Individual differences in nerve fiber myelination: A methodological study

Размещено на http://www.allbest.ru/

[Введите текст]

The Government of the Russian federation

Federal State Autonomous Educational Institution of Higher Professional Education

National Research University - Higher School of Economics

Faculty of Social Sciences, School of Psychology,

Master's program

“Cognitive sciences and technologies: from neuron to cognition”

Final qualifying work - MASTER THESIS

«Individual differences in nerve fiber myelination: A methodological study»

Student group № МКН 181

Matiulko, Irina

Last name, First name Middle name

Signature

Scientific adviser

Associate Professor, PhD

Position, Academic degree

Arsalidou, M.

Last, F. M./O.



Table of Contents

Chapter 1. Introduction

Chapter 2. Methods of myelination assessment

2.1 Introduction to the Principles of Magnetic Resonance Imaging

2.1.1 T1/T2 and T2*-weighted imaging

2.1.2 Functional Magnetic Resonance Imaging (MRI)

2.2 Diffusion MRI, or Diffusion Tensor Imaging (DTI)

2.3 Fast Macromolecular Proton Fraction (MPF) Mapping

Chapter 3. Individual differences in myelination

3.1 Myelination and age

3.2 Myelination and Gender

3.3 Myelination and Cognitive Abilities

3.3.1 Corpus Callosum (CC)

3.3.2 Superior Longitudinal Fasciculus (SLF) and Arcuate Fasciculus (AF)

3.3.3 Inferior Longitudinal Fasciculus (ILF)

3.3.4 Uncinate Fasciculus (UF)

3.3.5 Cingulum (C)

3.3.6 Inferior Fronto-Occipital Fasciculus (IFOF)

Chapter 4. Materials and methods

4.1 Participants

4.2 MRI Data Acquisition

4.3 Data Preprocessing

4.2 Data Processing

4.2 Statistical Analysis

Chapter 5. Results

5.1 Diffusion tensor imaging

5.1.1 Corpus callosum: fractional anisotropy

5.1.2 Corpus callosum: radial diffusivity

5.1.3 Corpus callosum: mean diffusivity

5.2.1 Inferior fronto-occipital fasciculus: fractional anisotropy

5.2.2 Inferior fronto-occipital fasciculus: radial diffusivity

5.2.3 Inferior fronto-occipital fasciculus: mean diffusivity

5.3.1 Inferior longitudinal fasciculus: fractional anisotropy

5.3.2 Inferior longitudinal fasciculus: radial diffusivity

5.3.3 Inferior longitudinal fasciculus: mean diffusivity

5.2 Myelin proton fraction mapping

Chapter 6. Discussion

6.1 Diffusion tensor imaging

6.1.1 Corpus callosum

6.1.2 Inferior fronto-occipital fasciculus

6.1.3 Inferior longitudinal fasciculus

6.2 Myelin proton fraction mapping

Conclusion

References



Chapter 1. Introduction

The brain is the most complex organ, which we use for thinking, acting and feeling. Neurons facilitate this process by connecting proximal and distant parts of cortical and subcortical regions. The speed with which neurons communicate is associated with the degree of myelination (Chevalier et al., 2015) and is affected by individual differences related to age (Zhao et al., 2019), gender (Bourisly et al., 2017; De Bellis et al., 2001), health (Coupй et al., 2019; Gorelick et al., 2017), and various physiological and environmental factors (Martin et al., 2020; Petrill et al., 2004; Tabbarah et al., 2002). The most popular indirect imaging method of assessing white matter integrity is diffusion tensor imaging (DTI; Alexander et al., 2007; Basser et al., 1994; Basser & Jones, 2002). Macromolecular proton fraction (MPF) is a novel technique used to image myelin integrity and myelination degree(Khodanovich et al., 2017; Yarnykh, Prihod'ko, et al., 2018; Yarnykh, 2012). The purpose of the current study was to compare and contrast normative values in DTI and MPF considering individual differences such as age and gender.This is the first investigation to report normative DTI indices in Russian children and adolescents and the first study worldwide to combine DTI and MPF to study white matter tracts in children (7-11 years), adolescents (12-15 years), and adults (20-30 years).

Structural Characteristics of the Neuron

Nerve fibers are a fundamental structural component of every neuron. A neuron (neural cell) is a functional unit of the brain, which is comprised of a cell body and two types of protoplasmic structures - dendrites and an axon, through which the information in the form of an electrochemical impulse is transmitted from one cell (a neuron or a receptor cell) to another until it reaches the target - another neuron, a muscle cell, or gland(Purves, 2004). Cell bodies of the neurons are organized in nuclei or layers that comprise the gray matter, whereas axons, which connect distributed or nearby regions of the gray matter, form the white matter of the brain(Wen & Chklovskii, 2005). The speed of electrical signal transmission along the axon is determined by myelination of the axonal membrane (i.e. how densely the axon is wrapped by myelin). For example, the velocity of signal transduction in unmyelinated axons ranges from 0.5 to 10 meters per second, while transduction velocity in myelinated axons reaches 150 meters per second(Purves, 2004). Myelin sheath is formed as an extension of the plasma membrane of specialized cells - oligodendrocytes in the central nervous system and Schwann cells in the peripheral nervous system. Myelin is a principal component of white matter and it accounts for approximately 50-60% of the white matter dry weight (Snaidero & Simons, 2014). The dry mass of myelin contains a high proportion of lipids (70-85%) and a lower proportion of proteins (15-30%). Cellular plasma membrane, in contrast, contains a greater percentage of proteins(Morell & Quarles, 1999), which form pores, channels, and transporters necessary for the transportation of ions and large polar molecules that cannot penetrate the lipid bilayer of the plasma membrane (Cooper, 2000). Myelin sheath insulates axonal membrane and makes it impenetrable to all molecules (Figure 1). Ionic channels required for generation and transmission of the action potential are clustered at myelin-free segments on the plasma membrane called the nodes of Ranvier - the gaps of myelin-free membrane segments located between two adjacent myelinated regions(Susuki & Rasband, 2008)such that transmembrane ionic flow is only possible through these sites. The conduction of electrical impulse through the channels located on these myelin-free regions of the axonal membrane is called saltatory. Therefore, the role of myelination is to increase the speed of electrical impulse conduction, which, in turn, leads to a decrease in reaction time and increase in information processing speed.



