University of Southern California Dissertations and Theses

Unveiling the white matter microstructure in 22q11.2 deletion syndrome with diffusion magnetic resonance imaging

(USC Thesis Other)

Unveiling the white matter microstructure in 22q11.2 deletion syndrome with diffusion magnetic resonance imaging

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Unveiling the White Matter Microstructure in 22q11.2 Deletion Syndrome with Diffusion Magnetic Resonance Imaging Julio Ernesto Villalón Reina Department of Biomedical Engineering Viterbi School of Engineering Faculty of the USC Graduate School University of Southern California Doctor of Philosophy August, 2019 Defense Committee Members Paul M. Thompson PhD. Professor of Ophthalmology, Neurology, Psychiatry and the Behavioral Sciences, Radiology and Engineering, Imaging Genetics Center, Mark and Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine of USC, University of Southern California. Natasha Lepore PhD. Associate Professor, Department of Radiology, Children’s Hospital Los Angeles, University of Southern California. Vasilis Marmarelis PhD. Dean's Professor of Biomedical Engineering and Professor of Biomedical Engineering, University of Southern California. Neda Jahanshad PhD. Assistant professor, Department of Neurology and Biomedical Engineering, Imaging Genetics Center, Mark and Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine of USC, University of Southern California. 1 The research presented here was performed at the USC Imaging Genetics Center (USC Mark and Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine of USC) in close collaboration with Professor Carrie E. Bearden of the Department of Psychiatry of the University of California Los Angeles (UCLA), the International 22q11.2 Brain Behavior Consortium (22q-IBBC) and the Enhancing Neuroimaging Genetics through Meta Analysis Consortium (ENIGMA). 2 Table of Contents 1. Introduction 5 2. Summary of Research 14 2.1. Related Publications 14 Journal papers 14 Conference paper 14 Conference abstracts 15 2.2. Additional Publications 16 Journal paper 16 Conference papers 16 Conference abstract 17 3. Altered White Matter Microstructure in 22q11.2 Deletion Syndrome: A Multi-Site Diffusion Tensor Imaging Study 18 3.1. Abstract 18 3.2. Introduction 19 3.3. Methods 20 3.3.1. Participants 20 3.3.2. Measurements of sample-specific phenotype characteristics 22 3.3.3. Image acquisition and processing 23 3.3.4. Statistical Analyses 24 Effects of 22q11DS and age on DTI-derived measures 24 Influence of psychotic disorder, deletion size and IQ on DTI measures 26 3.4. Results 28 3.4.1. Group differences across sites 28 3.4.2. Age-associated effects 29 3.4.3. Influence of psychosis 30 3.4.4. Comparison of WM Microstructure in 22q11DS-Psychosis to Idiopathic Schizophrenia 31 3.4.5. Influence of deletion type and IQ 33 3.5. Discussion 33 3.6. Acknowledgments and Funding Sources 38 3 4. Alternative diffusion anisotropy measures for the investigation of white matter alterations in 22q11.2 deletion syndrome 41 4.1. Abstract 41 4.2. Introduction 42 4.3. Methods 43 4.3.1. Participants 43 4.3.2. Image acquisition and processing 44 4.3.3. Diffusion Anisotropy measures 47 Tensor Distribution Function 47 Generalized Fractional Anisotropy from Q-ball Imaging 49 Anisotropic Power Map 50 4.3.4. Statistical Analysis 51 4.4. Results 51 4.5. Discussion 55 5. Advanced Microstructural Measures from Multi-Shell Diffusion MRI Acquisitions in 22q11.2 Deletion Syndrome 59 5.1. Introduction 59 5.2. Methods 62 5.2.1. Participants 62 5.2.2. Diffusion MRI models 63 Neurite Orientation Dispersion and Density Imaging (NODDI) 63 Constrained Spherical Deconvolution 65 5.2.3. Statistical Analysis 68 NODDI-derived Measures 68 Hindrance Modulated Orientational Anisotropy (OA) 69 5.3. Results 69 6. Conclusions 74 7. Supplements 75 Supplementary Figures (Chapter 3) 75 Supplementary Tables (Chapter 3) 82 References 4 1. Introduction 22q11.2 deletion syndrome (22q11DS; also known as Velocardiofacial or DiGeorge syndrome) results from a recurrent 1.5-3 megabase (Mb) microdeletion on locus 11.2 of the long arm of chromosome 22. It is estimated to result from de-novo mutations in 1 in 1000 fetuses. Since the appearance of genetic screening tests in the 1980’s, its prevalence has been estimated in 1 in 3000 to 6000 live births 1,2 . 22q11DS is associated with a range of characteristic abnormalities, including cardiac defects, craniofacial anomalies, and intellectual disability 1,3 . Notably, 22q11DS is among the conditions with increased prevalence of severe cardiovascular conditions, such as interrupted aortic arch type B (52%), truncus arteriosus (34%), tetralogy of Fallot (16%) and ventricular septal defects (5-10%). Particularly, it increases the risk for psychotic illness around 25-fold relative to the general population 2,4–6 . It is also associated with developmental disabilities (2-3%), cognitive decline, autism spectrum disorders (ASD) and attention deficit-hyperactivity disorder (ADHD) 5 , but the increased risk for psychosis in 22q11DS may be the most specific association, as it greatly exceeds the roughly 3-fold increased risk of psychosis associated with general developmental delay 7,8 . Markedly, mouse models of the 22q11.2 deletion show fewer neural progenitors of projection neurons in cortical layers 2/3, which leads to altered connectivity between cortical association areas 9 . Consequently, 22q11DS is a compelling model to study genetic causes and neural mechanisms underlying disorders of cortical circuit development such as schizophrenia. The non-invasive study of the brain’s anatomy and function in 22q11DS may help in discovering the altered patterns of connectivity underlying these disorders. The brain’s white matter holds a large proportion of the long distance axonal bundles connecting cortical association areas, particularly fronto-temporo-parietal long association tracts as well as inter-hemispheric commissural tracts. White matter microstructural properties can be quantified non-invasively in humans using diffusion weighted Magnetic Resonance Imaging (dMRI). dMRI is an evolving imaging technique that is at the core of the non-invasive characterization of the brain’s microstructure 10 . By unveiling the microstructure of the white matter and its histopathology non-invasively with dMRI, one may infer disrupted patterns of the connectivity of the brain. 5 dMRI is an MRI technique that is sensitive to the Brownian motion of water protons. In body tissues, diffusion is hindered by the cellular membranes 11 , a property that helps to reveal microstructural properties of brain tissue, offering tissue contrast not available with standard anatomical MRI such as T1- and T2-weighted images. Particularly in the brain, the presence of axonal membranes and myelin originating from oligodendrocytes tend to make diffusion anisotropic, due to the organized architecture of the brain following the connectivity patterns between brain areas with densely packed axonal bundles embedded in a myelin parenchyma. Diffusion anisotropy is perhaps the tissue property that has drawn the most attention since the early days of dMRI. When diffusion-sensitized gradients are applied to the brain, the magnetic resonance signal decays more rapidly when the gradient is aligned with the dominant direction of water diffusion, such as along the white matter fiber tracts 12 . This interesting attribute of the white matter coincides with the formal definition of anisotropy, which is the quality of any medium or substance of exhibiting different features when measured along axes with different directions. Precisely, there is anisotropy in dMRI when water diffusion is not equal in all directions (contrary to isotropy). This is clearly observed along the myelinated axonal bundles of the brain, more so than in any other tissue of the body 13 . How is diffusion measured with MRI? What are diffusion-sensitized gradients? Stejskal and Tanner proposed in 1965 a solution to the Bloch-Torrey partial differential equations for a symmetric set of pulsed gradients 14 . They realized that the introduction of a pair of pulsed gradients to the spin echo pulse sequence provides greater sensitivity to the diffusion of protons and is currently better known as the Pulsed Gradient Spin Echo (PGSE) sequence (see figure 1 below). Stejskal and Tanner’s solutions yielded the basic equations that are at the foundation of the most commonly used diffusion reconstruction models to this day: diffusion tensor imaging (DTI), constrained spherical deconvolution (CSD), q-ball imaging (QBI) to mention a few. The so-called Stejskal & Tanner formula states that: (1) where: (2) 6 Figure 1. Stejskal and Tanner pulse sequence, also known as Pulsed Gradient Spin Echo (PGSE) sequence (Figure taken from: Rutger Fick. Advanced dMRI Signal Modeling for Tissue Microstructure Characterization. Medical Imaging. Université Côte d’Azur; Inria Sophia Antipolis, 2017. https://hal.archives-ouvertes.fr/tel-01534104) In equations (1) and (2), is the signal deriving from a diffusion sensitized spin echo sequence, is the strength of the pair of gradients, each of a duration , and is the delay between them. See Figure 1 for reference. is the signal from an identical experiment but without the sensitized diffusion gradients. is commonly referred to as the b-value in the dMRI literature. is the gyromagnetic ratio (2.765x10 8 rad/s, T). is the diffusion constant that is obtained from two measurements (i.e. and ). , and are parameters that can be controlled by us and are all condensed in . Equation (1) determines an exponential decay of the signal as increases. As the gradient strength increases, the diffusion weighting increases and so does the signal attenuation. When repeating the PGSE experiment by applying the diffusion gradients along at least six non-collinear directions and one , it is possible to obtain 6 diffusion constants from equation (1). Due to the presence of membranes in the living tissue, diffusion does not follow a proper Gaussian profile as in free water at 37 ºC, hence the diffusion constant calculated is more formally called “Apparent Diffusion Constant” (ADC). the average ADC in the brain is typically 0.8-0.9x10 -3 mm 2 /s. The six ADC 7 measurements are combined in a second-order 3x3 symmetric tensor, also called the diffusion tensor: By diagonalizing the diffusion tensor one derives three eigenvalues and three eigenvectors from which a diffusion ellipsoid is characterized. The ellipsoid represents the isosurface of mean displacement of water protons. This is the foundation of Diffusion Tensor Imaging (DTI) 15,16 . The DTI ellipsoid was the first approach to describe anisotropic diffusion in the brain. Several assumptions are made in DTI, the most important one being the Gaussian profile of diffusion in living tissue and that anisotropic diffusion is elliptic. At current dMRI resolution most voxels of the brain contain multiple fiber configurations, such as crossing, bending and fanning, which may in turn result in non-Gaussian behavior of the diffusion signal, and the ellipsoid is insufficient to model multiple fiber compartments. Additionally, as the gradient strength (i.e. higher b-value) increases the signal decay increases in areas of low diffusion constant, especially fibers perpendicular to the axis of the measurement, and in the restricted diffusion inside axons. This sensitivity to multiple water compartments, i.e. intra-cellular and extra-cellular compartments, will also violate the Gaussian assumptions of diffusion. Since the early days of DTI many recognized the limitations of this model 16 . The DTI model is based on the assumption of Gaussian free diffusion, which is accurate only for homogeneous barrier-free pools of water, which is not the case for the brain beyond the ventricles of the brain. As exposed in the previous chapter, the external boundaries of cells in the brain offer multiple barriers to the free movement of water. These limitations become evident in two concrete scenarios. One of them is the existence of multiple fiber configurations across the brain that diverge from the idealized “single fiber bundle”, as is the case in the mid-sagittal corpus callosum. These fiber configurations include areas where two white matter bundles cross each other, kissing fibers, where two bundles brush and touch each other, diverging or “fanning” fibers and curving or “bending” fibers. All of these configurations generate multiple fiber directions when fitting a diffusion tensor, which will yield more isotropic tensors, and hence spuriously decrease the value of 8 the fractional anisotropy (FA). Moreover, at the current resolution of dMRI images, around 90% of the voxels have crossing fibers as demonstrated by Jeurissen (2012) 17 . The other scenario where DTI becomes insufficient to describe the microstructure is when the diffusion MRI (dMRI) is acquired with higher gradient strengths. When increasing the b-value above 1500 s/mm 2 , the signal decay of diffusion becomes non-mono-exponential. This is caused by the non-Gaussian properties of the restricted water, which corresponds to the slower component of the signal attenuation. In this case the assumption of Gaussianity as it applies to the hindered diffusion of the extra-cellular compartment does not apply. Several scalar indices can be derived from the tensor eigenvalues. These are fractional anisotropy (FA), mean diffusivity or trace (MD), axial diffusivity (AD), and radial diffusivity (RD). AD is defined as , which corresponds to the main eigenvector of the tensor. In a perfectly prolate tensor would correspond the average diffusion constant parallel to single axonal bundle. Radial or perpendicular diffusivity is defined as the mean of and : MD or average ADC is defined as one-third of the trace of the tensor : FA is defined as the normalized variance of the eigenvalues: 9 FA is the most widely used scalar measure to study the white matter microstructural organization and continues to be popular despite the known limitations of DTI. FA is normalized between 0 and 1, and in the presence of a single axonal bundle it will tend to be closer to 1 and it is seen as a marker of the cohesiveness of the axonal bundle wrapped in healthy myelin sheets. This interpretation fails with the increase of the complexity of the fiber configurations or fiber mixtures, where FA will tend to decrease, due to a more isotropic tensor ellipsoid. Other DTI indices, i.e. AD, RD and MD are used along with FA and it is also altered in a range of brain diseases 18 . Lower AD can reflect axonal damage and degeneration 19 , or reduced axonal diameter 20 . RD is associated with the amount of inter-axonal spacing (i.e., extra-cellular space) 20 . In animal models, induced demyelination and dysmyelination can lead to abnormally high RD 21–23 . MD is a generalized measure of the surface-to-volume ratio of cellular membranes 24 . AD, RD and MD are also greatly affected in the same manner as FA when violating the assumption of Gaussianity, such as fiber mixtures inside a voxel. Other proposed models overcome some limitations of DTI, including multi-tensor models, such as the tensor distribution function (TDF) 25,26 , q -ball imaging (QBI) 27,28 , and constrained spherical deconvolution 29 , among others. Each model leads to its own scalar anisotropy measure and many offer a richer understanding of WM microstructure than FA or other measures derived from the single tensor model (e.g. MD, RD and AD). Despite the numerous studies that have investigated DTI-derived measures in 22q11DS, there have been no conclusive findings, partly due to the small samples used (n<100, including patients and controls 30,31 ) or the heterogeneity of techniques used. Additionally, all reports to date have exclusively used DTI, which is not the most accurate dMRI technique to study the microstructure of the brain due to the limitations previously mentioned. The main hypothesis of this thesis is that subjects with 22q11DS have an altered white matter microstructure, caused by an altered axonal and myelin development that ultimately modify cortico-cortical connections becoming the substrate for psychiatric illness. Throughout this thesis I present a series of dMRI-based techniques coupled with modern statistical methods to describe the brain’s microstructure in 22q11DS and their correlation to cognitive and clinical markers. The outline of the thesis is as follows: 10 Chapter 3 – This chapter focuses on the study of DTI-derived measures on a large sample of 22q11DS subjects and age matched healthy controls. The main idea of this chapter is to tackle the lack of consensus of the findings across single site studies. We use the ENIGMA consortium to pool datasets from nine research centers around the world of 22q11DS and age matched healthy controls. ENIGMA is a global network of researchers in imaging and genomics to understand brain structure, function, and disease, based on brain imaging and genetic data ( http://enigma.ini.usc.edu/ ). All collaborating sites of the ENIGMA-22q working group shared imaging, genetic and clinical data in order to conduct this study. We preprocessed and statistically compared DTI-derived measures between groups by using the established ENIGMA-DTI protocol 32 . We also studied the influence of the clinical status (psychosis vs. no psychosis), the deletion type (i.e. longer vs. shorter microdeletions) and cognitive performance (measured by full scale IQ) on DTI-derived measures in the 22q11DS population. We hypothesized that widespread areas of the white matter would show an altered pattern of anisotropy and diffusivity compared to controls. Additionally, based on previous literature 30,33–35 , we hypothesized that 22q11DS probands have lower DTI diffusivity across the brain. We also hypothesized that the 22q11DS subjects with stronger cognitive decline, with psychotic disorder and with longer deletions have greater microstructural abnormalities than the ones without. Chapter 4 – Do alternative anisotropy measures, other than the standard DTI FA, yield different results when comparing 22q11DS probands and healthy controls? Are they more sensitive to changes in WM microstructure when comparing subgroups of 22q11DS? Here we analyze the ENIGMA multi-site dataset introduced in chapter 3 to examine alternative anisotropy measures that are derived from higher order reconstruction models that can be fit on the single-shell or one b-valued dMRI acquisitions. Specifically, the models we fit in this chapter are the tensor distribution function 26 , and Q-ball imaging 27 . From the tensor distribution function we derive a scalar anisotropy measure also known as TDF-FA. From the Q-ball imaging model we derive two measures, Generalized Fractional Anisotropy (GFA) and the Anisotropic Power Maps (AP). All these measures, as opposed to the single tensor model, take into account the effect of the mixture of fiber compartments or complex fiber configurations such as bending and fanning fibers. Here we use a similar statistical framework as in chapter 3 to study these measures in 22q11DS. Additionally, it is possible that due to the impaired axonal development in 22q11Ds, the higher FA values in DTI, compared to controls, may be caused by a decrease in the complexity of the WM, i.e. fewer crossing, dispersing, bending and kissing fibers. This would spuriously increase the FA in 11 22q11DS. The comparison of anisotropy measures from higher order models should help to address this question. Chapter 5 – Can higher order models from multi-shell (i.e multiple b-valued) diffusion MRI acquisitions better explain the elevated anisotropy as revealed by DTI? Can they aid in the explanation of the lower diffusivity in 22q11DS? In the previous chapters we have shown that by studying the white matter of 22q11DS with DTI we can replicate findings across multiple sites. Moreover, we have shown that the microstructure of the white matter is impaired in 22q11DS and that even though DTI measures give us a considerable insight into the microstructural changes, more advanced dMRI-based methods may be needed to better understand the underlying histological changes in the brain. Accordingly, biophysical dMRI models allow us to compartmentalize and analyze the changes in intra- and extra-cellular volume fractions (ICVF and ECVF, respectively) inside each voxel. Additional properties can be extracted from biophysical models, such as the dispersion of neurites (i.e. axons and dendrites), which is complementary to anisotropy. For instance, in this chapter we focus on the analysis of NODDI-derived measures. NODDI, which stands for Neurite Orientation and Density Imaging was initially proposed by Zhang et al 36 . We use the NODDI model to compare the ICVF as well as the dispersion of axons in the white matter between 22q11DS and healthy controls. Importantly, biophysical models such as NODDI are only feasible and reliable in dMRI acquisitions with multiple b-values, a.k.a. multi-shell dMRI. Here, we hypothesize that the ICVF is positively correlated with DTI FA and that the differences seen in DTI FA in 22q11DS are substantially caused by an increase in ICVF. More so, in the context of multiple fiber directions in one voxel, which occurs in 90% of the voxels 17 , it is possible to calculate the anisotropy value for each fiber compartment in a voxel. Each fiber compartment in a voxel is called a “fixel”. This type of anisotropy is most commonly known as Hindrance Modulated Orientational Anisotropy (OA) derived from the Constrained Spherical Deconvolution (CSD) model 37,38 . By using this information, it is possible to isolate in one voxel the anisotropy values of one fiber compartment without the interference of another crossing fiber bundle. For instance, in regions such as in the corona radiata three major systems of fibers cross, namely the longitudinal association fibers, the cortico-spinal tracts and callosal fibers. By using this technique, it is possible to extract a more accurate anisotropy measure of each of these tracts separately, without the bias of having multiple fiber compartments associated to only one anisotropy value for a voxel. We hypothesize that the 12 callosal fibers have a different orientational anisotropy (OA) than the cortico-fugal fibers, due to their distinct cortical origins, outer cortical layers and inner cortical layers, respectively. One additional note needs to be made here. In the previous chapters, the dMRI-derived measures (e.g. FA, TDF-FA, GFA) were suitable for an analysis based on the ENIGMA-DTI protocol. This protocol is a Region-of-Interest (ROI) based protocol. These measures are all rotation invariant and hence they are suitable for an analysis based on ROIs. OA instead is derived from CSD and is a directional anisotropy measure, meaning that each fixel will have a different anisotropy value. Hence, there is an additional orientational information, which requires another type of analysis, such as the one proposed by Raffelt et al (2017), who coined the term “fixel-based analysis” (FBA). We perform FBA along two major fiber systems (i.e. long association fibers and cortico-fugal fibers) and compare OA between 22q11DS subjects and healthy controls. 13 2. Summary of Research 2.1. Related Publications Journal papers Julio Villlalón-Reina & the ENIGMA 22q Consortium. ‘Altered White Matter Microstructure in 22q11.2 Deletion Syndrome: A Multi-Site Diffusion Tensor Imaging Study’ . Molecular Psychiatry (published: July 29, 2019) DOI: 10.1038/s41380-019-0450-0 Julio Villalón-Reina, Neda Jahanshad, Elliott A Beaton, Arthur W. Toga, Paul M. Thompson, Tony J. Simon. ‘White matter microstructural abnormalities in girls with chromosome 22q11.2 deletion syndrome, Fragile X or Turner syndrome as evidenced by diffusion tensor imaging’ . NeuroImage (2013); 81: 441–454. DOI: 10.1016/j.neuroimage.2013.04.028 Conference paper Julio Villalón-Reina & the ENIGMA 22q Consortium. Alternative diffusion anisotropy measures for the investigation of white matter alterations in 22q11.2 deletion syndrome. 14th International Symposium on Medical Information Processing and Analysis (Mazatlán, Mexico, October 24-26, 2018). Published in SPIE Proceedings Vol. 10975. Eduardo Romero; Natasha Lepore; Jorge Brieva (eds.). DOI: 10.1117/12.2513788 Conference abstracts Julio Villalón-Reina , Talia M. Nir, Neda Jahanshad, Leila Kushan, Carrie E. Bearden, Paul M. Thompson. ‘ Myelin and G-ratio Imaging in 22q11.2 Deletion syndrome: A Pilot Study’ . Society for Neuroscience (SfN’19), Chicago, Illinois, United States of America, October 19-23, 2019. Selected for a plenary talk. Julio Villalón-Reina , Talia M. Nir, Neda Jahanshad, Leila Kushan, Carrie E. Bearden, Paul M. Thompson. ‘ Altered Intracellular Volume Fraction and Neurite Dispersion in People with Chromosome 22q11.2 Copy Number Variants’ . The 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’19). Berlin, Germany, July 23-27, 2019. 14 Julio Villalón-Reina , Talia Nir, Neda Jahanshad, Leila Kushan, Christopher Ching, Carrie E. Bearden, Paul M. Thompson. ‘ Cortico-cortical vs Corticospinal Tract differences in 22q11.2 Deletion syndrome: A Fixel-based Analysis’ . Organization of Human Brain Mapping (OHBM’19), Rome, Italy, June 9-13, 2019. Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘ Tensor distribution function fractional anisotropy reveals microstructural disruption across all white matter in 22q11.2 deletion syndrome’ . Society for Neuroscience (SfN’18), San Diego, California, United States of America, November 3-7, 2018. Selected for a plenary talk. Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘Diffusion Tensor Imaging reveals highly atypical white matter in 22q11.2 Deletion Syndrome: Meta- and mega-analysis findings of the ENIGMA consortium’. The 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’18). Honolulu, Hawaii, United States of America, July 17-21, 2018. Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘Highly Atypical White Matter in 22q11.2 Deletion Syndrome: an ENIGMA-DTI Consortium Study’. Organization of Human Brain Mapping (OHBM’18), Singapore, June 17-21, 2018. Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘Diffusion MRI in 22q11.2 Deletion Syndrome: ENIGMA working group meta-analysis findings’. The 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC’17). Jeju, South Korea, July 11-17, 2017. Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘Diffusion Tensor Imaging in 22q11.2 Deletion Syndrome: ENIGMA working group meta-analysis findings’. Organization of Human Brain Mapping (OHBM’17), Vancouver, Canada, June 25-29, 2017. Selected for a plenary talk Julio Villalón-Reina & the ENIGMA 22q Consortium. ‘White Matter differences in 22q11.2 Deletion Syndrome: ENIGMA working group meta-analysis findings.’ Organization of Human Brain Mapping (OHBM’16), Geneva, Switzerland, June 26-30, 2016 Julio Villalón-Reina , Justin Galvis, Conor Corbin, Talia M. Nir, Leila Kushan, Paul M. Thompson, Carrie E. Bearden. ‘Exploring alternative diffusion tensor measures to study atypical development in 22q11.2 deletion syndrome’. Society for Neuroscience (SfN’15), Chicago, Illinois, United States of America, October 17-21, 2015. Julio Villalón-Reina , Kristian Eschenburg, Maria Jalbrzikowski, Leila Kushan, Talia M. Nir, Neda Jahanshad, Paul M. Thompson, Tony J. Simon, Carrie E. Bearden. ‘Structural small worldness is altered in 22q11.2 deletion syndrome: 15 validation using two datasets’. Organization of Human Brain Mapping (OHBM’14). Hamburg, Germany, June 8-12, 2014. Julio Villalón-Reina , Liang Zhan, Talia M. Nir, Kenia Martínez, Kristian Eschenburg, Maria Jalbrzikowski, Carolyn Chow, Carrie E. Bearden, Paul M. Thompson. ‘White matter microstructure in 22q11.2 deletion syndrome revealed by the tensor distribution function’ . Organization of Human Brain Mapping (OHBM’13). Seattle, Washington, United States of America, June 16-20, 2013. 2.2. Additional Publications Journal paper Talia M. Nir* , Julio E. Villalón-Reina* , Gautam Prasad, Neda Jahanshad, Shantanu H. Joshi, Arthur W. Toga, Matt A. Bernstein, Clifford R. Jack Jr., Michael W. Weiner, Paul M. Thompson, Alzheimer's Disease Neuroimaging Initiative (ADNI). ‘ Diffusion weighted imaging-based maximum density path analysis and classification of Alzheimer's disease’ . Neurobiology of Aging (2015); 36, Supplement 1: S132-S140. *Equal contribution DOI: 10.1016/j.neurobiolaging.2014.05.037 Conference papers Julio Villalón-Reina , Talia M. Nir, Boris A. Gutman, Neda Jahanshad, Clifford R. Jack Jr., Michael W. Weiner, Ofer Pasternak, Paul M. Thompson, and for the Alzheimer’s Disease Neuroimaging Initiative (ADNI). ‘Using Multiple Diffusion MRI Measures to Predict Alzheimer’s Disease with a TV-L1 Prior’ . A. Fuster et al. (eds.), Computational Diffusion MRI, Mathematics and Visualization, Springer International Publishing 2017. Selected for a plenary talk. DOI: 10.1007/978-3-319-54130-3_13 Julio Villalón-Reina , Talia M. Nir, Liang Zhan, K. L McMahon, G. I. de Zubicaray, M. j. Wright, Neda Jahanshad, Paul M. Thompson. "Reliability of Structural Connectivity Examined with Four Different Diffusion Reconstruction Methods at Two Different Spatial and Angular Resolutions". A. Fuster et al. (eds.), Computational Diffusion MRI, Mathematics and Visualization, Springer International Publishing 2016. DOI: 10.1007/978-3-319-28588-7_19 Conference abstract 16 Julio Villalón-Reina , Talia M. Nir, Neda Jahanshad, Matt Bernstein, Clifford Jack, Michael Weiner, Artemis Zavaliangos-Petropulu, Paul M. Thompson. ‘Classification of Alzheimer’s Disease using Nearest Shrunken Centroids on white matter DWI features’ . Organization of Human Brain Mapping (OHBM’15), Honolulu, Hawaii, United States of America, June 14-18, 2015. 17 Chapter 3 Altered White Matter Microstructure in 22q11.2 Deletion Syndrome: A Multi-Site Diffusion Tensor Imaging Study 3.1. Abstract 22q11.2 deletion syndrome (22q11DS) - a neurodevelopmental condition caused by a hemizygous deletion on chromosome 22 - is associated with an elevated risk of psychosis and other developmental brain disorders. Prior single-site diffusion magnetic resonance imaging (dMRI) studies have reported altered white matter (WM) microstructure in 22q11DS, but small samples and variable methods have led to contradictory results. Here we present the largest study ever conducted of dMRI-derived measures of WM microstructure in 22q11DS (334 22q11.2 deletion carriers and 260 healthy age- and sex-matched controls; age range 6-52 years). Using harmonization protocols developed by the ENIGMA-DTI working group, we identified widespread reductions in mean, axial and radial diffusivities in 22q11DS, most pronounced in regions with major cortico-cortical and cortico-thalamic fibers: the corona radiata, corpus callosum, superior longitudinal fasciculus, posterior thalamic radiations and sagittal stratum (Cohen’s d ’s ranging from -0.9 to -1.3). Only the posterior limb of the internal capsule (IC), comprised primarily of corticofugal fibers, showed higher axial diffusivity in 22q11DS. 22q11DS patients showed higher mean fractional anisotropy (FA) in callosal and projection fibers (IC and corona radiata) relative to controls, but lower FA than controls in regions with 18 predominantly association fibers. Psychotic illness in 22q11DS was associated with more substantial diffusivity reductions in multiple regions. Overall, these findings indicate large effects of the 22q11.2 deletion on WM microstructure, especially in major cortico-cortical connections. Taken together with findings from animal models, this pattern of abnormalities may reflect disrupted neurogenesis of projection neurons in outer cortical layers. 3.2. Introduction Disturbances in WM microstructural organization have been frequently reported in 22q11DS; however, studies to date have been relatively small, with highly variable findings. While many studies reported lower FA in 22q11DS compared to healthy controls (HC) in major WM tracts, including commissural, association and projection fibers 34,39–43 , several others reported higher overall FA 35,44,45 , or mixed findings across tracts 30,31,33,46–48 . Most studies reported consistent decreases in DTI-derived diffusivity measures (i.e., MD, RD, and AD), although some report mixed results 41 or higher WM diffusivity in 22q11DS 42,43 . Supplementary Table S1 summarizes prior findings. These contrasting reports have hindered conclusions regarding the nature of WM microstructural abnormalities in 22q11DS. Contrasting findings in prior studies may also be due to different analytical techniques, ranging from tract-based spatial statistics (TBSS 49 ) to voxel-wise analyses and tractometry. This technical variability makes it difficult to apply traditional meta-analytic approaches that attempt to combine summary statistics from prior publications. WM differences associated with psychosis are of interest in 22q11DS. Psychotic symptoms in 22q11DS have been associated with higher FA and lower WM diffusivities, but not always in the same regions across studies 30,33,35,43,50,51 . Additionally, there is variability in deletion breakpoints; 85-90% of individuals with the deletion have a ~3 Mb (A-D) deletion, containing 46 protein-coding genes, whereas ~10% of cases have a nested 1.5 Mb (A-B) deletion 1 . WM differences in 22q11DS may be due in part to variable deletion size, as deletion size impacts cortical surface area 52 . To address these uncertainties and determine the factors that affect WM abnormalities in 22q11DS, the 22q11DS Working Group of the Enhancing 19 Neuroimaging Genetics through Meta-analysis Consortium (ENIGMA-22q11DS) performed a coordinated analysis of the raw dMRI data from ten independent studies, and meta-analyzed group differences and their modulators. We addressed these questions: (1) Are there consistent group differences in WM microstructure between 22q11.2 deletion carriers and demographically-matched healthy controls? (2) Are there differential age effects between groups, suggesting altered WM development in 22q11DS? (3) Do 22q11DS participants with a psychotic disorder show more severe WM alterations, and do these differences overlap with those found in idiopathic schizophrenia? (4) Does deletion size impact DTI indices? (5) Is WM microstructure related to cognitive abilities, in 22q11DS and in HC? 3.3. Methods 3.3.1. Participants DMRI data were contributed from ten studies previously acquired by the ENIGMA-22q11DS working group. This analysis included 594 participants: 334 with 22q11DS (mean age: 16.88 ± 6.43, 153 females) and 260 healthy controls (HC; mean age: 16.55 ± 8.01, 123 females). Demographic characteristics are shown in Table 1 and Supplemental Table S2a-S2b . Psychotropic medication status at the time of scanning is included in Supplementary Table S2c . Individual study details are in Supplemental Table S3 . Institutional review boards at participating institutions approved all study procedures, and material transfer agreements approved any sharing of de-identified imaging data. Written informed consent was obtained from all study participants or a legal guardian. 20 Table 1. Demographic information of study participants. Demographic information of study participants, per site. SD = standard deviation; M = Male; F = Female, HC = Healthy Controls. (1) University of Pennsylvania/Children's Hospital of Philadelphia (PA, USA); (2) University California Los Angeles (CA, USA); (3) State University New York Upstate (NY, USA); (4) University of Newcastle (NSW, Australia); (5) Maastricht University (Netherlands); (6) Institute of Psychiatry (London, UK); (7) University California Davis #1 (CA, USA); (8) University California Davis #2 (CA, USA); (9) Cardiff Univ. (WAL, UK); (10) Utrecht Univ. (The Netherlands). Healthy Controls (HC) 22q11.2 DS Site N N (% by Sex) Mean Age (SD) Mean IQ (SD) N N (% by Sex) Mean Age (SD) Mean IQ (SD) Group Differences UPenn 49 30 (61.2%) M; 19 (38.8%) F 17.31 (3.22) — 43 26 (60%) M; 17 (40%) F 17.49 (3.13) 77.16 (10.96) Age: t = 0.27 ( p = 0.79) Sex: X 2 = 0.01 ( p = 0.94) IQ: NA UCLA 32 16 (50%) M; 16 (50%) F 12.59 (5.62) 111.97 (21.69) 49 25 (51%) M; 24 (49%) F 14.69 (5.59) 76.55 (12.61) Age: t = 1.60 ( p = 0.11) Sex: X 2 = 0.02 ( p = 0.89) IQ: t = -9.11 ( p < 0.0005) SUNY Upstate 11 5 (45.45%) M; 6 (54.5%) F 21.12 (2.01) 87.77 (16.25) 34 19 (55.8%) M; 15 (44.11%) F 20.85 (1.86) 78.31 (13.72) Age: t = -0.43 ( p = 0.67) Sex: X 2 = 0.49 ( p = 0.48) IQ: t = -1.9 ( p = 0.06) University of Newcastle 17 8 (47.1%) M; 9 (52.9%) F 17.06 (3.01) 106.63 (17.58) 16 6 (37.5%) M; 10 (62.5%) F 16.63 (2.75) 72.63 (13.45) Age: t = -0.43 ( p = 0.67) Sex: X 2 = 0.31 ( p = 0.58) IQ: t = -6.14 ( p <.0005) Maastrich t University 36 23 (63.8%) M; 13 (36.2%) F 29.97 (10.05) 105.13 (14.13) 24 11 (45.9%) M; 13 (54.1%) F 30.05 (7.86) 74.42 (9.76) Age: t = 0.04 ( p = 0.97) Sex: X 2 = 1.91 ( p = 0.17) IQ: t = -8.14 ( p <.0005) Institute of Psychiatry London 24 10 (41.7) % M; 14 (58.3%) F 18.36 (6.73) 115.92 (15.03) 24 13 (54.1%) M; 11 (45.9%) F 18.04 (6.88) 84.46 (14.15) Age: t = -0.16 ( p = 0.87) Sex: X 2 = 0.75 ( p = 0.39) IQ: t = -7.47 ( p <.0005) UC Davis #1 36 19 (52.7%) M; 17 (47.3%) F 10.22 (2.38) 116.06 (10.69) 31 16 (51.6%) M; 15 (48.4%) F 10.86 (2.14) 73.60 (14.41) Age: t = 1.14 ( p = 0.26) Sex: X 2 = 0.01 ( p = 0.92) IQ: t = -13.23 ( p <.0005) UC Davis #2 41 20 (48.8%) M; 21 (51.2%) F 11.05 (2.33) 115.16 (15.75) 46 21 (45.7%) M; 25 (54.3%) F 11.64 (2.53) 74.76 (13.83) Age: t = 1.11 ( p =. 027) Sex: X 2 = 0.09 ( p = 0.77) IQ: t = -12.45 ( p <.0005) Cardiff 14 6 (42.9%) M; 8 (57.1%) F 14.46 (1.79) 105.25 (9.55) 13 6 (46.2%) M; 7 (53.8%) F 16.03 (4.63) 80.92 (19.30) Age: t = 1.18 ( p = 0.25) Sex: X 2 = 0.03 ( p = 0.86) IQ: t = -3.91 ( p = 0.001) Utrecht — — — -- 54 38 (70.3%) M; 16 (29.7%) F 17.52 (4.22) 69.24 (7.66) NA 21 Total 260 137 (52.6%) M; 123 (47.3%) F 16.55 (8.01) 111.62 (16.16) 334 181 (54.1%) M; 153 (45.8%) F 16.88 (6.43) 75.14 (12.79) Age: t = 0.55 ( p = 0.57) Sex: X 2 = 0.04 ( p = 0.5) IQ: t = 25.9 ( p = 4.0e-79) 3.3.2. Measurements of sample-specific phenotype characteristics All sites conducted structured diagnostic interviews at the time of scanning to determine lifetime psychiatric diagnoses. Wechsler IQ assessments were used to assess cognitive function. Across sites, all cases received a molecularly confirmed diagnosis of 22q11.2 deletion. All 22q11DS subjects included in the psychotic disorder group had a DSM schizophrenia spectrum psychotic disorder diagnosis (schizophrenia, schizoaffective disorder, or psychosis not otherwise specified), as determined via structured diagnostic interview conducted by a trained mental health professional at each site, and supplemented by collateral information and medical records (see Supplementary Table S3 for details regarding study instruments and study inclusion/exclusion criteria). A cross-site reliability procedure was also undertaken, in which two investigators with clinical expertise independently reviewed a subset of representative cases from each site. References in Supplementary Table S3 provide additional detail regarding clinical characteristics of each study sample. Across sites, deletion size was determined using multiplex ligation-dependent probe amplification (MLPA) 53 . The large sample size here uniquely allowed for the comparison of the effects of the two most frequent deletion types, the longer A-D vs. the shorter A-B deletion, on DTI measures. From cases with available MLPA data, 206 subjects had the A-D deletion (89.9%), and 15 (6.5%) subjects had the A-B deletion (see Supplemental Table S3 ). 3.3.3. Image acquisition and processing Acquisition parameters of dMRI and T1-weighted MRI scans for each site are shown in Supplemental Table S4 and S5 . All raw data were pre-processed in an identical fashion at a single site. We denoised all dMRI images with the LPCA tool 54 and all volumes were skull-stripped using FSL’s BET tool 55 . Eddy correction was performed with FSL’s eddy_correct tool on all sites but Utrecht. T1-weighted images were bias field corrected with ANTs’ N4, denoised with the non-local 22 means algorithm 56 and skull-stripped with FreeSurfer 57 . Subsequently, the EPI (echo-planar imaging) distortion correction was performed by non-linearly aligning the non-diffusion sensitized volumes (b=0 s/mm 2 ) to the subjects’ corresponding preprocessed T1-weighted image. The non-linear registration was performed with ANTs 58 . The deformation fields were applied to all the diffusion sensitized volumes. For the scans from Utrecht, eddy and EPI distortion corrections were performed with FSL’s TOPUP and EDDY tools 59,60 . Thereafter, we computed DTI-FA maps which were used to register each subject linearly and nonlinearly to the ENIGMA DTI-FA common template 32 . After corroborating the correct alignment of each subject’s FA to the ENIGMA DTI-FA atlas, we concatenated the eddy_correct linear transformations with the linear transformations and nonlinear deformations to the ENIGMA DTI-FA template. This joint transformation and deformation field was applied to the skullstripped and denoised dMRI. By doing this we ensured that the original dMRI images were interpolated only once to the ENIGMA DTI template. With the dMRI in the ENIGMA template space we calculated the diffusion tensor with a nonlinear fitting and outlier detection for robust estimation 61 by using the DIPY package 62 . We computed four scalar maps from the fitted tensors: Fractional Anisotropy (FA), Mean Diffusivity (MD), Radial Diffusivity (RD and Axial Diffusivity (AD). The code for the ENIGMA-DTI protocol is freely available here: http://enigma.ini.usc.edu/ongoing/dti-working-group/ . FA, MD, RD, and AD maps were skeletonized as described in the ENIGMA DTI protocol 32,63 , based on the TBSS method 49 , ensuring that all data are normalized to the ENIGMA-DTI template. Mean values were calculated for each DTI measure along the skeleton within each ROI defined by the Johns Hopkins University WM atlas (JHU-ICBM-DTI-81) distributed by FSL 32,64 . For all analyses, we used the mean of the right and left values for bilateral ROIs, for each measure; we included the mean of all WM JHU-ICBM ROIs and we excluded the corticospinal tract, fornix and cingulum of the hippocampus as these ROIs are difficult to reliably register 65 . The ROIs included are shown in Figure 1 . 