In vivo histology of the myelin g-ratio with magnetic resonance imaging
Graphical abstract
Introduction
Myelin is a dielectric material that wraps around axons in the nervous system to provide efficient conduction of neural signals. Larger axons and a thicker myelin sheath contribute to faster conduction, but there is a trade-off between axon size and myelin thickness due to spatial constraints imposed by the brain dimensions. The relationship between axon size and myelin thickness is captured in a parameter called the myelin g-ratio, defined as the ratio of the inner (axon) to the outer (axon plus myelin) diameter of the fiber. The g-ratio that provides maximum conduction speeds has a broad optimal value around 0.6 (Rushton, 1951, Waxman, 1975).
In general, larger axons tend to have thicker myelin sheaths, but the relationship between the two is not linear (Berthold et al., 1983). Measurements across different vertebrate species indicate that the linear relationship breaks down for axons greater than 5 μm (Hildebrand and Hahn, 1978). In human peripheral nerves the axon caliber reaches its maximum by year five, but the myelin thickness increases until about 16 years of age, resulting in a gradual decrease of the g-ratio during development (Schröder et al., 1988). Recent studies have suggested that the sexual dimorphism in myelin content in white matter development is characterized by a higher g-ratio (relatively thinner myelin) in adolescent boys (Perrin et al., 2009, Paus and Toro, 2009).
In addition to changes in the myelin thickness during development, variations in the g-ratio have also been associated with cognition and disease (see Fields (2008) for a review). In particular, in multiple sclerosis remyelinated lesions are characterized by thinner myelin sheaths and a higher g-ratio (Albert et al., 2007). In vivo imaging of the myelin thickness in multiple sclerosis could provide a real-time tool for tracking myelination in lesions, facilitating the development and evaluation of new therapeutic agents that promote remyelination.
The g-ratio is an important parameter to measure with magnetic resonance imaging (MRI) because it provides information about microstructure not available from other imaging parameters. There are MRI protocols that are sensitive to the fiber volume fraction (FVF), but the FVF could vary widely, especially in disease, without the g-ratio changing. There are MRI protocols that are sensitive to myelin content, but myelin content would be expected to correlate with FVF to some extent. Computation of the g-ratio captures exactly how the myelin volume fraction changes with fiber volume fraction.
The g-ratio (Fig. 1) can be expressed as a function of the myelin volume fraction (MVF) and the axon volume fraction (AVF) or fiber volume fraction (FVF) (Stikov et al., 2011). Quantitative magnetic resonance imaging (qMRI) techniques such as magnetization transfer and diffusion imaging show great promise in measuring surrogates for the MVF and AVF. Given that there is a broad range of MRI biomarkers that could be used to infer the MVF and AVF, the following sections review MRI methods for quantitatively measuring myelin and axon content.
The myelin volume fraction is defined as the fraction of the voxel occupied by myelin sheaths. There is great interest in estimating the MVF, resulting in a number of MRI surrogate biomarkers with varying degrees of sensitivity and specificity to myelin. Some biomarkers are based on T1/T2 relaxometry measurements (Stüber et al., 2014, Mezer et al., 2013, MacKay et al., 2006, Whittall et al., 1997, MacKay et al., 1994). For example, the myelin water fraction (MWF) is sensitive to the short-T2 water trapped between the myelin sheaths, and has been correlated with absolute myelin content with histology in healthy (MacKay et al., 2006) and diseased tissue (Laule et al., 2006).
Unlike MWF, which is based on multicomponent T2 relaxometry, magnetization transfer (MT) sensitizes the MR sequence to hydrogen atoms bound to macromolecules, such as those found in myelin (Edzes and Samulski, 1977, Wolff and Balaban, 1989, Kucharczyk et al., 1994). The method is based on selective excitation of the hydrogen nuclei bound to macromolecules, which then interact with the hydrogen in free water, effectively reducing its net magnetization. MT reduces the free compartment magnetization wherever there is a significant macromolecular concentration. As only the free compartment can be directly imaged, the result is a unique type of contrast that makes regions with high macromolecular content appear darker than those with low macromolecular content.
