This chapter presents an investigation of the role of iron in GTS pathophysiology. Quanti-tative Susceptibility Mapping values extracted from subcortical regions served as surrogate measures of iron content. The accurate and non-biased estimation of subcortical suscep-tibility values was significantly improved by the preceding methodological investigation presented in Chapter 7. Blood samples were also acquired for serum ferritin quantia-tion to assess iron storage capacity. A multi-parametric approach was addiquantia-tionally used in which the relationship between subcortical susceptibility and spectroscopic measures of glutamatergic signalling were investigated. This work underwent a peer review process and was presented at the 2017 annual conference of the International Society of Mag-netic Resonance Imaging in Medicine in Honolulu, Hawaii, USA : Kanaan AS. et al., QSM meets MRS: The influence of subcortical iron on glutamatergic neurotransmission in a movement disorder population; Proc. Intl. Soc. Mag. Reson. Med. 25, 2017, Program No. 4649 [347].
9.1 Abstract
Gilles de la Tourette Syndrome (GTS) is a neuropsychiatric movement disorder funda-mentally characterized by tics and a complex genetic-environmental aetiological basis.
Previous work has indicated that the clinical manifestations of GTS are primarily driven by putative abnormalities in dopamine GABA, and glutamate. In view of the crucial role exhibited by the element iron in varied biochemical mechanisms sustaining devel-opmental processes and neurochemical pathways, we postulated that patients with GTS exhibit abnormalities in iron metabolism, which may influence mechanisms of subcor-tical neurotransmission. In this work, we used Quantitative Susceptibility Mapping, a recently established Magnetic Resonance technique, to estimate magnetic susceptibility
114
Iron 115
as a proxy measure for iron in 22 healthy controls and 22 patients with GTS for the first time. We additionally utilized previously acquired1H-MRS data to inspect the relation-ship between iron and glutamatergic signalling. We demonstrate that patients with GTS exhibit significant reductions of magnetic susceptibility in basal ganglia, brainstem and cerebellar nuclei. Reductions were specific to the striatum, substantia nigra, subthala-mic nucleus and the red nucleus and were mirrored by decreases in serum ferritin levels.
Significant correlations were observed between striatal – as well as total subcortical – magnetic susceptibility and striatal Gln:Glu. Our results indicate the presence of extant abnormalities in iron metabolism, which may exhibit an influence in GABA-Glu-Gln cycling. Perturbations in mechanisms regulated by iron-containing enzymes, may lead to disruptions in the spatio-temporal dynamics of excitatory, inhibitory and modulatory neurochemical systems, thus driving the manifesting clinical features in GTS.
9.2 Introduction
Iron is a trace element that is essential to the vitality of an organism as it is ideally suited for the catalysis of many biochemical reactions due to its ability to transition between two thermodynamically stable oxidation states [348]. Its flexibility in the traf-ficking of electrons renders it a crucial component of prosthetic groups (e.g. hemes and iron sulphur clusters) located within diverse proteins involved in oxygen transport, mito-chondrial energy metabolism, protein synthesis, in addition to neurotransmitter synthesis and transport [348]. Mediated by the blood-brain barrier, the acquisition of non-heme iron in the brain occurs in an age- and regionally-dependent manner, where its deposition during a “critical period" is necessary for normal development [349,350]. The deposition of non-heme iron is sharpest in subcortical grey-matter, where it overlaps with regions that contain dense proportions of the neurotransmitters dopamine, γ-aminobutyric acid (GABA) and glutamate (Glu) [351]. The atypical homeostasis of iron during different periods of development may influence the biochemical mechanisms that sustain typical neurochemical metabolism and myelination [161,352], providing a biological basis for the acquisition of abnormalities in motor and behavioural functions as exhibited by various developmental neuropsychiatric/movement disorders [151,153,353].