23 Figure 1. Depiction of the 18 regions of interest (ROIs) of the Johns Hopkins University (JHU-ICBM) white matter atlas [Mori et al. (2008)] that were analyzed in the present study. 3.3.4. Statistical Analyses Effects of 22q11DS and age on DTI-derived measures Group differences between 22q11DS and HC were investigated using two analytic approaches: a meta-analysis , which runs statistical comparisons for each site separately and combines the summary statistics across sites, and a mega-analysis , in which data are harmonized and pooled from individual subjects, and statistical analysis are run on the full group. The meta-analysis included 540 subjects: 278 22q11DS probands (mean age: 16.76 ± 6.78, 138 females) and 260 HC (mean age: 16.55 ± 8.01, 123 females) from nine independent datasets derived from eight sites ( Table 1) . Because Utrecht included only 22q11DS cases, it was not included in the case-control analyses. For each site, linear regressions were run, in which the mean DTI measure for each ROI was the dependent variable, diagnosis was the predictor of interest, and age, [age-mean(age)] 2 and sex were included as covariates. Given that DTI-derived measures tend to peak between 11 and 20 years for commissural and association fibers and in the early twenties for projection fibers 66,67 , we 24 included both the linear and quadratic effects of age in the model. The quadratic age term was centered to avoid collinearities with the linear age term. In addition, because females and males show different trajectories of DTI measures across development 68 , sex was accounted for in the model. Cohen's d effect sizes for diagnosis were computed. Subsequently, an inverse-variance weighted mixed-effect meta-analysis 69 to combine individual site effect sizes, as in Kelly et al. (2018) 65 . A pooled, or mega-analytic, approach was also conducted. As multiple factors can affect the distribution of DTI measures 70–72 , additional harmonization of DTI measures can be advantageous when conducting studies pooling dMRI data from different protocols. We used the COMBAT algorithm 73 to harmonize data across sites for each DTI measure (FA, MD, RD, AD) for each WM ROI. This algorithm uses an empirical Bayes framework to estimate additive and multiplicative site effects. It has been used previously for harmonization of multisite DTI data, and has been shown to perform better than several other methods for modeling and removing inter-site variability 73 . Next, group differences were assessed using the same model tested in the meta-analysis. Finally, the diagnosis-by-age interaction effect term was included in the mega-analytic model to test whether the effects of age differed in 22q11DS probands relative to HC. We used the Benjamini & Hochberg method to control for the family wise error rate 74 . The percentage of tolerated false positives was 5% (q<0.05). Critical p-values were calculated for each set of models, specifically: (1) meta-analysis; (2) mega-analysis; and (3) mega-analysis including diagnosis-by-age interaction. As noted in the Methods (2.4), covariates included age, [age-mean(age)] 2 and sex. For three sites (UCLA, Newcastle and Cardiff), an additional term for scanner type was included, as two scanners were used with the identical acquisition, and a random effect model was performed to take this covariate into account. Effect sizes for dichotomous variables (diagnosis and deletion type) were computed by converting t -values from the multiple linear regressions to Cohen’s d statistics according to the formula: 25 Where df is the number of degrees of freedom. For continuous variables (age, IQ), effect sizes were estimated by converting t -values from the multiple linear regressions to partial correlations according to the formula: where Res-df denotes the residual degrees of freedom. All statistical analyses were performed with the core statistical R packages for linear regression 75 . Additionally, given previous findings of non-linear trajectories of DTI-derived measures with respect to age in healthy individuals (5-82 years) 76 , we fit a Poisson non-linear model for age for each group separately (HC and 22q11DS) for each WM ROI and for each DTI-derived measure, to further investigate age effects. We used the previously harmonized data (see above COMBAT harmonization), to reduce site effects. We measured the age of peak FA and age of minimum MD, RD and AD as in Lebel et al. 76 and compared both groups using a two-tailed t-test for means with outlier removal ( ⍺=0.05). Thereafter, we calculated the percent changes of each DTI-measure for each ROI from age 6 (minimum age in both groups) to peak/minimum, and from peak/minimum to age 46 and 52 (maximum age for 22q11DS and HC, respectively). We compared the percent changes of each DTI-measure for all ROIs between 22q11DS and HC groups by using Yuen’s method with bootstrap-t for trimmed means ( ⍺=0.05). Influence of psychotic disorder, deletion size and IQ on DTI measures To assess potential differences in WM architecture as a function of clinical and genetic variability, we examined the effects of psychotic illness (35 with psychotic disorder vs. 191 without psychosis) and deletion size (206 AD vs. 15 AB) on DTI measures, within individuals with 22q11DS. Additionally, given that IQ is a group-associated variable, we examined partial correlations with IQ within the 22q11DS (N=304) and HC groups (N=102) separately. For these analyses the DTI measures for each ROI were included as dependent variables. Age, 26 [age-mean(age)] 2 and sex were included as covariates. FDR correction was performed as specified above (section 2.4.1). Given the strong association between age and psychosis onset 5 , as well as the differences in mean age between 22q11DS cases with and without psychosis (see Supplementary Table S2b ), to assess the effect of psychosis within the 22q11DS group we used a local nonparametric ANCOVA method 77 covarying for age, which allowed for a controlled test within age subgroups. The local nonlinear ANCOVA makes no parametric assumptions about how two variables are related and is robust to heteroscedasticity 77 . It approximates a regression line for each group using a running interval smoother and compares both groups at specific design points. In this case, it selects a specific age as a design point and compares the trimmed means of the DTI-by-ROI measure at all points close to the selected age (proximity calculated by the median absolute deviation (MAD)). If both groups have more than 12 subjects (based on the degrees of freedom required), the regression lines are comparable and a confidence interval is computed followed by a t-test. Dunnett’s T3 method is used to control the family-wise error 78 . We elected to use this approach for our comparison of 22q11DS cases with and without psychotic disorder, as this method was considered more robust to differences in sample size and mean age between groups (mean 22q11DS+Psychosis=23.87 years; mean 22qDS-No Psychosis= 17.99 years; t=-4.14, p=0.00016). Additionally, age variances were significantly different (F = 0.51, p=0.006) and non-normal (kurtosis=4.09 and 5.64; skewness=0.93 and 0.94, respectively) which prohibited an ordinary least squares linear regression analysis. An explanatory measure of effect size is also derived from this analysis 79 and reported in Supplementary Table S10 and Figure 4 . The code used here for the ANCOVA analysis is freely available at: https://dornsife.usc.edu/labs/rwilcox/software/ . Next, to determine whether the microstructural changes observed in 22q11DS-associated psychosis overlap with those seen in idiopathic schizophrenia, we compared our results for 22q11DS cases with and without psychosis to schizophrenia case-control results from the ENIGMA-Schizophrenia DTI Working Group 65 , analyzed using the same protocols as in our study. 27 3.4. Results 3.4.1. Group differences across sites We first investigated whether there were consistent group differences in WM microstructure between 22q11.2 deletion carriers and healthy controls, using a standardized processing pipeline. Equally important is to determine whether harmonization of the data would allow pooled analyses for further investigation of modulatory factors (psychosis, deletion size, and IQ). Figure 2 shows group differences for 22q11DS cases vs. HC, from the meta-analysis and mega-analysis: results were nearly identical, with similar effect sizes. Effect sizes for each site are shown in Supplementary Figure 1 . Most ROIs that significantly differed between 22q11DS and HC showed lower diffusivity values (MD, AD and RD) in 22q11DS subjects, but a mixed pattern for FA. Significantly higher FA in 22q11DS cases relative to HC was observed in the tapetum (TAP), genu (GCC), body and splenium of the corpus callosum (BCC/SCC), the anterior and posterior limb of the internal capsule (ALIC/PLIC), and posterior and superior corona radiata (PCR/SCR), with moderate to large effect sizes ( d ~0.3-0.8), for both analyses. In contrast, ROIs in association fibers - the superior longitudinal fasciculus (SLF), fornix/stria terminalis (FXST), and external/extreme capsules (EC) - showed significantly lower FA in 22q11DS relative to HC ( Supplementary Tables S6-S7 ). 22q11DS subjects had significantly lower MD than HC in almost all ROIs investigated, with greatest effects ( d ~1.0) in the PCR and posterior thalamic radiation (PTR); both contain mostly thalamocortical/corticothalamic and corticofugal fibers from posterior brain areas. For all 18 ROIs, MD was lower in 22q11DS, as was AD, for 15 of the 18 ROIs. Only the PLIC showed significantly higher AD in 22q11DS relative to HC. For RD, all ROIs showing significant differences (15 of 18 ROIs) were lower in 22q11DS than HC, with largest effects ( d ~0.7) in the corpus callosum and PCR ( Supplementary Tables S6-S7 ). 28 Figure 2. Results of meta- and mega-analysis including nine independent datasets from the ENIGMA-22q11DS working group. The bar graphs on the left side are organized based on the effect sizes for FA (positive to negative, from left to right). The brain maps on the right side are organized by rows, each one corresponding to respective bar graph on the left. These show the JHU-ICBM atlas white matter ROIs that passed multiple comparison correction after meta-analysis. The model tested was: DTI-ROI-measure=ß 0 + ß 1 Diagnosis + ß 2 Sex + ß 3 Age + ß 4 Age 2 centered . WM = Average of all white matter JHU-ICBM ROIs. 3.4.2. Age-associated effects Given the wide age range (6-52 years), we wanted to determine whether the development of WM is delayed or altered in 22q11DS. As shown in Supplementary Table S6, there were highly significant linear effects of age for all indices for the majority of ROIs. FA increased with age, while the opposite pattern was found for diffusivity values (MD, AD and RD). There were also significant quadratic effects for almost all ROIs for FA, MD and RD. AD showed fewer significant quadratic effects, in both the meta- and mega-analysis. However, no significant age-by-diagnosis interactions were observed ( Supplementary Table S8 ). Given the sparse representation of older adults, we also performed a mega-analysis with a subsample of subjects under 30 years old to explore potential age-by-diagnosis effects, which yielded similar results ( Supplementary Table S9 ). 29 We also investigated Poisson regression models to further evaluate the effects of age on WM development. These models did not provide a substantially better fit to the data than the linear regression model used above, as determined by the residual standard error of the fits (see Supplementary Tables S10, S11, S12 ). As such, we retained the linear regression models for our primary analyses, but report the additional trajectory information obtained from the Poisson models below. Scatterplots for the non-linear Poisson fits of age per ROI for each DTI-derived measure are displayed in Supplementary Figures S2-S5 . There were fewer ROIs with significant peak/minimum estimates in the 22q11DS group, across all DTI indices (see Supplementary Tables S13, S14 ). Generally, those ROIs without significant peak/minimum estimates have linear rather than exponential growth and decay trajectories. When comparing the mean age of peak FA (across ROIs) between HC and 22q11DS, average peak FA was significantly delayed in 22q11DS. We found significantly delayed mean age at minimum RD in 22q11DS, but no differences in minimum MD and AD. Mean rate of FA decrease (after peak FA), and mean rate of increase in RD and MD (after minimum RD/MD), were also significantly greater in HC vs. 22q11DS, with no differences in AD ( Supplementary Table S15 ). 3.4.3. Influence of psychosis Are the deletion-related WM changes more severe in those with psychotic disorder? Relative to 22q11DS subjects without psychosis, 22q11DS subjects with psychotic disorder showed overall lower diffusivity values, with significantly lower AD in the ALIC and PTR, both predominantly containing thalamic radiation fibers, in the cingulum of the cingulate gyrus (CGC) and the SLF, which mostly contain fronto-parietal and fronto-temporal association fibers, and the sagittal stratum (SS), which contains both posterior thalamic projection and temporal association fibers. 22q11DS-Psychosis was also associated with significantly lower RD and MD in the GCC, which contains callosal fibers, and significantly lower MD in the PLIC, where the superior thalamic radiation and cortico-pontine fibers are the major constituents. These differences were seen primarily between ages 20 and 26 for most ROIs; some ROIs (ALIC, PTR and SS) showed differences by age 16-17 ( Figure 3 and Supplementary Table S16) . Overall, these findings confirm that WM abnormalities detected by DTI diffusivity measures are more severe in 22q11DS patients with psychotic disorder, and are particularly evident in young adulthood. 30 3.4.4. Comparison of WM Microstructure in 22q11DS-Psychosis to Idiopathic Schizophrenia Next, we compared our results for 22q11DS cases with and without psychosis to schizophrenia case-control results (2359 HC vs. 1963 schizophrenia patients) 65 , plotted together for visualization purposes ( Figure 4 ). Effects for 22q11DS cases with and without psychosis differed markedly from those observed for idiopathic schizophrenia relative to HC. Specifically, while patients with 22q11DS-psychosis tended toward higher FA and lower diffusivity values compared to 22q11DS individuals without psychosis, patients with idiopathic schizophrenia showed overall lower FA across tracts and increased diffusivity values relative to HC, particularly for MD and RD. 3.4.5. Influence of deletion type and IQ Does the extent of the deletion affect WM microstructure? Subjects with the large A-D deletion showed a trend toward lower AD in the anterior corona radiata (ACR) and EC, and higher FA in the TAP; however, there were no statistically significant differences as a function of deletion size, after multiple comparison correction (see Supplementary Figure S6 and Supplementary Table S17). Additionally, regarding relationships between DTI indices and cognitive abilities, healthy controls showed trends toward positive correlations of MD, RD and AD in multiple ROIs with IQ, and a trend toward an inverse correlation of FA with IQ in the TAP. Within 22q11DS cases, findings were similar, but higher IQ was associated with significantly higher AD in the PTR, which contains mainly posterior cortico-thalamic and thalamo-cortical fibers. There was also a trend toward higher AD in average WM, genu of the CC, and SS being associated with higher IQ in 22q11DS ( Supplementary Figure S7 , Supplementary Table S18 ). While these relationships were not significant when corrected for multiple comparisons, the overall pattern of findings suggests that relationships between WM microstructure and cognition may differ in 22q11DS relative to typically developing controls. 31 Figure 3. Results from the local nonparametric ANCOVA analysis comparing 22q11DS subjects with psychotic disorder (N=35) vs. those with no lifetime history of psychotic symptoms (N=191). Shown here are the results for DTI indices that significantly differed between 22q-Psychosis vs. 22q-No Psychosis: AD in the ALIC, CGC, PTR, SLF and SS, RD in GCC, and MD in the GCC and PLIC. All analyses were performed on 25 design points corresponding to different age bands. Vertical red lines correspond to the ages at which these DTI measures (AD, MD, RD) significantly differed between subjects with 22q11DS with and without psychosis. 32 Figure 4. Comparison of Effect Sizes in this study, to those from the ENIGMA-Schizophrenia DTI Working Group using similar methods (2,359 healthy controls vs. 1,963 schizophrenia patients from 29 independent studies; Kelly et al. 2018; blue triangles) to 22q11DS probands with and without psychosis (red circles). Positive effect sizes: 22q-Psychosis > 22q-No Psychosis OR Schizophrenia Patients > Healthy Controls. Negative effect sizes: 22q-No Psychosis > 22q-Psychosis OR Healthy Controls > Schizophrenia Patients. 3.5. Discussion This is the largest study to date of WM microstructure in 22q11DS (334 22q11DS cases and 260 HC), assessed by DTI. Our analysis pipeline 32,65 allowed for coordinated prospective meta- and mega-analyses of the data across sites, unlike traditional meta-analyses that combine statistical results from the literature. This 33 approach addresses, for the first time, issues of low power due to small sample sizes and variable analysis protocols that contribute to heterogeneity and lack of clarity in DTI studies to date. In contrast to findings in many neuropsychiatric disorders 65,80 , our findings revealed overall lower DTI diffusivities (AD, RD and MD) in 22q11DS compared to HC, with regionally varying directions of effect for FA. Higher FA, lower RD and AD (and consequently, lower MD) appear to be the hallmark of microstructural alterations in the major WM tracts in 22q11DS, especially in the commissural fibers of the corpus callosum. While this may suggest greater myelination 21 , we must be cautious in applying this interpretation to our findings, given that dMRI cannot directly index the degree of myelination 81 . Anisotropy does not only depend on the presence of myelin in the WM, as it has been demonstrated in unmyelinated tracts 11 and is also sensitive to axonal density. RD is sensitive to axonal density and amount of extra-cellular space, and AD to axonal diameter and organization 20,82 . Moreover, since axonal density and myelination are correlated 11,83 , it is not possible to disentangle one from another when interpreting FA and RD differences between populations. We postulate that the observed group differences may result from an increase in the cumulative cellular membrane circumference 84 in 22q11DS (attributable to differences in axon composition, myelination and/or reactive astrocytes), which hinders diffusion perpendicularly to the WM tracts, hence increasing anisotropy and decreasing RD. Our findings of higher FA in 22q11DS relative to controls in ROIs in commissural tracts (TAP, GCC, BCC, SCC), no detectable differences in ROIs where projection fibers predominate (RLIC, SS, PTR, SFO), and lower FA in ROIs in long association tracts (EC, SLF, FXST) are consistent with findings in the mouse model of 22q11DS 9 . Specifically, this study found that proliferation of basal, but not apical progenitors is disrupted, and subsequently the frequency of projection neurons in layers 2/3, but not layers 5/6, is altered. Commissural and long association fibers originate primarily from projection neurons, i.e., pyramidal neurons in the outer layers 2/3, whereas corticofugal and cortico-thalamic projection fibers tend to originate from pyramidal cells in cortical layers 5/6. Moreover, our results suggest that the nature of WM disruptions may differ between callosal and long association fibers in 22q11DS, but advanced microstructural MRI techniques may be necessary to disentangle these differences. As such, these cross-species findings collectively suggest a potential neurobiological model in which haploinsufficiency at the 22q11.2 locus leads to disruptions of specific aspects of early brain 34 development, and subsequent changes in neural circuitry that likely elevate risk for neuropsychiatric disorders in 22q11DS patients. We speculate that our findings may be related to three types of histopathological alterations in WM of 22q11DS patients, all of which could reduce diffusivity. First, a recent neuropathology study of a 3-month old infant with 22q11DS reported decreased neuronal frequencies in outer cortical layers and increased neuronal frequencies in deeper cortical layers 9 . This is closely related to findings in the LgDel 22q11.2 mouse model mentioned above 9 . Pyramidal neurons of cortical layers 2/3 generate a substantial portion of the cortico-cortical axonal projections between association areas 85 . These axons are present in most of the WM ROIs included in this study. Consequently, target-to-origin signaling between cortical association areas (cortico-cortical projections) may be disrupted in 22q11DS, affecting the necessary cues to initiate proper axonal differentiation 86,87 , ultimately affecting the development of a typical distribution of axonal diameters 88–90 , and therefore altering RD and AD in WM bundles. Moreover, the PLIC was the only ROI showing higher AD in 22q11DS. AD has been associated with axonal diameter changes and axonal tortuosity in rats 20,82 . PLIC is the only ROI in this study that contains mostly corticofugal fibers, which primarily derive from cortical layers 5/6 64,85 , suggesting that the axonal size distribution within fiber bundles originating in the deeper cortical layers may differ from those originating in the outer cortical layers 88,90 . Further studies of animal models and post-mortem human brain tissue may shed light on this. Second, DTI abnormalities may also reflect gliotic changes secondary to microvascular insults. Post-mortem findings in 22q11DS adults indicate both deep WM gliosis associated with cerebrovascular changes 91 . Gliosis - occurring as the brain reacts to microvascular injuries - has been associated with increased anisotropy in a mouse brain injury model 92 . Third, DTI measures may be affected by ectopic neurons in WM that may result from neuronal migration defects during early development 93 . These have been reported in both neuropathologic 91,94,95 and neuroimaging studies of 22q11DS patients 96,97 . While we did not detect any heterotopias in our cohort, subtle microscopic ones may be detected only via histology. The age trajectories of FA MD, RD and AD, as well as peak and minimum age estimates of our control sample, were similar to those reported previously 76 . However, 22q11DS patients showed a delayed mean age of both peak FA and minimum RD; correspondingly, they also showed lower decrease and increase 35 rates for FA and MD after peak and minimum ages, respectively. As noted above, these findings may indicate a delay in maturation secondary to altered axonal diameters and organization in the deep WM, which could be precursors of a delayed myelination process. Conversely, lower rate of change after maturation (indicated by peak FA and minimum RD) may be indicative of underlying organizational changes in WM that abnormally hinder diffusion and may result from gliotic changes, as has been reported in adult post-mortem 22q11DS brain tissue 91 . Nevertheless, despite the harmonization protocol interpretive caution is warranted because the age distribution was variable across sites and data points were rather sparse in the higher age ranges. Consistent with some single-site studies suggesting inverse correlations between psychotic symptom severity in 22q11DS and diffusivity in the CC and long association tracts 30,31,33,35,50 , we found lower RD and MD in those with psychosis in the genu of the CC, and lower axial diffusivity in long association tracts such as the SLF and CGC. Interestingly, significantly lower AD was found in ROIs with predominantly cortico-thalamic and thalamo-cortical fibers such as the ALIC, SS and the PTR. A previous single-site tractometry study found significant associations between higher FA and lower RD in the ALIC with positive prodromal symptoms 31 . Future studies should prospectively investigate the role of the major thalamic projection tracts in the emergence and progression of psychotic symptoms in 22q11DS. Notably, WM microstructural alterations in 22q11DS with psychosis showed a largely opposite pattern from those seen in idiopathic schizophrenia, involving primarily FA increases rather than decreases and reductions (rather than increases) in diffusivity measures. A previous single site study of 22q11DS and youth at clinical high risk for psychosis reported this directionally opposite pattern as well 45 . This is in contrast to findings for cortical gray matter, in which 22q11DS patients with psychosis showed highly significant overlap with idiopathic schizophrenia, in terms of prominent cortical thinning in fronto-temporal regions 52 . Thus, our findings suggest that patterns of neuroanatomic overlap in 22q11DS-associated vs. idiopathic psychosis markedly differ for gray and WM, and suggest that different WM phenotypes may lead to similar downstream clinical outcomes. Our findings of altered AD in 22q11DS, more extreme in those with psychosis, may indicate altered axonal diameter and increased tortuosity of WM tracts 20,82 . Numerous smaller, tortuous axons in key connections between cortical association areas may lead to altered WM maturation, structural dysconnectivity and possibly psychosis. In idiopathic schizophrenia, WM degeneration 36 (demyelination and loss of axons with larger diameters) may also lead to disrupted axonal morphology that similarly results in structural dysconnectivity between cortical association areas. We did not see consistent effects of deletion size on WM architecture, and found little evidence that the relationship between WM microstructure and IQ differed between 22q11DS cases and HC. Sample size was quite limited for the A-B deletion type, and imaging protocols varied across sites, which may have affected our results. Additionally, given the highly variable psychotropic medications and medical comorbidities in 22q11DS patients, their effects could not be systematically investigated here. Previously, in a sample including many of the same participants as in the current analysis, we found that psychotropic medication was not significantly associated with cortical thickness or cortical surface area in 22q11DS patients 52 . Additionally, prior studies of patients with idiopathic schizophrenia found that WM changes detected by DTI were not attributable to antipsychotic medication 65,98 . Future studies with multi-shell acquisitions and novel biophysical models may resolve the contribution of the intra- and extra-axonal volume fractions and axonal diameters to these abnormalities 36,99 . Quantitative magnetization transfer 100 and perfusion MRI acquisitions 101 may help clarify any myelin abnormalities or underlying brain microvascular pathology in 22q11DS. Collectively, our findings indicate large effects of the 22q11.2 deletion on WM microstructure. Diffusivity was more consistently affected than FA. In animal models, disruptions to predominantly cortico-cortical and cortico-thalamic/thalamo-cortical connections in 22q11DS may be attributable to disrupted early neurogenesis. Future translational studies will help to determine the neurobiological underpinnings of these alterations. 3.6. Acknowledgments and Funding Sources I would like to thank all the ENIGMA-22q working group members and all collaborators of this work: Kenia Martínez 2 , Xiaoping Qu 1 , Christopher Ching 1,3 , Talia M. Nir 1 , Deydeep Kothapalli 1 , Conor Corbin 1 , Daqiang Sun 3,4 , Amy Lin 3 , Jennifer K. Forsyth 3,5 , Leila 37 Kushan 3 , Ariana Vajdi 3 , Maria Jalbrzikowski 6 , Laura Hansen 3 , Rachel K. Jonas 3 , Therese van Amelsvoort 8 , Geor Bakker 8 , Wendy R. Kates 9 , Kevin M. Antshel 10 , Wanda Fremont 9 , Linda E. Campbell 11 , Kathryn L. McCabe 12 , Eileen Daly 13 , Maria Gudbrandsen 13 , Clodagh Murphy 13 , Declan Murphy 13 , Michael Craig 14 , Beverly Emanuel 15 , Donna McDonald-McGinn 15 , Jacob Vorstman 16 , Ania Fiksinski 16 , Sanne Koops 16 , Kosha Ruparel 17 , David Roalf 17 , Raquel E. Gur 18 , J. Eric Schmitt 19 , Tony J. Simon 20 , Naomi J. Goodrich-Hunsaker 20,21 , Courtney A. Durdle 20 , Joanne Doherty 22,23 , Adam C. Cunningham 22 , Marianne van den Bree 22 , David E. J. Linden 22,23 , Michael Owen 22 , Hayley Moss 22 , Sinead Kelly 24 , Gary Donohoe 25 , Kieran C. Murphy 26 , Celso Arango 2 , Carrie E. Bearden 3,5 . 1) Imaging Genetics Center, Mark and Mary Stevens Neuroimaging & Informatics Institute, Keck School of Medicine of the University of Southern California, Marina del Rey, CA, USA 2) Department of Child and Adolescent Psychiatry, Hospital General Universitario Gregorio Marañón, Madrid, Spain; Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), Madrid, Spain; Universidad Europea de Madrid, Madrid, Spain 3) Department of Psychiatry and Biobehavioral Sciences, Semel Institute for Neuroscience and Human Behavior, University of California at Los Angeles, Los Angeles, CA, USA 4) Department of Mental Health, Veterans Affairs Greater Los Angeles Healthcare System, Los Angeles, CA, USA 5) Department of Psychology, University of California at Los Angeles, Los Angeles, CA USA 6) Department of Psychiatry, University of Pittsburgh, Pittsburgh, PA, USA 7) Departments of Neurology, Psychiatry, Radiology, Engineering, Pediatrics and Ophthalmology, University of Southern California, CA, USA 8) Department of Psychiatry & Neuropsychology, Maastricht University, Netherlands 9) Department of Psychiatry and Behavioral Sciences, State University of New York, Upstate Medical University, NY, USA 10) Department of Psychology, Syracuse University, Syracuse, NY, USA 11) Priority Research Centre GrowUpWell, University of Newcastle, Australia; School of Psychology, University of Newcastle, Australia 12) School of Psychology, University of Newcastle, Australia; UC Davis MIND Institute and Department of Psychiatry and Behavioral Sciences 13) Sackler Institute for Translational Neurodevelopment and Department of Forensic and Neurodevelopmental Sciences, King’s College London, Institute of 38 Psychiatry, Psychology & Neuroscience, London, UK 14) Sackler Institute for Translational Neurodevelopment and Department of Forensic and Neurodevelopmental Sciences, King’s College London, Institute of Psychiatry, Psychology & Neuroscience, London, UK & National Autism Unit, Bethlem Royal Hospital, UK 15) Division of Human Genetics, Children's Hospital of Philadelphia, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA 16) Department of Psychiatry, Brain Center Rudolf Magnus, University Medical Center Utrecht, The Netherlands 17) Department of Psychiatry, University of Pennsylvania, Philadelphia, PA, USA 18) Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania and Children's Hospital of Philadelphia, Philadelphia, PA, USA 19) Departments of Radiology and Psychiatry, University of Pennsylvania, Philadelphia, PA, USA 20) UC Davis MIND Institute and Department of Psychiatry and Behavioral Sciences, CA, USA 21) Brigham Young University, UT, USA 22) MRC Centre for Neuropsychiatric Genetics and Genomics, Division of Psychological Medicine and Clinical Neurosciences, Cardiff University, Cardiff, Wales, UK 23) The Cardiff University Brain Research Imaging Centre (CUBRIC), Cardiff University, Cardiff, Wales, UK 24) Department of Psychiatry, Beth Israel Deaconess Medical Center, Harvard Medical School, MA, USA 25) Centre for Neuroimaging and Cognitive Genomics (NICOG), Clinical Neuroimaging Laboratory, NCBES Galway Neuroscience Centre, National University of Ireland Galway, Galway, Ireland 26) Department of Psychiatry, Royal College of Surgeons in Ireland, Dublin, Ireland. The ENIGMA-22q working group gratefully acknowledges support from the NIH Big Data to Knowledge (BD2K) award (U54 EB020403 to Paul M. Thompson). This manuscript was also supported by grants from the National Institute of Mental Health: RO1 MH085953 and R01 MH100900 (Carrie E. Bearden), the Miller Family Endowed Term Chair at the UCLA Brain Research Institute (Carrie E. Bearden), Neurobehavioral Genetics Predoctoral Training Grant (5T32 MH073526) to Chris Ching and Amy Lin, NIMH grant 5U01 MH101723-02; NIH U01 MH087626, U01 MH101719 and MH089983 ( Raquel E. Gur ); U01 MH087636 ( Beverly Emanuel, Donna McDonald-McGinn ), the Wellcome Trust Institutional Strategic Support Fund 39 (Marianne van den Bree ), the Waterloo Foundation (WF 918-1234 to Marianne van den Bree ), the Baily Thomas Charitable Fund (2315/1 to Marianne van den Bree ), Wellcome Trust (102003/Z/13/Z to Joanne Doherty ), National Institute of Mental Health (NIMH 5UO1 MH101724 to Marianne van den Bree and Michael Owen ), Wellcome Trust (100202/Z/12/Z to Michael Owen ). We thank the participants and their families for being a part of our research. We also thank the ENIGMA-Schizophrenia Working Group for sharing their data for comparative analysis. 40 Chapter 4 Alternative diffusion anisotropy measures for the investigation of white matter alterations in 22q11.2 deletion syndrome 4.1. Abstract Diffusion MRI (dMRI) is widely used to study the brain’s white matter (WM) microstructure in a range of psychiatric and neurological diseases. As the diffusion tensor model has limitations in brain regions with crossing fibers, novel diffusion MRI reconstruction models may offer more accurate measures of tissue properties, and a better understanding of the brain abnormalities in specific diseases. Here we studied a large sample of 249 participants with 22q11.2 deletion syndrome (22q11DS), a neurogenetic condition associated with high rates of developmental neuropsychiatric disorders, and 224 age-matched healthy controls (HC) (age range: 8-35 years). Participants were scanned with dMRI at eight centers worldwide. Using a meta-analytic approach, we assessed the profile of group differences in four diffusion anisotropy measures to better understand the patterns of WM microstructural abnormalities and evaluate their consistency across alternative measures. When assessed in atlas-defined regions of interest, we found statistically significant differences for all anisotropy measures, all showing a widespread but not always coinciding pattern of effects. The tensor distribution function fractional anisotropy (TDF-FA) showed largest effect sizes all in the same direction (greater anisotropy in 22q11DS than HC). Fractional anisotropy based on the tensor model (FA) showed the second largest effect sizes after TDF-FA; some regions showed higher mean values in 22q11DS, but others lower. Generalized fractional anisotropy (GFA) showed the opposite pattern to TDF-FA with most 41 regions showing lower anisotropy in 22q11DS versus HC. Anisotropic power maps (AP) showed the lowest effect sizes also with a mixed pattern of effects across regions. These results were also consistent across skeleton projection methods, with few differences when projecting anisotropy values from voxels sampled on the FA map or projecting values from voxels sampled from each anisotropy map. This study highlights that different mathematical definitions of anisotropy may lead to different profiles of group differences, even in large, well-powered population studies. Further studies of biophysical models derived from multi-shell dMRI and histological validations may help to understand the sources of these differences. 22q11DS is a promising model to study differences among novel anisotropy/dMRI measures, as group differences are relatively large and there exist animal models suitable for histological validation. 4.2. Introduction Here, we compare the anisotropy measures derived from three different dMRI reconstruction models in a large sample of participants with 22q11.2 Deletion Syndrome (22q11DS) being analyzed by the ENIGMA Consortium’s 22q Working Group (ENIGMA-22q WG). 22q11DS is the most common neurogenetic syndrome that is caused by a chromosomal microdeletion and it increases the risk for psychotic illness around 25-fold relative to the general population 2,5 . Prior reports have shown that, compared to healthy controls, subjects with 22q11DS show higher FA in several WM regions with major tracts such as the corpus callosum, internal capsules and corona radiata 30 [12]. Though other studies have also reported higher FA in psychiatric and neurogenetic syndromes, it is rather counterintuitive, as FA has been classically regarded as a biomarker of higher WM coherence and “integrity”. Thence, we ask the question whether alternative anisotropy measures from other dMRI reconstruction models behave in a similar manner than FA in the context of a defined neurogenetic syndrome such as 22q11DS. The overall goal of the work is to better understand the microstructural basis of WM anomalies in the disorder, using a set of measures sensitive to different aspects of WM diffusion. Additionally, we test the stability of skeleton-based techniques for cross-subject comparisons of local dMRI parameters when using alternative anisotropy measures. Specifically, the ENIGMA-DTI protocol that is in turn based on Tract-Based Spatial Statistics (TBSS) is a global effort to harmonize DTI-based 42 measures across scans from multi-site studies 32,65 . TBSS was originally developed for analyses of FA; precisely, each subject’s voxel-values of FA are projected onto a WM skeleton that is based on the FA map. Hence, it is possible that the disease effects differ when projecting voxels to the skeleton sampled from anisotropy measures other than FA. As such, any effort to extend TBSS to study other dMRI measures should assess how the definition of the skeleton affects the results, as in theory it affects the spatial regions in which abnormalities are assessed. Exploring alternative anisotropy measures in the context of large, worldwide disease specific studies may help finding dMRI biomarkers that are robust and consistently sensitive to disease. This in turn could help identify patients with a poorer prognosis who are most likely to decline, as well as others with a better prognosis. If imaging biomarkers are found to be robust across multiple cohorts worldwide, they may serve as reliable targets for well-powered studies of factors that affect disease outcome and progression. 4.3. Methods 4.3.1. Participants Diffusion MRI data were contributed from eight studies previously analyzed by the ENIGMA-22q WG. This work included 473 participants: 249 with 22q11DS (mean age: 18.1 ± 5.4, 122 females) and 224 healthy controls (HC; mean age: 17.7 ± 5.9, 106 females). Demographic characteristics are shown in Table 1. Across sites, deletion size was determined using multiplex ligation-dependent probe amplification (MLPA) 53 . Institutional review boards at participating institutions approved all study procedures, and material transfer agreements approved any sharing of de-identified imaging data. Written informed consent was obtained from all study participants or from a legal guardian. 43 Table 1. Demographic information on study participants, per site. SD = standard deviation; M = Male; F = Female, HC = Healthy Controls. (1) University of Pennsylvania/Children's Hospital of Philadelphia (PA, USA); (2) University of California Los Angeles (CA, USA); (3) State University New York Upstate (NY, USA); (4) University of Newcastle (NSW, Australia); (5) Maastricht University (The Netherlands); (6) Institute of Psychiatry (London, UK); (7) University California Davis (CA, USA); (8) Cardiff University (Wales, UK). Healthy Controls (HC) 22q11.2 DS Site N N (% by Sex) Mean Age (SD) N N (% by Sex) Mean Age (SD) Group Differences U.Penn 49 30 (61.2%) M; 19 (38.8%) F 17.31 (3.22) 43 26 (60%) M; 17 (40%) F 17.49 (3.13) Age: t = 0.27 ( p = 0.79) Sex: X 2 = 0.01 ( p = 0.94) UCLA 32 16 (50%) M; 16 (50%) F 12.59 (5.62) 49 25 (51%) M; 24 (49%) F 14.69 (5.59) Age: t = 1.60 ( p = 0.11) Sex: X 2 = 0.02 ( p = 0.89) SUNY Upstate 11 5 (45.5%) M; 6 (54.5%) F 21.12 (2.01) 34 19 (55.9%) M; 15 (44.11%) F 20.85 (1.86) Age: t = -0.43 ( p = 0.67) Sex: X 2 = 0.49 ( p = 0.48) University of Newcastle 17 8 (47.1%) M; 9 (52.9%) F 17.06 (3.01) 16 6 (37.5%) M; 10 (62.5%) F 16.63 (2.75) Age: t = -0.43 ( p = 0.67) Sex: X 2 = 0.31 ( p = 0.58) Maastricht University 36 23 (63.8%) M; 13 (36.2%) F 29.97 (10.05) 24 11 (45.9%) M; 13 (54.1%) F 30.05 (7.86) Age: t = 0.04 ( p = 0.97) Sex: X 2 = 1.91 ( p = 0.17) Institute of Psychiatry , London 24 10 (41.7%) M; 14 (58.3%) F 18.36 (6.73) 24 13 (54.1%) M; 11 (45.9%) F 18.04 (6.88) Age: t = -0.16 ( p = 0.87) Sex: X 2 = 0.75 ( p = 0.39) UC Davis 41 20 (48.8%) M; 21 (51.2%) F 11.05 (2.33) 46 21 (45.7%) M; 25 (54.3%) F 11.64 (2.53) Age: t = 1.11 ( p =. 