Figure 1 - Scheme of a neuron. Black lines indicate the organelles and functional regions of the cell

White matter accounts for more than half of the total brain volume(Fields, 2010). It is composed of myelinated nerve fibers that link neurons in distributed brain areas. White matter is involved in complex brain processes and it is directly associated with cognitive(Desmond, 2002; Ohlhauser et al., 2018), sensory(Chang et al., 2016; Pryweller et al., 2014), and motor functions(Fleischman et al., 2015; Hollund et al., 2017; Sampaio-Baptista et al., 2013).Myelination of nerve fibers is experience-dependent, and it changes with age and learning, and these changes are specific to the functionally distinct brain regions(Fields, 2010; Filley & Fields, 2016). For example, brain regions involved in verbal working memory (i.e., maintaining and manipulating information) and verbal fluency (i.e., producing words in a specified category), and connected via the superior longitudinal fasciculus (SLF), inferior longitudinal fasciculus (ILF), inferior Fronto-Occipital Fasciculus (IFOF), and anterior thalamic radiation (AT), demonstrate a significant increase in DTI indices of white matter integrity with age(Peters et al., 2012).

Myelin sheath that insulates the axons is essential for fast transmission of the electrical signals that go along the nerve fibers(McDougall et al., 2018). The degree of myelination has been assessed with various post-mortem and invivo methods. Although there are many different techniques which allow in vivo investigation of the brain white matter, each method shows high degree of specificity to tissues and sensitivity(Hagiwara et al., 2018).



Chapter 2. Methods of Myelination Assessment

The classic methods used for myelination assessment and investigation of white matter integrity was postmortem histological staining with hematoxylin/eosin(Flechsig, 1901; Yakovlev & Lecours, 1967), the Luxol fast blue technique (Bodhireddy et al., 1994), and subsequent immunohistochemical staining for myelin basic protein and myelin-associated glycoprotein(Itoyama et al., 1980). Critically, these methods cannot be used for in vivo studies. Non-invasive magnetic resonance imaging (MRI) techniques that indirectly assess myelination indices have been proposed.

2.1 Introduction to the Principles of Magnetic Resonance Imaging

2.1.1 T1/T2 and T2*-weighted imaging

Structural MRI is based on the magnetization properties of the nuclei in the atoms. The hydrogen atoms (H+) of water or fat molecules in the tissues are characterized by spins (magnetic moments), which are normally randomly oriented in space. Application of a uniform external magnetic field (B0)excites protons, aligns them along the direction of the field, andcauses precession of the protons around their axis. The rate of precession is proportional to the strength of the external magnetic field, which in most clinical MRI scanners is 1.5 to 3 Tesla (T). Then, an electromagnetic pulse is applied at the frequency corresponding to the precession rate (radiofrequency pulse, RF), which results in deflection of the net magnetization vector(Berger, 2002; Paus et al., 2001). When the source of RF is removed, the nuclei return to their resting state (with the magnetization vector pointing along the B0) through the relaxation processes accompanied by energy emission that can be measured using a receiver coil. Spatial encoding of the MRI signal is achieved by applying gradient fields which distort the main magnetic field B0 resulting in variation of the resonance frequency of protons as a function of position along the gradient direction(Tobon, 2010).

The frequency information of the signal from each area in the imagined plane is then converted into the intensity using Fourier transformation. To build up the cross-sectional images, signal intensity is displayed on a gray scale arranged as a matrix of pixels. Different types of images may be obtained by varying the sequence of transmitted RF pulses introduced and then collected after theRF pulse is switched off. The image type and the target RF sequence depend on the type of tissue being examined: different tissues are characterized by different proton density (the number of hydrogen nuclei per unit of tissue volume) and, therefore, relax at different rates. Thus, the contrast in structural MRI images is created based on the difference in time needed for the protons to relax - the relaxation time. The longitudinal relaxation time (T1, spin-lattice relaxation) reflects the time taken for the vector of total magnetization to recover (return to its maximum value) and spinning protons to return to equilibrium and realign with the external magnetic field so that it becomes parallel to B0. This recovery is exponential, and it depends on the structural pattern of the surrounding environment (lattice) of each hydrogen nucleus since the energy of relaxation is transferred into nearby nuclei, atoms, and molecules(McRobbie, 2006). Therefore, a shorter T1 will correspond to the more structured tissues. The transverse relaxation time (T2, spin-spin relaxation time) determines the decay of transverse magnetization - the rate at which the spinning protons go out of phase with each other, i.e. losephase coherence among the nuclei spinning perpendicularly to B0(Figure 2). The transverse relaxation is related to the intrinsic field caused by the nearby protons. The individual nuclei within the tissue of interest precess in the transverse plane at slightly different rates so that their magnetic moments point to different directions and randomly interact with each other at the atomic or molecular levels leading to the transverse decay of the MR signal and irreversible dephasing of the transverse magnetization. Inhomogeneities of the main magnetic field also cause an irreversible field dephasing which is characterized by T2* relaxation time. T2 relaxation describes the true, predicted T2 of the tissue being examined, whereas T2* is referred to as observed relaxation since in real situations the decay of transverse magnetization is faster than predicted. Therefore, T2* is always smaller than or equal to T2.