The MT effect can be quantified using the magnetization transfer ratio (MTR). The MTR is defined as the fraction of magnetization left after an MT pulse is applied. MTR has been correlated with histological myelin measurements in animal models (Brochet and Dousset, 1999) and humans (Filippi et al., 1995, Schmierer et al., 2004).
The MTR metric includes contributions from several MR parameters (Sled and Pike, 2001, Henkelman et al., 2001), such as the longitudinal relaxation time (T1), the exchange rate (k), and the fractional pool size (F), defined as the fraction of exchanging protons that are bound to macromolecules. To decouple the individual contributions of T1, k and F from the MTR, a more exact mathematical description of the MT process is needed. A quantitative model for magnetization transfer has been developed by Henkelman et al. (1993). In this model, the evolution of the free and the bound pool magnetization is represented by a linear system with several tissue dependent and pulse sequence dependent parameters.
Based on this model, a number of quantitative magnetization transfer (qMT) methods have been proposed (Sled and Pike, 2001, Ramani et al., 2002, Yarnykh, 2002, Cercignani et al., 2005, Gochberg and Gore, 2007, Gloor et al., 2008). While the methodology and terminology used in the above methods differ significantly, their goal is similar — to determine the proportion of protons that are bound to macromolecules, and to quantify the rate of exchange of magnetization between the bound and free protons. Ex vivo histological validation has shown excellent specificity to myelin content, suggesting a linear relationship between F and MVF (Schmierer et al., 2008, Thiessen et al., 2013). Hence, an estimate of the MVF can be calculated from F using a coefficient of proportionality obtained from histological investigation. We denote this estimate MVFMRI.
There are numerous MRI acquisition and modeling approaches that can be used to estimate the axon volume fraction. The MRI signal comprises contributions from multiple tissue compartments. These include the free water (cerebrospinal fluid (CSF)), the intra-cellular compartments, including the intra-axonal compartment, the extra-cellular compartments, and the water between the myelin bilayers. The MRI signal can also be indirectly sensitive to the myelin itself, as described in the previous section. Diffusion MRI is well suited to probe the intra- and extra-cellular spaces and CSF compartments. Diffusion MRI is sensitive to the random thermal motion of water molecules. These water molecules impinge on cellular membranes, intracellular organelles, neurofilaments, and myelin. The diffusion displacement distribution therefore contains information about all of these structures.
Diffusion imaging alone is not sufficient to measure the axon volume fraction. There is negligible signal from myelin in typical diffusion MRI acquisitions. The transverse relaxation time T2 of myelin water is short (~ 10 ms (Whittall et al., 1997)) and the echo time necessary to achieve sufficient diffusion sensitization is long (~ 100 ms). As described above, the myelin water can be probed with techniques such as multicomponent T2 imaging (MacKay et al., 1994), and the myelin macromolecular structure can be probed by quantitative magnetization transfer (Henkelman et al., 1993). The lack of signal from the myelin compartment in diffusion imaging means that estimation of true volume fractions of the other compartments is difficult. Diffusion imaging itself allows us to probe the fraction of the non-myelin volume that is intra-axonal. The combination of myelin imaging and diffusion imaging makes axon volume fraction measurement possible, and this will be described below.
In recent years, many diffusion imaging and post-processing techniques have been developed to infer information about the tissue compartments from the diffusion signal. These techniques include the now-standard diffusion tensor model (Basser et al., 1994), ball-and-stick models (Behrens et al., 2007), “diffusion basis spectrum imaging” (Wang et al., 2011), “restriction spectrum imaging” (White et al., 2013), the “composite hindered and restricted model of diffusion” (CHARMED) (Assaf and Basser, 2005), “neurite orientation dispersion and density imaging” (NODDI) (Zhang et al., 2012), and double pulsed field gradient (dPFG) techniques (Zhou et al., 2013, Shemesh et al., 2012).