027) Sex: X 2 = 0.09 ( p = 0.77) Cardiff University 14 6 (42.9%) M; 8 (57.1%) F 14.46 (1.79) 13 6 (46.2%) M; 7 (53.8%) F 16.03 (4.63) Age: t = 1.18 ( p = 0.25) Sex: X 2 = 0.03 ( p = 0.86) 44 4.3.2. Image acquisition and processing Acquisition parameters of dMRI scans for each site are shown in Table 2. All raw data were pre-processed in an identical fashion at a single site (USC Imaging Genetics Center). We denoised all dMRI images with the LPCA tool 102 and all volumes were ‘skull-stripped’ using FSL’s BET tool 55 . Eddy correction was performed with FSL’s eddy_correct tool for all sites. T1-weighted images were bias field corrected with ANTS’ N4, denoised with the non-local means algorithm 54,56 , and skull-stripped with FreeSurfer 57 . Subsequently, EPI (echo-planar imaging) distortion correction was performed by non-linearly aligning the non-diffusion sensitized volumes (b=0 s/mm2) to the subjects’ corresponding preprocessed T1-weighted image 58 . The deformation fields were applied to all the diffusion-sensitized volumes. Diffusion gradient directions were rotated to accommodate linear registrations. Thereafter, we fitted tensors and computed FA maps which were used to register each subject linearly and nonlinearly to the FA-based ENIGMA-DTI common template 32 . After verifying the correct alignment of each subject’s FA to the ENIGMA-DTI template, we concatenated all linear transformations and nonlinear deformations. This joint transformation and deformation field was applied to the skull-stripped and denoised dMRI. By doing this, we ensured that the original dMRI images were interpolated only once to the ENIGMA-DTI template. With the dMRI in the ENIGMA-DTI template space we used a nonlinear fitting and outlier detection for robust estimation of the diffusion tensor 61 by using the DIPY package 62 . We also computed three additional anisotropy measures, namely the tensor distribution function fractional anisotropy (TDF-FA), the generalized fractional anisotropy (GFA) and the anisotropy power map (AP) (see next section for further details). Individual subjects’ FA maps were skeletonized as described in the ENIGMA DTI protocol, which has been used in several multi-site studies 32,65 . This protocol includes the ENIGMA-DTI template with its corresponding FA-based skeleton precomputed with the TBSS tool 49 . Thus, at each point in this skeleton, a given subject’s FA map is searched in the perpendicular direction to find the maximum FA value. This value is projected onto the ENIGMA-DTI skeleton. For the other three anisotropy measures (i.e., TDF-FA, GFA and AP) two approaches were used for skeletonization: 1) ‘FA-sampling’, the alternative anisotropy value of the previously selected voxel in the FA map was projected onto the skeleton; and 2) a new perpendicular search for the maximum value for each anisotropy measure was performed and projected onto the skeleton (‘TDF-sampling’, ‘GFA-sampling’ and 45 ‘AP-sampling’, respectively). This yielded two skeletonized maps per subject for each alternative anisotropy measure. Mean values were calculated for all anisotropy measures along the skeleton within each region of interest (ROI) defined by the Johns Hopkins University WM atlas (JHU-ICBM-DTI-81) distributed by FSL 64 . The ROIs included are shown in Figure 1. For all analyses, we used the mean of the right and left values for the bilateral ROIs, for each measure; we also included the average of all WM ROIs (Average-WM) and we excluded the corticospinal tract, fornix and cingulum of the hippocampus as these ROIs are difficult to register reliably 65 . Figure 1. Regions of Interest (ROIs) used in the present analysis derived from the JHU-ICBM-DTI-81 White Matter Atlas. 46 Table 2. Diffusion MRI acquisition parameters by site. Although some sites (UCLA, Newcastle and Cardiff) used two scanners to acquire data, acquisition parameters were identical across scanners at those sites. Site In-plane Resolution FOV (mm) Slice Thickness (mm) In-Plane Voxel Size (mm) No. of Slices No. of Diff. Directions b -value (s/mm² ) Field Strength Scanner Type TR (ms) TE (ms) Cardiff University 96 x 96 230 2.4 2.39 x 2.39 60 30 1000 3T GE Signa 18750 92.6 Maastricht University 112 x 112 230 3 2.05 x 2.05 38 32 1000 3T Phillips Intera 4834 94 SUNY Upstate 128 x 128 256 2 2 x 2 73 64 900 3T Siemens Tim Trio 10000 87 UCLA 96 x 96 190 2 1.97 x 1.97 50 64 1000 3T Siemens Tim Trio 7100 93 UPenn 128 x 128 240 2 1.87 x 1.87 70 64 1000 3T Siemens Tim Trio 8100 82 UC Davis 128 x 128 230 1.8 1.79 x 1.79 64 60 700 3T Siemens Tim Trio 6600 70 Institute of Psychiatry, London 96 x 96 210 2 2.18 x 2.18 52 64 1000 3T GE Signa 8000 95.6 University of Newcastle 104 x 104 250 2.4 2.40 x 2.40 65 64 1000 1.5T Siemens Avanto 8400 88 4.3.3. Diffusion Anisotropy measures Tensor Distribution Function The Tensor Distribution Function (TDF) represents the diffusion profile as a probabilistic mixture of tensors 26 allowing the reconstruction of multiple underlying fibers per voxel, together with a distribution of weights. We compute the voxel-wise TDF as the probability distribution function defined on all ( D) P feasible 3D Gaussian diffusion processes in tensor space : 47 Here is the measured intensity signal, where , , and are the ( q) S q δ G = r r δ G gyromagnetic ratio, the duration of the diffusion sensitization, and the applied magnetic gradient vector, respectively. The number of detected peaks is estimated by examining the local maxima of the tensor orientation distribution (TOD), defined in the unit sphere along directions : θ Here are the eigenvalues. The TDF-averaged eigenvalues of each fiber were λ calculated by computing the expected values along the principal direction of the fiber. From these eigenvalues a scalar TDF anisotropy measure (TDF-FA) is calculated as an extension of the standard FA formula (see Introduction): The intent of the TDF-FA measure is to replicate the standard FA measure in regions with a single dominant diffusion direction, but also adjust for the partial volume effects and weigh the relative contributions of different fiber pathways when there are multiple dominant directions of diffusion per voxel. The measure has been shown to help reduce the artificial drop in FA where fibers cross, and it can boost power to detect degeneration in Alzheimer’s disease 25 . 48 Generalized Fractional Anisotropy from Q-ball Imaging Q-ball imaging is a model-free approach that is compatible with non-Gaussian diffusion (i.e., due to crossing fibers or higher b -values, i.e. b ≥ 2000 s/mm 2 ) 27,28 . It is based on the q -space theory that establishes a Fourier relationship between the signal decay and the diffusion displacement profile, i.e., a probabilistic function that describes the probability for molecules to have diffused a certain distance at a given diffusion time: Here is the normalized signal attenuation measured at wave vector 𝑞 over E ( q, ) τ diffusion time 𝜏 , which can also be expressed as , where is the baseline S 0 S( q,τ) S 0 image acquired without diffusion sensitization; is the diffusion propagator P measured over the displacement ℝ and 𝜏 ; 𝑞 is defined as . The b -value is q = 2π rδ G related to 𝑞 and the diffusion time by . The diffusion orientation π q τ b = 4 2 2 distribution function (ODF) is approximated by estimating the radial integral of the diffusion propagator. This is accomplished by applying the Funk-Radon transform on the diffusion propagator. Here we defined the ODF with a Laplace-Beltrami regularization 28 . The Generalized Fractional Anisotropy (GFA) is defined as the variance across multiple directions of the ODF 27 : 49 Where and are the standard deviation and the root mean square of t d(Ψ) s m s r (Ψ) the ODF , respectively; and is the i th direction of the ODF; is the Ψ ( u ) Ψ i Ψ 〉 〈 mean of the ODF. Anisotropic Power Map The anisotropy of a high angular resolution diffusion imaging (HARDI) signal can be calculated by measuring the total power of the spherical harmonic (SH) coefficients encoding only the anisotropy information of the signal 103 . The ‘anisotropic power’ (AP) is a measure that is relatively robust to noise compared to measures derived directly from the ODFs (e.g. GFA). Here we used a QBI reconstruction, as described above, to derive the SH coefficients of order l =6. If are the SH of the HARDI signal normalized to the non-diffusion sensitized a l, m signal , the AP can be defined as the sum of the angular power spectrum of S 0 S= S D W I each SH of even order l ≥ 2 as: A neper scale is used to normalize the power signal with a reference power from a fully anisotropic tensor ([2.0, 0.0, 0.0] x 10 -3 mm 2 /s): 4.3.4. Statistical Analysis Group differences between 22q11DS and HC were investigated using a meta-analysis approach, which runs statistical comparisons for each site separately and combines the summary statistics across sites 65 . A multiple linear regression was run for each site including the average anisotropy measure for each ROI as the 50 dependent variable and diagnosis as the predictor of interest. Age, [age-mean(age)] 2 and sex were included as covariates. Given that DTI-derived measures tend to peak between 11 and 20 years for commissural and association fibers and in the early twenties for projection fibers 66 , we included both the linear and quadratic effects of age in the model. The quadratic age term was centered to avoid collinearities with the linear age term. In addition, because females and males show different trajectories of DTI measures across development 68 , sex was accounted for in the model. Cohen's d effect sizes for diagnosis were computed. Subsequently, an inverse-variance weighted mixed-effect meta-analysis (R metafor package 104 ) was used to combine individual site effect sizes. False discovery rate (FDR) correction for multiple comparisons was used (Matlab brainder package) 74,105 . The FDR threshold is computed for the vector of resulting p-values (i.e., for 19 ROIs) for each anisotropy measure. The percentage of tolerated false positives was set to 5% ( q < 0.05). 4.4. Results FA - the standard definition of diffusion anisotropy from the single-tensor model – showed a mixed pattern of both significant increases and decreases in 22q11DS vs. HC. The corpus callosum, the internal capsules and the superior and posterior sections of the corona radiata (SCR and PCR) showed higher FA, on average, in 22q11DS. The fornix/stria terminalis (FXST), the superior longitudinal fasciculus (SLF) and the external/extreme capsule (EC) showed lower FA in 22q11DS. Interestingly, TDF-FA was the only anisotropy measure that showed a consistent direction of significant effects: 22q11DS showed higher TDF-FA than HC in all ROIs. The other two anisotropy measures (GFA and AP) showed mixed patterns of effects, similar to FA. Most of the ROIs with significant differences in GFA showed negative effect sizes (lower values in 22q11DS subjects). For the GFA, only the tapetum (TAP), and the internal capsules (ALIC and PLIC) showed higher GFA in 22q11DS. The PCR, FXST, EC and the Average WM showed lower AP in 22q11DS, whereas the PLIC, SLF, TAP and GCC showed the largest effect sizes with higher AP in 22q11DS. Figures 2 and 3 , and Tables 3 and 4 summarize these results. Figure 4 shows the results overlaid on the JHU-WM atlas. 51 Figure 2. Results based on the FA-sampled voxels (first sampling scheme). Bar graphs showing the effect sizes (Cohen’s d ) for group differences in each respective measure, in the comparison between 22q11DS subjects and healthy controls after meta-analysis. Positive effect sizes represent higher anisotropy in 22q11DS (relative to controls) and negative effect sizes represent lower anisotropy in 22q11DS (relative to controls). Statistically significant results are highlighted by colored stars above and below the bars ( p <FDR significance threshold at q =0.05). The ROIs are sorted in order of effect size for group differences in the standard FA measure. Notably, the TDF-FA is never lower, on average, in 22q11DS than in matched controls. Notably, when comparing the two-voxel sampling schemes for GFA, AP and TDF-FA to project the anisotropy values to the ENIGMA-DTI skeleton there were no differences in terms of the directionality of the effects. For all the significant results within each anisotropy measure the effect sizes were in the same direction. Even so, when different sampling schemes were used, there were differences in the number of ROIs passing multiple comparison correction. For GFA, there were more significant results with the FA-sampling scheme -16 of 19 ROIs, as opposed to 12 of 19 with the GFA-sampling. For TDF-FA and AP the opposite was true. 15 of 19 ROIs were significant for TDF-FA with a TDF-sampling scheme, whereas 12 of 19 52 ROIs were significant when the voxels were sampled from the FA maps. For AP, 12 of 19 ROIs were significant when using the AP-sampling as opposed to 10 of 19 ROIs with the FA-sampling. Intuitively, one might expect that stronger effects would be seen when using a skeleton derived from the measure being considered, rather than always deriving the skeleton from FA, but in reality differences were minor. Table 3. Results based on the FA-sampled voxels. Yellow-colored cells represent the p -values that passed multiple comparisons correction. FDR thresholds were p =0.0263, 0.0298, 0.022, 0.0218, for FA, GFA, AP and TDF-FA, respectively. Abbreviations: d =Cohen’s d , p = p -value, SE=standard error, WM=Average WM. The ROIs are organized in order of the effect sizes for group differences in FA, from positive to negative. FA GFA AP TDF-FA ROI d p SE d p SE d p SE d p SE TAP 0.91 5.34E-20 0.10 0.57 3.51E-09 0.10 0.36 1.28E-04 0.09 0.92 1.90E-07 0.18 PLIC 0.80 3.21E-12 0.12 0.74 4.72E-07 0.15 0.40 5.71E-05 0.10 0.02 9.15E-01 0.15 ALIC 0.62 1.50E-10 0.10 0.39 8.37E-05 0.10 0.29 5.31E-02 0.15 0.52 6.03E-04 0.15 PCR 0.58 1.80E-09 0.10 -0.43 5.60E-06 0.10 -0.28 4.45E-03 0.10 1.33 1.22E-26 0.12 GCC 0.50 1.51E-07 0.10 -0.29 2.30E-03 0.09 0.35 7.38E-04 0.10 0.45 4.26E-02 0.22 SCC 0.41 1.84E-03 0.13 -0.23 1.63E-02 0.10 0.34 3.37E-02 0.16 0.19 2.10E-01 0.15 BCC 0.41 1.86E-05 0.10 -0.15 2.16E-01 0.12 0.17 7.61E-02 0.09 0.58 1.31E-09 0.10 SCR 0.28 3.23E-03 0.09 -0.16 9.29E-02 0.09 -0.06 5.48E-01 0.09 0.61 2.04E-10 0.10 RLIC 0.25 1.22E-02 0.10 -0.33 5.52E-04 0.09 0.03 8.16E-01 0.15 0.68 3.20E-06 0.15 CGC 0.19 7.21E-02 0.11 -0.44 1.70E-04 0.12 0.21 1.18E-01 0.14 0.17 1.44E-01 0.12 ACR 0.18 2.79E-01 0.17 -0.38 1.61E-02 0.16 0.37 2.20E-02 0.16 1.01 5.43E-24 0.10 SS 0.12 3.66E-01 0.13 -0.75 8.60E-10 0.12 0.00 9.89E-01 0.11 0.84 6.04E-08 0.15 WM 0.09 3.73E-01 0.10 -0.53 1.85E-07 0.10 -0.33 5.94E-04 0.09 0.55 9.13E-09 0.10 UNC 0.05 5.95E-01 0.09 -0.17 6.36E-02 0.09 -0.19 7.46E-02 0.11 0.34 2.18E-02 0.15 PTR -0.01 9.39E-01 0.13 -0.97 3.49E-22 0.10 0.27 1.24E-02 0.11 1.02 1.19E-08 0.18 SFO -0.19 3.31E-01 0.19 -0.35 4.67E-04 0.10 0.15 2.15E-01 0.12 -0.12 3.98E-01 0.14 FXST -0.3 0 1.92E-02 0.13 -0.76 5.38E-15 0.10 -0.34 3.39E-04 0.09 0.23 1.30E-01 0.15 SLF -0.32 7.28E-04 0.09 -0.92 9.82E-21 0.10 0.37 1.35E-04 0.10 0.54 4.19E-04 0.15 EC -0.41 8.86E-0 4 0.12 -0.68 2.11E-09 0.11 -0.34 2.80E-04 0.09 0.38 5.58E-05 0.09 In general, the three alternative anisotropy measures showed larger effect sizes when the skeletons were filled in with voxels sampled from their own maps as opposed to voxels sampled from the FA map (compare Tables 3 and 4 ). 53 Additionally, the TDF-FA had the highest effect sizes on average (across all ROIs) of all anisotropy measures (max=1.37, mean= 0.64). This is consistent with the intent of the TDF model, which is to replicate the effects of FA in areas with minimal fiber crossing but to adjust for depletion in FA that can result in regions with extensive fiber crossing. This boosting of the TDF-FA value relative to FA has been reported in prior studies 106,107 . FA computed from the standard tensor model had the second largest effect sizes (max=0.9, mean =0.22), GFA the third largest effect sizes (max=0.76, mean=-0.24), and AP had the lowest effect sizes (max=0.59, mean=-0.1). Figure 3. Results based on the second sampling scheme, where a search for the maximum value for each anisotropy measure is performed and projected onto the skeleton. The bar graphs show the effect sizes (Cohen’s d ) of the comparison between 22q11DS subjects and healthy controls after meta-analysis. Positive effect sizes represent higher anisotropy in 22q11DS and negative effect sizes represent lower anisotropy in 22q11DS. Colored stars above and below the bars highlight the statistically significant results ( p <FDR significance threshold at q =0.05). Despite some similarities, the biggest differences appeared between the anisotropy measures. Unlike FA, where the majority of the effect sizes showed higher values in 22q11DS, GFA showed lower values in 22q11DS for most of the ROIs. The only coinciding results between FA and GFA were in the TAP, ALIC and PLIC, with higher anisotropies in 22q11DS for both measures. In addition to this, GFA showed 54 significant negative effect sizes in ROIs where FA showed no significant differences - such as in the posterior thalamic radiation (PTR), the sagittal stratum (SS) and the overall WM ( d ~0.5-0.9). Opposite to GFA, TDF-FA showed positive effect sizes across all ROIs and also in many ROIs where FA did not show significant differences. The largest effects were seen in the TAP, PTR, all parts of the corona radiata (PCR, ACR and SCR) and the overall WM. Interestingly, ROIs known to have larger axons belonging to fiber bundles in one direction without other major crossing fibers, showed low non-significant effect sizes for the TDF measure (splenium, genu and the posterior limb of internal capsule). AP was the anisotropy measure with the most mixed pattern of results, often coinciding with the directionality of FA and TDF-FA, such as in the corpus callosum and the internal capsules, and other times coinciding with GFA (PCR, average WM, FXST and EC). 4.5. Discussion In this study, we applied several mathematical definitions of diffusion anisotropy in a large sample of the ENIGMA-22q WG. We compared 22q11DS subjects to HC, at the group level, and across multiple cohorts worldwide, by using a harmonized processing and meta-analysis approach across eight independent dMRI datasets. We found diverse patterns of group differences across the anisotropy measures, and besides the ALIC and the TAP, all other ROIs showed divergent results depending on the anisotropy measure. These results appear not to be dependent on the technique used to sample and project anisotropy values to the skeleton. Additionally, sampling methods based on the each anisotropy measure and not on FA seem to yield larger effect sizes especially for TDF-FA and AP. This feature is somewhat expected, in that the TBSS search algorithm always finds the locally highest anisotropy value perpendicular to the skeleton, Consequently the highest FA values will not necessarily correspond to the highest values for TDF-FA, AP and GFA. Inasmuch as these anisotropy measures take into account multiple fiber directions in a voxel, the highest anisotropy values may be located in voxels where FA is artificially low due to fiber crossings. Our findings might indicate that disease specific diffusion patterns are more detectable in regions with complex fiber configurations, which comprise 60-90% of the brain 17 , and TBSS solely based on FA is excluding those voxels from the analysis. This boosting of anisotropy has been particularly reported for TDF-FA 25,106 . 55 Figure 4 . Cohen’s d effect sizes overlaid on the JHU-ICBM-DTI-81 White Matter Atlas for each anisotropy measure. Only statistically significant effect sizes are displayed. The figures are based on the second sampling scheme, where a search for the maximum value for each alternative anisotropy measure (TDF-FA, GFA and AP) is performed and projected onto the skeleton. In a similar manner as FA and contrary to TDF-FA, GFA based on QBI with Laplace-Beltrami regularization 28 is known to be lower when there are crossing fibers, which yields a generalized lower anisotropy across the brain 108 . QBI is also more affected by higher signal-to-noise ratio (SNR) in simulation and phantom studies compared to FA 108 . This was evident in this study as there were fewer significant results when a GFA-sampling scheme was used, because it searches for the highest GFA values; GFA tends to be higher with increasing SNR in voxels with crossing fibers 108 . Nevertheless, GFA has been found to be less sensitive to 56 pathological changes in single fiber bundles but more sensitive to changes in areas of crossing fibers across the brain including studies using TBSS 109 , which we also confirmed. We found corresponding results with FA but with lower effect sizes in the tapetum of the corpus callosum and the internal capsules (i.e. single orientation fiber bundles), but also higher effect sizes in regions with more complex fiber configurations (crossing, diverging, bending or kissing fibers) such as the SFO, PTR, SS, FXST, SLF and EC. AP is the newest and less known anisotropy measure that we included in this comparison. Even though it showed a similar pattern as FA and TDF-FA, the results suggest that it is more sensitive to noise. The direction of effects was also the same as GFA for some ROIs (PCR, Average WM, FXST and EC), which may be caused by the QBI reconstruction used to derive the spherical harmonics. Further validation studies, focusing on the contribution of noise to the observed inter-subject variance in each dMRI measure as well as multi-shell biophysical compartmental models may help elucidate our findings. Finally, the microstructural changes related to 22q11DS are most likely better detected by TDF-FA. As has been shown before, TDF-FA boosts power when detecting disease specific changes compared to the single-tensor FA 25 . Our results show that higher TDF-FA does not only occur in single fiber bundles such as the corpus callosum but also in regions with crossing fibers, which TDF is able to capture due to its multi-tensor nature and hence detecting microstructural changes along all directions in each voxel. Interestingly, TDF-FA was not significant in the posterior limb of the internal capsule (PLIC), despite being mostly composed of cortifugal bundles with one dominant direction. This may be due to the different cortical origin of these fibers (i.e. deeper layers 5 and 6) compared to commissural and association fibers (outer layers 2 and 3), the latter being more compromised during early development in 22q11.2 deletion animal models 9 . Histological validation studies may be helpful to disentangle these histopathological changes. 57 Table 4. Results based on the projected voxels from each anisotropy measure. Blue-colored cells represent the p -values that passed multiple comparison correction. FDR thresholds were p =0.0263, 0.0305, 0.0227, 0.0282, for FA, GFA, AP and TDF-FA, respectively. Abbreviations: d =Cohen’s d , p = p -value, SE=standard error, WM=Average WM. The ROIs are organized in order of the effect sizes for group differences in FA, from positive to negative. FA GFA AP TDF-FA ROI d p SE d p SE d p SE d p SE TAP 0.91 5.34E-20 0.10 0.53 3.72E-08 0.10 0.42 1.20E-05 0.10 1.13 4.52E-14 0.15 PLIC 0.80 3.21E-12 0.12 0.76 4.06E-08 0.14 0.42 1.20E-05 0.10 0.01 9.27E-01 0.14 ALIC 0.62 1.50E-10 0.10 0.36 1.23E-04 0.09 0.38 3.08E-03 0.13 0.56 7.22E-05 0.14 PCR 0.58 1.80E-09 0.10 -0.18 5.36E-02 0.09 -0.31 1.05E-03 0.09 1.37 7.33E-32 0.12 GCC 0.50 1.51E-07 0.10 -0.22 1.75E-02 0.09 0.43 1.18E-04 0.11 0.62 1.43E-03 0.20 SCC 0.41 1.84E-03 0.13 -0.19 8.36E-02 0.11 0.24 9.46E-02 0.14 0.50 8.68E-04 0.15 BCC 0.41 1.86E-05 0.10 -0.16 1.52E-01 0.11 0.30 1.33E-03 0.09 0.73 6.05E-14 0.10 SCR 0.28 3.23E-03 0.09 -0.13 1.61E-01 0.09 -0.08 3.86E-01 0.09 0.67 5.73E-12 0.10 RLIC 0.25 1.22E-02 0.10 -0.24 9.52E-03 0.09 -0.13 2.49E-01 0.11 0.73 1.37E-07 0.14 CGC 0.19 7.21E-02 0.11 -0.14 1.84E-01 0.10 0.59 1.15E-09 0.10 0.28 6.66E-02 0.15 ACR 0.18 2.79E-01 0.17 -0.35 4.49E-02 0.17 0.35 2.68E-02 0.16 1.02 3.66E-24 0.10 SS 0.12 3.66E-01 0.13 -0.61 1.99E-07 0.12 -0.06 6.12E-01 0.11 0.90 3.65E-08 0.16 WM 0.09 3.73E-01 0.10 -0.36 4.28E-04 0.10 -0.30 1.26E-03 0.09 0.95 1.06E-21 0.10 UNC 0.05 5.95E-01 0.09 -0.19 4.51E-02 0.09 -0.19 1.01E-01 0.12 0.34 2.82E-02 0.15 PTR -0.01 9.39E-01 0.13 -0.91 1.78E-18 0.10 0.05 6.45E-01 0.11 1.08 7.52E-16 0.13 SFO -0.19 3.31E-01 0.19 -0.49 5.39E-05 0.12 0.35 2.17E-02 0.15 0.09 4.22E-01 0.11 FXST -0.30 1.92E-02 0.13 -0.56 1.52E-06 0.12 -0.42 1.12E-05 0.10 0.25 1.52E-01 0.17 SLF -0.32 7.28E-04 0.09 -0.89 1.77E-19 0.10 0.26 7.61E-03 0.10 0.55 4.31E-04 0.16 EC -0.41 8.86E-04 0.12 -0.65 1.37E-07 0.12 -0.42 7.94E-04 0.12 0.39 4.48E-05 0.10 58 Chapter 5 Advanced Microstructural Measures from Multi-Shell Diffusion MRI Acquisitions in 22q11.2 Deletion Syndrome 5.1. Introduction In this chapter we analyze a sample of subjects scanned with a state-of-the-art multi-shell diffusion MRI acquisition that allows for advanced reconstruction models that are better able to describe the brain’s nervous tissue. As a novelty in this chapter, we not only study the effects of the deletion of the 22q11.2 locus genes, but we also analyze their duplication. At a greater scale, both mutations happening in this locus (deletion and duplication), can be considered different versions of copy number variants of genes in 22q11.2. Interestingly, carriers of copy number variants (CNVs) at the 22q11.2 genetic locus show either risk - deletion of the 22q11.2 genes - or protective effects for psychotic illness -duplication of 22q11.2 genes. Rees et al. have previously shown that duplications of 22q11.2 genes protect to some extent against schizophrenia 110 . The results from chapters 3 and 4 showed that probands with 22q11.2 deletion syndrome (22q11Del) have wide areas of the white matter (WM) with higher diffusion tensor fractional anisotropy (DTI-FA) than healthy controls (HC), primarily the corpus callosum and the corona radiata, as well as some areas with lower FA, especially in long association fibers such as the superior longitudinal fasciculus. Additionally, when using a multi-tensor technique (i.e. the tensor distribution function - TDF), we found higher TDF-FA across all regions of the 59 white matter. Furthermore, only one region of the WM, that is the posterior limb of the internal capsule (PLIC), showed higher tensor axial diffusivity (AD) in 22q11Del than HC and no significant differences in TDF-FA. These results imply that although models based on the diffusion tensor might not detect the exact underlying microstructure, they are sensitive enough to detect overall changes in microstructure in brain disease. Even more so, tensor-based models are able to detect differences between WM tracts originating from different cortical layers. As mentioned in previous chapters, the PLIC contains a majority of corticofugal fibers that originate from deeper layers of the cortex (layers V and VI). These results motivate the search for alternative ways of reconstructing the diffusion MRI signal. The aim here is to process the diffusion MRI signal in such a way that it is possible to separate the tissue environment into discrete components, such as the intra-cellular and extra-cellular volume fractions for each voxel. A second aim is to separate the multiple fiber components inside each voxel. By extracting the fiber compartments (also known as “fixels”) in crossing fiber regions we may be able to analyze the microstructure along white matter fiber systems with different cortical origins. Traditionally, DTI-FA has been viewed as a biomarker of “fiber integrity”. Thus, the findings of the previous chapters seem at first counterintuitive. It remains unclear why there is higher FA in 22q11Del than in healthy age-matched controls, and that the microstructural changes that increase DTI-FA in 22q11Del worsen when there is cognitive and clinical decline (see chapter 3). This counterintuitive finding poses challenges for interpreting the developmental white matter abnormalities in subjects with 22q11.2 deletion. Notably, 22q11Del is not the only neurogenetic disease affecting the brain that has also shown increased DTI-FA in the white matter, as this has also been reported for Williams Syndrome 111 , for example. The main factor that contributes to the DTI-FA is the density of coherently organized cellular membranes, mostly derived from axons in the white matter. This principle has also been referred to as “cumulative cellular membrane circumference” 84 . Other factors modulating DTI-FA include several forms of crossing fiber configurations (i.e. dispersing, fanning, diverging, bending and kissing fibers), the intra-cellular volume fraction (ICVF), the extra-cellular volume fraction (ECVF) and the axonal diameter distribution. Each of these factors, alone or in combination, may affect anisotropy in DTI as well as in other tensor-based models. Nevertheless, multi-tensor models such as TDF are more robust to crossing fibers and can model better anisotropy in these scenarios. Recently, since the mid-2000’s, a myriad of models have been developed in order to separate the 60 different water pools of the brain tissue, namely the restricted and the hindered compartments. Restricted diffusion is found inside the axons and hindered diffusion is found in the extra-cellular space, i.e., the space in between the cells. Diffusion is highly restricted in the brain within the confines of the axonal membranes and, if correctly modeled, one can derive the dispersion of the axonal fibers. The dispersion of axons can be interpreted as a complement or opposite to anisotropy. The higher the dispersion of axons in the white matter the lower the anisotropy. Importantly, to disentangle these compartments, it is necessary to acquire dMRI with a multi-shell scheme. That is to say, a spherical gradient sample scheme acquired at multiple b-values in increasing order (e.g. b=1000, 2000, 3000 s/mm 2 and even higher). At higher b-values (> 1500 s/mm 2 ), the mono-exponential decay of the diffusion signal is violated and becomes bi-exponential or multi-exponential when the b-value exceeds 10,000 s/mm 2 . The intra-axonal restricted diffusion is responsible for the slow component of the exponential decay 112 . The slow decaying component also violates the Gaussianity assumption of the Stejskal–Tanner or DTI model of diffusion. Hence novel methods are needed to model the non-Gaussianity. A large family of diffusion MRI reconstruction algorithms that divide the tissue compartments make assumptions about the microstructural shape and organization of each compartment. They are not ‘model-free’ but rather depend on biophysically described objects such as tensors of different shapes, cylinders and sticks in order to model each compartment. Hence, these models are commonly referred to as ‘biophysical models’ . We use the Neurite Orientation Dispersion and Density Imaging (NODDI) model to derive these measures in our subject sample 36,113 . Specifically, we use NODDI-derived measures to determine the effects of gene dosage on the underlying white matter microstructure that may explain variation in clinical outcomes of 22q11.2 deletion (22q11Del) and 22q11.2 duplication (22q11Dup). We compare the ICVF and the neurite orientation dispersion index (ODI) between 22q11Del/22q11Dup and healthy controls. Lastly, our previous findings along with results from studies in the mouse model of 22q11Del suggest that in 22q11Del the microstructural features of corticofugal axonal projections differ from those of the cortico-cortical axonal projections. Cortico-cortical axonal fibers originate predominantly from outer cortical layers II and III, whereas corticofugal fibers, such as the corticospinal tract, originate predominantly from the inner layers V and VI. These cross-tract differences in 61 22q11Del may reflect a primary deficit in neurogenesis of the cortical projection neurons in layers II and III due to a disrupted proliferative capacity of a subset of cortical basal progenitors 9 . Here we introduce for the first time a fixel-based analysis (FBA) 115 to compare 22q11Del subjects and healthy controls in two cortico-cortical tracts, namely the corpus callosum and the superior longitudinal fasciculus (SLF), and the corticospinal tract (CST). FBA allows for an unbiased statistical analysis of the intravoxel fixels (intra-voxel fiber components) along a specific white matter fiber tract. Hence, FBA is dependent on a reconstruction model that can extract the several fiber compartments in the voxel. Noticeable, around 90% of the voxels in diffusion images at current resolution contain more than one fiber population 17 , hence several fixels can be extracted. Traditionally, the most commonly used technique to do so is the Constrained Spherical Deconvolution (CSD) 29 , which will be introduced in the Methods section below. 5.2. Methods 5.2.1. Participants Overall, 52 participants were scanned: 21 participants with 22q11Del (mean age: 20.3 years ± 8.7, 18 individuals with 22q11Dup (mean age: 21.4 years ± 13.5) and 13 healthy controls (HC) (mean age: 21.6 years ± 12.1). Diagnosis for deletion and duplication of the 22q11.2 locus genes was confirmed via multiplex ligation-dependent probe amplification (MLPA) 53 . Diffusion MRI (dMRI) data were acquired in a Siemens Prisma 3 Tesla scanner. Acquisition parameters were the following: voxel size=1.5 mm 3 , TE: 0.08 ms, TR: 3.2 ms. Spherical shells were acquired as follows: 3 b=200 s/mm 2 , 6 b=500 s/mm 2 , 46 b=1500 s/mm 2 and 46 b=3000 s/mm 2 . Short diffusion time or “small delta” (δ) = 13.7 ms and long diffusion time or “big delta” (Δ) = 43.4 ms. Images were corrected for eddy and EPI distortions using FSL’s TOPUP and EDDY tools 59 and normalized to the Illinois Institute of Technology white matter atlas (IIT atlas) using ANTs normalization tools 58,116 . Diffusion sensitized images were normalized to each subject’s corresponding T1-weighted images with BBR 117 . Diffusion tensor imaging (DTI) 15 was fitted with the DIPY software 62 . Watson and Bingham NODDI models were fit for all subjects 36,113 with the DMIPY tool 118 . Constrained Spherical Deconvolution 29 (CSD) was fitted with the MrTrix 62 software 119,120 . The DTI model has been previously explained in the Introduction section. Below we introduce the NODDI and CSD models. 5.2.2. Diffusion MRI models Neurite Orientation Dispersion and Density Imaging (NODDI) NODDI, as proposed by Zhang et al. 36 , is a composite model that takes into account three compartments that affect water diffusion in the brain and it assumes no net diffusion between these compartments. It models the cortico-spinal fluid (CSF) as a Gaussian isotropic ball, the intra-axonal restricted diffusion (ICVF) as a stick 121 (i.e. a cylinder of radius of zero) that represent the axons and dendrites of the brain tissue, and the extra-cellular hindered diffusion as a tensor with a zeppelin shape 112,122 . “Neurite” is a term that refers to the ensemble of axons and dendrites altogether. As opposed to the grey matter, in the white matter, axons are almost exclusively the only neurites found. Thus, “neurite” and “axonal” dispersion will be used interchangeably in this chapter because our analyses center on the microstructure of the white matter. The NODDI model can be expressed as: (1) Where the the total signal is divided into brain tissue signal and the A A t i s s u e free-water cerebrospinal fluid (CSF) signal , and the is the volume fraction A i s o v i s o of the CSF. Furthermore, the brain tissue signal can be formulated as: (2) Where and are the normalized signals of the intra-neurite (or A i c A e c intra-cellular) and extra-neurite (extra-cellular) compartments, respectively. The intra-cellular volume fraction is , and the extra-cellular volume fraction is v i c . The intra-cellular signal is modeled over the Orientation Distribution 1 ) ( − v i c A i c Function (ODF) : 63 (3) Here, q represents the gradient directions and b is the b-value of the diffusion weighting. The diffusion signal corresponds to the attenuation along the cylinder of orientation n with unrestricted intrinsic parallel diffusivity d || . The signal outside the neurites, that is the hindered extra-cellular space, is modeled as the signal attenuation due to anisotropic Gaussian distribution: (4) Where the is a cylindrically symmetric tensor or zeppelin. The dispersion of the neurites was initially modeled as a Watson distribution by Zhang et al. 36 . The Watson distribution has two parameters: . The ODF is modeled (κ, ) W μ with a Watson distribution: (5) Here, the distribution is constrained to be isotropic around the mean orientation 𝜇 , and M is Kummer’s confluent hypergeometric function. 𝜅 is called the concentration parameter. Once is estimated, the orientation dispersion index κ (ODI) is calculated as: (6) ODI goes from 0 to 1, the higher the value the more dispersed the neurites in a voxel. ODI, though complementary to anisotropy, may be more informative than DTI FA in areas with less organized patterns such as areas of multiple fiber crossings as well as towards the gray and white matter boundaries. 64 Recently, a newer version of NODDI was proposed using a Bingham distribution 113 , which can capture anisotropic orientation dispersion, as well as the (κ , , ) B 1 κ 2 μ i isotropic orientation dispersion, such as in the Watson distribution. The Watson distribution is a particular case of Bingham when . Because of this κ = κ 1 = κ 2 property the Bingham distribution is arguably better suited for modeling bending and fanning fibers. (7) Where represents the orientation with the largest dispersion about , and μ 2 μ 1 μ 3 represents the orientation with the least dispersion about . are μ 1 , , μ 1 μ 2 μ 3 orthogonal vectors to each other. Also, and represent de dispersion /(κ ) 1 − β /κ 1 about , along axes and , respectively. is a normalization constant μ 1 μ 2 μ 3 c B determined by Kummer's confluent hypergeometric function. The overall dispersion index (ODI) for the Bingham distribution is calculated as in equation (6) but along both, and . /(κ ) 1 − β /κ 1 Specifically, for both NODDI models (i.e. NODDI-Watson and NODDI-Bingham), parallel diffusivity for the stick and zeppelin were both fixed to 1.7x10 -9 m 2 /s and isotropic diffusivity was fixed to 3x10 -9 m 2 /s. A separate brute-force optimization for every voxel is done and then refined to a local minimum using gradient-based methods, as provided in the dMIPY package 118 . Tract Based Spatial Statistics (TBSS) 49 was used to sample values from FA and NODDI scalar maps (ODI and ICVF) and project them to the 17 IIT skeletonized WM regions (see Figure 1). We used the peak FA values of the white matter to sample the voxels to be projected onto the skeleton (see Chapter 4. for details). Constrained Spherical Deconvolution Constrained spherical deconvolution (CSD), initially proposed by Tournier et al. 29 , makes use of the basic principle of signal convolution applied to dMRI. Here, the MRI diffusion signal is conceived as the result of convolving a spherical distribution function , and convolution kernel corresponding 65 to the response function, where . The object of interest and the one to be estimated via deconvolution is . Formally, the convolution is written as: F E is the diffusion signal originating from the tissue, g is the integration variable. The shorthand on the right will be used from here on. The deconvolution can be expressed as: Deconvolution is an ill-posed problem that is very sensitive to noise and it greatly depends on the choice of the response function. In the dMRI case, this noise yields negative fiber orientations in the FOD, which is not physically possible. Hence, Tournier et. al introduced a non-negativity constraint by using a modified Tikhonov regularization method 29 . The CSD model works under the assumption that the diffusion signal of a single axon micro-environment is adequately represented by a single response function. Given this assumption, for one axon bundle along direction n , the “sharpness” of F along n , describes the “spread” or dispersion of that single axonal micro-environment. Ultimately, F can quantify the anisotropy of the axonal bundles along specific directions. The best mathematical way to represent F is by using the spherical harmonics (SH) basis functions 28,123 . In this context, F is called a fiber Orientation Distribution Function, also known as ‘fODF’ or ‘FOD’. Currently, there are many methods to calculate the response function for CSD and the approach depends on the type of analyses to be performed as well as the type of data at hand (single-shell or multi-shell). The core difference between the algorithms that exist to calculate a response function differ in the strategy used to determine the voxels that will be used for the estimation. One category of algorithms calculate the response function for either a single tissue (i.e. just the white matter), or for multiple tissues, i.e. the white matter, grey matter and corticospinal fluid. The convolution for the single tissue approach (ST), i.e. the white matter, is as follows: 66 Where l max is the order of the spherical harmonics basis functions. In its classic formulation, the white matter response function is a zeppelin shaped kernel estimated directly from the data by averaging the DTI eigenvalues of the voxels with a DTI FA above a certain threshold (FA>0.2) 123 . It can also be iteratively done, by doing CSD and estimating which voxels have truly only one fiber compartment, ultimately just selecting those voxels to average the response function 124 . Recently, a multi-tissue CSD has been proposed, where via an unsupervised approach, the three tissue kernels are estimated from the dMRI data itself 125,126 . The formulation would be: One of the advantages of CSD is that the FOD as opposed to the tensor is able to represent multiple fiber orientations in a voxel. This can also be performed by using single-shell HARDI data with low b-values (b=1000 s/mm 2 ) and at least 30 gradient directions, which makes it a suitable model for most of the clinical data. The other advantage is that, as explained above, one can derive an anisotropy value for each reconstructed fiber component, or ‘fixel’. By estimating the amplitude of each lobe of the FOD one can derive the Hindrance Modulated Orientational Anisotropy (OA), also known as Apparent Fiber Density (AFD), with the admonition that a closer approximation to the actual axonal fiber density is only possible when estimating the response function on diffusion images with larger b-values (b > 3000s/mm 2 ) 37,38 . Higher b-values will provide a slower exponentially decaying diffusion signal corresponding to the restricted diffusion inside the axons. Specifically, for our experiments we used the single tissue approach for the highest shell b=3000s/mm 2 . In order to calculate the response function we used the recursive estimation approach by Tax et al 124 . All algorithms have been implemented in the Mrtrix package 119,120 . We performed CSD for all 22q11Del probands as well as for healthy controls. 