Figure 2 - Exponential growth and decay of the longitudinal and transverse magnetization vectors respectively. T1 reflects the amount of time required for longitudinal magnetization (Mz) to achieve (1-1/e) or about 63% of M0 initial value. T2 reflects the amount of time required for transverse magnetization(Mxy) to decrease in e(~2.7) times

A short T1 corresponds to the tissues that recover faster than tissues with a long T1 and are characterized by greater Mz values. They appear brighter on T1-weightedimages. A short T2 corresponds to the tissues in which the signal decays very fast. The tissues with short T2 and are characterized by smaller values of the MR signals and appear darker on T2-weighted images than areas with long T2.

The MRI signal is measured repeatedly with a repetition time (TR) expressed as the amount of time between consecutive pulse sequences applied to the same slice of tissue at one time point - time-to-echo (TE) expressed as time between the introduction of the RF pulse and the receipt of the echo signal(Paus et al., 2001). Different tissues are characterized by different T1 and T2.

Differences in relaxation times result in image contrast. T1-weighted and T2-weighted scans are the MRI sequences most frequently used in structural studies since they clearly represent contrast between the tissues, especially in the pathological areas characterized by increased water content(Gideon et al., 1999; Tornheim & McLaurin, 1981). To obtain T1-weighted images, short TE and TR times are used. In contrast, T2-weighted images are obtained using longer TE and TR times. The ratio of the T1/T2 times is determined by the water content in a given tissue. High water content corresponds to a longer T1 and lower signal on a T1-weighted image. Cerebrospinal fluid (CSF) is characterized by the highest water content and, therefore, the longest T1: it is dark on T1-weighted scans and bright on T2-weighted scans. The lowest water content and shortest T1 are characteristic for white matter, which will appear bright on T1-weighted images and dark grey on T2-weighted images. The axons that are tightly bound will have even shorter T1 since such arrangement affects the interstitial water content. In addition, myelin lipids that comprise the white matter are also involved in magnetic interactions and they also influence T1. Conversely, gray matter has long T2 and it is bright (light grey) on T2-weighted images. T2 also depends on the iron content in the tissue since it may influence local inhomogeneities of the magnetic field: the higher the iron level, the shorter T2(Paus et al., 2001).

In clinical settings, a fluid attenuated inversion recovery (FLAIR) sequence is used to visualize pathologies and differentiate between cerebrospinal fluid, which appears dark on the scan and abnormal tissues which remain bright. The FLAIR sequence is analogous to T2-weighted images, but it uses very long TE and TR times(McRobbie, 2006). In the FLAIR image, CSF appears dark, white matter is dark grey, fat is light, and the cerebral cortex is light gray. Pathology, including inflammation, infection, and demyelination, would appear dark on T1-weighted, and bright on T2-weighted and FLAIR images (McRobbie, 2006).

T1/T2-weighted imaging is widely used in clinical studies. However, it gives only qualitative estimation of brain tissues for identification of pathology, such as tumors, trauma, inflammation, anddemyelinating lesions as the resolution of the images obtained is often limited. In addition, the indices of T1/T2-weighted imaging are weakly associated with actual myelin content determined using histological staining(Hagiwara et al., 2018). Several techniques have been developed and applied to visualize fiber tracts of interest and obtain qualitative characteristics of white matter structure.



2.1.2 Functional Magnetic Resonance Imaging (MRI)

Functional MRI (fMRI) is not a measure of myelination. As it is a popular MRI protocol, it is included as a part of the section for principles of MRI. fMRI measures task-induced or spontaneous (during resting state) changes in deoxyhemoglobin level in cerebral blood. fMRI allows indirect evaluation of neuronal activity - the input and subsequent processing of the neuronal information within a given area(Logothetis, 2002), by measuring the Blood Oxygenation Level Dependent (BOLD) signal, which reflects the changes in blood flow and blood oxygenation(Glover, 2011). Brain activity, including the maintenance of ionic gradients, formation, and propagation of action potentials, neurotransmitter release, synthesis, and transportation, as well as functioning and activity of the ionic transporters and other cellular structures, requires energy supply. The energy within the cell is synthesized in the form of adenosine triphosphate (ATP) by an enzyme called adenosine triphosphate synthase located in the inner membrane of the mitochondria, in the process of glycolytic oxygenation of glucose(Chaudhry & Varacallo, 2020). The oxygen is delivered via a special protein within the red blood cells - hemoglobin. When hemoglobin binds oxygen it becomes oxyhemoglobin, and when it loses oxygen it becomes deoxyhemoglobin. Oxyhemoglobin is diamagnetic, i.e. it has zero magnetic moment, whereas deoxyhemoglobin is paramagnetic, and it changes the magnetic susceptibility of blood resulting in local magnetic field distortions and a change in the net MRI signal(Buxton et al., 2014).

2.2 Diffusion MRI, or Diffusion Tensor Imaging (DTI)

DTI is a neuroimaging technique that is used to assess microstructure and orientation of myelinated white matter tracts in the brain. Diffusion MRI allows for investigation of the microstructure of the tissues by measuring diffusion of water molecules within the tissue considering their interaction with biological membranes and macromolecules.