The details of the acquisition and modeling in these diffusion imaging techniques for probing the non-myelin tissue compartments vary. As a starting point, diffusion can be either restricted or hindered in tissue, with restricted diffusion meaning there is a sharp falloff in the probability of displacement at a certain radius, and hindered meaning the diffusion profile is roughly Gaussian radially. Both restricted and hindered diffusion have a lower mean square displacement than free diffusion, and in both cases the displacement distribution can be either isotropic or anisotropic. In diffusion MRI modeling, restricted diffusion is usually attributed to the intra-cellular compartment and hindered diffusion to the extra-cellular compartment. Free water is neither restricted nor hindered. Due to the highly organized structure of white matter axons, the intra-axonal compartment is anisotropically restricted.
The diffusion tensor is one of the simplest diffusion signal models, wherein all compartments are amalgamated into one potentially anisotropic Gaussian diffusion displacement distribution. The anisotropy of this distribution is often characterized by the fractional anisotropy (FA). In previous work, we have shown with simulations that there is a quadratic relationship between the FA and the fiber volume fraction in the case of parallel, straight white matter fibers (Stikov et al., 2011). However, the FA is highly sensitive to geometry and this correlation breaks down in the case of partial volume averaging of fiber orientations, including crossing, curvature, and fanning of fibers. The CHARMED model of diffusion incorporates crossing fibers, and the restricted fraction it computes can therefore give a better measure of intra-axonal volume than the FA in the case of crossings. In this work, we use the NODDI model of diffusion. It models the fiber orientation distribution as a Watson distribution, thereby allowing for curving and fanning fibers, which the CHARMED model does not handle. It has been shown through simulations (Sveinsson and Dougherty, 2011) that the NODDI model is also robust to crossing fibers (Campbell et al., 2014), making it a good candidate for estimation of axon volume fraction in the presence of all sources of partial volume averaging of fiber orientations in the brain.
Section snippets
Theory
The myelin g-ratio is the ratio of the inner to the outer radius of the myelin sheath (Rushton, 1951) for a circular axon cross section. The myelin g-ratio is expected to vary across axons, meaning that for each imaging voxel, we have a g-ratio distribution. In healthy tissue, larger axons have higher g-ratios than smaller axons (Berthold et al., 1983). The g-ratio distribution can vary across voxels and change dramatically in disease. For instance, in amyotrophic lateral sclerosis, motor
Animal preparation
One cynomolgus (long-tailed) macaque was scanned in vivo and subsequently sacrificed for histological analysis. The Animal Use Protocol was reviewed and approved by the Montreal Neurological Institute Animal Care Committee, which follows the guidelines of the Canadian Council on Animal Care. The macaque was 4 years old, group-housed, drug and test-naive and weighed approximately 4 kg. As a pre-anesthetic the animal was given glycopyrrolate at a dose of 0.005 mg/kg i.m. followed by a 10 min wait,
Demonstration in cynomolgus macaque
Table 1 shows the measured axon volume fraction, myelin volume fraction, and g-ratio for eight MRI and electron microscopy segments of the monkey corpus callosum. One of a total of three segmented EM images is shown for each CC segment. Regression of the macromolecular pool size F with the myelin volume fraction measured with EM yielded a coefficient of proportionality of 1.6, which was used in the computation of the MVFMRI and AVFMRI from the imaging data. Note that the correlation between F
Discussion
We have presented a novel multimodal magnetic resonance imaging technique for computation of g-ratio maps in vivo in humans. We performed a demonstration of the method in the macaque, comparing to electron microscopy. In the human, we have demonstrated that full-brain maps of the g-ratio can be produced and are robust to the complex subvoxel fiber geometry that exists at the typical resolution of these MRI techniques. Note that there is a slight g-ratio decrease in the frontal region of the
Acknowledgments
The authors would like to thank Dr. Michael Petrides for providing the animal scanning resources.
This work was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC 17426-2012) and the Canadian Institutes of Health Research (CIHR MOP-43871). The granting agencies had no involvement in study design, data collection, analysis and interpretation of data, writing of the report, or the decision to submit the article for publication.