67 5.2.3. Statistical Analysis NODDI-derived Measures We performed linear regressions comparing fractional anisotropy (FA), Watson-ICVF, Watson-ODI, Bingham-ICVF and Bingham-ODI across 17 IIT skeletonized WM regions. Figure 1 shows the procedure to extract ROI-based microstructural measures for each subject in the IIT atlas space. We compared Healthy Controls vs. 22q11Del and HC vs. 22q11Dup groups. Diagnosis, age, age 2 and sex were included as independent variables, whereas the diffusion MRI index per ROI (17 x No. of measures) was included as the dependent variable for each regression. P-values were corrected for multiple comparisons with the Benjamini-Hochberg procedure ( q =0.05) 74 . Figure 1. Shows the processing pipeline to extract 17 Regions of Interest of the IIT White Matter Atlas for any microstructural map, e.g. Fractional Anisotropy, and NODDI-derived measures, etc. 68 Hindrance Modulated Orientational Anisotropy (OA) We followed the Fixel-Based Analysis protocol from Mrtrix 120 . After calculating whole-brain fiber orientation distributions (FODs) for all subjects, each FOD was dissected into up to three fixels per voxel. Subsequently, they were warped and reoriented to the IIT-HARDI template’s FODs. The IIT-HARDI template contains the mean spherical harmonic representation of the FODs of 72 healthy subjects. We chose this template because it is in the standard MNI coordinate space, it is population-based with high image sharpness (small white matter structures are visible), high SNR, and has FA values and spatial features that are similar to those of individual subjects 116 . The FODs of the IIT-HARDI template are used to reconstruct whole brain tractograms, from which the desired tracts are extracted in order to perform statistical analysis. Here, we manually segmented corticospinal tract (CST) and the superior longitudinal fasciculus (SLF) in the left and right hemispheres. The basic idea of the Fixel-Based-Analysis 115 , is to reorient each subject’s FODs to a given template FOD (e.g. to a standard template or a study-specific template) 130 . Once the fixels are reoriented, the fixels that are closest to the passing tracts of the white matter bundle under study will be selected and used for further statistical analysis. Specifically, the orientational anisotropy (OA) of the selected fixel will be used for statistical analysis. This is very suitable for regions of crossing fibers, such as in the Corona Radiata where long association fibers, like the SLF, and the CST are crossing. The Fixel-based analysis allows the study of the diffusion properties (i.e. anisotropy) along specific fixel without the contamination of the signal of other existing fixels. Connectivity-based fixel enhancement 131 was used to statistically compare the Hindrance modulated Orientational Anisotropy (OA) 37,38 between both populations (22qDel and HC). The family-wise error was controlled via permutation testing (5000 permutations). Only fixels intersecting the reconstructed tracts (CST and SLF) were analyzed. 5.3. Results 69 The Orientation Dispersion Index (ODI) did not show significant differences between groups (Healthy Controls vs. 22qDel/22qDup). Conversely, ICVF showed significant differences for both NODDI models (Watson and Bingham) showing a high degree of correspondence. Regions with significant FA differences also showed significant ICVF changes, in the same direction as FA. Nevertheless, 22qDel and 22qDup showed opposite directions of effects for FA and ICVF. Significant ICVF differences were found in 10 out of 17 WM regions in 22qDel and 22qDup (See Figures 2 and 3 ). These results may indicate that the abnormalities in FA associated with 22q11.2 CNVs (deletion and duplication) may be primarily related to changes in intra-cellular volume fraction rather than the dispersion pattern of the axons. While ICVF was more sensitive than ODI and FA, we did not detect an advantage of Bingham vs. Watson dispersion modeling in the current sample. Opposite effects of ICVF and FA in 22qDel vs. 22qDup suggest gene dosage effects on the white matter microstructure. The fixel based analysis comparing 22qDel probands and healthy controls shows that the majority of fixels within the Superior Longitudinal Fasciculus (SLF) on both sides of the brain have significant differences in Orientational Anisotropy (OA) (see Figure 4 ), whereas scattered noisy fixels were found to be statistically different along the Corticospinal tract (CST). Our results suggest that a fixel-based dMRI analysis may be able to separate the intravoxel fiber populations, offering a better understanding of the microstructural changes in distinct families of long range white matter ber tracts. Fiber crossing occurring in the deep white matter (Corona Radiata) may often involve the intersection of three tracts, e.g., callosal fibers, fronto-parieto-temporal longitudinal fibers and cortico-fugal projection fibers. These crossing fibers can make it hard to interpret single-valued voxel-based dMRI measures, such as the DTI-derived measures. Here we were able to distinguish OA along the SLF vs. the OA along the CST, in line with our hypothesis that cortico-fugal and cortico-cortical fibers are differentially affected by the deletion of 22q11.2 locus genes. 70 Figure 2. Shows the result of the linear regressions comparing 22q11.2 Deletion Syndrome vs. Healthy Controls. It shows the results for Fractional Anisotropy (FA), NODDI-Watson and NODDI-Bingham Intra-cellular Volume Fraction (ICVF) and Orientation Dispersion Index. Stars indicate the significant regions passing multiple comparisons (False Discovery Rate, q =0.05). Regions of Interest: Fmajor=forceps major, Fminor=forceps minor, Fornix, Lcing2=left cingulum hippocampal portion, Lcing=left cingulum, Lcst=left corticospinal tract, Lifo=left inferior-fronto-occipital fasciculus, Lilf=left inferior longitudinal fasciculus, Lslf=left superior longitudinal fasciculus, Lunc=left uncinate, Rcing2=right cingulum hippocampal portion, Rcing=right cingulum, Rcst=right corticospinal tract, Rifo=right inferior-fronto-occiptal fasciculus, Rilf=right inferior longitudinal fasciculus, Rslf=right superior longitudinal fasciculus, Runc=right uncinate. 71 Figure 3. Shows the result of the linear regressions comparing 22q11.2 Duplication Syndrome vs. Healthy Controls. It shows the results for Fractional Anisotropy (FA), NODDI-Watson and NODDI-Bingham Intra-cellular Volume Fraction (ICVF) and Orientation Dispersion Index. Stars indicate the significant regions passing multiple comparisons (False Discovery Rate, q =0.05). Regions of Interest: Fmajor=forceps major, Fminor=forceps minor, Fornix, Lcing2=left cingulum hippocampal portion, Lcing=left cingulum, Lcst=left corticospinal tract, Lifo=left inferior-fronto-occipital fasciculus, Lilf=left inferior longitudinal fasciculus, Lslf=left superior longitudinal fasciculus, Lunc=left uncinate, Rcing2=right cingulum hippocampal portion, Rcing=right cingulum, Rcst=right corticospinal tract, Rifo=right inferior-fronto-occiptal fasciculus, Rilf=right inferior longitudinal fasciculus, Rslf=right superior longitudinal fasciculus, Runc=right uncinate. 72 Figure 4. Shows the result of the Fixel-Based Analysis (FBA) comparing 22q11.2 Deletion Syndrome vs. Healthy Controls in the Superior Longitudinal Fasciculus. It shows the results when comparing fixel values of Orientational Anisotropy (OA). Hot color indicate the level of significance after multiple comparison corrections 131 . Upper right: Shows the segmented Superior Longitudinal Fasciculus in one brain hemisphere. Upper left and bottom row : Shows in hot colors the fixels along the SLF with significant p-values. 73 6. Conclusions Diffusion MRI is a valuable tool to study the microstructure of the brain and unveil the underpinnings of neurological illness. Despite many of the limitations of early diffusion MRI reconstruction models, such as diffusion tensor imaging and multi-tensor models, these models are still able to provide a global view of the regions of the brain compromised by disease. Precisely,the work shown here on diffusion tensor imaging provided an overall perspective of white matter abnormalities in 22q11.2 Deletion Syndrome. Specifically, we found the basic patterns of abnormalities of anisotropy and diffusivity across the brain in 22q11.2 Deletion Syndrome compared to healthy age-matched controls. Our results derive from the largest neuroimaging dataset to date from individuals with 22q11.2 deletion syndrome, which was crucial to resolve previous conflicting findings from smaller studies. Additionally, with the multi-tensor approach of the tensor distribution function we were able to overcome the crossing-fiber limitation of the single tensor model and were able to show a consistent pattern of increased diffusion anisotropy across the brain in 22q11.2 Deletion Syndrome. After identifying the general pattern of brain microstructural abnormalities caused by the deletion - especially in those regions with an unexpected direction of effects - we used advanced diffusion MRI reconstruction models with multi-shell dMRI acquisitions. These analyses revealed further specific abnormalities in the white matter that were not evident with tensor based approaches. Biophysical models such as Neurite Orientation Dispersion and Density Imaging (NODDI) revealed a larger intra-cellular compartment in 22q11.2 Deletion Syndrome. Interestingly, the opposite pattern was revealed in subjects with 22q11.2 duplication. These findings are encouraging and establish a foundation for future work linking gene dosage of copy number variants on more specific microstructural features such as axonal dispersion, free-water content due to neuroinflammation, the axonal diameter distribution, and the diffusion dynamics between intra- and extra-cellular compartments. Multi-shell high angular resolution dMRI data was also able to distinguish microstructural features on specific white matter tracts without the bias of crossing tracts. Current techniques allow for the parcellation of microstructural measures along discrete fiber orientations and perform population analysis, which ultimately allowed us to confirm previous findings reported in animal models of 22q11.2 Deletion Syndrome. 74 7. Supplements Supplementary Figures (Chapter 3) Supplementary Figure 1. Meta-analysis results. Results of the meta-analysis, including nine independent datasets from the ENIGMA-22q11DS working group. Effect sizes for each dataset are shown separately as colored dots. The model tested was: DTI-ROI-measure=ß 0 + ß 1 Diagnosis+ß 2 Sex+ß 3 Age+ß 4 Age 2 centered. WM = Average of all white matter JHU-ICBM ROIs. 75 Supplementary Figure 2. Nonlinear fits of age for Fractional Anisotropy (FA) . Nonlinear Poisson fits of age for FA indices, for typically developing healthy controls (HC; blue) and 22q11.2 Deletion Syndrome (22q11DS; red). 76 Supplementary Figure 3. Nonlinear fits of age for Mean Diffusivity (MD) . Nonlinear Poisson fits of age for MD indices for HC (blue) and (22q11DS (red). 77 Supplementary Figure 4. Nonlinear fits of age for Axial Diffusivity (AD) . Nonlinear Poisson fits of age for AD for HC(blue) and 22q11DS (red). 78 Supplementary Figure 5. Nonlinear fits of age for Radial Diffusivity (RD) . Nonlinear Poisson fits of age for RD for HC (blue) and 22q11DS (red). 79 Supplementary Figure 6. Modulating effects of deletion type. Effect of deletion type on DTI measures for each ROI. Analyses compared 22q11DS subjects with A-D deletion type (N=206) against subjects with A-B deletion type (N=15). 80 Supplementary Figure 7. Correlation between DTI measures and IQ for each ROI . Analyses were run separately for 22q11DS and Healthy Controls. Age, [Age-mean(Age)] 2 , and sex were included as covariates. 81 Supplementary Tables (Chapter 3) Supplementary Table S1. Previous published DTI studies on 22q11DS. Summary of previous published DTI studies comparing probands with 22q11DS and healthy controls. Author (Year) MRI Magnet Strength N (22q11DS / Healthy Controls) 22q11DS Age [mean (SD)] Healthy Age [mean (SD)] FA effects MD effects AD effects RD effects Brief Study Findings Barnea-Goraly et al. 2003 1.5T 38 (19/19) 12.2 (3.9) 14.4 (4.2) Mixed NA^ NA NA Lower FA in Frontal, Temporal and Parietal lobes. Higher FA in Occipital Lobes (Splenium) Simon et al. 2005 1.5T 36 (18/18) 9.8 (1.4) 10.4 (1.9) Mixed NA NA NA Lower FA in Corpus Callosum; Higher FA in Cingulate gyrus & Parietal lobe and Corpus Callosum Simon et al.* 2008 1.5T 36 (18/18) 9.8 (1.4) 10.4 (1.9) 22q> HC HC>22 q HC>2 2q HC>22 q Higuer FA and lower RD in Fronto- occipital fasciculus nd Superior Longitudinal Fasciculus; Lower RD in Parietal lobe, Tapetum, Optic Radiation, Forceps Major; *Reanalysis of Simon et al. 2005 Sundram et al. 2008 1.5T 23 (11/12) 12 (2.2) 13 (2.5) Mixed NA NA NA Lower FA in PLIC, superior Corona Radiata, Tapetum, PTR, Arcuate fasciculus; Higher FA in Genu of Corpus Callosum, ALIC, anterior Corona Radiata. Only lower FA with IQ included as covariate. Negative correlation between psychotic symptoms and FA. Author (Year) da Silva Alves et al. 2011 3T 70 (27/31) 29.8 (7.6) 32.4 (9.7) HC>2 2q* NA NA NA Lower FA throughout cortex (frontal, parietal, parahippocampal) in full sample and SZ+ subsample; Lower FA in parietal cortex in SZ- sample; *Mixed FA results w/o IQ as covariate with increased FA in Cyngulate gyrus. Negative correlation between pshychotic symptoms and FA Kikinis et al. 2012 1.5T 18 (9/9) 27.3 (7.1) 27.2 (6.9) HC>2 2q NA HC>2 2q No signific ant results Lower FA and lower AD in Parietal lobe: intersection of IFO, SFL, ILF , Cingulum and ATR. Radoeva et al. 2012 1.5T 49 (33/16) 17.7 (1.8) 18 (1.7) HC>2 2q NA HC>2 2q HC>22 q Lower FA in Uncinate Fasciculus. Higher AD in SCP , PTR, ACR, SCR, PCR, CGC, SLF , IFO, SS, EC, RLIC. RD changes in Corona Radiata. ^Unaffected siblings only in comparison sample. Villalon-Reina et al. 2013 3T 39 (19/20) 10.7 (1.8) 10.1 (2.2) HC>2 2q Mixed Mixed Mixed Lower FA in Superior Temporal gyri & Superior Corona Radiata. *Female only sample; Other genetic disorders included Perlstein et al.ª 2014 1.5T 99 (52/47) 18 (2.2) 18.1 (1.6) Mixed NA HC>2 2q HC>22 q Lower FA, AD in Fornix; Higher FA, lower RD in ALIC. Lower AD, RD in Uncinate. ªFollow- up study to Radioeva, 2012 - Healthy group= HC+unaffected siblings MRI Magnet Strength N (22q11DS / Healthy Controls) 22q11DS Age [mean (SD)] Healthy Age [mean (SD)] FA effects MD effects AD effects RD effects Brief Study Findings Author (Year) Jalbrzikowski et al. 2014 3T 65* (36/29) 16.3 (4.3) 15.5 (3.8) Mixed NA HC>2 2q HC>22 q Higher FA, lower AD and RD throughout white matter (whole-brain analysis). Lower FA in Cingulate Gyrus proximal to Hippocampus (ROI- analysis). *Multi- scanner study Deng et al. 2015 3T 81* (43/38) 10.83 (1.9) 10 (2.3) HC>2 2q 22q>H C No signifi cant results HC>22 q Lower FA in Fornix; Fornix was only ROI reported. *Tractography study of the Fornix - some 22q data failed tract estimation Kates et al.º 2015 1.5T 97 (51/46) 18 (2.3) 18 (1.6) Mixed NA HC>2 2q HC>22 q Lower FA in anterior Cingulum (CB), higher FA in posterior CB; Lower RD and AD in CB. Lower FA in CB in 22q not treated with antipsychotic or mood stabilizers; ºFollow-up study to Radioeva 2012 and Perlstein, 2014 - healthy group=HC+unaffect ed siblings; *Tractography study of the Cingulum. Kikinis et al.^ 2016 1.5T 97 (50/47) 18.1 (2.3) 18 (1.6) No signifi cant result s HC>22 q HC>2 2q HC>22 q Lower MD, AD, RD in Corpus Callosum, SLF , Corona Radiata. Lower AD in high risk for psychosis 22q patients. ^Follow-up study to Radioeva 2012 and Perlstein, 2014 and Kates, 2015 - healthy group=HC+unaffect ed siblings; MRI Magnet Strength N (22q11DS / Healthy Controls) 22q11DS Age [mean (SD)] Healthy Age [mean (SD)] FA effects MD effects AD effects RD effects Brief Study Findings Author (Year) Bakker et al.* 2016 3T 54 (21/33) 26.2 (3.6) 26.2 (5.5) 22q> HC HC>22 q HC>2 2q HC>22 q Higher FA in Corpus Callosum, FOF , and Thalamic Radiations. Lower AD, RD, MD in Corpus Callosum, Thalamic Radiations, SLF , ILF , FOF . *Study also compared 22qDS to non-22q psychosis ultra- high-risk subjects. Olszewski et al. 2017 3T 87 (57/30*) 20.8 (2.29) 20.9 (1.46) 22q> HC NA No signifi cant results HC>22 q Higher FA and lower RD in IFO, Cingulum, ATR. Lower RD in ILF . Prodromal symptoms and psychosis related to higher FA and lower RD in IFO. *12 subjects were unaffected siblings. Roalf 2017 et al. 2017 3T 78 (39/39) 19.8 (4.25) 19.9 (1.73) HC>2 2q HC>22 q HC>2 2q 22q>H C Lower FA in CGC and CGH. Lower MD in ILF . Lower AD in Forceps major, CGC, IFO, ILF , SLF . Higuer RD in CGH. MRI Magnet Strength N (22q11DS / Healthy Controls) 22q11DS Age [mean (SD)] Healthy Age [mean (SD)] FA effects MD effects AD effects RD effects Brief Study Findings Author (Year) Supplementary Table S2a. Demographics of 22q11DS probands. Demographics of 22q11DS probands: deletion size and psychosis diagnosis. AD = typical deletion type A-D; AB = deletion type A-B; Other = Other atypical deletions. Site Deletion Type Psychotic Disorders UPenn 36 AD; 4 AB; 2 Other 4 Yes; 39 No UCLA 40 AD; 4 AB; 4 Other 5 Yes; 44 No SUNY Upstate 17 AD; 1 AB 2 Yes; 32 No University of Newcastle 9 AD — Maastricht University 13 AD; 4 AB 10 Yes; 14 No Institute of Psychiatry London 10 AD; 1 AB 3 Yes; 19 No UC Davis #2 40 AD; 1 Other — UC Davis #1 10 AD — Cardiff University 5 AD — Utrecht University 26 AD; 1 AB; 1 Other 11 Yes; 43 No Total 206 (89.9%) AD; 15 (6.5%) AB; 8 (3.5%) Other 35 (15.5%) Yes; 191 (84.5%) No Table S2b. Demographic information of the 22q11DS participants with and without psychosis. Age distribution for both groups is non-normal (Shapiro-Wilk test). Test statistic and p-values are shown in the table. A percentile bootstrap test was used to estimate the group differences of age. 95% confidence intervals (CI) are shown (reject H 0 if CI contains zero). SD = Standard deviation. 22q11.2 DS with Psychosis 22q11.2 DS without Psychosis Group differences Number of participants 35 (15.5%) 191 (84.5%) Total participants: 226 Age (mean, SD) mean: 23.87; SD: 8.02 W = 0.92 p = 0.026 mean:17.99; SD: 5.78 W = 0.94 p = 1.004e-06 Non-equal means: 95% CI = (-8.711487 -3.069191) Equal variances: 95% CI = (-82.57228 22.51956) Sex 15 (43%) F; 20 (57%) M 81 (42%) F; 110 (58%) M X 2 = 5.11 (p = 0.023) Supplementary Table S2c. Psychotropic Medications of the 22q11DS Participants across Sites at the time of data acquisition. Site Typical Antipsychoti c Atypical Antipsychoti c Anti- convulsant Psycho- stimulant Anti- depressant Lithium UCLA 3 2 3 10 19 0 Davis_1 0 1 1 6 1 1 Davis_2 0 2 1 11 3 0 IoP 0 4 0 0 4 0 Maastrich t 0 10 1 2 5 2 Newcastl e 0 2 1 1 2 0 SUNY 0 5 1 3 8 0 UPenn 2 6 2 6 15 0 Utrecht 1 5 2 5 2 1 Total 6 37 12 44 59 4 Supplementary Table S3. Clinical characteristics of study participants. Clinical characteristics of 22q11DS and control participants, across sites: inclusion/ exclusion criteria and instruments for diagnosing psychotic disorder and psychotic symptom severity, by site. References indicate representative publications for each study sample. Site Study Inclusion and Exclusion criteria Instrument for psychiatric diagnosis / Rating psychotic symptoms severity Instrument for IQ Assessment Citations UCLA (1,2) Inclusion Criteria 22q11DS and Controls: 1) no significant abuse of drugs or alcohol during the last 6 months or prior abuse/dependence likely to lead to central nervous system impairment; 2) between 5 and 50 years of age; 3) sufficient acculturation and fluency in the English language to avoid invalidating research measures Inclusion Criteria 22q11DS Only: 1) Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) 2) no evidence of a comorbid neurological disorder (e.g., uncontrolled epilepsy, encephalitis); 3) Verbal IQ >= 60 in order to complete the Structured Interview for Prodromal Syndromes Inclusion Criteria Controls Only: 1) no evidence of current or past significant psychopathology; 2) self- and parent report of no prior treatment for psychiatric disorder; 3) no evidence of traumatic brain injury, or other neurological disorder or impairment; 4) no history of significant medical complications likely to affect cognitive functioning (e.g., Type I diabetes, cancer, neural tube defects, etc.) 5) no first-degree relative has been diagnosed with or treated for psychotic disorder; and 8) sex, age, race, and parental educational level comparable to that of the patient participants 6) verbal IQ >= 70 SCID 1 interview (over 10) C-DISC 2 (18 & under) SIPS 3 BPRS 4 WASI 57 Ho et al., 2012 14 Jalbrzikowski et al., 2012 15 , 2013 16 , 2014 17 Jonas et al., 2015 18 Schreiner et al., 2013 19 , 2017 20 Site State University of New York at Upstate (SUNY) Inclusion Criteria 22q11DS and Controls: 1) Between the ages of 9 and 15 years of age at the 1st timepoint, or 12 and 18 years if they entered the study at the 2nd timepoint. 2) No orthodontia or paramagnetic implants. 3) Birthweight Over 2500 grams 4) No traumatic brain injury or loss of consciousness for > 15 minutes 5) No fetal exposure to drugs or alcohol Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria Controls Only: 1) History of severe psychiatric disorder in self or 1st degree relatives 2) Placement in a gifted or special education classroom 3) Seizure or other neurological / genetic disorder SCID 1 SIPS 3 BPRS 4 K-SADS-PL 5 WISC-III 58 Radoeva et al., 2014 21 , Antshel et al., 2013 22 , Radoeva et al., 2012 23 , Kunwar et al.,2012 24 , Kates et al., 2011 25 , Kates et al., 2011 26 , Coman et al., 2010 27 , Roizen et al., 2010 28 , Antshel et al., 2010 29 , Antshel et al., 2008 30 , Kates et al., 2007 31 , Antshel et al., 2007 32 , Kates et al., 2007 33 , Antshel et al., 2007 34 , Aneja et al., 2007 35 , Antshel et al., 2006 36 , Kates et al., 2006 37 UC Davis Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria 22q11DS cases and Controls: 1) Brain infarct, CNS infection, head injury, other focal neurologic abnormality 2) Current or past use of antipsychotic medications 3) Existing diagnosis of psychosis SCID 1 SIPS 3 WISC-IV 59 WASI 57 Scott et al., 2016 38 , Deng et al., 2015 39 , Stephenson et al., 2014 40 Study Inclusion and Exclusion criteria Instrument for psychiatric diagnosis / Rating psychotic symptoms severity Instrument for IQ Assessment Citations Site University of Pennsylvania/ Children’s Hospital of Philadelphia Inclusion Criteria 22q11DS and Controls: 1) Age ≥8 2) Ability to provide informed consent/assent 3) English proficiency 4) Ambulatory and stable medical status 5) Estimated IQ >70 Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria 22q11DS and Controls: 1) Pervasive developmental disorder per medical records or mental retardation (IQ < 70) 2) Medical or neurological disorders that may affect brain function (e.g., uncontrolled seizures, head trauma, CNS tumor, and infection) or visual performance (e.g., blindness). K-SADS-PL 5 SIPS 3 WISC 59 WAIS 60 Yi et al., 2013 41 , Niarchou et al., 2017 42 , Tang et al., 2017 43 , Tang et al., 2017 44 Utrecht University Medical Center, Netherlands Inclusion Criteria 22q11DS: 1) Confirmed diagnosis of 22q11.2 deletion 2) age >=12 3) VIQ >55 K-SADS-PL 5 Dutch version of: WISC-III 58 or WISC-R 61 ; WAIS –III 60 Fiksinski et al., 2017 45 Study Inclusion and Exclusion criteria Instrument for psychiatric diagnosis / Rating psychotic symptoms severity Instrument for IQ Assessment Citations Site Kings College (Institute of Psychiatry), London Inclusion Criteria 22q11DS and Controls: 1) no significant abuse of drugs or alcohol during the last 6 months or prior abuse/dependence likely to lead to central nervous system impairment; 2) between 5 and 50 years of age; 3) sufficient acculturation and fluency in the English language to avoid invalidating research measures Inclusion Criteria 22q11DS Only: 1) Confirmed diagnosis of 22q11.2 microdeletion by FISH test 2) no evidence of a comorbid neurological disorder (e.g., uncontrolled epilepsy, encephalitis); 3) Verbal IQ >= 60 in order to complete the Structured Interview for Prodromal Syndromes Inclusion Criteria Controls Only: 1) no evidence of current or past significant psychopathology or psychiatric disorders; 2) no evidence of traumatic brain injury, or other neurological disorder or impairment; 3) no history of significant medical complications likely to affect cognitive functioning (e.g., Type I diabetes, cancer, neural tube defects, etc.) 4) no first-degree relative has been diagnosed with or treated for psychotic disorder; and 5) sex, age, race, and parental educational level comparable to that of the patient participants 6) verbal IQ >= 70 SCID 1 C-DISC 2 SIPS 3 BPRS 4 WASI 57 In preparation Maastricht Inclusion Criteria 22q11DS and Controls: 1) age >18 years Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria 22q11DS and Controls: Present substance use or history of abuse or dependency, neurological affliction, or pregnancy. PANSS 6 MINI 7 PAS-ADD 8 Dutch version of WAIS –III 60 Bakker et al., 2016 46 Da Silva Alves et al., 2011 47 Study Inclusion and Exclusion criteria Instrument for psychiatric diagnosis / Rating psychotic symptoms severity Instrument for IQ Assessment Citations Site Cardiff Inclusion Criteria 22q11DS and Controls: 1) Age ≥ 10 Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria 22q11.2DS and Controls: Contraindications for MRI scanning (e.g. MRI incompatible implants or prostheses) CAPA 9 SIPS 3 Adults only: SIPS 3 PANSS6 PAS-ADD 8 SCID-II 10 SAPS 11 SANS 12 SPI-A 13 WASI 57 Monks et al., 2014 48 , Niarchou et al., 2014 49 , 2015 50 , Chawner et al., In press 51 Newcastle Inclusion Criteria 22q11DS and Control: 1) English language fluency Inclusion Criteria 22q11DS Only: Confirmed diagnosis of 22q11.2 microdeletion (by FISH or microarray) Exclusion Criteria 22q11DS and Controls: Clinically detectable medical disorder known to affect brain structure (e.g., hypertension), or a history of head injury. Exclusion Criteria Controls Only: 1) The presence of a genetic disorder, mental health problems, a history of severe head injury, seizure disorder, or other ocular, neurological or major medical problems that could influence task performance. K-SADS-PL 5 SCID 1 WISC-III 58 WASI 57 Campbell et al., 2010 52 , 2015 53 , McCabe et al., 2012 54 , 2013 55 ,2014 56 Study Inclusion and Exclusion criteria Instrument for psychiatric diagnosis / Rating psychotic symptoms severity Instrument for IQ Assessment Citations Site Supplementary Table S4. DMRI acquisition parameters. DMRI sequence acquisition parameters by site. Although some sites (UCLA, Newcastle, Cardiff, Utrecht) used two scanners to acquire data, acquisition parameters were identical across scanners for each site. Site In-plane Resolution (acquisition) In-plane Resolution (upsampled) FOV (mm) Slice Thickness (mm) In-Plane Voxel Size (acquisition) (mm) In-Plane Voxel Size (upsampled) (mm) Number of Slices No. of Diffusion Directions b-value (s/mm² ) Field Strength Scanner Type TR (ms) TE (ms) Cardiff 96 x 96 256 x 256 230 2.4 2.39 x 2.39 1.79 x 1.79 60 30 1000 3T GE Signa 18750 92.6 Maastricht 112 x 112 256 x 256 230 3 2.05 x 2.05 0.89 x 0.89 38 32 1000 3T Phillips Intera 4834 94 SUNY 128 x 128 -- 256 2 2 x 2 -- 73 64 900 3T Siemens Tim Trio 10000 87 UCLA 96 x 96 -- 190 2 1.97 x 1.97 -- 50 64 1000 3T Siemens Tim Trio 7100 93 UPenn 128 x 128 -- 240 2 1.87 x 1.87 -- 70 64 1000 3T Siemens Tim Trio 8100 82 Davis_1 128 x 128 -- 220 3 1.71 x 1.71 -- 40 12 1000 3T Siemens Tim Trio 6700 99 Davis_2 128 x 128 256 x 256 230 1.8 1.79 x 1.79 0.89 x 0.89 64 60 700 3T Siemens Tim Trio 6600 70 IoP 96 x 96 256 x 256 210 2 2.18 x 2.18 0.82 x 0.82 52 64 1000 3T GE Signa 8000 95.6 Newcastle 104 x 104 -- 250 2.4 2.40 x 2.40 -- 65 64 1000 1.5T Siemens Avanto 8400 88 Utrecht 128 x 99 128 x 128 240 2 1.87 x 2.4 1.87 x 1.87 75 30 1000 3T Phillips Ingenia 7011 68 Supplementary Table S5. T1-weighted acquisition parameters. Acquisition parameters of the T1-weighted scans, by site. Site Scanner vendor and type Sequence Field Strength Acquisition Direction # of Slices Slice Thickness (mm) Voxel Size (mm3) TI (ms) TE (ms) TR (ms) Flip Angle UCLA 1 Siemens Tim Trio MPRAGE 3T sagittal 160 1.2 1.0x1.0x1.2 900 2.86 2300 9 UCLA 2 Siemens Tim Trio MPRAGE 3T sagittal 160 1.2 1.0x1.0x1.2 900 2.91 2300 9 SUNY Siemens Tim Trio MPRAGE_SAG_B WM 3T sagittal 176 1 1.0x1.0x1.0 1100 3.31 2530 7 UC Davis 1 Siemens Tim Trio MPRAGE_SAGMa gdeburgIPAT 3T axial 160 1 1.0x1.0x1.0 1100 2.93 1820 12 UC Davis 2 Siemens Tim Trio MPRAGE_0.9mm iso w/flow comp 3T sagittal 192 0.9 0.9x0.9x0.9 1100 4.37 2200 7 University of Penn Siemens MPRAGE 3T axial 160 1 0.937x 0.937x1.0 1100 3.51 1810 9 Utrecht Philips Achieva, since March 2016 Philips Ingenia 3D T1TFE 3T axial 160 1 0.875x 0.875x1.0 821.92 4.6 9960 8 IoP GE MPRAGE 3T sagittal 166 1.2 1.0x1.0x1.0 650 2.9 6900 8 Maastricht Philips Intera MPRAGE 3T axial 120 1.2 1.17×1.17×1.20 807.89 4.6 9800 8 Cardiff GE 3D FSPGR 3T oblique-axial 172 1 1.0x1.0x1.0 450 3 7900 20 Cardiff Siemens Prisma MPRAGE 3T sagittal 176 1 1.0x1.0x1.0 850 3.06 2300 9 Newcastle Siemens Avanto MPRAGE 1.5T sagittal 176 1 0.98x0.98x1.0 1100 4.3 1980 15 Supplementary Table S6. Meta-analysis results: 22q11DS vs. Controls. Cohen’s d effect sizes for Diagnosis and Sex and partial correlation r-values of Age and [Age- mean(Age)]2 for each ROI, by DTI measure. The model was tested on each 22q11DS case-control dataset (excluding Utrecht), and effect sizes and correlations were further meta-analyzed. The model tested was: DTI-ROI-measure=ß0+ß1Diagnosis+ß2Sex+ß3Age+ß4Age 2 centered. Blue-shadowed cells indicate a statistically significant result that passed the False Discovery Rate threshold at a q-value of 0.05. JHU- ROI FA Diagnosis Sex Age [Age-mean(Age)] 2 d p st.error d p r p r p ACR 0.228 0.133 0.151 0.002 0.986 0.179 0.004 0.181 0.004 ALIC 0.642 1.04E-12 0.090 0.030 0.816 0.332 1.06E-07 0.177 0.019 Average WM 0.094 0.287 0.088 0.429 0.001 0.331 0.001 0.243 3.58E-04 BCC 0.372 2.69E-05 0.089 0.181 0.159 0.230 0.003 0.247 0.001 CGC 0.199 0.036 0.095 0.410 0.002 0.310 0.001 0.230 1.68E-04 EC -0.466 1.10E-04 0.121 -0.331 0.010 0.305 1.92E-06 0.176 0.005 FXST -0.300 0.006 0.110 0.598 5.42E-06 0.180 0.004 0.123 0.055 GCC 0.582 4.19E-09 0.099 0.182 0.158 0.139 0.029 0.167 0.008 UNC 0.034 0.698 0.088 -0.189 0.142 0.206 0.001 0.168 0.007 PCR 0.520 5.83E-09 0.089 0.507 1.04E-04 0.317 1.78E-05 0.162 0.010 PLIC 0.809 1.93E-15 0.102 0.379 0.003 0.150 0.017 0.185 0.034 PTR -0.008 0.943 0.116 0.085 0.507 0.237 0.001 0.152 0.016 RLIC 0.195 0.065 0.105 0.326 0.012 0.298 1.16E-06 0.112 0.081 SCC 0.440 2.18E-04 0.119 -0.074 0.566 0.181 0.004 0.120 0.060 SCR 0.263 0.003 0.088 0.513 8.49E-05 0.203 0.004 0.115 0.074 SFO -0.117 0.513 0.180 0.133 0.301 0.221 0.001 0.184 0.003 SLF -0.324 2.42E-04 0.088 0.113 0.379 0.316 0.002 0.178 0.005 SS 0.084 0.488 0.121 0.300 0.020 0.328 2.28E-07 0.140 0.027 TAP 0.864 6.00E-21 0.092 0.943 3.97E-12 0.262 6.03E-05 0.175 0.005 JHU- ROI MD Diagnosis Sex Age [Age-mean(Age)] 2 d p st.error d p r p r p ACR -1.025 7.80E-28 0.094 0.517 7.68E-05 0.307 4.44E-05 0.208 0.001 ALIC -0.337 0.007 0.126 0.038 0.768 0.312 9.83E-08 0.159 0.012 Average WM -0.166 0.268 0.150 -0.494 1.52E-04 0.147 0.037 0.143 0.024 BCC -0.765 6.66E-12 0.111 -0.477 2.531E-04 0.234 0.004 0.272 0.004 CGC -0.663 3.45E-08 0.120 -0.492 1.642E-04 0.339 1.73E-05 0.177 0.005 EC -0.383 1.32E-04 0.100 -0.109 0.394 0.329 5.83E-06 0.183 0.004 FXST -0.447 0.002 0.144 -0.893 3.83E-11 0.278 2.72E-06 0.084 0.197 GCC -1.039 2.69E-15 0.131 -0.125 0.329 0.205 0.001 0.140 0.028 UNC -0.357 0.006 0.130 -0.204 0.114 0.236 0.001 0.243 0.003 PCR -1.341 8.56E-26 0.128 -0.317 0.014 0.344 5.75E-06 0.218 3.71E-04 PLIC -0.173 0.415 0.212 -0.455 4.676E-04 0.263 1.05E-05 0.200 0.027 PTR -1.137 9.85E-19 0.129 -0.877 7.994E-11 0.232 1.55E-04 0.131 0.040 RLIC -0.771 2.04E-10 0.121 -1.202 1.107E-17 0.263 0.001 0.134 0.036 SCC -0.884 1.16E-13 0.119 -0.709 8.976E-08 0.280 2.61E-06 0.180 0.004 SCR -0.724 2.11E-05 0.170 -0.319 0.014 0.309 1.47E-04 0.175 0.005 SFO -0.462 2.23E-07 0.089 -0.279 0.031 0.251 3.65E-05 0.171 0.007 SLF -0.907 3.67E-10 0.145 -0.723 5.22E-08 0.338 6.43E-05 0.212 0.001 SS -0.912 6.65E-10 0.148 -1.328 1.44E-20 0.251 3.05E-05 0.123 0.054 TAP -0.873 7.56E-09 0.151 -0.365 0.005 0.335 9.64E-06 0.151 0.017 JHU- ROI AD Diagnosis Sex Age [Age-mean(Age)] 2 d p st.error d p r p r p ACR -0.870 3.81E-21 0.092 0.494 1.52E-04 0.252 2.80E-05 0.079 0.220 ALIC 0.170 0.172 0.125 -0.030 0.812 0.124 0.052 0.107 0.096 Average WM -0.293 0.148 0.202 -0.437 0.001 0.152 0.016 0.122 0.056 BCC -0.512 1.06E-08 0.089 -0.899 2.99E-11 0.165 0.009 0.210 0.046 CGC -0.544 1.25E-08 0.096 -0.174 0.177 0.156 0.013 0.136 0.034 EC -0.926 1.20E-23 0.092 -0.586 8.07E-06 0.167 0.008 0.095 0.137 FXST -0.677 2.88E-09 0.114 -0.248 0.054 0.146 0.021 0.097 0.134 GCC -0.747 6.49E-16 0.092 0.226 0.079 0.241 7.08E-05 0.092 0.155 UNC -0.317 3.42E-04 0.088 -0.431 0.001 0.156 0.014 0.110 0.097 PCR -1.244 9.00E-29 0.112 0.315 0.015 0.226 1.92E-04 0.162 0.011 PLIC 0.572 0.001 0.166 -0.195 0.130 0.214 0.001 0.168 0.008 PTR -1.336 3.41E-13 0.184 -0.822 8.95E-10 0.188 0.003 0.100 0.119 RLIC -0.746 2.20E-16 0.091 -0.761 1.13E-08 0.186 0.003 0.118 0.064 SCC -0.619 5.63E-12 0.090 -0.844 3.45E-10 0.228 1.95E-04 0.173 0.006 SCR -0.344 4.89E-04 0.099 0.113 0.381 0.186 0.003 0.129 0.044 SFO -0.503 6.96E-07 0.101 -0.270 0.036 0.193 0.002 0.157 0.013 SLF -1.332 1.84E-15 0.168 -0.577 1.10E-05 0.152 0.016 0.134 0.036 SS -0.983 1.20E-10 0.153 -0.887 4.98E-11 0.127 0.047 0.126 0.048 TAP -0.060 0.557 0.103 0.309 0.017 0.218 4.09E-04 0.086 0.183 JHU- ROI RD Diagnosis Sex Age [Age-mean(Age)] 2 d p st.error d p r p r p ACR -0.765 3.46E-11 0.115 0.238 0.065 0.256 3.51E-04 0.235 1.36E-04 ALIC -0.628 3.19E-12 0.090 0.030 0.816 0.356 8.31E-09 0.174 0.006 Averag e WM -0.193 0.038 0.093 -0.511 9.17E-05 0.318 2.21E-05 0.238 1.02E-04 BCC -0.594 3.78E-11 0.090 -0.294 0.023 0.264 0.002 0.295 0.001 CGC -0.268 0.002 0.088 -0.435 0.001 0.353 5.98E-05 0.227 2.23E-04 EC 0.142 0.211 0.113 0.199 0.122 0.348 8.33E-07 0.181 0.004 FXST -0.094 0.388 0.109 -0.759 1.22E-08 0.268 2.65E-05 0.117 0.068 GCC -0.812 7.06E-19 0.092 -0.282 0.029 0.120 0.059 0.154 0.015 UNC -0.147 0.168 0.106 0.105 0.414 0.227 1.82E-04 0.241 0.004 PCR -0.953 8.50E-25 0.093 -0.519 7.17E-05 0.339 2.30E-05 0.202 0.001 PLIC -0.694 3.03E-07 0.136 -0.483 2.12E-04 0.173 0.006 0.160 0.107 PTR -0.659 8.47E-08 0.123 -0.469 3.23E-04 0.295 2.45E-04 0.182 0.004 RLIC -0.471 3.79E-06 0.102 -0.801 2.19E-09 0.294 2.95E-05 0.107 0.094 SCC -0.691 1.55E-12 0.098 -0.234 0.069 0.228 4.29E-04 0.145 0.022 SCR -0.613 1.02E-11 0.090 -0.536 4.25E-05 0.257 0.002 0.138 0.030 SFO -0.254 0.037 0.122 -0.207 0.108 0.232 4.33E-04 0.155 0.014 SLF -0.278 0.002 0.088 -0.475 2.68E-04 0.339 3.39E-04 0.200 0.001 SS -0.560 2.54E-05 0.133 -0.963 1.54E-12 0.357 4.20E-07 0.145 0.022 TAP -0.965 1.57E-10 0.151 -0.805 1.84E-09 0.313 1.40E-05 0.187 0.003 Supplementary Table S7. Mega-analysis results: 22q11DS vs. Controls. Cohen’s d effect sizes, Student’s t-values, and corresponding p-values for effects of Diagnosis, Sex, Age, and [Age-mean(Age)]2 for each ROI, by DTI measure. The model was tested on the entire ENIGMA-DTI 22q11DS case-control sample (excluding Utrecht) using the harmonized data. The model tested was: DTI-ROI-measure=ß0+ß1Diagnosis+ß2Sex+ß3Age+ß4Age 2 centered. Blue-shadowed cells highlight the p-values passing the False Discovery Rate threshold at a q-value of 0.05. JHU-ROI FA Diagnosis Sex Age [Age-mean(Age)] 2 d t p t p t p t p ACR 0.212 2.450 0.015 0.340 0.731 5.430 8.46E-08 -5.080 5.16E-07 ALIC 0.614 7.100 3.93E-12 1.840 0.066 11.830 5.00E-16 -4.430 1.16E-05 Average WM 0.070 0.804 4.22E-01 2.538 0.011 12.365 5.00E-16 -7.410 5.07E-13 BCC 0.289 3.340 0.001 1.430 0.154 5.420 9.10E-08 -4.500 8.49E-06 CGC 0.176 2.040 0.042 2.420 0.016 10.820 5.00E-16 -6.220 9.75E-10 EC -0.487 -5.640 2.80E-08 2.490 0.013 9.230 5.00E-16 -4.700 3.26E-06 FXST -0.311 -3.600 3.51E-04 0.530 0.598 4.310 1.96E-05 -2.860 0.004 GCC 0.550 6.360 4.38E-10 2.480 0.013 2.310 0.021 -3.610 3.34E-04 UNC 0.041 0.470 0.638 2.490 0.013 3.730 2.10E-04 -2.470 0.014 PCR 0.477 5.520 5.32E-08 2.270 0.024 8.560 1.11E-16 -6.470 2.27E-10 PLIC 0.797 9.220 5.00E-16 2.750 0.006 4.880 1.42E-06 -3.210 0.001 PTR -0.049 -0.560 0.575 0.760 0.446 5.370 1.19E-07 -5.240 2.26E-07 RLIC 0.151 1.750 0.080 1.330 0.184 6.730 4.25E-11 -5.070 5.48E-07 SCC 0.434 5.020 7.10E-07 -0.170 0.864 4.340 1.71E-05 -4.210 2.98E-05 SCR 0.264 3.050 0.002 1.640 0.101 6.530 1.57E-10 -4.500 8.45E-06 SFO -0.079 -0.910 0.362 1.270 0.203 6.510 1.76E-10 -3.520 4.62E-04 SLF -0.326 -3.770 1.79E-04 1.320 0.189 11.080 5.00E-16 -7.140 2.98E-12 SS 0.039 0.450 0.654 0.320 0.752 8.880 5.00E-16 -6.610 9.39E-11 TAP 0.837 9.680 5.00E-16 1.280 0.201 8.240 1.33E-15 -5.020 7.02E-07 JHU-ROI MD Diagnosis Sex Age [Age-mean(Age)] 2 d t p t p t p t p ACR -0.912 -10.540 5.00E-16 -2.060 0.040 -11.330 5.00E-16 5.550 4.54E-08 ALIC -0.319 -3.690 2.47E-04 -1.690 0.092 -9.320 5.00E-16 2.460 0.014 Average WM -0.641 -7.410 4.91E-13 -0.515 0.606 -0.057 9.55E-01 -0.124 9.01E-01 BCC -0.524 -6.060 2.55E-09 -2.130 0.034 -6.180 1.27E-09 4.050 5.86E-05 CGC -0.571 -6.600 9.70E-11 -2.820 0.005 -11.870 5.00E-16 6.760 3.54E-11 EC -0.356 -4.110 4.53E-05 -1.830 0.068 -11.710 5.00E-16 5.750 1.47E-08 FXST -0.433 -5.010 7.43E-07 -1.180 0.237 -9.310 5.00E-16 3.740 2.00E-04 GCC -0.870 -10.060 5.00E-16 -2.680 0.008 -6.120 1.78E-09 3.160 1.67E-03 UNC -0.294 -3.410 7.10E-04 -0.640 0.525 -6.420 2.98E-10 3.810 1.54E-04 PCR -1.132 -13.090 5.00E-16 -1.360 0.176 -12.610 5.00E-16 7.170 2.47E-12 PLIC -0.236 -2.730 0.01 -1.300 0.194 -9.120 5.00E-16 3.890 1.12E-04 PTR -0.936 -10.830 5.00E-16 -0.170 0.866 -8.730 5.00E-16 5.440 8.24E-08 RLIC -0.697 -8.060 4.88E-15 -1.040 0.297 -9.610 5.00E-16 5.580 3.91E-08 SCC -0.737 -8.520 1.11E-16 -1.940 0.053 -9.040 5.00E-16 4.230 2.75E-05 SCR -0.579 -6.690 5.60E-11 -1.570 0.117 -11.190 5.00E-16 6.550 1.37E-10 SFO -0.354 -4.100 4.83E-05 -0.140 0.889 -6.790 3.05E-11 3.500 0.001 SLF -0.722 -8.350 5.55E-16 -0.930 0.351 -13.100 5.00E-16 7.610 1.22E-13 SS -0.812 -9.390 5.00E-16 0.650 0.515 -8.910 5.00E-16 5.450 7.59E-08 TAP -0.731 -8.450 2.22E-16 -0.950 0.343 -11.970 5.00E-16 7.380 6.11E-13 JHU-ROI AD Diagnosis Sex Age [Age-mean(Age)] 2 d t p t p t p t p ACR -0.849 -9.820 5.00E-16 -1.580 0.114 -8.220 1.55E-15 2.320 0.021 ALIC 0.127 1.470 0.142 0.280 0.781 -0.400 0.693 -1.220 0.223 Average WM -0.543 -6.280 6.94E-10 -0.446 0.655 0.393 0.695 -0.529 0.597 BCC -0.421 -4.870 1.51E-06 -0.600 0.550 -2.610 0.009 0.800 0.426 CGC -0.528 -6.100 1.98E-09 0.620 0.535 2.180 0.030 -1.500 0.135 EC -0.905 -10.470 5.00E-16 0.400 0.690 -3.760 1.87E-04 2.010 0.046 FXST -0.655 -7.580 1.55E-13 -0.540 0.592 -4.710 3.23E-06 1.090 0.274 GCC -0.672 -7.770 3.99E-14 -1.330 0.185 -5.970 4.40E-09 0.900 0.371 UNC -0.258 -2.990 0.003 1.620 0.107 -3.030 0.003 2.030 0.042 PCR -1.129 -13.060 5.00E-16 0.760 0.450 -7.950 1.13E-14 3.390 0.001 PLIC 0.455 5.260 2.04E-07 1.510 0.131 -7.620 1.19E-13 1.920 0.055 PTR -0.924 -10.680 5.00E-16 -0.100 0.922 -4.840 1.73E-06 2.170 0.031 RLIC -0.707 -8.180 2.11E-15 0.510 0.613 -4.340 1.71E-05 1.360 0.175 SCC -0.531 -6.140 1.60E-09 -1.690 0.092 -5.760 1.46E-08 1.360 0.176 SCR -0.329 -3.810 1.56E-04 0.080 0.934 -6.860 1.87E-11 3.050 0.002 SFO -0.480 -5.550 4.53E-08 1.420 0.156 -3.250 0.001 1.340 0.182 SLF -1.145 -13.240 5.00E-16 0.760 0.447 -5.190 2.93E-07 2.770 0.006 SS -0.863 -9.980 5.00E-16 1.090 0.278 -1.340 0.182 0.180 0.861 TAP -0.019 -0.220 0.824 0.090 0.928 -6.350 4.73E-10 4.770 2.36E-06 JHU-ROI RD Diagnosis Sex Age [Age-mean(Age)] 2 d t p t p t p t p ACR -0.677 -7.820 2.75E-14 -1.600 0.111 -9.790 5.00E-16 5.940 5.26E-09 ALIC -0.579 -6.700 5.30E-11 -2.380 0.018 -12.710 5.00E-16 4.490 8.83E-06 Average WM -0.144 -1.661 9.73E-02 -1.895 0.059 -12.540 5.00E-16 7.080 4.44E-12 BCC -0.391 -4.520 7.46E-06 -1.730 0.084 -6.060 2.58E-09 4.470 9.59E-06 CGC -0.222 -2.570 0.010 -3.190 0.001 -13.420 5.00E-16 7.310 9.47E-13 EC 0.134 1.540 0.123 -2.350 0.019 -12.320 5.00E-16 6.040 2.87E-09 FXST -0.061 -0.710 0.480 -1.390 0.164 -9.090 5.00E-16 4.370 1.48E-05 GCC -0.712 -8.230 1.44E-15 -2.640 0.009 -3.700 2.39E-04 3.560 4.06E-04 UNC -0.156 -1.810 0.072 -2.020 0.044 -5.040 6.32E-07 3.020 0.003 PCR -0.828 -9.570 5.00E-16 -2.200 0.028 -11.930 5.00E-16 7.460 3.46E-13 PLIC -0.581 -6.710 4.86E-11 -2.080 0.038 -6.380 3.75E-10 3.610 3.31E-04 PTR -0.556 -6.430 2.88E-10 -0.700 0.486 -8.990 5.00E-16 6.710 5.00E-11 RLIC -0.407 -4.710 3.13E-06 -1.310 0.192 -9.070 5.00E-16 6.060 2.54E-09 SCC -0.568 -6.570 1.18E-10 -0.800 0.426 -7.930 1.29E-14 5.270 1.99E-07 SCR -0.530 -6.130 1.69E-09 -1.740 0.082 -9.590 5.00E-16 6.070 2.45E-09 SFO -0.221 -2.560 0.011 -1.040 0.299 -7.580 1.58E-13 4.060 5.68E-05 SLF -0.220 -2.550 0.011 -1.550 0.122 -14.060 5.00E-16 8.180 2.00E-15 SS -0.479 -5.540 4.87E-08 -0.250 0.802 -11.330 5.00E-16 7.410 4.98E-13 TAP -0.851 -9.850 5.00E-16 -1.090 0.275 -10.470 5.00E-16 6.180 1.24E-09 Supplementary Table S8. Mega-analysis results: 22q11DS vs. Controls, including Age-by-Diagnosis interaction. Mega-analysis results when including the Age-by-Diagnosis interaction term comparing 22q11DS vs. Controls. Student’s t and p-values are provided for the effects of Diagnosis and covariates on each ROI, by DTI measure. The model tested was: DTI-ROI-measure=ß 0 +ß 1 Diagnosis+ß 2 Sex+ß 3 Age+ß 4 Age 2 centered +ß 5 (AgeXDiagnosis). This model was tested on the entire ENIGMA-DTI 22q11DS case-control sample (excluding Utrecht) using the harmonized data. No p-values passed the FDR threshold at a q-value of 0.05 for the variable of interest Age-by-Diagnosis. JHU- ROI FA Diagnosis Sex Age [Age- mean(Age)] 2 Age-by- Diagnosis t p t p t p t p t p ACR 0.760 0.450 4.090 4.88E-05 -4.950 9.80E-07 0.360 0.722 0.800 0.804 ALIC 3.060 0.002 9.400 5.00E-16 -4.390 1.34E-05 1.820 0.069 0.820 0.816 Average WM 0.590 0.557 9.850 5.00E-16 -7.330 8.33E-13 2.520 0.012 0.770 0.773 BCC 1.210 0.226 4.150 3.85E-05 -4.400 1.33E-05 1.430 0.153 0.890 0.888 CGC 1.740 0.083 9.100 5.00E-16 -6.300 6.07E-10 2.360 0.019 0.320 0.316 EC -3.110 0.002 6.650 7.40E-11 -4.460 9.85E-06 2.530 0.012 0.350 0.353 FXST -3.040 0.002 2.300 0.022 -2.520 0.012 0.620 0.533 0.080 0.083 GCC 2.910 0.004 2.050 0.040 -3.620 3.23E-04 2.460 0.014 0.690 0.695 UNC 0.030 0.979 2.810 0.005 -2.400 0.017 2.500 0.013 0.860 0.859 PCR 1.350 0.177 6.110 1.88E-09 -6.200 1.16E-09 2.320 0.021 0.350 0.347 PLIC 3.030 0.003 3.360 0.001 -3.030 0.003 2.790 0.005 0.470 0.466 PTR 0.310 0.754 4.570 6.18E-06 -5.260 2.06E-07 0.730 0.466 0.560 0.556 RLIC -0.290 0.773 4.600 5.29E-06 -4.800 2.08E-06 1.390 0.165 0.280 0.279 SCC 4.830 1.81E-06 5.320 1.50E-07 -4.720 3.04E-06 -0.340 0.735 5.00E-16 0.002 SCR 0.630 0.527 4.710 3.24E-06 -4.310 1.95E-05 1.670 0.095 0.520 0.519 SFO -0.120 0.907 5.260 2.10E-07 -3.510 4.81E-04 1.260 0.209 0.790 0.786 SLF -2.260 0.024 8.170 2.22E-15 -6.880 1.64E-11 1.360 0.175 0.420 0.417 SS -0.140 0.887 6.730 4.48E-11 -6.440 2.73E-10 0.330 0.738 0.730 0.726 TAP 4.030 6.43E-05 6.550 1.40E-10 -4.960 9.26E-07 1.270 0.206 0.870 0.873 JHU- ROI MD Diagnosis Sex Age [Age- mean(Age)] 2 Age-by- Diagnosis t p t p t p t p t p ACR -4.130 4.24E-05 -8.800 5.00E-16 5.440 8.31E-08 -2.070 0.039 0.910 0.913 ALIC -0.830 0.405 -6.860 1.94E-11 2.300 0.022 -1.720 0.085 0.480 0.480 Average WM -3.500 4.95E-04 -0.410 0.685 -0.020 0.985 -0.480 0.630 0.560 0.562 BCC -1.250 0.212 -4.040 6.23E-05 3.760 1.89E-04 -2.200 0.028 0.200 0.196 CGC -3.480 0.001 -9.850 5.00E-16 6.820 2.53E-11 -2.760 0.006 0.360 0.364 EC -1.350 0.179 -8.960 5.00E-16 5.600 3.45E-08 -1.840 0.066 0.740 0.740 FXST -0.940 0.350 -6.560 1.30E-10 3.480 0.001 -1.250 0.213 0.240 0.240 GCC -4.370 1.51E-05 -5.010 7.25E-07 3.170 0.002 -2.660 0.008 0.720 0.717 UNC -2.260 0.024 -5.630 2.87E-08 3.920 9.83E-05 -0.580 0.561 0.330 0.331 PCR -5.170 3.32E-07 -9.810 5.00E-16 7.040 6.08E-12 -1.360 0.175 0.930 0.930 PLIC -3.420 0.001 -8.750 5.00E-16 4.300 2.04E-05 -1.170 0.244 0.010 0.012 PTR -3.610 3.36E-04 -6.330 5.12E-10 5.210 2.75E-07 -0.210 0.832 0.420 0.421 RLIC -3.000 0.003 -7.360 7.05E-13 5.440 8.24E-08 -1.050 0.292 0.800 0.799 SCC -4.820 1.89E-06 -8.030 6.33E-15 4.440 1.11E-05 -1.850 0.064 0.130 0.129 SCR -1.450 0.147 -7.930 1.32E-14 6.210 1.07E-09 -1.640 0.101 0.180 0.178 SFO 0.340 0.731 -3.980 7.95E-05 3.070 0.002 -0.260 0.796 0.030 0.030 SLF -2.630 0.009 -9.760 5.00E-16 7.350 7.54E-13 -0.970 0.330 0.430 0.429 SS -2.290 0.022 -5.970 4.22E-09 5.090 5.07E-07 0.560 0.574 0.110 0.105 TAP -2.920 0.004 -9.040 5.00E-16 7.170 2.60E-12 -0.970 0.330 0.600 0.604 JHU- ROI AD Diagnosis Sex Age [Age- mean(Age)] 2 Age-by- Diagnosis t p t p t p t p t p ACR -3.930 9.51E-05 -6.420 2.99E-10 2.280 0.023 -1.580 0.115 0.990 0.994 ALIC 1.520 0.129 0.320 0.747 -1.380 0.167 0.220 0.825 0.310 0.310 Average WM -3.000 0.003 -0.020 0.986 -0.430 0.669 -0.420 0.677 0.600 0.602 BCC -0.740 0.459 -1.220 0.223 0.550 0.583 -0.670 0.503 0.190 0.186 CGC -2.570 0.011 1.630 0.105 -1.450 0.148 0.630 0.532 0.900 0.897 EC -4.700 3.38E-06 -3.280 0.001 2.070 0.039 0.430 0.669 0.590 0.588 FXST -2.590 0.010 -3.380 0.001 0.990 0.323 -0.560 0.574 0.620 0.623 GCC -3.330 0.001 -4.810 1.98E-06 0.920 0.358 -1.310 0.190 0.820 0.821 UNC -2.290 0.023 -3.110 0.002 2.210 0.027 1.680 0.094 0.240 0.236 PCR -5.470 6.83E-08 -6.380 3.91E-10 3.380 0.001 0.770 0.443 0.800 0.797 PLIC -1.320 0.187 -8.390 4.44E-16 2.580 0.010 1.730 0.084 0.000 1.75E-04 PTR -2.520 0.012 -2.590 0.010 1.790 0.074 -0.200 0.839 0.050 0.053 RLIC -3.900 1.07E-04 -3.820 1.50E-04 1.460 0.145 0.540 0.588 0.500 0.498 SCC -1.810 0.070 -4.060 5.60E-05 1.210 0.228 -1.730 0.085 0.480 0.478 SCR -0.460 0.649 -4.650 4.28E-06 2.790 0.005 0.020 0.985 0.240 0.242 SFO 0.240 0.807 -0.870 0.384 0.840 0.401 1.280 0.201 0.010 0.007 SLF -5.020 7.17E-07 -3.860 1.26E-04 2.670 0.008 0.740 0.459 0.750 0.748 SS -2.190 0.029 0.190 0.846 -0.180 0.856 0.980 0.329 0.050 0.046 TAP 0.200 0.838 -4.760 2.45E-06 4.640 4.49E-06 0.070 0.942 0.750 0.749 JHU- ROI RD Diagnosis Sex Age [Age- mean(Age)] 2 Age-by- Diagnosis t p t p t p t p t p ACR -3.040 0.003 -7.590 1.46E-13 5.820 1.04E-08 -1.600 0.110 0.910 0.911 ALIC -2.530 0.012 -9.840 5.00E-16 4.380 1.41E-05 -2.390 0.017 0.870 0.868 Average WM -0.470 0.638 -9.680 5.00E-16 6.930 1.24E-11 -1.900 0.058 0.830 0.831 BCC -1.200 0.229 -4.330 1.81E-05 4.280 2.24E-05 -1.770 0.078 0.500 0.504 CGC -1.560 0.118 -10.860 5.00E-16 7.300 1.07E-12 -3.160 0.002 0.560 0.561 EC 1.100 0.272 -9.310 5.00E-16 5.850 8.69E-09 -2.380 0.018 0.600 0.600 FXST 0.900 0.368 -6.310 5.71E-10 4.080 5.27E-05 -1.460 0.144 0.200 0.196 GCC -3.590 3.59E-04 -3.090 0.002 3.560 4.11E-04 -2.610 0.009 0.750 0.752 UNC -0.810 0.416 -4.000 7.10E-05 2.990 2.95E-03 -2.010 0.045 0.920 0.922 PCR -3.570 3.84E-04 -9.150 5.00E-16 7.290 1.15E-12 -2.220 0.027 0.770 0.771 PLIC -3.010 0.003 -5.210 2.68E-07 3.610 3.29E-04 -2.050 0.041 0.730 0.727 PTR -3.080 0.002 -7.370 6.45E-13 6.700 5.46E-11 -0.670 0.506 0.580 0.584 RLIC -1.270 0.204 -6.680 6.02E-11 5.840 8.92E-09 -1.340 0.180 0.500 0.499 SCC -4.660 3.94E-06 -7.600 1.34E-13 5.590 3.53E-08 -0.680 0.498 0.030 0.028 SCR -1.570 0.116 -6.900 1.45E-11 5.800 1.14E-08 -1.790 0.074 0.330 0.333 SFO 0.320 0.751 -5.020 6.95E-07 3.740 2.06E-04 -1.120 0.264 0.140 0.142 SLF -0.270 0.785 -10.490 5.00E-16 7.900 1.55E-14 -1.590 0.113 0.410 0.414 SS -1.490 0.137 -8.370 5.55E-16 7.150 2.92E-12 -0.290 0.769 0.420 0.425 TAP -3.620 3.28E-04 -7.960 1.03E-14 6.020 3.31E-09 -1.110 0.267 0.720 0.715 Supplementary Table S9. Mega-analysis results: 22q11DS vs. Controls age 30 and under, including Age-by-Diagnosis interaction. Mega-analysis results when including only subjects aged 30 or younger (N=245 22q11DS and 269 Controls), and including the Age-by-Diagnosis interaction term. Student’s t and p- values are provided for the effects of Diagnosis and covariates on each ROI, by DTI measure. The model tested was: DTI-ROI-measure=ß 0 +ß 1 Diagnosis+ß 2 Sex+ß 3 Age+ß 4 Age 2 centered +ß 5 (AgeXDiagnosis). This model was tested on the case-control sample (excluding Utrecht) using the harmonized data. There was no p-value controlling the FDR at a q-value of 0.05 for the Age-by-Diagnosis term. JHU- ROI FA Diagnosis Age [Age-mean(Age)] 2 Sex Age-by-Diagnosis t p t p t p t p t p ACR 1.260 0.208 4.140 4.08E-05 -0.680 0.496 0.430 0.668 0.580 0.583 ALIC 2.180 0.030 7.920 1.53E-14 -2.760 0.006 1.680 0.094 0.910 0.910 Average WM 0.330 0.739 8.450 3.33E-16 -3.790 1.67E-04 2.300 0.022 0.950 0.945 BCC 0.450 0.650 3.320 0.001 -2.650 0.008 1.240 0.216 0.510 0.506 CGC 1.190 0.234 7.680 8.39E-14 -2.400 0.017 2.170 0.031 0.600 0.604 EC -2.090 0.037 5.880 7.32E-09 -1.430 0.153 2.190 0.029 0.790 0.794 FXST -2.200 0.028 2.260 0.024 -1.120 0.263 0.330 0.741 0.320 0.324 GCC 1.810 0.072 1.320 0.187 -1.090 0.277 2.220 0.027 0.750 0.752 UNC 1.090 0.276 3.110 0.002 -0.030 0.978 2.280 0.023 0.290 0.287 PCR 1.420 0.157 5.830 1.01E-08 -3.500 4.99E-04 2.380 0.018 0.790 0.792 PLIC 2.650 0.008 3.300 0.001 -0.470 0.638 2.930 0.004 0.810 0.810 PTR 0.760 0.448 4.640 4.52E-06 -2.680 0.008 0.580 0.562 0.290 0.291 RLIC -0.260 0.793 4.350 1.66E-05 -2.350 0.019 1.460 0.145 0.430 0.431 SCC 4.190 3.33E-05 5.010 7.43E-07 -2.100 0.036 -0.080 0.936 0.010 0.010 SCR 1.000 0.317 4.590 5.57E-06 -3.040 0.002 1.850 0.065 0.910 0.908 SFO 0.970 0.333 5.630 3.06E-08 -1.400 0.161 1.370 0.170 0.150 0.154 SLF -1.340 0.180 7.450 4.06E-13 -3.660 2.81E-04 1.360 0.175 0.950 0.949 SS 0.230 0.821 6.360 4.57E-10 -3.510 4.89E-04 0.390 0.700 0.870 0.866 TAP 3.120 0.002 5.490 6.23E-08 -2.210 0.027 1.050 0.293 0.980 0.983 JHU- ROI MD Diagnosis Age [Age-mean(Age)] 2 Sex Age-by-Diagnosis t p t p t p t p t p ACR -3.100 0.002 -7.310 1.03E-12 1.640 0.101 -2.070 0.039 0.770 0.770 ALIC -0.250 0.800 -5.500 6.02E-08 1.840 0.067 -1.820 0.069 0.320 0.319 Average WM -1.730 0.085 0.660 0.507 -1.560 0.119 -0.230 0.814 0.380 0.380 BCC -0.530 0.596 -3.240 0.001 2.860 0.004 -2.310 0.022 0.140 0.145 CGC -2.500 0.013 -8.310 8.88E-16 3.600 3.48E-04 -2.900 0.004 0.720 0.722 EC -0.420 0.676 -7.290 1.16E-12 2.940 0.003 -1.910 0.056 0.320 0.320 FXST -1.280 0.200 -6.210 1.09E-09 1.560 0.119 -1.230 0.219 0.840 0.840 GCC -3.150 0.002 -4.000 7.38E-05 0.140 0.890 -2.540 0.011 0.880 0.881 UNC -1.590 0.112 -4.600 5.46E-06 3.590 3.61E-04 -0.250 0.806 0.620 0.622 PCR -3.750 1.95E-04 -8.230 1.67E-15 3.120 0.002 -1.470 0.141 0.650 0.652 PLIC -3.180 0.002 -7.950 1.20E-14 1.090 0.278 -1.200 0.232 0.020 0.018 PTR -3.380 0.001 -6.060 2.62E-09 2.840 0.005 -0.430 0.667 0.990 0.989 RLIC -2.190 0.029 -6.410 3.31E-10 2.530 0.012 -1.170 0.241 0.690 0.688 SCC -4.150 3.84E-05 -6.990 8.49E-12 0.790 0.428 -2.010 0.045 0.170 0.168 SCR -0.970 0.331 -6.780 3.34E-11 3.040 0.002 -1.930 0.054 0.240 0.241 SFO -0.310 0.758 -3.990 7.43E-05 2.270 0.024 -0.660 0.513 0.370 0.371 SLF -2.030 0.043 -8.390 4.44E-16 3.710 2.27E-04 -1.150 0.249 0.540 0.535 SS -1.760 0.079 -5.360 1.24E-07 2.560 0.011 0.400 0.690 0.220 0.221 TAP -2.800 0.005 -8.300 9.99E-16 2.320 0.021 -0.980 0.329 0.830 0.831 JHU- ROI AD Diagnosis Age [Age-mean(Age)] 2 Sex Age-by-Diagnosis t p t p t p t p t p ACR -2.210 0.027 -4.640 4.52E-06 1.200 0.232 -1.570 0.117 0.290 0.291 ALIC 1.550 0.123 0.750 0.451 -0.550 0.583 0.060 0.951 0.240 0.245 Average WM -1.550 0.121 0.870 0.384 -1.950 0.051 -0.160 0.870 0.500 0.497 BCC -0.870 0.383 -1.290 0.197 1.370 0.172 -1.070 0.285 0.510 0.514 CGC -1.920 0.055 1.540 0.124 0.620 0.537 0.340 0.737 0.970 0.969 EC -2.610 0.009 -2.160 0.031 2.010 0.045 0.080 0.937 0.350 0.346 FXST -2.490 0.013 -3.240 0.001 0.620 0.534 -0.720 0.471 0.900 0.897 GCC -2.450 0.015 -3.950 8.77E-05 -0.460 0.642 -1.360 0.176 0.930 0.927 UNC -0.810 0.419 -1.890 0.060 2.890 0.004 1.830 0.068 0.790 0.787 PCR -3.260 0.001 -4.460 1.00E-05 0.370 0.709 0.710 0.479 0.280 0.281 PLIC -1.340 0.182 -7.280 1.25E-12 0.390 0.698 1.790 0.074 0.000 0.002 PTR -2.320 0.021 -2.490 0.013 0.960 0.339 -0.640 0.524 0.260 0.263 RLIC -2.960 0.003 -3.090 0.002 0.830 0.405 0.420 0.674 0.770 0.775 SCC -1.540 0.124 -3.270 0.001 -1.320 0.189 -1.760 0.079 0.640 0.637 SCR 0.470 0.637 -3.320 0.001 0.830 0.408 -0.140 0.892 0.070 0.070 SFO 0.180 0.856 -0.690 0.492 1.840 0.067 0.890 0.375 0.050 0.053 SLF -3.710 0.000 -3.060 0.002 1.030 0.304 0.500 0.620 0.560 0.559 SS -1.670 0.096 0.220 0.826 -0.360 0.720 0.780 0.438 0.120 0.120 TAP -0.900 0.370 -5.310 1.65E-07 0.480 0.631 -0.290 0.772 0.330 0.327 JHU- ROI RD Diagnosis Age [Age-mean(Age)] 2 Sex Age-by-Diagnosis t p t p t p t p t p ACR -2.690 0.007 -6.690 5.75E-11 1.390 0.166 -1.620 0.107 0.820 0.824 ALIC -1.630 0.103 -8.160 2.66E-15 2.900 0.004 -2.390 0.017 0.580 0.580 Average WM -0.170 0.863 -8.210 1.89E-15 2.880 0.004 -1.930 0.054 0.700 0.704 BCC -0.320 0.751 -3.370 0.001 2.980 0.003 -1.670 0.095 0.230 0.234 CGC -0.930 0.350 -9.060 5.00E-16 3.030 0.003 -3.060 0.002 0.930 0.932 EC 1.010 0.314 -7.870 2.09E-14 2.530 0.012 -2.290 0.022 0.580 0.583 FXST 0.260 0.793 -5.990 3.99E-09 1.570 0.118 -1.320 0.188 0.700 0.701 GCC -2.540 0.011 -2.380 0.018 0.640 0.524 -2.400 0.017 0.890 0.885 UNC -1.150 0.250 -3.640 2.96E-04 2.110 0.035 -1.640 0.102 0.520 0.520 PCR -2.930 0.004 -8.140 3.11E-15 3.840 1.40E-04 -2.340 0.020 0.930 0.927 PLIC -2.850 0.004 -5.010 7.51E-07 0.620 0.537 -2.170 0.030 0.440 0.436 PTR -3.040 0.003 -7.100 4.16E-12 3.800 1.60E-04 -0.660 0.507 0.270 0.271 RLIC -0.780 0.433 -5.910 6.28E-09 2.690 0.007 -1.410 0.159 0.470 0.468 SCC -4.360 0.000 -7.030 6.61E-12 2.110 0.035 -0.880 0.381 0.020 0.022 SCR -1.480 0.141 -6.280 7.13E-10 3.320 0.001 -2.070 0.039 0.660 0.665 SFO -0.690 0.492 -5.220 2.59E-07 1.760 0.078 -1.450 0.147 0.980 0.984 SLF -0.270 0.789 -9.130 5.00E-16 4.240 2.64E-05 -1.700 0.091 0.610 0.605 SS -1.290 0.197 -7.610 1.34E-13 3.980 7.99E-05 -0.410 0.684 0.680 0.679 TAP -2.960 0.003 -6.920 1.40E-11 2.430 0.016 -0.920 0.358 0.880 0.885 Supplementary Table S10. Age analysis - linear model of age. Table shows the residual standard error (RSE) of the linear model of age for FA, MD, RD and AD: DTI-ROI-measure = ß 0 + ß 1 *Age + ß 2 *Age 2 centered . JHU-ROI 22q11.2DS Healthy Controls RSE FA RSE MD RSE RD RSE AD RSE FA RSE MD RSE RD RSE AD ACR 0.028 3.16E-05 3.52E-05 4.75E-05 0.027 3.20E-05 3.45E-05 4.72E-05 ALIC 0.030 3.05E-05 3.16E-05 5.39E-05 0.026 2.79E-05 2.86E-05 4.98E-05 Average WM 0.022 2.54E-05 2.84E-05 3.48E-05 0.020 2.40E-05 2.65E-05 3.34E-05 BCC 0.039 5.09E-05 5.98E-05 7.62E-05 0.032 4.91E-05 5.23E-05 6.91E-05 CGC 0.040 3.18E-05 4.52E-05 6.77E-05 0.040 3.10E-05 4.47E-05 6.08E-05 EC 0.027 2.40E-05 2.92E-05 3.59E-05 0.023 2.24E-05 2.63E-05 3.42E-05 FXST 0.036 3.42E-05 3.46E-05 6.35E-05 0.033 3.02E-05 3.34E-05 5.33E-05 GCC 0.033 4.13E-05 4.43E-05 6.79E-05 0.032 4.05E-05 4.28E-05 6.89E-05 UNC 0.047 3.97E-05 5.26E-05 6.51E-05 0.039 3.60E-05 4.28E-05 6.61E-05 PCR 0.028 3.29E-05 3.61E-05 4.32E-05 0.029 3.14E-05 3.66E-05 4.51E-05 PLIC 0.030 3.24E-05 3.78E-05 5.07E-05 0.028 2.90E-05 3.63E-05 4.20E-05 PTR 0.031 3.92E-05 3.60E-05 8.19E-05 0.029 3.88E-05 3.78E-05 7.48E-05 RLIC 0.029 3.13E-05 3.52E-05 4.96E-05 0.029 2.86E-05 3.47E-05 4.56E-05 SCC 0.028 3.87E-05 3.64E-05 6.97E-05 0.027 3.52E-05 3.38E-05 6.56E-05 SCR 0.030 2.93E-05 3.51E-05 4.51E-05 0.025 2.67E-05 3.07E-05 3.88E-05 SFO 0.037 5.11E-05 5.28E-05 6.51E-05 0.031 3.47E-05 3.43E-05 5.83E-05 SLF 0.025 2.78E-05 3.11E-05 3.99E-05 0.026 2.68E-05 3.02E-05 4.13E-05 SS 0.033 3.55E-05 3.81E-05 6.21E-05 0.029 3.33E-05 3.64E-05 5.76E-05 TAP 0.051 5.37E-05 6.89E-05 7.20E-05 0.046 4.86E-05 6.17E-05 7.04E-05 Supplementary Table S11. Age analysis - non-linear exponential regression model (Poisson fit). This table shows residual standard error (RSE) for the non- linear model for FA, MD, RD and AD: DTI-ROI-measure = ß 0 + ß 1 *Age*exp(-ß 2 *Age). JHU-ROI 22q11.2DS Healthy Controls RSE FA RSE MD RSE RD RSE AD RSE FA RSE MD RSE RD RSE AD ACR 0.028 3.16E-05 3.53E-05 4.74E-05 0.027 3.21E-05 3.47E-05 4.72E-05 ALIC 0.030 3.05E-05 3.15E-05 5.40E-05 0.026 2.79E-05 2.86E-05 4.96E-05 Average WM 0.022 2.53E-05 2.84E-05 3.48E-05 0.020 2.40E-05 2.65E-05 3.34E-05 BCC 0.039 5.07E-05 5.97E-05 7.62E-05 0.032 4.91E-05 5.23E-05 6.91E-05 CGC 0.040 3.16E-05 4.52E-05 6.78E-05 0.040 3.11E-05 4.47E-05 6.08E-05 EC 0.027 2.39E-05 2.92E-05 3.56E-05 0.023 2.24E-05 2.63E-05 3.42E-05 FXST 0.036 3.41E-05 3.46E-05 6.35E-05 0.033 3.02E-05 3.33E-05 5.33E-05 GCC 0.033 4.13E-05 4.44E-05 6.79E-05 0.032 4.06E-05 4.28E-05 6.89E-05 UNC 0.047 3.96E-05 5.26E-05 6.48E-05 0.040 3.60E-05 4.28E-05 6.61E-05 PCR 0.028 3.28E-05 3.60E-05 4.32E-05 0.028 3.14E-05 3.65E-05 4.51E-05 PLIC 0.030 3.23E-05 3.78E-05 5.07E-05 0.028 2.90E-05 3.63E-05 4.20E-05 PTR 0.031 3.92E-05 3.60E-05 8.19E-05 0.029 3.88E-05 3.76E-05 7.49E-05 RLIC 0.029 3.12E-05 3.52E-05 4.96E-05 0.029 2.87E-05 3.47E-05 4.56E-05 SCC 0.028 3.87E-05 3.64E-05 6.97E-05 0.027 3.54E-05 3.40E-05 6.57E-05 SCR 0.030 2.92E-05 3.51E-05 4.51E-05 0.025 2.67E-05 3.07E-05 3.88E-05 SFO 0.037 5.11E-05 5.28E-05 6.51E-05 0.031 3.46E-05 3.42E-05 5.83E-05 SLF 0.025 2.78E-05 3.10E-05 3.99E-05 0.026 2.69E-05 3.02E-05 4.13E-05 SS 0.033 3.55E-05 3.80E-05 6.21E-05 0.029 3.33E-05 3.62E-05 5.76E-05 TAP 0.052 5.37E-05 6.90E-05 7.20E-05 0.046 4.86E-05 6.17E-05 7.03E-05 Supplementary Table S12. Linear vs. Non-linear age models. T-test results comparing the RSE of the linear model of age vs the non-linear Poisson model of age. Linear model: DTI-ROI-measure = ß 0 + ß 1 *Age + ß 2 *Age 2 . Non-linear Poisson model: DTI-ROI-measure = ß 0 + ß 1 *Age*exp(-ß 2 *Age). Group / DTI-measure 22q11.2 Deletion Syndrome p-value (⍺=0.05) Healthy Controls p-value (⍺=0.05) FA 0.132 0.850 MD 0.001 0.083 RD 0.029 0.359 AD 0.289 0.792 Supplementary Table S13. Age of FA peak and MD, RD and AD minima for healthy control group. The peak FA and minima of MD, RD, AD of the Poisson fits of age were calculated as described in Lebel et al. 2012, by calculating the derivative of the fit: FA peak (or MD minimum ) = 1/ß 2 . The full model is: DTI-ROI-measure = ß 0 + ß 1 *Age*exp(-ß 2 *Age). P-values and standard errors (SE) of the fitted ß 2 coefficient are also shown. Colored cells indicate the significant p-values below the threshold that controls the false discovery rate (FDR). JHU- ROI Healthy Controls FA MD RD AD peak p SE min p SE min p SE min p SE ACR 21.15 9.8E-26 4.0E-03 36.32 5.7E-09 4.6E-03 28.65 5.6E-16 4.0E-03 97.71 2.5E-01 8.8E-03 ALIC 43.96 3.0E-07 4.3E-03 70.85 5.3E-02 7.3E-03 47.51 3.4E-06 4.4E-03 15.98 3.6E-08 1.1E-02 Aver. WM 25.58 2.8E-30 3.0E-03 31.76 2.2E-18 3.3E-03 28.46 2.8E-27 2.9E-03 51.71 3.5E-02 9.1E-03 BCC 23.69 8.9E-14 5.4E-03 26.21 7.0E-07 7.5E-03 24.90 1.2E-10 6.0E-03 38.27 4.0E-01 3.1E-02 CGC 31.64 3.2E-15 3.8E-03 33.19 8.0E-14 3.8E-03 33.40 2.5E-16 3.4E-03 24.36 9.5E-04 1.2E-02 EC 29.67 5.6E-14 4.2E-03 36.28 1.7E-09 4.4E-03 33.30 1.3E-14 3.7E-03 76.20 5.3E-01 2.1E-02 FXST 21.02 1.7E-12 6.4E-03 38.38 5.1E-06 5.6E-03 31.84 5.3E-09 5.2E-03 67.95 2.7E-01 1.3E-02 GCC 18.79 6.2E-12 7.4E-03 39.06 3.8E-03 8.8E-03 22.99 2.9E-08 7.6E-03 169.94 6.2E-01 1.2E-02 UNC 24.16 1.0E-10 6.1E-03 31.15 2.7E-07 6.1E-03 26.91 6.1E-10 5.8E-03 50.53 3.6E-01 2.2E-02 PCR 22.83 9.4E-29 3.5E-03 31.06 1.4E-18 3.4E-03 27.26 9.5E-26 3.1E-03 54.34 2.5E-02 8.2E-03 PLIC 26.39 7.0E-06 8.3E-03 42.24 4.2E-07 4.6E-03 31.05 8.9E-07 6.4E-03 62.90 2.1E-03 5.1E-03 PTR 21.60 4.4E-28 3.7E-03 26.31 5.2E-19 3.9E-03 24.52 4.3E-30 3.1E-03 29.55 1.6E-03 1.1E-02 RLIC 22.93 3.0E-16 5.0E-03 30.92 5.3E-12 4.5E-03 26.69 1.5E-15 4.4E-03 52.94 8.5E-02 1.1E-02 SCC 25.85 6.1E-15 4.7E-03 38.27 9.8E-08 4.8E-03 28.84 6.1E-18 3.7E-03 85.67 3.7E-01 1.3E-02 SCR 23.60 8.8E-22 4.0E-03 29.39 5.6E-16 3.9E-03 26.26 8.2E-22 3.6E-03 48.75 4.2E-02 1.0E-02 SFO 26.90 1.5E-16 4.2E-03 26.11 1.3E-14 4.7E-03 26.19 1.5E-21 3.7E-03 25.11 5.0E-02 2.0E-02 SLF 25.96 4.4E-26 3.3E-03 29.91 2.1E-20 3.3E-03 29.04 1.9E-25 3.0E-03 36.58 1.3E-02 1.1E-02 SS 23.68 1.4E-33 3.0E-03 26.86 7.8E-15 4.5E-03 26.13 5.1E-28 3.1E-03 2.97 7.9E-01 1.3E+00 TAP 29.38 3.1E-13 4.4E-03 26.95 1.4E-30 2.8E-03 28.43 6.4E-21 3.4E-03 22.72 2.3E-22 4.1E-03 Supplementary Table S14. Age of FA peak and MD, RD and AD minima for 22q11DS. The peak FA and minima of MD, RD, AD of the Poisson fits of age were calculated as described in Lebel et al. 2012, by calculating the derivative of the fit: FA peak (or MD minimum ) = 1/ß 2 . The full model is: DTI-ROI-measure = ß 0 + ß 1 *Age*exp(-ß 2 *Age). P-values and standard errors (SE) of the fitted ß 2 coefficient are also shown. Colored cells indicate the significant p-values below the threshold that controls the false discovery rate (FDR), indicating that the ß 2 coefficient is significantly different from zero. JHU- ROI 22q11.2DS FA MD RD AD peak p SE min p SE min p SE min p SE ACR 38.03 1.0E-01 1.6E-02 38.80 2.9E-05 6.1E-03 39.52 1.2E-03 7.7E-03 37.45 3.4E-04 7.4E-03 ALIC 44.69 1.2E-03 6.8E-03 56.35 3.7E-02 8.5E-03 45.94 4.7E-04 6.1E-03 8.38 2.1E-01 9.5E-02 Aver. WM 33.00 2.5E-06 6.3E-03 36.04 3.8E-08 4.9E-03 34.44 1.4E-08 5.0E-03 49.71 5.0E-02 1.0E-02 BCC 23.86 4.2E-10 6.5E-03 28.50 1.1E-07 6.4E-03 25.46 2.2E-10 6.0E-03 -104.68 7.8E-01 3.4E-02 CGC 30.55 9.4E-09 5.5E-03 27.92 1.1E-15 4.2E-03 30.94 4.0E-13 4.2E-03 43.77 5.4E-01 3.7E-02 EC 55.63 6.7E-02 9.8E-03 32.70 1.5E-09 4.9E-03 40.13 4.1E-05 6.0E-03 20.95 2.2E-17 5.3E-03 FXST 58.49 3.0E-01 1.6E-02 57.37 4.8E-02 8.8E-03 48.75 1.2E-02 8.1E-03 327.09 8.9E-01 2.2E-02 GCC 18.85 4.2E-08 9.4E-03 34.08 6.3E-03 1.1E-02 24.17 4.8E-04 1.2E-02 73.88 3.6E-01 1.5E-02 UNC -147.09 8.8E-01 4.6E-02 25.33 7.7E-08 7.1E-03 48.29 3.3E-01 2.1E-02 19.34 3.6E-19 5.4E-03 PCR 29.84 5.4E-08 6.0E-03 33.18 1.5E-09 4.8E-03 32.38 1.4E-09 4.9E-03 35.26 1.8E-04 7.5E-03 PLIC 30.83 3.2E-03 1.1E-02 31.62 2.2E-04 8.4E-03 33.37 4.8E-03 1.1E-02 40.11 1.2E-01 1.6E-02 PTR 23.48 2.9E-07 8.1E-03 43.54 8.7E-03 8.7E-03 29.01 1.2E-07 6.3E-03 1422.14 9.7E-01 1.9E-02 RLIC 28.74 3.6E-07 6.7E-03 31.22 7.4E-08 5.8E-03 29.02 2.0E-10 5.2E-03 54.64 4.1E-01 2.2E-02 SCC 17.23 3.4E-12 8.0E-03 42.11 2.5E-02 1.1E-02 29.06 6.4E-04 1.0E-02 131.68 6.4E-01 1.6E-02 SCR 37.48 2.0E-02 1.1E-02 34.48 6.1E-08 5.2E-03 34.79 2.1E-05 6.6E-03 36.28 8.3E-04 8.2E-03 SFO 75.71 4.8E-01 1.9E-02 83.98 3.7E-01 1.3E-02 85.05 4.0E-01 1.4E-02 277.85 8.5E-01 1.9E-02 SLF 34.69 3.6E-07 5.5E-03 33.57 1.1E-10 4.4E-03 34.23 7.8E-11 4.3E-03 33.05 3.3E-03 1.0E-02 SS 30.26 7.0E-07 6.5E-03 36.31 2.6E-05 6.4E-03 31.71 3.3E-10 4.8E-03 -36.34 5.1E-01 4.2E-02 TAP 35.71 1.9E-03 8.9E-03 41.02 2.9E-04 6.6E-03 40.22 9.0E-04 7.4E-03 44.50 1.2E-01 1.4E-02 Supplementary Table S15. Group Differences in Peak FA and Minimum Diffusivity Values - Non-linear Poisson fits. Top Panel: Differences in mean peak FA/minimum MD, RD, AD across all ROIs between 22q11DS participants and Healthy Controls. Bottom Panel: Differences in mean percent FA decrease (after peak age) and mean percent change of increase (after minimum age) of MD, AD and RD (alpha=0.05). *p<.05 Peak FA / Minimum MD, RD, AD differences DTI Measure FA MD RD AD 22q11DS Mean Peak/Minimum 30.48 34.40 35.07 38.25 Healthy Controls Mean Peak/ Minimum 24.30 32.24 28.04 46.90 Differences in Mean t = -3.032 p-value = 0.007 t = -1.230 p-value = 0.227 t = -3.784 p-value = 0.001 t = 1.079 p-value = 0.289 Differences in percent changes of FA decrease; MD, RD, AD increase DTI Measure FA MD RD AD Mean percent change 22q11DS -0.89 0.56 5.25 0.092 Mean percent change Healthy Controls -3.94 1.66 1.37 0.039 Mean percent change differences t = -6.05 CI = (-3.93, -2.16) t = 2.96 CI = (0.39, 1.81) t = 4.86 CI = (2.44, 5.31) t = -0.59 CI = (-0.21, 0.10) Supplementary Table S16. 22q11DS+Psychosis vs. 22q11DS-No Psychosis. Results of the local non-parametric ANCOVA comparing 22q11DS subjects with psychotic disorder (N=35) vs. those with no history of psychotic symptoms (N=191). Twenty-five design points (or age bands) were selected. The critical p-value is 0.0056. Tables show results for DTI indices that significantly differed between 22q- Psychosis vs. 22q-No Psychosis for at least one design point: A) Axial Diffusivity (AD) in the Anterior Limb of the Internal Capsule (ALIC); B) AD in the Cingulum of the Cingulate Gyrus (CGC); C) AD in the Posterior Thalamic Radiation (PTR); D) AD in the Superior Longitudinal Fasciculus (SLF); E) AD in the Sagittal Stratum (SS); F) Mean Diffusivity (MD) in the Genu of Corpus Callosum (GCC); G) MD in the Posterior Limb of the Internal Capsule (PLIC); H) Radial Diffusivity (RD) in the Genu of the Corpus Callosum (GCC). Color filled cells indicate the design points that are significantly different between groups. N1 and N2 refer to the sample sizes at each design point; DIF = estimated difference between the means; TEST = resulting test statistic; SE = standard error; ci.low = lower bound of the confidence interval; ci.hi = upper bound of the confidence interval. A) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -4.34E-05 1.90 2.28E-05 -1.98E-05 1.07E-04 0.0876 -0.34 13.47 101 12 -4.29E-05 1.90 2.26E-05 -1.97E-05 1.05E-04 0.0892 -0.36 14.11 111 12 -4.14E-05 1.85 2.24E-05 -2.07E-05 1.04E-04 0.0979 -0.32 14.75 117 15 -6.59E-05 2.98 2.21E-05 4.65E-06 1.27E-04 0.0134 -0.50 15.39 121 19 -4.36E-05 2.02 2.16E-05 -1.63E-05 1.04E-04 0.0618 -0.33 16.04 131 19 -4.43E-05 2.05 2.16E-05 -1.55E-05 1.04E-04 0.0579 -0.35 16.68 132 18 -5.13E-05 2.39 2.15E-05 -8.24E-06 1.11E-04 0.0318 -0.40 17.32 134 18 -5.03E-05 2.35 2.15E-05 -9.08E-06 1.10E-04 0.0344 -0.44 17.96 132 20 -5.50E-05 2.68 2.05E-05 -1.73E-06 1.12E-04 0.0177 -0.45 18.60 131 22 -5.94E-05 3.18 1.87E-05 7.64E-06 1.11E-04 0.0052 -0.47 19.24 126 23 -5.59E-05 3.14 1.78E-05 6.59E-06 1.05E-04 0.0052 -0.49 19.88 123 24 -5.61E-05 3.33 1.69E-05 9.46E-06 1.03E-04 0.0030 -0.48 20.52 112 25 -5.79E-05 3.44 1.68E-05 1.12E-05 1.04E-04 0.0024 -0.54 Age (Design Points) B) 21.16 106 28 -4.94E-05 2.91 1.70E-05 2.36E-06 9.65E-05 0.0073 -0.47 21.81 106 27 -4.82E-05 2.71 1.78E-05 -1.07E-06 9.74E-05 0.0123 -0.41 22.45 96 27 -4.69E-05 2.61 1.80E-05 -2.88E-06 9.67E-05 0.0151 -0.43 23.09 82 25 -5.30E-05 2.65 2.00E-05 -2.41E-06 1.08E-04 0.0151 -0.48 23.73 74 25 -5.62E-05 2.77 2.03E-05 -2.10E-09 1.12E-04 0.0113 -0.50 24.37 62 24 -5.73E-05 2.85 2.01E-05 1.68E-06 1.13E-04 0.0084 -0.46 25.01 56 24 -5.64E-05 2.78 2.02E-05 3.13E-07 1.12E-04 0.0098 -0.47 25.65 49 22 -5.54E-05 2.56 2.16E-05 -4.43E-06 1.15E-04 0.0178 -0.48 26.29 38 21 -6.29E-05 2.84 2.21E-05 1.59E-06 1.24E-04 0.0102 -0.64 26.93 35 21 -6.90E-05 3.25 2.12E-05 1.03E-05 1.28E-04 0.0035 -0.68 27.58 26 20 -6.30E-05 2.67 2.36E-05 -2.23E-06 1.28E-04 0.0137 -0.66 28.22 21 20 -6.07E-05 2.44 2.49E-05 -8.26E-06 1.30E-04 0.0232 -0.62 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -3.34E-05 1.78 1.88E-05 -1.87E-05 8.54E-05 0.1012 -0.31 13.47 101 12 -3.43E-05 1.87 1.83E-05 -1.65E-05 8.50E-05 0.0886 -0.30 14.11 111 12 -3.44E-05 1.89 1.82E-05 -1.61E-05 8.48E-05 0.0869 -0.33 14.75 117 15 -4.71E-05 2.86 1.65E-05 1.45E-06 9.27E-05 0.0137 -0.48 15.39 121 19 -3.53E-05 2.04 1.73E-05 -1.26E-05 8.32E-05 0.0559 -0.34 16.04 131 19 -3.81E-05 2.20 1.73E-05 -9.91E-06 8.60E-05 0.0408 -0.38 16.68 132 18 -4.59E-05 2.69 1.71E-05 -1.31E-06 9.31E-05 0.0147 -0.41 17.32 134 18 -4.45E-05 2.59 1.72E-05 -3.12E-06 9.22E-05 0.0180 -0.45 17.96 132 20 -4.90E-05 2.71 1.81E-05 -9.66E-07 9.90E-05 0.0138 -0.45 18.60 131 22 -4.95E-05 2.91 1.70E-05 2.42E-06 9.65E-05 0.0072 -0.40 19.24 126 23 -4.76E-05 2.93 1.63E-05 2.57E-06 9.26E-05 0.0064 -0.41 Age (Design Points) C) 19.88 123 24 -4.80E-05 3.09 1.55E-05 4.99E-06 9.10E-05 0.0039 -0.45 20.52 112 25 -4.72E-05 2.95 1.60E-05 2.93E-06 9.16E-05 0.0057 -0.42 21.16 106 28 -4.54E-05 3.02 1.50E-05 3.78E-06 8.71E-05 0.0042 -0.38 21.81 106 27 -4.53E-05 3.00 1.51E-05 3.43E-06 8.72E-05 0.0046 -0.37 22.45 96 27 -4.37E-05 2.76 1.58E-05 -1.06E-07 8.75E-05 0.0086 -0.39 23.09 82 25 -4.76E-05 2.78 1.71E-05 2.06E-07 9.50E-05 0.0090 -0.43 23.73 74 25 -5.19E-05 2.94 1.77E-05 3.06E-06 1.01E-04 0.0058 -0.42 24.37 62 24 -6.26E-05 3.50 1.79E-05 1.30E-05 1.12E-04 0.0012 -0.52 25.01 56 24 -6.69E-05 3.61 1.85E-05 1.57E-05 1.18E-04 0.0009 -0.54 25.65 49 22 -6.54E-05 3.20 2.04E-05 8.86E-06 1.22E-04 0.0031 -0.47 26.29 38 21 -6.60E-05 3.19 2.07E-05 8.75E-06 1.23E-04 0.0032 -0.54 26.93 35 21 -6.01E-05 2.69 2.23E-05 -1.64E-06 1.22E-04 0.0114 -0.46 27.58 26 20 -6.80E-05 3.15 2.16E-05 8.26E-06 1.28E-04 0.0042 -0.57 28.22 21 20 -6.21E-05 2.88 2.16E-05 2.38E-06 1.22E-04 0.0085 -0.51 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -4.23E-05 1.85 2.29E-05 -2.10E-05 1.06E-04 0.0949 -0.33 13.47 101 12 -4.52E-05 1.96 2.31E-05 -1.87E-05 1.09E-04 0.0785 -0.35 14.11 111 12 -4.19E-05 1.86 2.25E-05 -2.04E-05 1.04E-04 0.0951 -0.34 14.75 117 15 -5.51E-05 2.92 1.89E-05 2.79E-06 1.07E-04 0.0136 -0.48 15.39 121 19 -4.83E-05 2.85 1.70E-05 1.38E-06 9.53E-05 0.0104 -0.47 16.04 131 19 -5.03E-05 2.97 1.70E-05 3.38E-06 9.73E-05 0.0081 -0.48 16.68 132 18 -5.32E-05 3.21 1.66E-05 7.25E-06 9.91E-05 0.0051 -0.56 17.32 134 18 -5.07E-05 3.07 1.65E-05 4.94E-06 9.64E-05 0.0070 -0.53 17.96 132 20 -4.59E-05 3.04 1.51E-05 4.16E-06 8.76E-05 0.0064 -0.51 Age (Design Points) D) 18.60 131 22 -4.72E-05 3.39 1.39E-05 8.72E-06 8.58E-05 0.0021 -0.48 19.24 126 23 -4.17E-05 3.13 1.33E-05 4.86E-06 7.86E-05 0.0037 -0.46 19.88 123 24 -4.01E-05 2.87 1.40E-05 1.39E-06 7.88E-05 0.0073 -0.45 20.52 112 25 -4.32E-05 3.26 1.32E-05 6.56E-06 7.99E-05 0.0024 -0.46 21.16 106 28 -3.85E-05 2.84 1.36E-05 9.07E-07 7.62E-05 0.0069 -0.38 21.81 106 27 -3.53E-05 2.53 1.39E-05 -3.25E-06 7.38E-05 0.0154 -0.34 22.45 96 27 -3.06E-05 2.14 1.43E-05 -9.07E-06 7.04E-05 0.0392 -0.30 23.09 82 25 -3.05E-05 2.03 1.50E-05 -1.12E-05 7.22E-05 0.0514 0.32 23.73 74 25 -2.47E-05 1.63 1.51E-05 -1.71E-05 6.65E-05 0.1124 -0.28 24.37 62 24 -2.73E-05 1.72 1.58E-05 -1.66E-05 7.11E-05 0.0925 -0.31 25.01 56 24 -2.74E-05 1.67 1.64E-05 -1.80E-05 7.28E-05 0.1027 -0.30 25.65 49 22 -2.63E-05 1.41 1.86E-05 -2.52E-05 7.77E-05 0.1661 -0.26 26.29 38 21 -1.25E-05 0.69 1.83E-05 -3.80E-05 6.31E-05 0.4970 -0.14 26.93 35 21 -1.53E-05 0.78 1.95E-05 -3.87E-05 6.93E-05 0.4388 -0.15 27.58 26 20 -9.58E-06 0.43 2.21E-05 -5.16E-05 7.08E-05 0.6685 -0.12 28.22 21 20 -9.34E-06 0.41 2.28E-05 -5.37E-05 7.23E-05 0.6855 -0.10 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -2.63E-05 1.49 1.77E-05 -2.27E-05 7.54E-05 0.1673 -0.32 13.47 101 12 -2.85E-05 1.63 1.75E-05 -2.00E-05 7.69E-05 0.1353 -0.33 14.11 111 12 -2.90E-05 1.68 1.73E-05 -1.87E-05 7.68E-05 0.1257 -0.37 14.75 117 15 -3.45E-05 2.46 1.40E-05 -4.33E-06 7.34E-05 0.0290 -0.44 15.39 121 19 -1.94E-05 1.45 1.34E-05 -1.76E-05 5.65E-05 0.1635 -0.26 16.04 131 19 -2.08E-05 1.55 1.34E-05 -1.62E-05 5.78E-05 0.1378 -0.26 16.68 132 18 -2.76E-05 2.09 1.32E-05 -8.93E-06 6.41E-05 0.0524 -0.38 Age (Design Points) E) 17.32 134 18 -2.61E-05 1.99 1.31E-05 -1.03E-05 6.25E-05 0.0640 -0.36 17.96 132 20 -3.03E-05 2.56 1.18E-05 -2.49E-06 6.31E-05 0.0197 -0.44 18.60 131 22 -2.87E-05 2.67 1.08E-05 -1.11E-06 5.84E-05 0.0132 -0.40 19.24 126 23 -2.70E-05 2.58 1.05E-05 -1.94E-06 5.60E-05 0.0148 -0.36 19.88 123 24 -2.69E-05 2.71 9.95E-06 -6.23E-07 5.45E-05 0.0105 -0.43 20.52 112 25 -3.00E-05 3.12 9.60E-06 3.42E-06 5.66E-05 0.0032 -0.43 21.16 106 28 -2.39E-05 2.15 1.11E-05 -6.82E-06 5.47E-05 0.0379 -0.33 21.81 106 27 -2.23E-05 1.96 1.14E-05 -9.14E-06 5.38E-05 0.0574 -0.29 22.45 96 27 -2.23E-05 1.99 1.12E-05 -8.80E-06 5.34E-05 0.0559 -0.28 23.09 82 25 -2.41E-05 2.13 1.13E-05 -7.19E-06 5.54E-05 0.0419 -0.31 23.73 74 25 -2.16E-05 1.85 1.17E-05 -1.08E-05 5.40E-05 0.0743 -0.29 24.37 62 24 -2.34E-05 1.86 1.26E-05 -1.15E-05 5.82E-05 0.0726 -0.33 25.01 56 24 -2.09E-05 1.64 1.27E-05 -1.44E-05 5.62E-05 0.1102 -0.28 25.65 49 22 -2.27E-05 1.81 1.26E-05 -1.21E-05 5.75E-05 0.0805 -0.29 26.29 38 21 -2.02E-05 1.51 1.33E-05 -1.67E-05 5.71E-05 0.1413 -0.29 26.93 35 21 -1.77E-05 1.22 1.45E-05 -2.24E-05 5.78E-05 0.2326 -0.24 27.58 26 20 -1.04E-05 0.65 1.61E-05 -3.42E-05 5.50E-05 0.5231 -0.16 28.22 21 20 -6.67E-06 0.38 1.74E-05 -4.14E-05 5.48E-05 0.7045 -0.07 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -1.86E-05 1.17 1.59E-05 -2.55E-05 6.27E-05 0.2668 -0.24 13.47 101 12 -2.21E-05 1.42 1.56E-05 -2.11E-05 6.54E-05 0.1855 -0.27 14.11 111 12 -2.19E-05 1.43 1.53E-05 -2.04E-05 6.42E-05 0.1831 -0.28 14.75 117 15 -3.54E-05 2.26 1.57E-05 -7.97E-06 7.87E-05 0.0455 -0.37 15.39 121 19 -3.66E-05 2.47 1.48E-05 -4.39E-06 7.76E-05 0.0238 -0.40 Age (Design Points) F) 16.04 131 19 -3.90E-05 2.64 1.48E-05 -1.83E-06 7.99E-05 0.0168 -0.39 16.68 132 18 -4.47E-05 3.12 1.43E-05 5.09E-06 8.43E-05 0.0066 -0.49 17.32 134 18 -4.56E-05 3.15 1.45E-05 5.51E-06 8.58E-05 0.0059 -0.51 17.96 132 20 -4.94E-05 3.30 1.50E-05 8.00E-06 9.08E-05 0.0045 -0.57 18.60 131 22 -4.81E-05 3.31 1.46E-05 7.81E-06 8.84E-05 0.0033 -0.44 19.24 126 23 -4.55E-05 3.28 1.39E-05 7.05E-06 8.39E-05 0.0031 -0.45 19.88 123 24 -4.60E-05 3.48 1.32E-05 9.42E-06 8.25E-05 0.0017 -0.46 20.52 112 25 -4.76E-05 3.49 1.36E-05 9.88E-06 8.53E-05 0.0018 -0.51 21.16 106 28 -4.34E-05 3.27 1.33E-05 6.61E-06 8.03E-05 0.0027 -0.47 21.81 106 27 -4.15E-05 2.98 1.39E-05 2.99E-06 8.00E-05 0.0057 -0.40 22.45 96 27 -4.29E-05 3.03 1.42E-05 3.66E-06 8.21E-05 0.0050 -0.44 23.09 82 25 -4.82E-05 3.35 1.44E-05 8.33E-06 8.81E-05 0.0023 -0.49 23.73 74 25 -4.82E-05 3.24 1.49E-05 7.04E-06 8.93E-05 0.0028 -0.46 24.37 62 24 -5.56E-05 3.50 1.59E-05 1.16E-05 9.95E-05 0.0013 -0.53 25.01 56 24 -5.19E-05 3.17 1.64E-05 6.59E-06 9.73E-05 0.0031 -0.48 25.65 49 22 -5.49E-05 3.41 1.61E-05 1.04E-05 9.93E-05 0.0016 -0.51 26.29 38 21 -5.01E-05 2.90 1.73E-05 2.28E-06 9.80E-05 0.0068 -0.47 26.93 35 21 -4.69E-05 2.45 1.91E-05 -6.08E-06 1.00E-04 0.0210 -0.42 27.58 26 20 -2.96E-05 1.49 1.99E-05 -2.56E-05 8.48E-05 0.1507 -0.32 28.22 21 20 -2.81E-05 1.36 2.07E-05 -2.93E-05 8.55E-05 0.1881 -0.29 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -4.08E-05 2.90 1.41E-05 1.79E-06 7.99E-05 0.0099 -0.45 13.47 101 12 -4.05E-05 2.94 1.38E-05 2.34E-06 7.86E-05 0.0096 -0.44 14.11 111 12 -3.99E-05 2.92 1.37E-05 2.05E-06 7.77E-05 0.0102 -0.44 Age (Design Points) G) 14.75 117 15 -4.32E-05 2.80 1.55E-05 4.44E-07 8.60E-05 0.0145 -0.44 15.39 121 19 -3.11E-05 2.13 1.46E-05 -9.40E-06 7.17E-05 0.0459 -0.34 16.04 131 19 -3.10E-05 2.16 1.44E-05 -8.74E-06 7.07E-05 0.0438 -0.32 16.68 132 18 -3.37E-05 2.65 1.27E-05 -1.56E-06 6.89E-05 0.0154 -0.42 17.32 134 18 -3.26E-05 2.53 1.29E-05 -3.05E-06 6.83E-05 0.0193 -0.44 17.96 132 20 -3.45E-05 2.62 1.32E-05 -1.94E-06 7.09E-05 0.0163 -0.39 18.60 131 22 -4.13E-05 2.72 1.52E-05 -8.06E-07 8.35E-05 0.0132 -0.40 19.24 126 23 -3.91E-05 2.68 1.46E-05 -1.24E-06 7.94E-05 0.0133 -0.43 19.88 123 24 -4.23E-05 3.01 1.41E-05 3.35E-06 8.13E-05 0.0060 -0.46 20.52 112 25 -4.38E-05 3.13 1.40E-05 5.02E-06 8.26E-05 0.0043 -0.49 21.16 106 28 -4.26E-05 3.23 1.32E-05 6.12E-06 7.90E-05 0.0027 -0.47 21.81 106 27 -4.40E-05 3.20 1.37E-05 5.99E-06 8.20E-05 0.0030 -0.43 22.45 96 27 -4.34E-05 3.25 1.33E-05 6.47E-06 8.04E-05 0.0027 -0.47 23.09 82 25 -4.49E-05 3.24 1.39E-05 6.56E-06 8.33E-05 0.0030 -0.49 23.73 74 25 -4.61E-05 3.13 1.47E-05 5.29E-06 8.68E-05 0.0036 -0.48 24.37 62 24 -4.78E-05 3.24 1.48E-05 6.96E-06 8.87E-05 0.0027 -0.51 25.01 56 24 -5.07E-05 3.38 1.50E-05 9.15E-06 9.23E-05 0.0018 -0.49 25.65 49 22 -5.08E-05 3.21 1.58E-05 7.00E-06 9.46E-05 0.0030 -0.49 26.29 38 21 -5.28E-05 2.99 1.77E-05 3.90E-06 1.02E-04 0.0054 -0.50 26.93 35 21 -5.46E-05 3.01 1.82E-05 4.32E-06 1.05E-04 0.0055 -0.52 27.58 26 20 -6.35E-05 3.21 1.98E-05 8.78E-06 1.18E-04 0.0037 -0.60 28.22 21 20 -4.85E-05 2.49 1.95E-05 -5.35E-06 1.02E-04 0.0205 -0.52 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -3.49E-05 2.23 1.56E-05 -8.35E-06 7.81E-05 0.0486 -0.43 Age (Design Points) H) 13.47 101 12 -3.10E-05 2.02 1.54E-05 -1.15E-05 7.35E-05 0.0721 -0.38 14.11 111 12 -3.00E-05 1.99 1.51E-05 -1.18E-05 7.18E-05 0.0779 -0.42 14.75 117 15 -3.20E-05 2.22 1.44E-05 -7.85E-06 7.18E-05 0.0490 -0.43 15.39 121 19 -2.10E-05 1.71 1.22E-05 -1.29E-05 5.48E-05 0.1045 -0.30 16.04 131 19 -2.26E-05 1.85 1.23E-05 -1.13E-05 5.66E-05 0.0819 -0.32 16.68 132 18 -2.74E-05 2.27 1.20E-05 -5.97E-06 6.07E-05 0.0370 -0.37 17.32 134 18 -2.85E-05 2.38 1.20E-05 -4.67E-06 6.16E-05 0.0304 -0.41 17.96 132 20 -3.26E-05 2.72 1.20E-05 -6.30E-07 6.58E-05 0.0153 -0.45 18.60 131 22 -3.18E-05 2.81 1.13E-05 4.42E-07 6.31E-05 0.0106 -0.41 19.24 126 23 -2.69E-05 2.50 1.08E-05 -2.94E-06 5.67E-05 0.0197 -0.39 19.88 123 24 -2.75E-05 2.71 1.01E-05 -5.93E-07 5.56E-05 0.0119 -0.44 20.52 112 25 -3.35E-05 3.24 1.03E-05 4.91E-06 6.20E-05 0.0033 -0.49 21.16 106 28 -2.83E-05 2.88 9.85E-06 1.05E-06 5.56E-05 0.0070 -0.41 21.81 106 27 -2.67E-05 2.71 9.84E-06 -6.00E-07 5.39E-05 0.0110 -0.39 22.45 96 27 -2.31E-05 2.35 9.83E-06 -4.10E-06 5.03E-05 0.0245 -0.32 23.09 82 25 -2.62E-05 2.46 1.06E-05 -3.28E-06 5.56E-05 0.0201 -0.40 23.73 74 25 -2.87E-05 2.67 1.08E-05 -1.08E-06 5.85E-05 0.0122 -0.43 24.37 62 24 -3.48E-05 3.04 1.15E-05 3.09E-06 6.65E-05 0.0044 -0.47 25.01 56 24 -3.26E-05 2.83 1.15E-05 6.86E-07 6.46E-05 0.0076 -0.48 25.65 49 22 -3.23E-05 2.74 1.18E-05 -2.91E-07 6.49E-05 0.0097 -0.46 26.29 38 21 -2.73E-05 2.27 1.20E-05 -5.96E-06 6.06E-05 0.0313 -0.41 26.93 35 21 -2.74E-05 2.15 1.27E-05 -7.88E-06 6.26E-05 0.0402 -0.38 27.58 26 20 -1.64E-05 1.03 1.58E-05 -2.75E-05 6.02E-05 0.3116 -0.22 28.22 21 20 -1.45E-05 0.93 1.56E-05 -2.86E-05 5.76E-05 0.3617 -0.21 N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size Age (Design Points) Age (Design Points) N1 N2 DIF TEST SE ci.low ci.hi p-value Effect Size 12.83 87 12 -3.25E-05 2.28 1.43E-05 -7.04E-06 7.21E-05 0.0290 -0.37 13.47 101 12 -3.67E-05 2.70 1.36E-05 -9.15E-07 7.44E-05 0.0109 -0.44 14.11 111 12 -3.57E-05 2.71 1.32E-05 -7.88E-07 7.21E-05 0.0111 -0.42 14.75 117 15 -3.77E-05 2.65 1.42E-05 -1.64E-06 7.70E-05 0.0138 -0.37 15.39 121 19 -2.20E-05 1.62 1.36E-05 -1.56E-05 5.95E-05 0.1152 -0.31 16.04 131 19 -2.41E-05 1.79 1.34E-05 -1.31E-05 6.13E-05 0.0828 -0.25 16.68 132 18 -2.46E-05 1.92 1.28E-05 -1.08E-05 6.01E-05 0.0645 -0.34 17.32 134 18 -2.65E-05 2.04 1.30E-05 -9.41E-06 6.23E-05 0.0502 -0.40 17.96 132 20 -2.84E-05 2.30 1.23E-05 -5.75E-06 6.25E-05 0.0275 -0.36 18.60 131 22 -3.45E-05 2.48 1.39E-05 -4.08E-06 7.30E-05 0.0189 -0.31 19.24 126 23 -3.53E-05 2.65 1.33E-05 -1.56E-06 7.22E-05 0.0119 -0.39 19.88 123 24 -3.78E-05 2.88 1.31E-05 1.43E-06 7.41E-05 0.0063 -0.45 20.52 112 25 -3.99E-05 3.10 1.29E-05 4.25E-06 7.56E-05 0.0033 -0.48 21.16 106 28 -4.01E-05 3.08 1.30E-05 4.06E-06 7.61E-05 0.0033 -0.50 21.81 106 27 -4.13E-05 3.07 1.34E-05 4.07E-06 7.85E-05 0.0035 -0.43 22.45 96 27 -3.96E-05 3.01 1.31E-05 3.21E-06 7.60E-05 0.0043 -0.43 23.09 82 25 -3.69E-05 2.60 1.42E-05 -2.35E-06 7.62E-05 0.0136 -0.40 23.73 74 25 -3.55E-05 2.35 1.51E-05 -6.24E-06 7.72E-05 0.0236 -0.41 24.37 62 24 -3.76E-05 2.47 1.52E-05 -4.46E-06 7.96E-05 0.0175 -0.42 25.01 56 24 -4.09E-05 2.63 1.56E-05 -2.20E-06 8.39E-05 0.0120 -0.44 25.65 49 22 -4.17E-05 2.31 1.81E-05 -8.38E-06 9.18E-05 0.0265 -0.43 26.29 38 21 -3.72E-05 2.22 1.68E-05 -9.24E-06 8.36E-05 0.0336 -0.39 26.93 35 21 -4.14E-05 2.06 2.01E-05 -1.42E-05 9.69E-05 0.0482 -0.39 27.58 26 20 -4.22E-05 1.79 2.36E-05 -2.31E-05 1.08E-04 0.0853 -0.44 28.22 21 20 -3.58E-05 1.52 2.35E-05 -2.93E-05 1.01E-04 0.1416 -0.36 Supplementary Table S17. Modulating effects of deletion type. Cohen’s d, Student’s t and p-values for the effects of deletion type on each ROI-by-DTI measure. The model was tested on all samples (Utrecht included) using the harmonized data as a mega-analysis. There was no p-value controlling the FDR at a q-value of 0.05. Orange-shadowed cells correspond to uncorrected significant results (p < 0.05). The model included age, [age-mean(age)] 2 and sex as covariates. JHU-ROI FA Dele.on_type d t p ACR -0.139 -1.010 0.312 ALIC 0.054 0.390 0.695 Average WM 0.145 1.056 0.292 BCC 0.079 0.580 0.566 CGC 0.180 1.310 0.192 EC -0.114 -0.830 0.407 FXST 0.081 0.590 0.556 GCC 0.023 0.170 0.868 UNC -0.035 -0.260 0.797 PCR 0.139 1.010 0.313 PLIC 0.203 1.480 0.140 PTR -0.065 -0.470 0.638 RLIC 0.186 1.350 0.178 SCC 0.064 0.460 0.643 SCR 0.157 1.150 0.253 SFO -0.090 -0.650 0.514 SLF 0.040 0.290 0.770 SS -0.007 -0.050 0.958 TAP 0.278 2.020 0.044 JHU-ROI MD Dele.on_type d t p ACR -0.154 -1.120 0.264 ALIC -0.006 -0.040 0.968 Average WM 0.130 0.947 0.345 BCC 0.032 0.230 0.817 CGC -0.037 -0.270 0.790 EC -0.069 -0.500 0.615 FXST -0.078 -0.570 0.571 GCC -0.126 -0.910 0.362 UNC 0.042 0.310 0.760 PCR -0.232 -1.690 0.093 PLIC -0.043 -0.320 0.753 PTR -0.083 -0.600 0.548 RLIC -0.111 -0.810 0.420 SCC 0.049 0.360 0.720 SCR -0.095 -0.690 0.492 SFO -0.042 -0.300 0.761 SLF 0.020 0.150 0.883 SS -0.029 -0.210 0.831 TAP -0.190 -1.380 0.169 JHU-ROI AD Dele.on_type d t p ACR -0.336 -2.440 0.015 ALIC -0.119 -0.860 0.388 Average WM 0.136 0.988 0.324 BCC -0.053 -0.380 0.702 CGC 0.027 0.190 0.846 EC -0.300 -2.180 0.030 FXST -0.088 -0.640 0.523 GCC -0.265 -1.930 0.055 UNC -0.114 -0.830 0.407 PCR -0.236 -1.720 0.087 PLIC -0.070 -0.510 0.610 PTR -0.137 -1.000 0.318 RLIC -0.073 -0.530 0.596 SCC -0.048 -0.350 0.726 SCR -0.003 -0.020 0.984 SFO -0.154 -1.120 0.263 SLF 0.012 0.090 0.930 SS -0.079 -0.570 0.568 TAP 0.049 0.350 0.724 JHU-ROI RD Dele.on_type d t p ACR -0.022 -0.160 0.874 ALIC 0.021 0.160 0.877 Average WM 0.012 0.087 0.931 BCC -0.032 -0.230 0.817 CGC -0.118 -0.860 0.391 EC 0.080 0.580 0.561 FXST -0.079 -0.580 0.564 GCC -0.012 -0.080 0.933 UNC 0.070 0.510 0.611 PCR -0.197 -1.430 0.154 PLIC -0.150 -1.090 0.275 PTR -0.056 -0.410 0.682 RLIC -0.183 -1.330 0.184 SCC -0.008 -0.060 0.952 SCR -0.163 -1.190 0.236 SFO 0.041 0.300 0.764 SLF -0.033 -0.240 0.811 SS -0.018 -0.130 0.898 TAP -0.267 -1.950 0.053 Supplementary Table S18. Correlation between DTI measures and IQ for each ROI. Partial correlations, Student’s t and p-values for the effects of IQ on each ROI-by-DTI measure for 22q11DS and healthy controls. Regressions included age, [age-mean(age)] 2 and sex as covariates. Blue-shadowed cells indicate a statistically significant result that passed the False Discovery Rate threshold at a q-value of 0.05. Orange-shadowed cells correspond to uncorrected significant results (p < 0.05). JHU-ROI FA IQ - 22q11.2DS IQ - Healthy Controls r t p (FDR p < 0.000014) r t p ACR -0.059 -0.980 0.327 0.071 0.720 0.475 ALIC 0.053 0.890 0.376 -0.060 -0.610 0.544 Average WM 0.036 0.600 0.549 -0.110 -1.120 0.265 BCC -0.005 -0.090 0.926 -0.010 -0.100 0.920 CGC -0.031 -0.520 0.602 -0.047 -0.480 0.632 EC 0.087 1.460 0.145 0.091 0.920 0.360 FXST 0.080 1.330 0.186 0.020 0.200 0.842 GCC -0.051 -0.850 0.396 -0.037 -0.370 0.710 UNC 0.007 0.120 0.908 0.124 1.260 0.208 PCR -0.057 -0.950 0.341 -0.122 -1.240 0.216 PLIC -0.025 -0.410 0.684 -0.118 -1.200 0.232 PTR 0.142 2.390 0.017 -0.095 -0.960 0.338 RLIC 0.023 0.380 0.702 -0.060 -0.610 0.545 SCC -0.017 -0.280 0.782 -0.100 -1.020 0.311 SCR -0.063 -1.050 0.293 -0.092 -0.930 0.352 SFO 0.012 0.200 0.838 0.009 0.090 0.928 SLF 0.089 1.480 0.141 0.043 0.430 0.671 SS 0.061 1.020 0.309 -0.004 -0.040 0.965 TAP -0.077 -1.290 0.198 -0.256 -2.680 0.008 JHU-ROI MD IQ - 22q11.2DS IQ - Healthy Controls r t p r t p ACR 0.067 1.120 0.263 0.170 1.740 0.083 ALIC -0.018 -0.300 0.763 0.221 2.290 0.023 Average WM 0.109 1.830 0.068 -0.017 -0.170 0.864 BCC -0.049 -0.820 0.415 0.100 1.020 0.309 CGC 0.040 0.660 0.508 0.070 0.710 0.480 EC 0.005 0.090 0.927 0.126 1.280 0.203 FXST 0.035 0.580 0.561 0.058 0.590 0.559 GCC 0.116 1.940 0.049 0.172 1.760 0.080 UNC -0.068 -1.130 0.258 0.006 0.060 0.953 PCR -0.001 -0.010 0.988 0.250 2.610 0.010 PLIC -0.034 -0.570 0.572 0.062 0.630 0.527 PTR 0.102 1.710 0.088 0.232 2.410 0.017 RLIC -0.054 -0.900 0.370 0.189 1.940 0.054 SCC -0.011 -0.180 0.854 0.199 2.050 0.042 SCR -0.033 -0.550 0.583 0.173 1.770 0.079 SFO 0.034 0.560 0.573 0.268 2.810 0.006 SLF -0.061 -1.010 0.316 0.209 2.160 0.032 SS 0.052 0.860 0.390 0.204 2.100 0.037 TAP -0.019 -0.320 0.751 0.243 2.530 0.012 JHU-ROI AD IQ - 22q11.2DS IQ - Healthy Controls r t p r t p ACR 0.064 1.060 0.292 0.220 2.280 0.024 ALIC 0.075 1.250 0.211 0.203 2.090 0.039 Average WM 0.125 2.100 0.037 -0.100 -1.010 0.316 BCC -0.003 -0.050 0.963 0.089 0.900 0.372 CGC 0.026 0.440 0.657 0.037 0.370 0.714 EC 0.090 1.500 0.135 0.270 2.830 0.005 FXST 0.112 1.870 0.062 0.057 0.580 0.560 GCC 0.142 2.380 0.018 0.175 1.790 0.075 UNC -0.018 -0.300 0.766 0.156 1.590 0.114 PCR 0.000 0.000 0.999 0.256 2.670 0.008 PLIC -0.004 -0.070 0.944 -0.009 -0.090 0.926 PTR 0.257 4.420 1.37E-05 0.132 1.350 0.177 RLIC 0.011 0.180 0.856 0.243 2.530 0.012 SCC 0.009 0.150 0.884 0.159 1.630 0.105 SCR -0.025 -0.410 0.681 0.073 0.740 0.458 SFO 0.052 0.860 0.391 0.247 2.570 0.011 SLF 0.072 1.200 0.230 0.205 2.110 0.036 SS 0.162 2.730 0.006 0.287 3.030 0.003 TAP -0.064 -1.070 0.288 -0.135 -1.380 0.170 JHU-ROI RD IQ - 22q11.2DS IQ - Healthy Controls r t p r t p ACR 0.076 1.270 0.205 0.093 0.940 0.347 ALIC -0.064 -1.060 0.291 0.132 1.340 0.181 Average WM -0.031 -0.510 0.611 0.176 1.810 0.072 BCC -0.038 -0.630 0.530 0.023 0.230 0.818 CGC 0.024 0.400 0.688 0.048 0.490 0.622 EC -0.043 -0.710 0.480 -0.035 -0.350 0.725 FXST -0.031 -0.520 0.604 0.034 0.340 0.736 GCC 0.078 1.300 0.195 0.089 0.900 0.368 UNC -0.048 -0.800 0.426 -0.096 -0.970 0.332 PCR -0.004 -0.060 0.954 0.185 1.900 0.060 PLIC -0.021 -0.350 0.728 0.054 0.550 0.585 PTR -0.061 -1.010 0.312 0.210 2.170 0.031 RLIC -0.062 -1.030 0.304 0.126 1.280 0.203 Supplementary Table S3 references 1 Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition. (SCID-I/ P). First MB, Spitzer RL, Gibbon M, Williams JBW. New York: Biometrics Research, New York State Psychiatric Institute; 2002. 