In biological tissues, water exists in two main forms - free and bound to other molecules, including proteins(Kerch, 2018). Free water molecules move stochastically in the solution, colliding with each other and diffusing from the regions of higher water concentration to the regions of lower concentration. Einstein(1956) explained the phenomenon of diffusion proposed by Adolf Fick in the context of Brownian motion defined as the thermally-driven irregular motion of particles dissolved in a liquid medium caused by collision with other molecules(Beaulieu, 2002; Jakobsson & Chiu, 1988). In a homogenous medium, such as a glass of water or ventricles in the brain, diffusion is isotropic (i.e., equal in all directions). In contrast, within the heterogeneous medium, which includes various cellular and subcellular structures and compartments that may impede free movement of water, diffusion is anisotropic , i.e. the motion of water molecules to one direction is greater than the motion to other directions(Basser & Jones, 2002; Roberts et al., 2013). This form of diffusion depends on tissue orientation and structure.

Diffusion of the molecules can be characterized by the diffusion coefficient, D, (mm2/s), which links diffusive flux and concentration gradient(Doran et al., 1996; Johansen-Berg & Behrens, 2009). Diffusion of the molecules in a solution may be influenced by intermolecular interactions, temperature, molecular weight, and various active processes within the tissue(Beaulieu, 2002). Isotropic diffusion is described by a single apparent diffusion coefficient which characterizes the interaction between the diffusing molecules and cellular structures(Basser& Jones, 2002, Beaulieu, 2002). To describe anisotropic diffusion, a diffusion tensormay be used instead of a scalar diffusion coefficient(Basser & Jones, 2002). In the brain white matter, especially in highly myelinated regions, water diffusion is anisotropic (i.e. it is restricted by axonal membranes; Beaulieu, 2002). Thus, the main direction of anisotropic water diffusion within white matter reflects fiber orientation and characteristics of water diffusion may be used to determine structural characteristics of a given brain region.

Anisotropic diffusion of water molecules in tissues is described by a diffusion tensor which replaces the scalar diffusion coefficient(Basser & Jones, 2002). The displacement of the molecules is modeled using three-dimensional Gaussian distribution(O'Donnell & Westin, 2011). The diffusion tensor model is fitted using multi-linear regression; to estimate diffusion tensor, diffusion gradients must be applied at least along six noncollinear, noncoplanar directions(Basser et al., 1994). The diffusion tensor determines three orthogonal eigenvectors and associated eigenvalues (л1, л2, л3), which describe the diffusivity in the direction of each eigenvector. The direction of the main eigenvector corresponds to the direction of the fastest diffusion and determines the axis of a given fiber tract in the regions where the white matter tracts do not branch and cross (O'Donnell & Westin, 2011). Diffusion tensor may be geometrically represented as an ellipsoid whose axes are of the length and are aligned with the eigenvectors (Basser et al., 1994). The eigenvalues of the diffusion tensor can be used to calculate the widely-used diffusion parameters, including 1) the mean diffusivity (MD), which reflects the total diffusion in a voxel, 2) longitudinal or axial diffusivity (, assuming л1?л2?л3?0), which is equal to the eigenvalue corresponding to the longest axis, 3) transverse or radial diffusivity reflecting the diffusion in the transverse plane), and 4) fractional anisotropy (FA) which characterizes the eccentricity of the diffusion ellipsoid and provides quantitative measure of the shape of the diffusion ranging from 0 to 1 (Basser& Jones, 2002; O'Donnell & Westin, 2011). MD is an inverse measure of membrane density and spacing between axons, and it reflects the directionally invariant overall diffusion rate (Beaulieu, 2014). Notably, MD is similar in white and gray matter (Basser & Pierpaoli, 1996; Beaulieu, 2014). FA reflects the difference in the shape of the tensor ellipsoid between a perfect sphere and an ellipsoid that describes anisotropic diffusion. Therefore, an increase in FA in a given region of the brain reflects an increase in size and density of axons, whereas a decline in radial diffusivity also corresponds to higher axonal myelination (Beaulieu, 2014). Mathematically, FA is expressed as a normalized variance of the eigenvalues of the diffusion tensor matrix:



		(1)

The diffusion of water molecules is measured using the pulsed-gradient, spin echo pulse sequence with a single-shot, echo planar imaging(Alexander et al., 2007; Stejskal & Tanner, 1965; Westin et al., 2002). The diffusion coefficient in a given directioncan be estimated by measuring the attenuation of signal intensity (Sk) caused by the phase dispersion in relation to the original signal (S0) according to the Stejskal-Tanner equation (Figure 3;Basser, 1995):

		(2)

where S0: original intensity of a signal measured without diffusion-sensitizing gradient;

Si: intensity of diffusion-weighted signal after application of diffusion-sensitizing gradient in the direction i;

b: diffusion-sensitizing “gradient factor” that depends on the gyromagnetic ratio, the gradient strength, and pulse sequence.

A system of equations solved for D is used to estimate the diffusion tensor from a set of diffusion-weighted images. Since the number of independent coefficients in tensor (i.e., the number of degrees of freedom) is 6, a minimum of 6 diffusion-weighted images (one for each direction of the gradient pulses) are required and one additional baseline image without diffusion-sensitizing gradients S0 (i.e., the original signal intensity).

The value ofthe Le Bihan's factor b(Le Bihan, 1991) or diffusion-sensitizing “gradient factor” is a function of the gyromagnetic ratio for a given nucleus. It describes the gradient strength and pulse sequence (i.e., timings of the diffusion-sensitizing gradients; Alexander et al., 2007; Beaulieu, 2002; O'Donnell & Westin, 2011). The b factor is calculated as follows (Westin et al., 2002):

		(3)

where г: gyromagnetic ratio of the nucleus of interest (for hydrogen nucleus, г = MHz/Tesla);

G: strength of the diffusion sensitizing gradient pulses;

д: duration of the diffusion gradient pulses;

Д: time between diffusion gradient RF pulses.