Gilles de la Tourette syndrome (GTS) presents an example of a disorder with motor and behavioural deficits as a result of fundamental alterations in the dynamics of cortico-striatal circuitry [28]. In essence, GTS is characterized by the presence of multiple motor and vocal tics and a high incidence of comorbid features, which may include Atten-tion Deficit/Hyperactivity Disorder (ADHD), Obsessive-Compulsive Behavior/Disorder (OCB/D), depression, and anxiety [9, 18]. Though the underlying pathophysiological
Iron 116
mechanisms of GTS have not been completely elucidated, current data suggests that acquired abnormalities in habit formation systems [280,282] may be driven by deficits in subcortical neurochemical signalling [31, 274], thus leading to a burst-like disinhibition of thalamo-cortical output [76, 78]. Methodologically varied work has indicated that patients with GTS exhibit abnormalities in (a) D2 receptor binding; (b) dopamine ac-tive transporter density/binding; and(c) phasic dopamine transmission, thus suggesting putative abnormalities in the functional dynamics of tonic and phasic dopaminergic sig-nalling [31]. Given the significant role exhibited by excitatory and inhibitory afferents on the dopaminergic nuclei [118,121], these effects may be driven or further compounded by alterations in the GABAergic and glutamatergic neurotransmitter systems. Along this line, abnormalities in the GABAergic system in cortical and subcortical regions have been demonstrated via 1H-Magnetic Resonance Spectroscopy (1H-MRS) [130, 286], positron emission tomography (PET) [127] and quantitative post-mortem studies [125,126]. More recently, our group has demonstrated that GTS patients exhibit abnormalities in gluta-matergic metabolism, where reductions of Gln, Glu+Gln (Glx), and the Gln:Glu ratio were observed in the striatum via 1H-MRS [274]. Put together, this data implies the presence of abnormalities in GABA-Glu-Gln cycling, which may lead to alterations in the spatio-temporal dynamics of excitatory, inhibitory and modulatory subcortical neu-rochemical signalling.
One unifying feature exhibited by excitatory, inhibitory and modulatory neurotransmit-ters is that the enzymes involved their metabolism and the production of their receptors and transporters require iron for typical function. The effects of iron on the dopaminergic system are well recognized as its deficiency during specific periods of development has been demonstrated to lead to deficits in the synthesis, catabolism, transport and uptake of dopamine [161]. More specifically, iron deficiency has been shown to lead to alter-ations in the function of(a)D1/D2 receptors;(b)dopamine active transporters; and(c) the dopamine catabolic enzymes tyrosine hydroxylase and monoamine oxidase B [161].
Although work linking iron to Glu/GABA metabolism is less extensive, studies have indicated that iron plays an important functional role in both excitatory and inhibitory neurotransmitter metabolism and transport. Specifically, dietary iron deficiency at differ-ent periods of prenatal and postnatal developmdiffer-ent has been demonstrated to lead to(a) downregulations in the key GABA-Glu-Gln cycle enzymes Glu decarboxylase, Glu dehy-drogenase, GABA transaminase and isocitrate dehydrogenase [354–356]; (b) reductions of GABA content in pallidal, striatal and hippocampal regions [357,358];(c) reductions in the binding of Glu to vesicular membranes [355, 359]; and (d) the suppression of glutamatergic neurotransmission in striatal and hippocampal regions [360,361].
Given this data, it seems plausible to hypothesize that patients with GTS may exhibit an abnormality in the cerebral homeostasis of iron. In general, a link between iron
Iron 117
deficiency and an increased risk of contracting neurodevelopmental psychiatric disorders (e.g. ADHD, autism spectrum disorder) is well established [362, 363]. The notion of disturbed iron homeostasis in GTS is supported by preliminary work indicating that the patients exhibit reductions in serum ferritin [163, 316, 364, 365]. Ferritin is a stable and reliable indicator of total iron content in the body, as it is the main intracellular iron sequestering protein [366]. However, because serum ferritin levels do not necessarily reflect iron concentrations in specific organs, estimates of regional iron content by non-invasive and in-vivo magnetic resonance imaging (MRI) techniques are of paramount interest.
In essence, non-heme iron sequestered in ferritin exhibits paramagnetic properties that lead to the induction of magnetic moments when applied to an external magnetic field.