Table of abbreviations
- AVF(MRI,hist)
- axon volume fraction (estimated from MRI, histology)
- MVF(MRI,hist)
- myelin volume fraction (estimated from MRI, histology)
- FVF(MRI,hist)
- fiber volume fraction (estimated from MRI, histology)
- qMRI
- quantitative magnetic resonance imaging
- NODDI
- neurite orientation dispersion and density imaging
- CHARMED
- combined hindered and restricted model of diffusion
- T2
- spin–spin relaxation time
- T1
- spin–lattice relaxation time
- qMT
- quantitative magnetization transfer
- MTR
- magnetization transfer ratio
- MWF
- myelin water
References (64)
- et al.
How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging
NeuroImage
(2003) - et al.
Composite hindered and restricted model of diffusion (charmed) mr imaging of the human brain
NeuroImage
(2005) - et al.
Estimation of the effective self-diffusion tensor from the NMR spin echo
J. Magn. Reson.
(1994) - et al.
Probabilistic diffusion tractography with multiple fibre orientations: what can we gain?
NeuroImage
(2007) - et al.
Three-dimensional quantitative magnetisation transfer imaging of the human brain
NeuroImage
(2005) White matter in learning, cognition and psychiatric disorders
Trends Neurosci.
(2008)- et al.
Rapid calculation of using variable flip angle gradient refocused imaging
Magn. Reson. Imaging
(1987) - et al.
Relation between myelin sheath thickness and axon size in spinal cord white matter of some vertebrate species
J. Neurol. Sci.
(1978) - et al.
Insights into brain microstructure from the distribution
Magn. Reson. Imaging
(2006) - et al.
Quantitative MRI in the very preterm brain: assessing tissue organization and myelination using magnetization transfer, diffusion tensor and T1 imaging
NeuroImage
(2013)
Sex differences in the growth of white matter during adolescence
NeuroImage
Precise estimate of fundamental in vivo MT parameters in human brain in clinically feasible times
Magn. Reson. Imaging
An automated tool for detection of flair-hyperintense white-matter lesions in multiple sclerosis
NeuroImage
Bound pool fractions complement diffusion measures to describe white matter micro and macrostructure
NeuroImage
Myelin and iron concentration in the human brain: a quantitative study of MRI contrast
NeuroImage
Integrative properties and design principles of axons
Int. Rev. Neurobiol.
Axon diameter mapping in the presence of orientation dispersion with diffusion MRI
NeuroImage
NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain
NeuroImage
Extensive cortical remyelination in patients with chronic multiple sclerosis
Brain Pathol.
Characterization of cerebral white matter properties using quantitative magnetic resonance imaging stains
Brain Connect.
Ax-Caliber: a method for measuring axon diameter distribution from diffusion MRI
Magn. Reson. Med.
Axon diameter and myelin sheath thickness in nerve fibres of the ventral spinal root of the seventh lumbar nerve of the adult and developing cat
J. Anat.
Pathological correlates of magnetization transfer imaging abnormalities in animal models and humans with multiple sclerosis
Neurology
Combined NODDI and qMT for full-brain g-ratio mapping with complex subvoxel microstructure
What is the optimal value of the g-ratio for myelinated fibers in the rat CNS? A theoretical approach
PLoS One
Multiexponential T2, magnetization transfer, and quantitative histology in white matter tracts of rat spinal cord
Magn. Reson. Med.
Cross relaxation and spin diffusion in the proton NMR or hydrated collagen
Nature
Multiplexed echo planar imaging for Sub-Second whole brain FMRI and fast diffusion imaging
PLoS One
A magnetization transfer imaging study of normal-appearing white matter in multiple sclerosis
Neurology
Quantitative magnetization transfer imaging using balanced SSFP
Magn. Reson. Med.
Quantitative magnetization transfer imaging via selective inversion recovery with short repetition times
Magn. Reson. Med.
Quantitative Magnetization Transfer Imaging: Validation and Analysis Tool Development
Cited by (0)
- 1
These authors contributed equally to this work.