Therefore, the values of b factor summarize the attenuating effect of all diffusion gradients in one direction in the resulting MR signal (Figure 3). Based on the b-values, a symmetric b-matrix is constructed for each diffusion-weighted image. The b-matrix summarizes the attenuating effect of all gradients introduced in three directions and is used to assess the diffusion tensor using multivariate linear regression(Basser et al., 1994; Basser & Jones, 2002; Mattiello et al., 1997).

Figure 3 - (A) The scheme of the Stejskal-Tanner spin-echo diffusion weighted imaging sequence (adapted from Mori & Zhang, 2006). Abbreviations: TR - repetition time; TE - time-to-echo. (B) The parameters of the diffusion weighted acquisition sequence (adapted fromWestin et al., 2002). g - the strength of the diffusion sensitizing gradient pulses; д - the duration of the gradients; - the time between diffusion gradient pulses

The orientation of the main eigenvector may be visualized using color-coding in which blue color corresponds to superior to inferior direction, red color reflects left to right orientation, and green denotes anterior to posterior direction(Pajevic & Pierpaoli, 1999). FA values are usually reflected by the brightness of the color (O'Donnell & Westin, 2011). The trajectories of the fiber tracts are assessed using DT-MRI fiber tractography(Basser et al., 2000)given that the orientation of the major eigenvector of the diffusion tensor is parallel to the fiber tracts. Tractography starts at a “seed” point, that is a specified location at which the direction of the principal eigenvector is measured followed by moving a small fixed distance (? 1mm) in that direction (tract integration). Evaluation of the fiber orientation is repeated and re-evaluated for successive small steps until the tract is terminated(Alexander et al., 2007). DTI-based tractography has a number of limitations, including the susceptibility of estimates of eigenvector directions to thermal noise, physiologic fluctuations, and image artifacts. In addition, this method is based on the assessment of the principal eigenvector, thus making it impossible to study the direction of the fiber tracts that branch or cross (Alexander et al., 2007; O'Donnell & Westin, 2011).

2.3 Fast Macromolecular Proton Fraction (MPF) Mapping

Fast macromolecular proton fraction (MPF) mapping is a quantitative MRI technique used for clinical assessment of myelin content in brain tissues(Yarnykh & Yuan, 2004). MPF refers toa biophysical parameter, which describes the amount of macromolecular protons in tissues which are involved in magnetic cross-relaxation with free water protons; this results in the magnetization transfer (MT) effect(Eng et al., 1991; Wolff & Balaban, 1989).Magnetization transfer imaging is based on magnetization exchange between spins via chemical exchange or cross-relaxation(Edzes & Samulski, 1978; Fung, 1986)caused by spin diffusion - the quantum mechanical process by which spins can flip in a rigid soil medium and which is associated with coupling between spin-lattice relaxations of different nuclei, such as the macromolecular protons and the water protons, in the presence of motion(Derome, 1987). Due to ultra-short transverse relaxation time T2 of macromolecular protons and their lower mobility compared to liquid protons, relaxation time of macromolecular spins cannot be directly visualized using MRI(Henkelman et al., 2001). Coupling between the mobile water protons and protons associated with macromolecules results in exchange, i.e. magnetization transfer, which allows for the indirect imaging of these less mobile spins within the macromolecular proton fraction(Tozer et al., 2003).

The effect of cross relaxation in a heterogenous biological system is traditionally studied in the context of a two-pool model, which implies that a given biological system is comprised of two pools - the water protons (w;i.e., free pool), and protons associated with macromolecules (m; i.e., bound pool). These two pools differ in their transverse relaxation time, with the transverse relaxation time of macromolecular protons and protons associated with biological membranes to be much smaller (less than 1 ms) than that of water (greater than 10 ms). When two protons are close to each other, cross-relaxation between them may take place, resulting in simultaneous flips of spins via dipolar magnetic interaction. The spin-lattice relaxation in each pool is determined by a sum of two exponential decays which are, in turn, characterized by two apparent relaxation rates and (Edzes & Samulski, 1978). The time dependence of magnetization is described by the following equations:

	mi(t) = -(Mi(t) - Mi?)/2Mi? dmi/dt = - R1imi(t) - kimi(t) + kimj(t) mi(t) = exp(-t) + exp(-t)	(4)



where mi: reduced magnetization obtained from the normalization of the magnetization in each pool i;

Mi(t): magnetization at time t;

Mi?: equilibrium magnetization;

R1i : spin-lattice relaxation rate of the pool i in the absence of cross relaxation;

ki: the cross-relaxation rate constant defined for magnetization transfer from free (water) to bound (macromolecular fraction) pools;

and : normalized intensities of the two relaxation components in the relaxation curve of the protons in the pool i, which depend on initial magnetization mi(0) and mj(0);

pi: proton fraction in the i pool.

The equilibrium magnetization Mi? includes two components - the longitudinal magnetization of free (M0(1 - f)) and bound (M0f) pools, where f is the macromolecular proton fraction (MPF), i.e. the fraction of bound spins expressed via their concentration (f = [B]/ ([B] + [F]); the longitudinal relaxation rates R1iof free (R1F) and bound (R1B) spins can be expressed as R1F,B = 1/T1F,B , where T1F and T1B are longitudinal relaxation times of the free and bound pools respectively, and R1B can therefore be expressed as follows:

	R1B = R1 -	(5)

The effective (i.e., time-averaged) saturation rates WF,B of the two pools can be expressed as follows:

	WF,B = рщ21rmsgF,B(Д, T2F,B)	(6)

where щ1rms: saturation power, i.e., root-mean-square amplitude of the saturation pulse averaged over the pulse duration;

gF,B(Д, T2F,B): the absorption lineshapes of free and bound pools determined by the offset frequency Д and intrinsic T2 of each pool.