The dominant magnetic susceptibility of subcortical iron rich structures induces field perturbations that can be non-invasively measured via magnetic resonance (MR) phase imaging techniques. Quantitative Susceptibility Mapping (QSM) is a recently established MRI technique that estimates the (relative) intrinsic magnetic susceptibility of tissue us-ing the MR signal phase followus-ing(a) the estimation of the magnetic field distribution, (b) the elimination of background field contributions, and (c) solving the inverse prob-lem from magnetic field perturbation to susceptibility [192,367]. Considering that the paramagnetic properties of sequestered iron render it as the most dominant contribu-tor to magnetic susceptibility in deep grey matter, a strong linear relationship between subcortical iron and magnetic susceptibility has been demonstrated and validated via multiple techniques [225,227,368,369].
Consequently, the primary aim of this work was to investigate whether patients with GTS exhibit reductions in cerebral iron content as indexed by lower magnetic susceptibility.
We focused on estimating magnetic susceptibility in a set of subcortical nuclei that have been implicated in GTS pathophysiology and relating these susceptibility measures to serum ferritin levels. Given the important role exhibited by iron containing enzymes that are integral to the tri-carboxylic acid cycle, we additionally postulated that iron will exhibit an association with excitatory signalling, which, to the best of our knowledge, is a relationship that has not been investigated simultaneously in vivo. To this end, we use the combination of QSM and 1H-MRS to investigate whether subcortical iron levels exhibit an association with glutamatergic neurotransmission as indexed by the Gln:Glu ratio.
Iron 118
9.3 Materials and Methods
9.3.1 Population Sampling
The study was approved by the local ethics committees and all participants gave written informed for their participation. A total of 43 right-handed adult patients with GTS and 40 age/gender matched healthy controls were recruited as part of a larger study that included the acquisition of structural, functional, and spectroscopic Magnetic Res-onance (MR) data (see Chapter 8 [274]). Given that the pole-artifact in the vendor provided phase maps discussed in Chapter 7 was only observed after the beginning of data acquisition, the susceptibility weighted gradient-echo sequence was optimized to output multi-channel data for optimal coil-combination off-line. As such, the combi-nation of T1-weighted, susceptibility-weighted (multi-channel), and spectroscopic MR data were acquired from a subsample that included 28 patients (4 female, 18-65 years) and 22 controls (5 female, 18-65 years). Patients using any psychoactive substances un-derwent a four-week washout period before participation and were deemed ineligible if they exhibited severe tics to the head and face, a history of other neurological disorders, current abuse of drugs and alcohol and MRI contraindications. All participants were di-agnosed based on DSM-5 criteria and underwent a thorough clinical assessment battery as described in Chapter 5. Age- and gender-matched healthy control subjects without a history of neurological, psychiatric and tic disorders were recruited and assessed in a similar manner as the patients. All subjects were instructed to (a) not drink coffee or tea and to abstain from smoking for at least 2h before the examination and (b) 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 acquisition conducted between 10AM and 4PM.
9.3.2 Measurement of serum Ferritin
A 10ml blood sample was collected from the majority of the subjects for the in-vitro quantitative determination of serum ferritin levels as a representative measure of the body’s iron reserves. The sample was first centrifuged at 24,000 rpm for a period of 10 minutes to separate hematocrit from plasma, which was subsequently stored in 1000ml aliquots at -70◦C. Serum ferritin levels were quantified based on the electochemilumi-nescence immunoassay, in which a voltage applied to a sample containing tagged ferritin molecules induces chemiluminescent emissions that are measured by a photomultiplier (Elecsys 2010, Roche Diagnostics GmbH, Mannheim, Germany).
Iron 119
9.3.3 Magnetic Resonance Imaging and Spectroscopy
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 nothing in particular. Magnetization-Prepared 2 Rapid Gradient Echo (MP2RAGE) and 1 H-MRS data were acquired as described in Chapter8 [274]. High quality spectra localized to the left-striatal voxel were obtained using a careful acquisition and processing scheme that incorporated (a) an automated voxel (re-)localization technique; (b) the removal of motion corrupted outlier signals;(c) frequency and phase drift correction in the time domain; (d) absolute metabolite quantitation with the consideration of within voxel compartmentation; and(e)a semi-automated quality assessment protocol. Susceptibility weighted data were acquired using a 3D-flow compensated, spoiled, gradient recalled echo sequence with the parameters: TR=30s; TE=17ms; 256×256 matrix; flip-angle=13◦; 0.8mm isotropic nominal resolution.