2 Diagnostic Interview Schedule for Children. Shaffer D, Schwab-Stone M, Fisher P, Cohen P, Piacentini J, Davies M et al. The Diagnostic Interview Schedule for Children–Revised Version (DISC-R), I: preparation, field testing, interrater reliability, and acceptability. J Am Acad Child Adolesc Psychiatry 1993; 32:643–650. 3 Structured Interview for Prodromal Syndromes. McGlashan TH. Structured Interview for Prodromal Syndromes (SIPS). Yale University: New Haven, 2001. 4 Brief Psychiatric Rating Scale. Ventura J, Green MF, Shaner A, Liberman RP. Training and quality assurance with the Brief Psychiatric Rating Scale: The drift busters. Int J Methods Psychiatr Res 1993;3: 221–244. 5 Kiddie Schedule for Affective Disorders and Schizophrenia- Present and Lifetime Version. Kaufman J, Birmaher B, Brent D, Rao U, Flynn C, Moreci P et al. Schedule for affective disorders and schizophrenia for school-age children-present and lifetime version (K-SADS-PL): initial reliability and validity data. J Am Acad Child Adolesc Psychiatry 1997; 36:980–988. 6 Positive and Negative Syndrome Scale. Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull 1987;13:261–276. 7 Mini International Neuropsychiatric Interview. Sheehan DV, Lecrubier Y, Harnett-Sheehan K, Janavs J, Weiller E, Keskiner A et al. The validity of the Mini International Neuropsychiatric Interview (MINI) according to the SCID-P and its reliability. Eur Psychiatry 1997; 12:232–241. 8 Psychiatric Assessment Schedules for Adults with Developmental Disabilities. Moss S, Patel P, Prosser H, Goldberg D, Simpson N, Rowe S et al. Psychiatric morbidity in older people with moderate and severe learning disability. I: Development and reliability of the patient interview (PAS ADD). Br J Psychiatry 1993; 163:471–480. SCC 0.001 0.010 0.992 0.149 1.520 0.129 SCR 0.002 0.040 0.971 0.171 1.750 0.082 SFO 0.021 0.350 0.724 0.175 1.800 0.074 SLF -0.106 -1.770 0.079 0.091 0.920 0.359 SS -0.032 -0.530 0.598 0.106 1.080 0.284 TAP 0.022 0.360 0.723 0.314 3.340 0.001 9 Child and Adolescent Psychiatric Assessment (CAPA) version 4.2. Angold A, Prendergast M, Cox A, Harrington R, Simonoff E, Rutter M. The Child and Adolescent Psychiatric Assessment (CAPA). Psychol Med 1995; 25:739–753. 10 Structured Clinical Interview for DSM_IV Axis II Personality Disorders. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured Clinical Interview for DSM-IV Axis I Disorders: Patient Edition Biometrics Research. Biometrics Research, New York State Psychiatric Institute: New York, 2002. 11 Scale for the Assessment of Positive Symptoms. Andreasen NC. Scale for the assessment of positive symptoms. University of Iowa; Iowa City, 1984. 12 Scale for the Assessment of Negative Symptoms. Andreasen NC. Scale for the assessment of negative symptoms. University of Iowa; Iowa City, 1984. 13 Schizophrenia Proneness Instrument-Adult. Schultze-Lutter F, Addington J, Ruhrmann S, Klosterkötter J. Schizophrenia Proneness Instrument (SPI-A). Giovanni Fioriti: Rome, Italy, 2007. 14 Ho JS, Radoeva PD, Jalbrzikowski M, Chow C, Hopkins J, Tran W et al. Deficits in mental state attributions in individuals with 22q11.2 deletion syndrome (velo-cardio-facial syndrome). Autism 2012; 5: 407-418. 15 Jalbrzikowski M, Carter C, Senturk D, Chow C, Hopkins JM, Green MF et al. Social cognition in 22q11.2 microdeletion syndrome: Relevance to psychosis. Schizophr Res 2012; 142(0): 99-107. 16 Jalbrzikowski M, Jonas R, Senturk D, Patel A, Chow C, Green MF et al. Structural abnormalities in cortical volume, thickness, and surface area in 22q11.2 microdeletion syndrome: Relationship with psychotic symptoms. NeuroImage: Clin 2013; 3: 405–415. 17 Jalbrzikowski M, Villalon-Reina JE, Karlsgodt KH, Senturk D, Chow C, Thompson PM et al. Altered white matter microstructure is associated with social cognition and psychotic symptoms in 22q11.2 microdeletion syndrome. Front Behav Neurosci 2014; 8: 393. 18 Jonas RK, Jalbrzikowski M, Montojo CA, Patel A, Kushan L, Chow CC et al. Altered brain structure-function relationships underlie executive dysfunction in 22q11.2 deletion syndrome. Mol Neuropsychiatry 2015; 1: 235– 246. 19 Schreiner MJ, Lazaro MT, Jalbrzikowski M, Bearden CE. Converging levels of analysis on a genomic hotspot for psychosis: Insights from 22q11.2 Deletion Syndrome. Neuropharmacology 2013; 68: 157-173. 20 Schreiner M, Forsyth JK, Karlsgodt KH, Anderson AE, Hirsh N, Kushan L et al. Intrinsic connectivity network-based classification and detection of psychotic symptoms in youth with 22q11.2 deletions. Cereb Cortex 2017; 27: 3294-3306. 21 Radoeva PD, Coman IL, Salazar CA, Gentile KL, Higgins AM, Middleton FA et al. Association between autism spectrum disorder in individuals with velocardiofacial (22q11.2 deletion) syndrome and PRODH and COMT genotypes. Psychiatr Genet 2014; 24:269-272. 22 Antshel KM, Hendricks K, Shprintzen R, Fremont W, Higgins AM, Faraone SV et al. The longitudinal course of attention deficit/hyperactivity disorder in velo-cardio-facial syndrome. J Pediatr 2013;163:187-193. 23 Radoeva PD, Coman IL, Antshel KM, Fremont W, McCarthy CS, Kotkar A et al. Atlas-based white matter analysis in individuals with velo-cardio facial syndrome (22q11.2 deletion syndrome) and unaffected siblings. Behav Brain Funct 2012;8:38. 24 Kunwar A, Ramanathan S, Nelson J, Antshel KM, Fremont W, Higgins AM, et al. Cortical gyrification in velo-cardio-facial (22q11.2 deletion) syndrome: a longitudinal study. Schizophr Res 2012;137:20-25. 25 Kates WR, Antshel KM, Faraone SV, Fremont WP, Higgins AM, Shprintzen RJ et al. Neuroanatomic predictors to prodromal psychosis in velocardiofacial syndrome (22q11.2 deletion syndrome): a longitudinal study. Biol Psychiatry 2011;69:945-952. 26 Kates WR, Bansal R, Fremont W, Antshel KM, Hao X, Higgins AM et al. Mapping cortical morphology in youth with velocardiofacial (22q11.2 deletion) syndrome. J Am Acad Child Adolesc Psychiatry 2011;50:272-282. 27 Coman IL, Gnirke MH, Middleton FA, Antshel KM, Fremont W, Higgins AM et al. The effects of gender and catechol O-methyltransferase (COMT) Val108/158Met polymorphism on emotion regulation in velo-cardio-facial syndrome (22q11.2 deletion syndrome): An fMRI study. Neuroimage 2010;53:1043-1050. 28 Roizen NJ, Higgins AM, Antshel KM, Fremont W, Shprintzen R, Kates WR. 22q11.2 deletion syndrome: are motor deficits more than expected for IQ level? J Pediatr 2010;157:658-661. 29 Antshel KM, Shprintzen R, Fremont W, Higgins AM, Faraone SV, Kates WR. Cognitive and psychiatric predictors to psychosis in velocardiofacial syndrome: a 3-year follow-up study. J Am Acad Child Adolesc Psychiatry 2010;49:333-344. 30 Antshel KM, Peebles J, AbdulSabur N, Higgins AM, Roizen N, Shprintzen R et al. Associations between performance on the Rey-Osterrieth Complex Figure and regional brain volumes in children with and without velocardiofacial syndrome. Dev Neuropsychol 2008;33:601-622. 31 Kates WR, Antshel KM, Fremont WP, Shprintzen RJ, Strunge LA, Burnette CP et al. Comparing phenotypes in patients with idiopathic autism to patients with velocardiofacial syndrome (22q11 DS) with and without autism. Am J Med Genet A 2007;143A:2642-2650. 32 Antshel KM, Aneja A, Strunge L, Peebles J, Fremont WP, Stallone K et al. Autistic spectrum disorders in velo-cardio facial syndrome (22q11.2 deletion). J Autism Dev Disord 2007;37:1776-1786.. 33 Kates WR, Krauss BR, Abdulsabur N, Colgan D, Antshel KM, Higgins AM et al. The neural correlates of non-spatial working memory in velocardiofacial syndrome (22q11.2 deletion syndrome). Neuropsychologia 2007;45:2863-2873. 34 Antshel KM, Stallone K, Abdulsabur N, Shprintzen R, Roizen N, Higgins AM et al. Temperament in velocardiofacial syndrome. J Intellect Disabil Res 2007;51:218-227. 35 Aneja A, Fremont WP, Antshel KM, Faraone SV, AbdulSabur N, Higgins AM et al. Manic symptoms and behavioral dysregulation in youth with velocardiofacial syndrome (22q11.2 deletion syndrome). J Child Adolesc Psychopharmacol 2007;17:105-114. 36 Antshel KM, Fremont W, Roizen NJ, Shprintzen R, Higgins AM, Dhamoon A et al. ADHD, major depressive disorder, and simple phobias are prevalent psychiatric conditions in youth with velocardiofacial syndrome. J Am Acad Child Adolesc Psychiatry 2006;45:596-603. 37 Kates WR, Antshel KM, Abdulsabur N, Colgan D, Funke B, Fremont W et al. A gender-moderated effect of a functional COMT polymorphism on prefrontal brain morphology and function in velo-cardio-facial syndrome (22q11.2 deletion syndrome). Am J Med Genet B Neuropsychiatr Genet 2006;141B:274-280. 38 Scott JA, Goodrich-Hunsaker N, Kalish K, Lee A, Hunsaker MR, Schumann CM et al. The hippocampi of children with chromosome 22q11.2 deletion syndrome have localized anterior alterations that predict severity of anxiety. J Psychiatry Neurosci 2016;41:203-213. 39 Deng Y, Goodrich-Hunsaker NJ, Cabaral M, Amaral DG, Buonocore MH, Harvey D et al. Disrupted fornix integrity in children with chromosome 22q11.2 deletion syndrome. Psychiatry Res 2015;232:106-114. 40 Stephenson DD, Beaton EA, Weems CF, Angkustsiri K, Simon TJ. Identifying patterns of anxiety and depression in children with chromosome 22q11.2 deletion syndrome: comorbidity predicts behavioral difficulties and impaired functional communications. Behav Brain Res 2015;276:190-198. 41 Yin JJ, Tang SX, McDonald-McGinn DM, Calkins ME, Whinna DA, Souders MC et al. Contribution of congenital heart disease to neuropsychiatric outcome in school-age children with 22q11.2 deletion syndrome. Am J Med Genet B Neuropsychiatr Genet 2014; 0:137–147. 42 Niarchou M, Moore TM, Tang SX, Calkins ME, McDonald-McGuinn DM, Zackai EH et al.The dimensional structure of psychopathology in 22q11.2 Deletion Syndrome. J Psychiatr Res 2017;92:124-131. 43 Tang SX, Moore TM, Calkins ME, Yi JJ, McDonald-McGinn DM, Zackai EH et al. Emergent, remitted and persistent psychosis-spectrum symptoms in 22q11.2 deletion syndrome. Transl Psychiatry 2017;7:e1180. 44 Tang SX, Moore TM, Calkins ME, Yi JJ, Savitt A, Kohler CG et al. The Psychosis Spectrum in 22q11.2 Deletion Syndrome Is Comparable to That of Nondeleted Youths. Biol Psychiatry 2017;82:17-25. 45 Fiksinski AM, Breetvelt EJ, Duijff SN, Bassett AS, Kahn AS, Vorstman JA. Autism spectrum and psychosis risk in the 22q11.2 deletion syndrome. Findings from a prospective longitudinal study. Schizophr Res 2017; e- pub ahead of print 21 January 2017; doi: 10.1016/j.schres.2017.01.032. 46 Bakker G, Caan MWA, Vingerhoets WAM, da Silva-Alves F, de Koning M, Boot E, et al. Cortical morphology differences in subjects at increased vulnerability for developing a psychotic disorder: A comparison between subjects with ultra-high risk and 22q11.2 deletion syndrome. PLoS ONE 2016; 11: e0159928. 47 Da Silva-Alves F, Schmitz N, Bloemen O, can der Meer J, Meijer J, Boot E et al. White matter abnormalities in adults with 22q11 deletion syndrome with and without schizophrenia. Schizophr Res 2011;132: 75-83. 48 Monks S, Niarchou M, Davies AR, Walters JT, Williams N, Owen MJ et al. Further evidence for high rates of schizophrenia in 22q11.2 deletion syndrome. Schizophr Res 2014;153: 231-236. 49 Niarchou M, Zammit S, van Goozen SHM, Thapar A, Tierling HM, Owen MJ et al. Psychopathology and cognition in children with 22q11.2 deletion syndrome. Br J Psychiatry 2014; 204: 46–54. 50 Niarchou M, Martin J, Thapar A, Owen MJ, van den Bree MBM. The clinical presentation of attention deficit-hyperactivity disorder (ADHD) in children with 22q11.2 deletion syndrome. Am J Med Genet 2015; 168: 730–738. 51 Chawner S, Doherty JL, Moss H, Niarchou M, Walters J, Owen MJ et al. Case-control study finds no evidence that 22q11.2 Deletion Syndrome is associated with cognitive deterioration. Br J Psychiatry (in press) 52 Campbell LE, Azuma R, Ambery F, Stevens A, Smith A, Morris RG et al. Executive functions and memory abilities in children with 22q11.2 deletion syndrome. Aust N Z J Psychiatry 2010; 44: 364-371. 53 Campbell LE, McCabe KL, Melville JL, Strutt PA, Schall U. Social cognition dysfunction in adolescents with 22q11.2 deletion syndrome (velo-cardio-facial syndrome): relationship with executive functioning and social competence/functioning. J Intellect Disabil Res 2015; 59: 845-859. 54 McCabe KL & Carr VJ. Relationship between childhood adversity and clinical and cognitive features in schizophrenia. J Psychiatr Res 2012; 46: 600-607. 55 McCabe KL, Melville JL, Rich D, Strutt PA, Cooper G, Loughland CM et al. Divergent patterns of social cognition performance in autism and 22q11.2 deletion syndrome (22q11DS). J Autism Dev Disord 2013; 43: 1926-1934. 56 McCabe KL, Atkinson RJ, Cooper G, Melville JL, Harris J, Schall U et al. Pre-pulse inhibition and antisaccade performance indicate impaired attention modulation of cognitive inhibition in 22q11.2 deletion syndrome. J Neurodev Disord 2014; 6:38. 57 Wechsler Abbreviated Scale of Intelligence Wechsler D. Wechsler abbreviated scale of intelligence. Harcourt Brace & Company. New York, NY: The Psychological Corporation; 1999. 58 Wechsler Intelligence Scale for Children – Third Edition Wechsler D. Wechsler Intelligence Scale for Children.3rd ed. San Antonio, TX: Psychological Corp; 1991. 59 Wechsler Intelligence Scale for Children –Fourth Edition Wechsler D. Wechsler Intelligence Scale for Children.4th ed. San Antonio, TX: Psychological Corp; 2003. 60 Wechsler Adult Intelligence Scale – Third Edition Wechsler D. Wechsler adult intelligence scale - 3rd ed. San Antonio,TX: Psychological Corp; 1997. 61 Wechsler Intelligence Scale for Children—Revised Wechsler D. Manual of the Wechsler intelligence scale for children-revised. New York: Psychological Corp; 1974. 62 Gudbrandsen M, Daly E, Murphy CM, Wichers RH, Stoencheva V, Perry E, Andrews D, Blackmore CE, Rogdaki M, Kushan L, Bearden CE, Murphy DGM, Craig MC, Ecker C. The Neuroanatomy of Autism Spectrum Disorder Symptomatology in 22q11.2 Deletion Syndrome. Cereb Cortex (2018). References 1 McDonald-McGinn DM, Sullivan KE, Marino B, Philip N, Swillen A, Vorstman JAS et al. 22q11.2 deletion syndrome. Nat Rev Dis Primers 2015; 1 : 15071. 2 Oskarsdóttir S, Vujic M, Fasth A. Incidence and prevalence of the 22q11 deletion syndrome: a population-based study in Western Sweden. Arch Dis Child 2004; 89 : 148–151. 3 Botto LD, May K, Fernhoff PM, Correa A, Coleman K, Rasmussen SA et al. A Population-Based Study of the 22q11.2 Deletion: Phenotype, Incidence, and Contribution to Major Birth Defects in the Population. Pediatrics 2003; 112 : 101–107. 4 Murphy KC, Owen MJ. Velo-cardio-facial syndrome: a model for understanding the genetics and pathogenesis of schizophrenia. Br J Psychiatry 2001; 179 : 397–402. 5 Schneider M, Debbané M, Bassett AS, Chow EWC, Fung WLA, van den Bree M et al. Psychiatric disorders from childhood to adulthood in 22q11.2 deletion syndrome: results from the International Consortium on Brain and Behavior in 22q11.2 Deletion Syndrome. Am J Psychiatry 2014; 171 : 627–639. 6 Bassett AS, Chow EWC, AbdelMalik P, Gheorghiu M, Husted J, Weksberg R. The schizophrenia phenotype in 22q11 deletion syndrome. Am J Psychiatry 2003; 160: 1580–1586. 7 Chow EW, Bassett AS, Weksberg R. Velo-cardio-facial syndrome and psychotic disorders: implications for psychiatric genetics. Am J Med Genet 1994; 54 : 107–112. 8 Morgan VA, Leonard H, Bourke J, Jablensky A. Intellectual disability co-occurring with schizophrenia and other psychiatric illness: population-based study. Br J Psychiatry 2008; 193 : 364–372. 9 Meechan DW, Tucker ES, Maynard TM, LaMantia A-S. Diminished dosage of 22q11 genes disrupts neurogenesis and cortical development in a mouse model of 22q11 deletion/DiGeorge syndrome. Proc Natl Acad Sci U S A 2009; 106 : 16434–16445. 10 Cercignani M, Bouyagoub S. Brain microstructure by multi-modal MRI: Is the whole greater than the sum of its parts? Neuroimage 2018; 182 : 117–127. 11 Beaulieu C. The basis of anisotropic water diffusion in the nervous system--a technical review. NMR Biomed 2002; 15 : 435–455. 12 Moseley ME, Cohen Y, Kucharczyk J, Mintorovitch J, Asgari HS, Wendland MF et al. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990; 176 : 439–445. 83 13 Beaulieu C. The Biological Basis of Diffusion Anisotropy. Diffusion MRI. 2014; : 155–183. 14 Stejskal EO. Use of Spin Echoes in a Pulsed Magnetic-Field Gradient to Study Anisotropic, Restricted Diffusion and Flow. J Chem Phys 1965; 43 : 3597–3603. 15 Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectroscopy and imaging. Biophys J 1994; 66 : 259–267. 16 Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffusion anisotropy. Magn Reson Med 1996; 36 : 893–906. 17 Jeurissen B, Leemans A, Tournier J-D, Jones DK, Sijbers J. Investigating the prevalence of complex fiber configurations in white matter tissue with diffusion magnetic resonance imaging. Hum Brain Mapp 2013; 34 : 2747–2766. 18 Alexander AL, Lee JE, Lazar M, Field AS. Diffusion tensor imaging of the brain. Neurotherapeutics 2007; 4 : 316–329. 19 Kinoshita Y, Ohnishi A, Kohshi K, Yokota A. Apparent diffusion coefficient on rat brain and nerves intoxicated with methylmercury. Environ Res 1999; 80: 348–354. 20 Schwartz ED, Cooper ET, Fan Y, Jawad AF, Chin C-L, Nissanov J et al. MRI diffusion coefficients in spinal cord correlate with axon morphometry. Neuroreport 2005; 16 : 73–76. 21 Song S-K, Sun S-W, Ju W-K, Lin S-J, Cross AH, Neufeld AH. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage 2003; 20 : 1714–1722. 22 Harsan LA, Poulet P, Guignard B, Parizel N, Skoff RP, Ghandour MS. Astrocytic hypertrophy in dysmyelination influences the diffusion anisotropy of white matter. J Neurosci Res 2007; 85 : 935–944. 23 Nair G, Tanahashi Y, Low HP, Billings-Gagliardi S, Schwartz WJ, Duong TQ. Myelination and long diffusion times alter diffusion-tensor-imaging contrast in myelin-deficient shiverer mice. Neuroimage 2005; 28 : 165–174. 24 Latour LL, Svoboda K, Mitra PP, Sotak CH. Time-dependent diffusion of water in a biological model system. Proc Natl Acad Sci U S A 1994; 91 : 1229–1233. 25 Nir TM, Jahanshad N, Villalon-Reina JE, Isaev D, Zavaliangos-Petropulu A, Zhan L et al. Fractional anisotropy derived from the diffusion tensor distribution function boosts power to detect Alzheimer’s disease deficits. Magn Reson Med 2017; 78 : 2322–2333. 26 Leow AD, Zhu S, Zhan L, McMahon K, de Zubicaray GI, Meredith M et al. The tensor distribution function. Magn Reson Med 2009; 61 : 205–214. 27 Tuch DS. Q-ball imaging. Magn Reson Med 2004; 52 : 1358–1372. 84 28 Descoteaux M, Angelino E, Fitzgibbons S, Deriche R. Regularized, fast, and robust analytical Q-ball imaging. Magn Reson Med 2007; 58 : 497–510. 29 Tournier J-D, Calamante F, Connelly A. Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution. Neuroimage 2007; 35 : 1459–1472. 30 Jalbrzikowski M, Villalon-Reina JE, Karlsgodt KH, Senturk D, Chow C, Thompson PM et al. Altered white matter microstructure is associated with social cognition and psychotic symptoms in 22q11.2 microdeletion syndrome. Front Behav Neurosci 2014; 8 : 393. 31 Perlstein MD, Chohan MR, Coman IL, Antshel KM, Fremont WP, Gnirke MH et al. White matter abnormalities in 22q11.2 deletion syndrome: preliminary associations with the Nogo-66 receptor gene and symptoms of psychosis. Schizophr Res 2014; 152 : 117–123. 32 Jahanshad N, Kochunov PV, Sprooten E, Mandl RC, Nichols TE, Almasy L et al. Multi-site genetic analysis of diffusion images and voxelwise heritability analysis: a pilot project of the ENIGMA-DTI working group. Neuroimage 2013; 81 : 455–469. 33 Kates WR, Olszewski AK, Gnirke MH, Kikinis Z, Nelson J, Antshel KM et al. White matter microstructural abnormalities of the cingulum bundle in youths with 22q11.2 deletion syndrome: associations with medication, neuropsychological function, and prodromal symptoms of psychosis. Schizophr Res 2015; 161 : 76–84. 34 Kikinis Z, Asami T, Bouix S, Finn CT, Ballinger T, Tworog-Dube E et al. Reduced fractional anisotropy and axial diffusivity in white matter in 22q11.2 deletion syndrome: a pilot study. Schizophr Res 2012; 141 : 35–39. 35 Olszewski AK, Kikinis Z, Gonzalez CS, Coman IL, Makris N, Gong X et al. The social brain network in 22q11.2 deletion syndrome: a diffusion tensor imaging study. Behav Brain Funct 2017; 13 : 4. 36 Zhang H, Schneider T, Wheeler-Kingshott CA, Alexander DC. NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage 2012; 61 : 1000–1016. 37 Dell’Acqua F, Simmons A, Williams SCR, Catani M. Can spherical deconvolution provide more information than fiber orientations? Hindrance modulated orientational anisotropy, a true-tract specific index to characterize white matter diffusion. Hum Brain Mapp 2013; 34 : 2464–2483. 38 Raffelt D, Tournier J-D, Rose S, Ridgway GR, Henderson R, Crozier S et al. Apparent Fibre Density: a novel measure for the analysis of diffusion-weighted magnetic resonance images. Neuroimage 2012; 59 : 3976–3994. 39 da Silva Alves F, Schmitz N, Bloemen O, van der Meer J, Meijer J, Boot E et al. White matter abnormalities in adults with 22q11 deletion syndrome with and without 85 schizophrenia. Schizophr Res 2011; 132 : 75–83. 40 Radoeva PD, Coman IL, Antshel KM, Fremont W, McCarthy CS, Kotkar A et al. Atlas-based white matter analysis in individuals with velo-cardio-facial syndrome (22q11.2 deletion syndrome) and unaffected siblings. Behav Brain Funct 2012; 8 : 38. 41 Villalon-Reina J, Jahanshad N, Beaton E, Toga AW, Thompson PM, Simon TJ. White matter microstructural abnormalities in girls with chromosome 22q11.2 deletion syndrome, Fragile X or Turner syndrome as evidenced by diffusion tensor imaging. Neuroimage 2013; 81 : 441–454. 42 Deng Y, Goodrich-Hunsaker NJ, Cabaral M, Amaral DG, Buonocore MH, Harvey D et al. Disrupted fornix integrity in children with chromosome 22q11.2 deletion syndrome. Psychiatry Res 2015; 232 : 106–114. 43 Roalf DR, Eric Schmitt J, Vandekar SN, Satterthwaite TD, Shinohara RT, Ruparel K et al. White matter microstructural deficits in 22q11.2 deletion syndrome. Psychiatry Res Neuroimaging 2017; 268 : 35–44. 44 Simon TJ, Wu Z, Avants B, Zhang H, Gee JC, Stebbins GT. Atypical cortical connectivity and visuospatial cognitive impairments are related in children with chromosome 22q11.2 deletion syndrome. Behav Brain Funct 2008; 4 : 25. 45 Bakker G, Caan MWA, Schluter RS, Bloemen OJN, da Silva-Alves F, de Koning MB et al. Distinct white-matter aberrations in 22q11.2 deletion syndrome and patients at ultra-high risk for psychosis. Psychol Med 2016; 46 : 2299–2311. 46 Barnea-Goraly N, Menon V, Krasnow B, Ko A, Reiss A, Eliez S. Investigation of white matter structure in velocardiofacial syndrome: a diffusion tensor imaging study. Am J Psychiatry 2003; 160 : 1863–1869. 47 Simon TJ, Ding L, Bish JP, McDonald-McGinn DM, Zackai EH, Gee J. Volumetric, connective, and morphologic changes in the brains of children with chromosome 22q11.2 deletion syndrome: an integrative study. Neuroimage 2005; 25 : 169–180. 48 Sundram F, Murphy DG, Murphy KC. White matter microstructure in children with Velocardiofacial Syndrome: A Diffusion Tensor Imaging and Voxel Based Morphometry study. J Intellect Disabil Res 2008; 52 : 812–812. 49 Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE et al. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 2006; 31 : 1487–1505. 50 Kikinis Z, Cho KIK, Coman IL, Radoeva PD, Bouix S, Tang Y et al. Abnormalities in brain white matter in adolescents with 22q11.2 deletion syndrome and psychotic symptoms. Brain Imaging Behav 2017; 11 : 1353–1364. 51 Tylee DS, Kikinis Z, Quinn TP, Antshel KM, Fremont W, Tahir MA et al. Machine-learning classification of 22q11.2 deletion syndrome: A diffusion tensor 86 imaging study. Neuroimage Clin 2017; 15 : 832–842. 52 Sun D, Ching CRK, Lin A, Forsyth JK, Kushan L, Vajdi A et al. Large-scale mapping of cortical alterations in 22q11.2 deletion syndrome: Convergence with idiopathic psychosis and effects of deletion size. Mol Psychiatry 2018. doi: 10.1038/s41380-018-0078-5 . 53 Sørensen KM, Agergaard P, Olesen C, Andersen PS, Larsen LA, Ostergaard JR et al. Detecting 22q11.2 deletions by use of multiplex ligation-dependent probe amplification on DNA from neonatal dried blood spot samples. J Mol Diagn 2010; 12 : 147–151. 54 Coupe P, Yger P, Prima S, Hellier P, Kervrann C, Barillot C. An Optimized Blockwise Nonlocal Means Denoising Filter for 3-D Magnetic Resonance Images. IEEE Transactions on Medical Imaging. 2008; 27 : 425–441. 55 Smith SM. Fast robust automated brain extraction. Hum Brain Mapp 2002; 17 : 143–155. 56 Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA et al. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging 2010; 29 : 1310–1320. 57 Ségonne F, Dale AM, Busa E, Glessner M, Salat D, Hahn HK et al. A hybrid approach to the skull stripping problem in MRI. Neuroimage 2004; 22 : 1060–1075. 58 Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration with cross-correlation: evaluating automated labeling of elderly and neurodegenerative brain. Med Image Anal 2008; 12 : 26–41. 59 Andersson JLR, Skare S, Ashburner J. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. Neuroimage 2003; 20 : 870–888. 60 Andersson JLR, Sotiropoulos SN. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage 2016; 125 : 1063–1078. 61 Chang L-C, Jones DK, Pierpaoli C. RESTORE: robust estimation of tensors by outlier rejection. Magn Reson Med 2005; 53 : 1088–1095. 62 Garyfallidis E, Brett M, Amirbekian B, Rokem A, van der Walt S, Descoteaux M et al. Dipy, a library for the analysis of diffusion MRI data. Front Neuroinform 2014; 8 : 8. 63 DTI Protocols « ENIGMA. http://enigma.ini.usc.edu/protocols/dti-protocols/ (accessed 31 Jul2018). 64 Mori S, Oishi K, Jiang H, Jiang L, Li X, Akhter K et al. Stereotaxic white matter atlas based on diffusion tensor imaging in an ICBM template. Neuroimage 2008; 40 : 570–582. 65 Kelly S, Jahanshad N, Zalesky A, Kochunov P, Agartz I, Alloza C et al. Widespread white matter microstructural differences in schizophrenia across 4322 individuals: 87 results from the ENIGMA Schizophrenia DTI Working Group. Mol Psychiatry 2018; 23 : 1261–1269. 66 Lebel C, Walker L, Leemans A, Phillips L, Beaulieu C. Microstructural maturation of the human brain from childhood to adulthood. Neuroimage 2008; 40: 1044–1055. 67 Kochunov P, Glahn DC, Lancaster JL, Winkler AM, Smith S, Thompson PM et al. Genetics of microstructure of cerebral white matter using diffusion tensor imaging. Neuroimage 2010; 53 : 1109–1116. 68 Simmonds DJ, Hallquist MN, Asato M, Luna B. Developmental stages and sex differences of white matter and behavioral development through adolescence: a longitudinal diffusion tensor imaging (DTI) study. Neuroimage 2014; 92 : 356–368. 69 Homepage [The metafor Package]. http://www.metafor-project.org/doku.php (accessed 31 Jul2018). 70 Zhu T, Hu R, Qiu X, Taylor M, Tso Y, Yiannoutsos C et al. Quantification of accuracy and precision of multi-center DTI measurements: a diffusion phantom and human brain study. Neuroimage 2011; 56 : 1398–1411. 71 Zhan L, Leow AD, Jahanshad N, Chiang M-C, Barysheva M, Lee AD et al. How does angular resolution affect diffusion imaging measures? Neuroimage 2010; 49 : 1357–1371. 72 Jahanshad N, Zhan L, Bernstein MA, Borowski BJ, Jack CR, Toga AW et al. Diffusion tensor imaging in seven minutes: Determining trade-offs between spatial and directional resolution. In: 2010 IEEE International Symposium on Biomedical Imaging: From Nano to Macro . 2010, pp 1161–1164. 73 Fortin J-P, Parker D, Tunç B, Watanabe T, Elliott MA, Ruparel K et al. Harmonization of multi-site diffusion tensor imaging data. Neuroimage 2017; 161 : 149–170. 74 Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J R Stat Soc Series B Stat Methodol 1995; 57 : 289–300. 75 CRAN - Package CORE. https://cran.r-project.org/web/packages/CORE/index.html (accessed 28 Apr2019). 76 Lebel C, Gee M, Camicioli R, Wieler M, Martin W, Beaulieu C. Diffusion tensor imaging of white matter tract evolution over the lifespan. Neuroimage 2012; 60: 340–352. 77 Wilcox R. Chapter 11 - More Regression Methods. In: Wilcox R (ed). Introduction to Robust Estimation and Hypothesis Testing (Fourth Edition) . Academic Press, 2017, pp 585–691. 78 Dunnett CW. Pairwise Multiple Comparisons in the Unequal Variance Case. J Am Stat Assoc 1980; 75 : 796–800. 79 Wilcox RR, Tian TS. Measuring effect size: a robust heteroscedastic approach for two 88 or more groups. J Appl Stat 2011; 38 : 1359–1368. 80 Vederine F-E, Wessa M, Leboyer M, Houenou J. A meta-analysis of whole-brain diffusion tensor imaging studies in bipolar disorder. Prog Neuropsychopharmacol Biol Psychiatry 2011; 35 : 1820–1826. 81 Laule C, Vavasour IM, Kolind SH, Li DKB, Traboulsee TL, Moore GRW et al. Magnetic resonance imaging of myelin. Neurotherapeutics 2007; 4 : 460–484. 82 Takahashi M, Ono J, Harada K, Maeda M, Hackney DB. Diffusional anisotropy in cranial nerves with maturation: quantitative evaluation with diffusion MR imaging in rats. Radiology 2000; 216 : 881–885. 83 Mottershead JP, Schmierer K, Clemence M, Thornton JS, Scaravilli F, Barker GJ et al. High field MRI correlates of myelin content and axonal density in multiple sclerosis--a post-mortem study of the spinal cord. J Neurol 2003; 250 : 1293–1301. 84 Concha L, Livy DJ, Beaulieu C, Wheatley BM, Gross DW. In vivo diffusion tensor imaging and histopathology of the fimbria-fornix in temporal lobe epilepsy. J Neurosci 2010; 30 : 996–1002. 85 Pandya DN, Yeterian EH. Architecture and Connections of Cortical Association Areas. In: Peters A, Jones EG (eds). Association and Auditory Cortices . Springer US: Boston, MA, 1985, pp 3–61. 86 Innocenti GM, Vercelli A, Caminiti R. The diameter of cortical axons depends both on the area of origin and target. Cereb Cortex 2014; 24 : 2178–2188. 87 Berbel P, Innocenti GM. The development of the corpus callosum in cats: a light- and electron-microscopic study. J Comp Neurol 1988; 276 : 132–156. 88 LaMantia A-S, Rakic P. Cytological and quantitative characteristics of four cerebral commissures in the rhesus monkey. J Comp Neurol 1990; 291 : 520–537. 89 LaMantia AS, Rakic P. Axon overproduction and elimination in the corpus callosum of the developing rhesus monkey. J Neurosci 1990; 10: 2156–2175. 90 Aboitiz F, Scheibel AB, Fisher RS, Zaidel E. Fiber composition of the human corpus callosum. Brain Res 1992; 598 : 143–153. 91 Kiehl TR, Chow EWC, Mikulis DJ, George SR, Bassett AS. Neuropathologic features in adults with 22q11.2 deletion syndrome. Cereb Cortex 2009; 19 : 153–164. 92 Budde MD, Janes L, Gold E, Turtzo LC, Frank JA. The contribution of gliosis to diffusion tensor anisotropy and tractography following traumatic brain injury: validation in the rat using Fourier analysis of stained tissue sections. Brain 2011; 134 : 2248–2260. 93 Meechan DW, Tucker ES, Maynard TM, LaMantia A-S. Cxcr4 regulation of interneuron migration is disrupted in 22q11.2 deletion syndrome. Proc Natl Acad Sci U 89 S A 2012; 109 : 18601–18606. 94 Wu P, Teot L, Murdoch G, Monaghan-Nichols AP, McFadden K. Neuropathology of 22q11 deletion syndrome in an infant. Pediatr Dev Pathol 2014; 17 : 386–392. 95 Baharnoori M, Mandell DM, Andrade DM, Chow EWC, Bassett AS, Kiehl T-R. Periventricular nodular heterotopia and bilateral intraventricular xanthogranulomas in 22q11.2 deletion syndrome. Human Pathology: Case Reports 2017; 9 : 55–57. 96 van Kogelenberg M, Ghedia S, McGillivray G, Bruno D, Leventer R, Macdermot K et al. Periventricular heterotopia in common microdeletion syndromes. Mol Syndromol 2010; 1 : 35–41. 97 Andrade DM, Krings T, Chow EWC, Kiehl T-R, Bassett AS. Hippocampal malrotation is associated with chromosome 22q11.2 microdeletion. Can J Neurol Sci 2013; 40 : 652–656. 98 Karlsgodt KH, van Erp TGM, Poldrack RA, Bearden CE, Nuechterlein KH, Cannon TD. Diffusion tensor imaging of the superior longitudinal fasciculus and working memory in recent-onset schizophrenia. Biol Psychiatry 2008; 63 : 512–518. 99 Kaden E, Kruggel F, Alexander DC. Quantitative mapping of the per-axon diffusion coefficients in brain white matter. Magn Reson Med 2016; 75 : 1752–1763. 100 Geeraert BL, Lebel RM, Mah AC, Deoni SC, Alsop DC, Varma G et al. A comparison of inhomogeneous magnetization transfer, myelin volume fraction, and diffusion tensor imaging measures in healthy children. Neuroimage 2017. doi: 10.1016/j.neuroimage.2017.09.019 . 101 Telischak NA, Detre JA, Zaharchuk G. Arterial spin labeling MRI: clinical applications in the brain. J Magn Reson Imaging 2015; 41 : 1165–1180. 102 Manjón JV, Coupé P, Concha L, Buades A, Collins DL, Robles M. Diffusion weighted image denoising using overcomplete local PCA. PLoS One 2013; 8 : e73021. 103 Dell’Acqua F, Lacerda L, Catani M, Simmons A. Anisotropic Power Maps: A diffusion contrast to reveal low anisotropy tissues from HARDI data. In: Proceedings Joint Annual Meeting ISMRMESMRMB, ISMRM2014, Milan . 2014, p 0730. 104 Viechtbauer W. Metafor: Meta-analysis package for R version 1.9-2. Computer software 2015. https://cran.r-project.org/web/packages/metafor/index.html . 105 Winkler AM. Brainder. https://brainder.org . 106 Zhan L, Leow AD, Zhu S, Barysheva M, Toga AW, McMahon KL et al. Analyzing multi-fiber reconstruction in high angular resolution diffusion imaging using the tensor distribution function. In: 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro . 2009, pp 1402–1405. 107 Zhan L, Leow AD, Aganj I, Lenglet C, Sapiro G, Yacoub E et al. Differential information content in staggered multiple shell hardi measured by the tensor 90 distribution function. In: 2011 IEEE International Symposium on Biomedical Imaging: From Nano to Macro . 2011, pp 305–309. 108 Fritzsche KH, Laun FB, Meinzer H-P, Stieltjes B. Opportunities and pitfalls in the quantification of fiber integrity: what can we gain from Q-ball imaging? Neuroimage 2010; 51 : 242–251. 109 Corbo D, Caiazzo G, Trojsi F, Monsurrò MR, Gallo A, Bonavita S et al. Advantages of QBI in TBSS analyses. Magn Reson Imaging 2014; 32 : 184–189. 110 Rees E, Kirov G, Sanders A, Walters JTR, Chambert KD, Shi J et al. Evidence that duplications of 22q11.2 protect against schizophrenia. Mol Psychiatry 2014; 19 : 37–40. 111 Hoeft F, Barnea-Goraly N, Haas BW, Golarai G, Ng D, Mills D et al. More is not always better: increased fractional anisotropy of superior longitudinal fasciculus associated with poor visuospatial abilities in Williams syndrome. J Neurosci 2007; 27 : 11960–11965. 112 Assaf Y, Basser PJ. Composite hindered and restricted model of diffusion (CHARMED) MR imaging of the human brain. Neuroimage 2005; 27 : 48–58. 113 Tariq M, Schneider T, Alexander DC, Gandini Wheeler-Kingshott CA, Zhang H. Bingham-NODDI: Mapping anisotropic orientation dispersion of neurites using diffusion MRI. Neuroimage 2016; 133 : 207–223. 114 Stikov N, Campbell JSW, Stroh T, Lavelée M, Frey S, Novek J et al. In vivo histology of the myelin g-ratio with magnetic resonance imaging. Neuroimage 2015; 118 : 397–405. 115 Raffelt DA, Tournier J-D, Smith RE, Vaughan DN, Jackson G, Ridgway GR et al. Investigating white matter fibre density and morphology using fixel-based analysis. Neuroimage 2017; 144 : 58–73. 116 Zhang S, Arfanakis K. Evaluation of standardized and study-specific diffusion tensor imaging templates of the adult human brain: Template characteristics, spatial normalization accuracy, and detection of small inter-group FA differences. Neuroimage 2018; 172 : 40–50. 117 Greve DN, Fischl B. Accurate and robust brain image alignment using boundary-based registration. Neuroimage 2009; 48 : 63–72. 118 Fick R, Deriche R, Wassermann D. Dmipy: An Open-source Framework for Reproducible dMRI-Based Microstructure Research . 2018 doi: 10.5281/zenodo.1188268 . 119 Tournier J-D, Calamante F, Connelly A. MRtrix: Diffusion tractography in crossing fiber regions. Int J Imaging Syst Technol 2012; 22 : 53–66. 120 MRtrix3. http://www.mrtrix.org (accessed 3 May2019). 121 Behrens TEJ, Woolrich MW, Jenkinson M, Johansen-Berg H, Nunes RG, Clare S et al. Characterization and propagation of uncertainty in diffusion-weighted MR imaging. Magnetic Resonance in Medicine: An Official Journal of the International Society for 91 Magnetic Resonance in Medicine 2003; 50: 1077–1088. 122 Panagiotaki E, Schneider T, Siow B, Hall MG, Lythgoe MF, Alexander DC. Compartment models of the diffusion MR signal in brain white matter: a taxonomy and comparison. Neuroimage 2012; 59 : 2241–2254. 123 Tournier J-D, Calamante F, Gadian DG, Connelly A. Direct estimation of the fiber orientation density function from diffusion-weighted MRI data using spherical deconvolution. Neuroimage 2004; 23 : 1176–1185. 124 Tax CMW, Jeurissen B, Vos SB, Viergever MA, Leemans A. Recursive calibration of the fiber response function for spherical deconvolution of diffusion MRI data. Neuroimage 2014; 86 : 67–80. 125 Jeurissen B, Tournier J-D, Dhollander T, Connelly A, Sijbers J. Multi-tissue constrained spherical deconvolution for improved analysis of multi-shell diffusion MRI data. Neuroimage 2014; 103 : 411–426. 126 Dhollander T, Raffelt D, Connelly A. Unsupervised 3-tissue response function estimation from single-shell or multi-shell diffusion MR data without a co-registered T1 image. In: ISMRM Workshop on Breaking the Barriers of Diffusion MRI . 2016 https://www.researchgate.net/profile/Thijs_Dhollander/publication/307863133_U nsupervised_3-tissue_response_function_estimation_from_single-shell_or_multi-shel l_diffusion_MR_data_without_a_co-registered_T1_image/links/57cfb5c708ae83b3746 23e5a.pdf . 127 Glasser MF, Van Essen DC. Mapping human cortical areas in vivo based on myelin content as revealed by T1- and T2-weighted MRI. J Neurosci 2011; 31 : 11597–11616. 128 Ganzetti M, Wenderoth N, Mantini D. Whole brain myelin mapping using T1- and T2-weighted MR imaging data. Front Hum Neurosci 2014; 8 : 671. 129 HCPpipelines . Github https://github.com/Washington-University/HCPpipelines (accessed 6 May2019). 130 Raffelt D, Tournier J-D, Fripp J, Crozier S, Connelly A, Salvado O. Symmetric diffeomorphic registration of fibre orientation distributions. Neuroimage 2011; 56 : 1171–1180. 131 Raffelt DA, Smith RE, Ridgway GR, Tournier J-D, Vaughan DN, Rose S et al. Connectivity-based fixel enhancement: Whole-brain statistical analysis of diffusion MRI measures in the presence of crossing fibres. Neuroimage 2015; 117 : 40–55. 92