The saturation power щ1rms allows for the approximation of the time-dependent amplitude of a shaped saturation pulse and it is proportional to the effective flip angle of the saturation pulse (FAMT).

The transfer of saturation from the macromolecular spins (which are 106 time more sensitive to off-resonance irradiation than water spins) to spins of liquid pool occurs in response to exposure of the tissue to an off-resonance radio frequency pulse(Henkelman et al., 2001). The excitation of the proton spins bound to macromolecules by a RF pulse results in the transfer of a certain amount of their energy to the nearby proton spins, which is referred to as the magnetization transfer (MT) effect (Henkelman et al., 2001). Traditionally, the MT effect has been assessed using an MT ratio (MTR) which characterizes this effect based on measuring a relative decrease in the observed signal in saturation experiment with an off-resonance RF pulse (Yarnykh, 2012), and allows the evaluation of myelin content in white matter, although this index is heavily dependent on experimental conditions, including scanner type and scanning sequence (Tozer et al., 2003). In addition, due to the interaction with other physical parameters of a two-pool model, such as R1, which also correlates with myelin content and is sensitive to a number of physiological parameters, including iron and calcium content and axon count and size, the association between MTR and myelin content in tissue is nonlinear (Henkelman et al., 2001).

One of the quantitative two-pool model parameters which demonstrates strong correlation with myelin density is macromolecular proton fraction which reflects the total macromolecular content in neural tissue(Yarnykh, 2012). In addition, the MPF allows the assessment of a relative amount of protons that demonstrate semi-solid molecular motion and are involved in magnetic cross-relaxation with water spins (Yarnykh, 2016). Whole-brain MPF mapping can be performed via a single-point approach using one off-resonance saturation data point (i.e., single MT-weighted image; Yarnykh, 2012). A recent study by Yarnykh (2016) suggests a new high-resolution whole-brain MPF mapping method which is based on the single-point approach described above and allows for the reduction of scanning time and utilization of only a small number (n=3) of source images.

Intensity of the signal measured after the introduction of an off-resonance saturation pulse depends on the offset frequency and power of the saturation pulse (pulsed Z-spectra) and is proportional to the longitudinal magnetization of the free pool before an excitation pulse (MZF):

	SMT = MZFexp(-TE/T2*)sinб	(7)

The SMT/Srefratio, where Srefis the intensity obtained using the same sequence without off-resonance saturation, defines a normalized pulsed Z-spectrum and allows for the elimination of the factors associated with coil sensitivity, M0, and T2*:

	mz(Д,FAMT) = SMT/Sref= (WF(Д, FAMT), WB(Д, FAMT))/WF=0, WB=0)	(8)

After normalization of pulsed Z-spectrum to the intensity measured using the same sequence but without off-resonance saturation and exclusion of the longitudinal relaxation rates R1F,B , assuming that the relaxation rates of the two pools are equal to one another and to the observed rate R1 (R1F= R1B = R1), four adjustable parameters f, k, T2F, and T2B are left in the model. The two-pool model parameters can be therefore determined using the parameter fitting approach. To reduce the number of measurements, several model parameters can be fixed in order to fit the other set of parameters. As such, assuming that T2B is constant, T2F is constrained by the constant product of two other constants R1F T2F, and the forward rate constant k is constrained by a fixed value of the reverse rate constant determined for magnetization transfer from the bound pool to the free pool, the MPF parameter f can be determined from the single-parameter fit:

	R = k(1 - f)/f	(9)

Using a single off-resonance measurement and the constraints described above, f can also be determined by a non-linear iterative method.

The data can be processed in several ways, but the final goal of image processing is the reconstruction of MPF maps. The first step is the reconstruction of the maps of the model parameters f, k, T2F, and T2B from the four-parameter fit of the pulsed MT model using all Z-spectroscopic data points obtained for each subject and a complementary R1 map.In addition, based on the maps of primary fitted parameters, the maps of the R1T2F and R= k(1 - f)/fshould also bereconstructed. The next step is the reconstruction of the histograms of R, R1T2F, and T2Bfor the scanned brain volume and determination of their values by calculating the average of histogram median values across all participants (Yarnykh, 2016).

The single-point MPF mapping method implies reconstruction of f maps for each combination of Д and FAMT by iterative solving the pulsed MT matrix equation by the Gauss-Newton method with fixed standardized values of non-adjustable parameters of the two-pool model (R, R1T2F, and T2B) and corrections for inhomogeneities of the B1 (the radiofrequency field which is perpendicular to the main magnetic field B0) and B0 fields using R1 maps obtained with the four-parameter fit approach. Therefore, to measure MPF by a single-point method, the following need to be acquired: a spoiled gradient echo (GRE) MT-weighted image with off-resonance saturation; a reference GRE image without off-resonance saturation which will be used to normalize the data; R1 (1/T1) map acquired independently by the two-point variable flip angle approach (Deoni et al., 2003)which requires T1-weighted and a proton density (PD)-weighted spoiled GRE images. R1 and PD maps reconstructed from two FA images with B1 correction will then be used to produce a synthetic reference GRE image. The final step is the reconstruction of a MPF map from an MT-weighted image normalized to the synthetic GRE reference image and an R1 map using B0 and B1 corrections(Henkelman et al., 2001; Yarnykh, 2012, 2016; Yarnykh & Yuan, 2004).