9.3.4 Quantitative Susceptibility Mapping
High-quality phase maps were reconstructed from the multi-channel complex signals using an automated, data-driven coil combination method. Specifically, a conjugate virtual-body-coil map was reconstructed by computing the singular value decomposition (SVD) across the channels, and then taking the dominant singular vector as the body coil reference. SVD compressed data were sorted in decreasing order of virtual channel eigen-values, and input into ESPIRiT for the estimation of coil sensitivities [262,270]. QSM was computed using the superfast dipole inversion approach which employs(a) sophisti-cated harmonic artifact reduction for phase (SHARP) to eliminate background field con-tributions and threshold k-space division (TKD) for calculating magnetic susceptibility [273]. All QSM data were referenced to median cerebrospinal fluid susceptibility, which was computed within a subject specific mask of the lateral ventricles [219]. Motivated by previous morphometric, functional, spectroscopic, genetic and post-mortem work, carefully delineated masks of deep grey matter nuclei were generated as described the following section. Target regions of interest included the striatum (caudate-putamen), pallidum, thalamus, substantia nigra, subthalamic nucleus, red nucleus, dentate nucleus.
An overview the data processing scheme is illustrated in Figure9.1.
Iron 120
Figure 9.1: Processing and analysis framework utilized to obtain high-quality quantitative susceptibility maps and 1H-MR spectra.
9.3.5 Masking of Subcortical Matter Nuclei
Masks of the striatum (caudate-putamen), globus pallidus and thalamus were obtained via the FSL-FIRST Bayesian model-based subcortical segmentation algorithm [297], which was applied on optimized hybrid-contrast MP2RAGE-QSM images [370]. Robust co-registration between skull-stripped MP2RAGE and Fast Low-Angle SHot (FLASH) data was achieved using rigid-body linear transformation of the T1-weighted data onto
Iron 121
N4-bias field corrected FLASH magnitude data. Given the difficulty of segmenting brain stem and cerebellar nuclei on T1-weighted data due to lack of contrast, in addition to the infeasibility of performing manual segmentation of multiple nuclei in many sub-jects, we utilized an atlas-based registration approach to achieve accurate delineations of brainstem/cerebellar nuclei. Specifically, the diffeomorphic Greedy-SyN ANTS non-linear transformation model (https://github.com/stnava/ANTs) was used to compute a non-linear transformation warp between MP2RAGE and MNI space, which was used to map each subjects QSM data into standard space for subsequent calculation of a population-specific average image. The standardized QSM template exhibited high con-trast in brainstem/cerebellar regions and was used to carefully delineate masks of the subthalamic nucleus, substantia nigra, the red nucleus and the dentate nucleus. All masks were delineated by the same operator and were subsequently warped back into native QSM space. The same atlas-based registration procedure was applied to obtain subject specific masks of the lateral ventricles, which were used for referencing the QSM data to CSF. All masks were thresholded at 0.5 to ensure maximal inclusion of grey-matter tissue while limiting partial volume effects. Following visual inspection of all the masks for quality, median susceptibility values from all regions of interest were computed for further analysis.
9.3.6 Quality Control
Given that patients with GTS are ultimately characterized by movement, head motion during MR data acquisition may influence voxel intensities and bias group comparisons.