Abstract (if available)

Abstract Diffusion MRI is a valuable tool to study the microstructure of the brain and unveil the underpinnings of neurological illness. Despite many of the limitations of early diffusion MRI reconstruction models, such as diffusion tensor imaging and multi-tensor models, these models are able to provide a global view of the regions of the brain compromised by disease. Precisely, the work shown here on diffusion tensor imaging provides an overall perspective of white matter abnormalities in 22q11.2 Deletion Syndrome. Specifically, we found the basic patterns of abnormalities of anisotropy and diffusivity across the brain in 22q11.2 Deletion Syndrome compared to healthy age-matched controls. Our results derive from the largest neuroimaging dataset to date from individuals with 22q11.2 deletion syndrome, which was crucial to resolve previous conflicting findings from smaller studies. Additionally, with the multi-tensor approach of the tensor distribution function we were able to overcome the crossing-fiber limitation of the single tensor model and were able to show a consistent pattern of increased diffusion anisotropy across the brain in 22q11.2 Deletion Syndrome. After identifying the general pattern of brain microstructural abnormalities caused by the deletion—especially in those regions with an unexpected direction of effects—we used advanced diffusion MRI reconstruction models with multi-shell dMRI acquisitions. These analyses revealed specific abnormalities in the white matter that were not evident with tensor based approaches. Biophysical models such as Neurite Orientation Dispersion and Density Imaging (NODDI) revealed a larger intra-cellular compartment in 22q11.2 Deletion Syndrome. Interestingly, the opposite pattern was revealed in subjects with 22q11.2 duplication. Multi-shell high angular resolution dMRI data was also able to distinguish microstructural features on specific white matter tracts without the bias of crossing tracts. Current novel techniques allow for the parcellation of microstructural measures along discrete fiber orientations and perform population analysis, which ultimately allowed us to confirm previous findings reported in animal models of 22q11.2 Deletion Syndrome. These findings are encouraging and establish a foundation for future work linking gene dosage of copy number variants on more specific microstructural features such as axonal density and dispersion, free-water content due to neuroinflammation, the axonal diameter distribution, and the diffusion dynamics between intra- and extra-cellular compartments.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Pattern detection in medical imaging: pathology specific imaging contrast, features, and statistical models