To date, neuroimaging has become one of the most effective and promising tools to study brain function and structure. MR-based techniques are currently the only methods which allow non-invasive in vivo investigation of white matter microstructure. DTI has been widely used in cognitive and psychophysiological studies(Basser & Pierpaoli, 1996; Jang et al., 2016; Kantarci et al., 2011; Moseley et al., 2002). MPF, a novel MRI sequence, demonstrated its efficiency in assessing myelination pathologies, including demyelination in multiple sclerosis and traumatic brain injury(Yarnykh et al., 2018; Yarnykh, 2002). However, there are no available studies assessing individual age- and gender-related variations in myelination of fibers associated with complex cognitive functions using these two methods in the same participants. Importantly, this is the first investigation to compare these methods as a function of age, in school-age children, adolescents and adults.



Chapter 3. Individual Differences in Myelination

3.1 Myelination and age

The process of myelination follows a complex topographical and temporal pattern across development (Figure 4) as demonstrated by histological studies(Yakovlev & Lecours, 1967).The changes in myelination are proposed to be synchronized among functionally related fiber networks(Valk & Knaap, 2013). The speed and timing with which a region myelinates during ontogenesis reflects its position in the functional hierarchy (Yakovlev & Lecours, 1967). For example, myelination of the motor root fibers is faster and shorter than that of sensory roots(Nelson et al., 2001).

Figure 4 - Myelination cycle (adapted from Yakovlev and Lecours, 1967)

Chronological pattern of nerve fiber myelination in the CNS. The approximate age range within which myelination of a given fiber tracts terminates is indicated in thick black arrows. Extending ends of the bars correspond to increasing staining intensity and density of myelinated nerve fibers.

The most rapid growth of the brain occurs during the first three years of life(Dekaban, 1978). An MRI study(Hedman et al., 2011)shows that, between 10 and 12 years of age, brain weight reaches its adult values. During childhood, an increase in white matter volume exceeds that of the grey matter(Groeschel et al., 2010). In DTI studies, white matter maturation is traditionally examined by assessing the changes in several metrics, including AD and FA, which increase with age, and RD and MD, which are known to decrease as myelin thickness increases (Tamnes et al., 2018). Findings on age-related changes in DTI metrics, such as radial diffusivity, which reflects the rate of water diffusion in the direction perpendicular to the main axis of the axon (Johansen-Berg & Behrens, 2009; Kumar et al., 2013), indicate that the most active maturation of the association and projection pathways which maintain cortical and brainstem integration occurs during adolescence, when an pronounced increase in reaction time is observed, and continues into late adolescence and adulthood (Asato et al., 2010). This period of white mater maturation is accompanied by the maturation of a number of critical cognitive functions, including executive control and interhemispheric communication (Asato et al., 2010).The majority of fiber tracts, including the corpus callosum, superior longitudinal fasciculus, inferior longitudinal fasciculus, superior and inferior fronto-occipital fasciculi, uncinate fasciculus, and cingulum exhibit nonlinear pattern of maturation, as indicated by nonlinear (quadratic) increases in FA and MD (Tamnes et al., 2018). Early development is characteristic for the CC and fornix, whereas pathways which form frontal-temporal connections, including cingulum, uncinate fasciculus, and superior longitudinal fasciculus, demonstrate a more protracted maturation cycle (Lebel et al., 2012).

Critically, there isno period in the development when white matter microstructure in the brain is static. During childhood, adolescence, and early adulthood, white matter maturation is fast and dynamic and shows pronounced increase in fiber density and myelination as indicated by an increase in FA and decrease in MD and RD; during mid-adulthood, white matter structure is relatively stable, and in late adulthood, changes in white matter integrity become accelerating again, but they reflect degenerative processes, including decline in myelination which, which may be identified by a sharp reduction of the FA and an increase in MD and RD (Lebel et al., 2012; Schmithorst et al., 2002; Sexton et al., 2014; Westlye et al., 2010).

In an adult brain, the following three myeloarchitectonic regions, which myelinate with different cycles as distinct tectogenetic units are identified: a median periventricular zone, the hippocampus, and the hippocampal rudiment and septal area; a paramedian (limbic) zone; a supralimbic zone. Myelination of the fornix, as well as septal region, olfactory trigone, the periventricular gray, hypothalamus, and the median thalamus, begins in the fourth postnatal month and terminates within the second or third years of postnatal life(Yakovlev & Lecours, 1967).

It is important to highlight that the normative values for DTI indices may differ across studies. For example, some studies report normative FA in the CC that range between 0.74-0.78, depending on the region of the tract(Kim et al., 2008), whereas others provide FA values in CC ranging from 0.40 to 0.65 (Bengtsson et al., 2005; Johansen-Berg & Behrens, 2009). Moreover, different imaging methods may provide different estimates of white matter microstructure (Scholz et al., 2009). For example, some DTI studies show thatthe CC is more myelinated in males than in females (Shin et al., 2005; Westerhausen et al., 2003), while T1-weighted imaging demonstrates and opposite pattern (Scholz et al., 2009). Notably, values of DTI metrics may also be affected by selection of the plane in which DTI was obtained, in turn, affects thickness and size of the region of interest selected for analysis (Kim et al., 2008). Taking into account the disagreements which exist in the literature regarding the variations in microstructure of the CC, as well as the effects of age and gender on maturation of different parts of the CC, this fiber tract will be considered in the present study. Also, there may be inconsistencies in identifying periods of white matter maturation. For example, some earlier studies show that the peak of myelination in the ILF occurs at 11 years of age (Lebel et al., 2008), whereas recent findings demonstrate that the maturation of this fiber tract continues until age of 24 years (Lebel & Beaulieu, 2011; Lebel & Deoni, 2018). Therefore, it is still not clear when maturation of the ILF is complete and to what degree the rate of its maturation and myelination can vary. Maturation pattern in some fiber tracts is still poorly understood due to disagreements which exist regarding anatomical structure of the tract. A comprehensive study by Wu and colleagues (2016) suggest that the inferior fronto-occipital fasciculus (IFOF), which is one of the major association white matter pathways in the brain, has at least five subcomponents characterized by different connectivity patterns, albeit other researchers divide this fasciculus into three (Conner et al., 2018) or two subdivisions (Martino et al., 2010), or consider it as a whole tract (Rofes et al., 2014).

It is also important to emphasize that different studies use different approaches for setting age limits for each age group, and even if the groups are similar, normative values may differ among studies. As a result, there are still disagreements regarding maturation patterns in various white matter pathways, which makes the study of the individual differences in nerve fiber myelination very important for developmental biology and neuroscience, making a great contribution to existing theories of brain development. The next section discusses differences in white matter maturation and structure which are associated with gender.

3.2 Myelination and Gender

Histological methods greatly contributed to the understanding of temporal pattern and regional specificity of the white matter, however, data on gender specificity of myelination pattern have been mostly obtained using MR-based techniques. Although, male and female brains show many similarities, some studies highlight differences between the genders (e.g.,Kanaan et al., 2012; Xin et al., 2019), yet others suggest that these differences are driven by environmental and cultural factors(Jдncke, 2018), and that the two genders only differ in prevalence of certain rare types of brains(Joel et al., 2018).

Analysis of FA maps from 1065 young healthy adult individuals using a 3D Convolutional Neural Network method shows sex differences in FA in the whole-brain range(Xin et al., 2019), and the proton density and T2-weighted MRI data corrected for total intracranial volume demonstrate that females exhibit a higher proportion of gray matter, while males are characterized by a greater percentage of total white matter and cerebrospinal fluid(Gur et al., 1999), albeit some studies indicate that these differences are insignificant(Bourisly et al., 2017). Critically, the absolute differences in brain size and regional differences in the proportions of gray and white matter between males and females mostly correspond to differences in white matter volume(Allen et al., 2003; Passe et al., 1997; Schmithorst et al., 2008). In addition, after correcting for total cranial volume, no significant differences in the volumes of amygdala, hippocampus, and the dorsal prefrontal cortex between men and women are reported, although adult females are found to have significantly greater volume of orbitofrontal cortical region(Gur et al., 2002).

Several MRI studies(De Bellis et al., 2001; den Braber et al., 2013; Ritchie et al., 2018)demonstrate that the pattern of white and gray matter maturation in certain brain regions differs in males and females. The total volumes of the cerebral gray and white matter and corpus callosum show a pronounced sex-age interaction with males exhibiting a more significant age-related reduction of the gray matter and an increase in the white matter volume and corpus callosal region (De Bellis et al., 2001). According to the results of voxel-based morphometry analysis carried out by Bourislyand colleagues(2017), white matter volume in the frontal, temporal, parietal, occipital, and insular regions of the brain is larger in men compared to women, and only white matter of the postcentral gyrus in the right parietal lobe is greater in women compared to men. However, the authors of the study did not control for the total brain volume, although they claim that these differences cannot be explained by an assumption that the brain volume of men is, in general, bigger than that of women. The data on gender differences in the white matter in frontal regions are contradictory. A study by Szeszkoand colleagues(2003)reports that adult women have higher FA values in the left frontal lobe compared to men. In addition, females exhibit a leftward frontal asymmetry in FA (i.e., higher FA values in the left hemisphere), which correlates with a better comprehension of verbal constructions and memory functioning in women, whereas males show no such asymmetry. These differences may be explained by the fact that the maturation of the white matter in the frontal lobes continues into late adulthood with a peak at about 45-50 years of age(Bartzokis et al., 2001; Schmithorst et al., 2008). A DTI study by Kanaan and colleagues(2012)found that adult males have higher FA values in the cerebellum and at the anterior portion of the left superior longitudinal fasciculus, whereas females show greater FA values in the corpus callosum, interpreted to contribute to better interhemispheric communication and intellectual performance(Luders et al., 2007).

An MRI study of healthy children and adolescents (6-17 years; Blanton et al., 2004) shows that boys exhibit significant right>left asymmetry in the total cerebral volume, total cerebral white matter, and white matter of the middle and superior frontal gyri, whereas girls are characterized by the same asymmetry pattern in the total cerebral white matter and white matter of the superior frontal gyrus. In addition, white matter in the left inferior frontal gyrus shows age-dependent increase in boys, but not in girls. The authors speculate that these differences may be associated with gender effects in the development of speech and language lateralization. A DTI study by Schmithorst and colleagues(2008)shows that girls (mean age ~12 years) have higher FA in the splenium of the corpus callosum, whereas boys are characterized by greater FA values in bilateral frontal white matter, in the right arcuate fasciculus, and in the left parietal and occipito-parietal white matter. Interestingly, in the left frontal lobe, boys exhibit a positive association between age and FA values and girls are characterized by a negative correlation between age and FA. In contrast, in the right arcuate fasciculus, the FA indices positively correlate with age in girls and negatively in boys. In the right frontal lobe and right occipito-temporo-parietal white matter girls exhibit a positive association between FA values and age, whereas no significant correlation is found in boys.

...