬁcant reductions were observed in RVTRS (t13= 2.40, P=0.03) and YGTSS global impairment (t13=3.60, P=0.003).

A comprehensive analysis of the clinical measures of the entire sample is presented in Chapter10.

8.4.2 Test-Retest Reliability

The Sørensen–Dice metric quantifying the spatial overlap of test-retest localization yielded means and SDs of 0.80±0.10 for the aMCC, 0.83±0.10 for the thalamus, and 0.80±0.09 for the striatum, indicating high repeatability of our localization technique (Figure8.2).

Assessing the reliability of the spectral measurements (Table 8.3), we found no sig-niﬁcant diﬀerences in (a) spectral quality parameters (FWHM, SNR) and (b) intra-voxel tissue proportions (WM, GM, CSF). Detectable metabolites (as deﬁned above) included total N-acetylaspartatyl compounds (tNAA; i.e. N-acetylaspartete, NAA, plus N-acetylaspartylglutamate, NAAG), (phospho)creatine (Cre), choline compounds (Cho), Glu, Gln, glutamate plus glutamine (Glx), myo-inositol (m-Ins) and GABA in all voxels, in addition to Aspartate (Asp), Lactate (Lac) and Taurine (Tau) in the aMCC. Con-centration estimates were consideredunreliable if they exhibited a signiﬁcant test-retest group diﬀerence in the control cohort. Metabolites that failed this criterion included Asp and Lac in the aMCC and m-Ins in the striatum, and were excluded from further analyses.

Glutamate 87

Table 8.3: Test-Retest Reliability of absolute metabolite quantitation. Sum-mary of spectral quality parameters, absolute metabolite concentrations and statistical

results of test-retest healthy control 1H-MRS measurements

N Baseline Retest COV % Difference D 95 % CI Statistic P-Value

aMCC Metabolite concentrations are reported in mmol/L units.

Abbreviations: aMCC = anterior mid-cingulate cortex; Asp = aspartate; CI =confidence interval; Cho = choline com-pounds; COV = coefficient of variation; Cre = (phospoho)creatine; CSF = cerebrospinal fluid (%); D = Cohen’s D; FWHM

= full width at half maximum (Hz); GABA =γ-aminobutyric acid; GM = grey matter (%);Gln = glutamine; Glu = gluta-mate; Glx = glutamate plus glutamine; Lac = lactate; m-Ins = myo-inositol; NAA = N-acetylaspartate with an additional contribution from N-acetylaspartylglutamate; SNR = signal-to-noise ratio; Tau = taurine; WM = white matter (%).

Glutamate 88

Figure 8.2: Spatial overlap of test-retest voxel localization Representative images illustrating the extent of spatial overlap (purple) achieved between visit 1 (red) and visit 2 (blue) using the auto-align re-localization technique for the(a) aMCC,(b) bi-lateral thalamic and(c) the left corpus striatal single voxel spectroscopy regions of

interest.

8.4.3 Degree of tic-urges and tic-suppression during MR data acquisi-tion

All patients were instructed to remain still without actively suppressing their tics and thinking of “nothing” in particular during MR data acquisition. Following the scans, the urge-to-tic and the active suppression of tics were veriﬁed via self-report. Speciﬁcally, the patients were asked: (a) “How strong was your urge to tic during scanning?” (0=no urge at all, 10=unbearable urge to tic) and(b) “How much effort did you expend to suppress tics?” (0=no eﬀort was made to suppress tics; 10=maximal eﬀort/attention was made to suppress tics). Median baseline values of the urge-to-tic and the eﬀort expended to suppress tics were 3.5 and 3.0, respectively. In the on-treatment condition, similar

Glutamate 89

median values were observed (urges=3.0 and suppression=2.5), indicating that intra-group comparisons of patients’ spectral measures negligible bias from active suppression of tics. Given the low observed median values of tic suppression at baseline, it seems further plausible to suggest that the mild suppression of tics may not have had a strong inﬂuence on inter-group comparisons.

8.4.4 Group Differences in Metabolite Concentrations

Group comparison of left striatal baseline measurements revealed signiﬁcant decreases in Gln concentrations (t60=2.54, P=0.0119), Glx concentrations (t60=2.54, P=0.017) and the Gln:Glu ratio (t60=2.22, P=0.03) in GTS compared to controls (Table 8.4). The thalamus additionally exhibited decreases of Glx concentrations in GTS compared to controls at baseline (t61=2.44,P=0.018). Cohen’s eﬀect sizes were relatively high (stria-tum: DGln=0.64;DGlx=0.63; DGln:Glu=0.56; thalamus: DGlx=0.61) indicating practical signiﬁcance. In the subset of patients that underwent aripiprazole treatment, paired-group comparisons revealed signiﬁcant increases in striatal Glu (t8= -2.30, P=0.047) and Glx concentrations (t8= -3.0,P=0.015) in on-treatment patients compared to base-line GTS (Table8.5). In addition, we observed trends for increases in striatal Gln (t8= -1.843, P=0.098) in thalamic Glx (t8= -2.133,P =0.064). Comparing baseline measure-ments of the control sample with GTS patients on-treatment, we did not observe any signiﬁcant diﬀerences in Gln, Glu, Glx concentrations and the Gln:Glu ratio. Consider-ing all other metabolites, we only observed a diﬀerence in thalamic Cre concentrations (t61= 2.39, P=0.02) when comparing controls to patients at baseline. For the aMCC, no diﬀerences were observed for any metabolites. Sample distributions and statistical results for Gln, Glu and Glx are illustrated in Figure 8.3. Representative spectra from ten subjects per voxel are illustrated in Figures 8.4–8.6.

Glutamate 90

Figure 8.3: Spectral localization, fitting and statistical analysis.Left-panel:

Sagittal, coronal and axial images illustrating the localization of the cingular (aMCC), thalamic (THA) and corpus striatal (STR) regions of interest. The reconstructed masks were generated based on geometric information extracted from the raw file header. Mid panel: Exemplary spectra illustrating LCModel fits, baselines and residual signals of frequency- and phase-drift-corrected data. The inset images demonstrate the location of voxels with respect to the GM, WM and CSF compartments, which were used to calculate within-voxel tissue proportions for absolute quantitation. A combination of segmentation outputs from SPM12 and FSL-FIRST was used for accurate masking of subcortical nuclei. Right panel: Plots illustrating the distribution of Gln, Glu and Glx concentrations in controls (green), GTS patients at baseline (red) and GTS patients fol-lowing treatment with aripiprazole (blue); ** denotes significance at p<0.05; * denotes

a trend for significance (p<0.1).

Glutamate 91

Table 8.4: Control vs. GTS group comparison of absolute metabolite con-centrations. Summary of spectral quality parameters, absolute metabolite concentra-tions and statistical results (GTS patients vs. controls) of 1H-MRS measurements.

N (HC/GTS) Controls Patients % Difference D CI(95 %) Statistic P-Value

aMCC

Metabolite concentrations are reported in mmol/L units.

Abbreviations: aMCC = anterior mid-cingulate cortex; Asp = aspartate; CI =confidence interval; Cho = choline compounds; COV = coefficient of variation; Cre = (phospoho)creatine; CSF = cerebrospinal fluid (%); D = Cohen’s D; FWHM = full width at half maximum (Hz); GABA = Î¸s-aminobutyric acid; GM = grey matter (%);Gln = glutamine; Glu = glutamate; Glx = glutamate plus glutamine; Lac = lactate; m-Ins = myo-inositol; NAA = N-acetylaspartate with an additional contribution from N-acetylaspartylglutamate; SNR = signal-to-noise ratio; Tau = taurine;

WM = white matter (%).

Glutamate 92

Table 8.5: GTS Off- and On-treatment group comparison of absolute metabolite concentrations. Summary of spectral quality parameters, absolute metabolite concentrations and statistical results of 1H-MRS measurements in GTS

patients at baseline and during treatment

N GTS GTS-APZ % Difference D CI(95 %) Statistic P-Value

aMCC Metabolite concentrations are reported in mmol/L units.

Abbreviations: aMCC = anterior mid-cingulate cortex; Asp = aspartate; CI =confidence interval; Cho = choline compounds; COV = coefficient of variation; Cre = (phospoho)creatine; CSF = cerebrospinal fluid (%); D = Cohen’s D; FWHM = full width at half maximum (Hz); GABA =γ-aminobutyric acid; GM = grey matter (%);Gln = glutamine; Glu = glutamate; Glx = glutamate plus glutamine; Lac = lactate; m-Ins = myo-inositol; NAA = N-acetylaspartate with an additional contribution from N-acetylaspartylglutamate; SNR = signal-to-noise ratio; Tau = taurine; WM = white matter (%).

Glutamate 93

Figure 8.4: Representative aMCC ¹H-MRS spectra of the frequency and phase-drift corrected data (red) and LCModel fits (black)

Glutamate 94

Figure 8.5: Representative THA ¹H-MRS spectra of the frequency and phase-drift corrected data (red) and LCModel fits (black)

Glutamate 95

Figure 8.6: Representative STR ¹H-MRS spectra of the frequency and phase-drift corrected data (red) and LCModel fits (black)

Glutamate 96

8.4.5 Correlation of Metabolite Concentrations with Clinical Variables

The multiple regression model using age and gender as covariates revealed a signiﬁcant negative correlation between GTS baseline measurements of striatal Gln concentrations and RVTRS (r=-0.52,P=0.012) (Figure8.7A). The analysis also revealed a signiﬁcant negative correlation between thalamic Glu and PUTS (R=-0.47,P=0.017) (Figure8.7B).

Figure 8.7: Correlation of absolute metabolite concentrations with clinical measures. The multiple linear regression model revealed significant negative correla-tions between(a) left corpus striatal Gln concentrations and post-scan measurements of actual tic severity (modified Rush Video-based Tic Rating Scale; RVTRS), in addi-tion to(b)bi-lateral thalamic Glu concentrations and pre-monitory urges (Premonitory

Im Dokument Elemental and neurochemical based analysis of the pathophysiological mechanisms of Gilles de la Tourette syndrome (Seite 100-135)