PDF

Diffusion MRI of the human brain: signal modeling and quantitative analysis

PDF

Pediatric magnetic resonance image processing: applications to posterior fossa cancer and normal development

PDF

Age differences in diffusivity in the locus coeruleus and its ascending noradrenergic tract

Asset Metadata

Creator Villalón Reina, Julio Ernesto (author)

Core Title Unveiling the white matter microstructure in 22q11.2 deletion syndrome with diffusion magnetic resonance imaging

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 08/15/2019

Defense Date 05/22/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag 22q11.2 deletion syndrome,22q11.2 duplication,brain imaging,copy number variants,diffusion magnetic resonance imaging,diffusion MRI,diffusion tensor imaging,dMRI,DTI,neuroimaging,OAI-PMH Harvest,psychosis,schizophrenia

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Thompson, Paul M. (committee chair), Jahanshad, Neda (committee member), Lepore, Natasha (committee member), Marmarelis, Vasilis Z. (committee member)

Creator Email jevillalonr@gmail.com,villalon@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-215514

Unique identifier UC11663194

Identifier etd-VillalnRei-7793.pdf (filename),usctheses-c89-215514 (legacy record id)

Legacy Identifier etd-VillalnRei-7793.pdf

Dmrecord 215514

Document Type Dissertation

Format application/pdf (imt)

Rights Villalón Reina, Julio Ernesto

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA