Iron-related gene expression associated with magnetic susceptibility reductions: Application to the pathophysiology of a movement disorder population
Ahmad Seif Kanaan1,2, Alfred Anwander1, Riccardo Metere1, Andreas Schäfer3, Torsten Schlumm1, Jamie Near4, Berkin Bilgic5, Kirsten Müller-Vahl2, and Harald Möller1

1Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Department of Psychiatry, Hannover Medical School, Hannover, Germany, 3Siemens Healthcare, Erlangen, Germany, 4Douglas Mental Health Institute, McGill University, Montreal, QC, Canada, 5Massachusetts General Hospital, Harvard Medical School, Charlestown, MA, United States


We employ a genetic-imaging approach to examine the underlying genetic basis of magnetic susceptibility reductions at a major locus of pathophysiology in Gilles de la Tourette syndrome (GTS). Voxel-wise statistical differences of motor-striatal susceptibility exhibited significant associations with the expression profile of iron-related gene-sets extracted from the Allen Human Brain Atlas, thus suggesting that the expression of iron-related genes coincides with patterns of susceptibility reductions in GTS. This work supports previous studies relating magnetic susceptibility to brain iron and provides an example of an analytic strategy in which valuable insights can be gleaned by exploring associations between gene-expression and image-derived phenotypes.


Gilles de la Tourette syndrome (GTS) is a neuropsychiatric movement disorder characterized by tics with reported abnormalities in the neurotransmission of dopamine, GABA and glutamate (1, 2). Given that iron plays an integral role in varied biochemical processes involved in neurotransmitter synthesis and transport (3), we hypothesized that iron exhibits a role in GTS pathophysiology. Utilizing Quantitative Susceptibility Mapping (QSM) as a surrogate measure of iron, we showed that GTS patients exhibit magnetic susceptibility reductions in subcortical regions implicated in disease pathophysiology (Fig. 1) (4). To explore the underlying genetic basis of these reductions, we employed an imaging-genetic approach to assess relationships between voxel-wise, nucleus specific, susceptibility differences with the default expression profile of iron-related gene-sets extracted from the Allen Human Brain Atlas (AHBA) (5). Given that genetic transcriptional profiles are known to exhibit distinct expression patterns in the brain, we aimed to investigate spatially specific relationships exhibited between susceptibility and gene expression patterns to glean further insights into pathophysiological mechanisms of iron-related changes. To explore the relevance of these reductions to the clinical population, we additionally employed a machine learning approach to investigate associations with clinical symptomatology.


QSM and MP2RAGE data were acquired from 28 GTS patients and 26 age/gender matched healthy controls on a 3T Siemens MAGNETOM Verio using a 32-channel head coil. A 10ml blood sample was collected from each subject for the quantitation of serum Ferritin, in addition to a comprehensive clinical assessment battery. Susceptibility-weighted data were acquired using FLASH (TR=30ms; TE=17ms; flip-angle=13°; 0.8mm isotropic nominal resolution). High-quality phase maps were reconstructed using data-driven coil combination (6) and QSM images were computed using the SDI approach (7) with referencing to lateral ventricle CSF (8). To investigate associations between magnetic susceptibility and clinical symptoms, we first decomposed the clinical data into a set of clinical scores using Principal Component Analysis (PCA). To interrogate genetic mechanisms that may drive abnormalities in iron levels, we employed a cross-correlation approach to examine the relationship between susceptibility differences with gene expression profiles extracted from the AHBA. Voxel-wise susceptibility difference statistical maps were calculated via nonparametric permutation testing while accounting for age, gender and image quality (FSL-randomise, 10,000 permutations). Transcriptional levels of genes incorporated within four iron-related gene sets (iron-homeostasis, iron-deficiency; iron transport and uptake and Iron-storage (9, 10)) were then extracted from loci of pathophysiology at specific coordinates sampled in the AHBA. For each gene-set, principal components were extracted and cross-correlated to statistical values of striatal susceptibility differences in same coordinate space (MNI). To evaluate whether the observed correlations were significant, a permutation-based approach was implemented in which the null distribution was constructed using a re-sampling based approach (10,000 permutations). For each permutation, Pearson correlation was calculated between magnetic susceptibility differences and the average gene-expression value of random set of genes with an equal size to the gene set of interest (alpha level of 0.05).


The correlation matrix between all the acquired clinical variables revealed sufficient complementary for data-reduction using PCA (Fig 2A). PCA yielded a set 4 components explaining 77% of the variance that were interpreted as representing scores for (i) depression/anxiety; (ii) motor-tics, (iii) obsessions/compulsions, (iv) attention-deficits/hyperactivity (Fig. 2B). Regression analysis between the motor-tic score and surrogate measures of iron revealed a trend with serum ferritin levels and a significant negative association with striatal susceptibility (Fig. 2C). Driven by these results, iron-related gene expression profiles were extracted within three functionally distinct sub-territories of the striatum (motor, associative, limbic) and cross-correlated with statistical maps of magnetic susceptibility reductions at the same coordinates. Permutation based inference revealed significant positive associations between striatal-motor susceptibility and the principal components of the iron-related gene-sets (Fig. 3). Inspection of associations between the mean expression profile of the iron-related gene-sets and magnetic susceptibility statistical values, revealed similar findings. These results indicate that iron-related abnormalities in the motor sub-division of the striatum exhibit a major role in the pathophysiology of GTS.