Consequently, we used a step-wise, multivariate outlier detection approach to remove low-quality data based on (a) structural image quality indices calculated on the mag-nitude image and (b) susceptibility values extracted from subcortical nuclei. Specifi-cally, a multivariate robust squared Mahalanobis distance framework was implemented to detect outlier datasets based on the Shannon entropy focus criterion (EFC), which is an index for image ghosting and blurring [371]; the Quality Index 1 (QI1), which is an index for image degradation resulting from bulk motion, residual magnetization, incomplete spoiling and ghosting [372], and the smoothness of voxels calculated as the full-width half maximum (FWHM) of the spatial distribution of image intensity values in voxel units (https://github.com/preprocessed-connectomes-project). This step was implemented on the whole sample and identified one severely affected dataset which was marked for removal (Figure9.2A). To ensure that the remaining datasets did not contain further outliers, multivariate robust squared Mahalanobis distance outlier detection was additionally performed on vectors of median susceptibility values extracted from the sub-cortical masks for each sample separately. This procedure identified four outlier datasets
Iron 122
Table 9.1: Statistical comparisons of magnitude image data quality metrics
Controls GTS CI (95%) Statistic P-Value SNR -0.03±0.47 0.0±0.41 -0.31 to 0.24 U42=229 0.385 CNR 0.18±0.14 0.22±0.16 -0.14 to 0.05 U42=209 0.223 FBER 2.25±0.84 2.14±0.88 -0.42 to 0.65 U42=215 0.267 EFC 0.45±0.07 0.46±0.05 -0.04 to 0.03 U42=217 0.283 QI-1 0.08±0.05 0.1±0.05 -0.05 to 0.01 U42=200 0.165 FWHM 2.1±0.13 2.13±0.18 -0.13 to 0.07 U42=211 0.237
Abbreviations: SNR = Signal-to-noise ratio; CNR = Contrast-to-noise ratio; FBER
= Foreground-to-background ratio; EFC = Entropy Focus Criterion; QI-1 = Quality Index 1; FWHM = Voxel Smoothness
within the patient sample which were marked for removal (Figure9.2B). Following qual-ity control, group comparison of magnitude image qualqual-ity metrics (signal-to-noise ratio, contrast-to-noise ratio, voxel smoothness EFC, QI1) revealed no significant differences between patients and controls (Table 9.1).
9.3.7 Statistical Analysis
Statistical analysis was performed in the Python programming language (Scipy v.0.15.1 and Statsmodels v.0.6.1). The normality of distribution and homogeneity of variance were assessed via the Kolomgrov-Simrnov and Levene’s tests, respectively. Group differ-ences in serum ferritin values were assessed using Welche’s test given the inhomogeneous variance. Susceptibility data tended to exhibit non-parametric distributions, and as such, group differences of median susceptibility values for each region-of-interest were assessed using Mann-Whitney-Wilcoxon rank sum tests. The significance threshold for initial exploratory analyses of susceptibility values from combined regions of interest (basal ganglia, brainstem, all nuclei) was set to P<0.05 uncorrected. Group comparisons of magnetic susceptibility data from distinct nuclei were corrected using False-discovery-rate multiple comparison correction. A multiple linear regression model accounting for age, gender and two indices of image quality (EFC, QI1) was used to examine the rela-tionship exhibited by susceptibility with (a) ferritin, (b) Gln:Glu ratio and (b) clinical scores. The variance inflation factor was used to assess multi-collinearity between pre-dictor variables.
Iron 123
Figure 9.2: QQ plots of the multivariate outlier detection technique imple-mented via squared Mahalanobis distance. (A) Multivariate outlier detection implemented on the magnitude image quality metrics of the whole sample. (B) Multi-variate outlier detection implemented on the susceptibility values of the patient sample.
Iron 124
9.4 Results
Twenty-two healthy control subjects and 28 patients with GTS took part in the study.
Multi-channel phase data acquired using a FLASH sequence at 3T were combined using a data driven method to generate high-quality phase maps that were used to reconstruct the QSM images. Median values of relative magnetic susceptibility were estimated from a set of carefully delineated subcortical masks to index iron in the brain. Collection of susceptibility weighted data from one patient was not completed due to claustrophobia.
All data underwent rigorous quality control to ensure the reliability of the data input into the statistical models (see section 9.3.6). The remaining samples, which included 22 healthy controls (18-65 years, 5 female) and 22 patients with GTS (18-65 years, 4 female), were comparable in terms of age (t42=0.85,P=0.40), gender (odds ratio= 0.76
All data underwent rigorous quality control to ensure the reliability of the data input into the statistical models (see section 9.3.6). The remaining samples, which included 22 healthy controls (18-65 years, 5 female) and 22 patients with GTS (18-65 years, 4 female), were comparable in terms of age (t42=0.85,P=0.40), gender (odds ratio= 0.76