We demonstrate a link between magnetic susceptibility reductions and default expression profiles of iron-related genes within a major locus of pathophysiology in GTS. These findings suggest that the expression profiles of iron-related genes coincide with patterns of susceptibility reductions in patients with GTS, thus providing a link between disrupted iron homeostasis and GTS pathophysiology. This work supports previous studies relating magnetic susceptibility to brain iron and provides an example of an analytic strategy in which valuable insights on disease pathophysiology can be achieved by exploring associations between genetic transcriptional profiles and image derived phenotypes.


This work was funded by the FP7 Marie Curie Actions of the European Commission (“TS-EUROTRAIN FP7-PEOPLE-2012-ITN, Grant No. 316978”) and, in part, by the Helmholtz Alliance “ICEMED: Imaging and Curing EnvironmentalMetabolic Diseases”.


  1. Kanaan AS, Gerasch S, García-García I, Lampe L, Pampel A, Anwander A, et al. (2017): Pathological glutamatergic neurotransmission in Gilles de la Tourette syndrome. BRAIN. 140: 218–234.
  2. Singer H (2013): The Neurochemistry of Tourette Syndrome. In: Martino D, Leckman JF, editors. Tourette Syndr. Oxford University Press, pp 276–297.
  3. Bianco L, Unger E, Beard J (2010): Iron Deficiency and Neuropharmacology. In: Yehuda S, Mostofsky DI, editors. Iron Defic Overload From Basic Biol to Clin Med. Totowa, NJ: Humana Press, pp 141–158.
  4. Kanaan AS, Anwander A, Schäfer A, Bilgic B, Schlumm T, Near J, et al. (2017): QSM meets MRS: The influence of subcortical iron on glutamatergic neurotransmission in a movement disorder population. Proc 25th Annu Meet ISMRM. p 4649.
  5. Hawrylycz MJ, Lein ES, Guillozet-Bongaarts AL, Shen EH, Ng L, Miller JA, et al. (2012): An anatomically comprehensive atlas of the adult human brain transcriptome. Nature. 489: 391–399.
  6. Bilgic B, Polimeni JR, Wald LL, Setsompop K, States U (2016): Automated tissue phase and QSM estimation from multichannel data. Proc Intl Soc Mag Reson Med 24. Singapore, Singapore. doi: 10.1016/j.neuroimage.2012.05.067.4.
  7. Schweser F, Deistung A, Sommer K, Reichenbach JR (2013): Toward online reconstruction of quantitative susceptibility maps: Superfast dipole inversion. Magn Reson Med. 69: 1582–1594.
  8. Straub S, Schneider TM, Emmerich J, Freitag MT, Ziener CH, Schlemmer HP, et al. (2016): Suitable reference tissues for quantitative susceptibility mapping of the brain. Magn Reson Med. doi: 10.10. doi: 10.1002/mrm.26369.
  9. Clardy S, Wang X, Zhao W, Liu W, Chase G, Beard J, et al. (2006): Acute and chronic effects of developmental iron deficiency on mRNA expression patterns in the brain. J Neural Transm. 71: 173–96.
  10. Hower V, Mendes P, Torti FM, Laubenbacher R, Akman S, Torti S V (2009): A general map of iron metabolism and tissue-specific subnetworks w. 422–443.


Fig1. Representative images illustrating the quality of (A) the raw magnitude and phase single-channel data, (B) signal phase following multi-channel coil combination via SVD-ESPIRiT, and (C-G) QSM following estimation using the SDI approach. Basal Ganglia masks were generated using FSL-FIRST performed on QSM-MP2RAGE hybrid contrast images. Brainstem/cerebellar masks were generated by non-linear transformation of atlas-based masks that were carefully delineated on a QSM group average template in MNI space. The quality of segmentation outputs is illustrated in the mid-panel. Group comparison statistics of nucleus susceptibility means revealed significant reductions in GTS (lower panel). * P-FDR <0.05; ** P-FDR <0.1.

Fig2. (A) PCA was implemented to reduce the dimensionality of the acquired clinical battery. The correlation matrix between all the variables revealed sufficient complementary for data-reduction using PCA. (B) PCA yielded a set 4 components representing (PC1) depression/anxiety; (PC2) motor-tics, (PC3) obsessions/compulsions, (PC4) attention-deficits/hyperactivity, as shown by polar plot representing the absolute component loadings. (C) Regression analysis between the motor-tic score and surrogate measures of iron reveled a trend for negative association with serum ferritin levels and a significant negative association with striatal susceptibility.

Fig 3. (A) Illustration of the motor, limbic and executive striatal sub-divisions and the location of tissue samples extracted by AHBA. (B) Polar plots illustrating the absolute component loadings of iron-related gene-sets. (C) Gene expression scores revealed strong correlations with statistical values of striatal-motor susceptibility differences extracted from the same coordinates of tissue samples extracted by AHBA. (D) A permutation based approach was implemented to assess significance by constructing a null-distribution between magnetic susceptibility and the average gene-expression profile of a random set of genes (equal in size to the gene-set of interest) using a resampling approach with 10,000 permutations.

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)