Myelin water imaging to detect demyelination and remyelination and its validation in pathology

Abstract

Damage to myelin is a key feature of multiple sclerosis (MS) pathology. Magnetic resonance imaging (MRI) has revolutionized our ability to detect and monitor MS pathology in vivo. Proton density, T₁ and T₂ can provide qualitative contrast weightings that yield superb in vivo visualization of central nervous system tissue and have proved invaluable as diagnostic and patient management tools in MS. However, standard clinical MR methods are not specific to the types of tissue damage they visualize, and they cannot detect subtle abnormalities in tissue that appears otherwise normal on conventional MRIs. Myelin water imaging is an MR method that provides in vivo measurement of myelin. Histological validation work in both human brain and spinal cord tissue demonstrates a strong correlation between myelin water and staining for myelin, validating myelin water as a marker for myelin. Myelin water varies throughout the brain and spinal cord in healthy controls, and shows good intra‐ and inter‐site reproducibility. MS plaques show variably decreased myelin water fraction, with older lesions demonstrating the greatest myelin loss. Longitudinal study of myelin water can provide insights into the dynamics of demyelination and remyelination in plaques. Normal appearing brain and spinal cord tissues show reduced myelin water, an abnormality which becomes progressively more evident over a timescale of years. Diffusely abnormal white matter, which is evident in 20%–25% of MS patients, also shows reduced myelin water both in vivo and postmortem, and appears to originate from a primary lipid abnormality with relative preservation of myelin proteins. Active research is ongoing in the quest to refine our ability to image myelin and its perturbations in MS and other disorders of the myelin sheath.

MYELIN IN MULTIPLE SCLEROSIS

The most obvious pathologic feature of multiple sclerosis (MS) are multiple white matter plaques, characterized by demyelination with varying degrees of remyelination, inflammation, and axonal loss 54, 95, 110, 131, 133. As is true of all pathologic processes in the central nervous system (CNS) MS plaques also show gliosis, comprised of astrocytes ranging in morphology, depending on the inflammatory demyelinative activity of the lesion, from marked acute reactive hyperplastic forms to chronic fibrillary gliosis, the latter imparting the “sclerotic” texture which is responsible for the name of this disorder 110. Subsequently, it was recognized that demyelinated plaques also occur in cortical 13, 59, 156 and deep gray matter 53, 56, 181. However, from the very first descriptions of the pathology of MS 17, 18, 21, 22, 26, the white matter demyelinated plaque has been the most prominently emphasized and consistent feature of MS, making it the prototypic “demyelinating disease.” While it is becoming increasingly obvious that axonal damage occurs in MS 41, 167 and the relentless degeneration of axons is probably the most important contributor to clinical progression 164, the overwhelming majority of MS plaques manifest a greater degree of loss of myelin than axons. This would indicate that demyelination, which manifests clinically as focal deficits resulting from conduction block of the action potential 152, must be an important primary pathogenic event in MS.

Thus, early on in the 1980s when magnetic resonance imaging (MRI) was first introduced and it became clear that this was an exquisitely sensitive tool for the demonstration of MS plaques in vivo 127, 160, one of the main objectives was to determine the MRI features that could be attributable to each of the histopathologic features of MS, but most particularly demyelination, which at that time was the major feature thought to be responsible for MS symptomatology. However, it soon became apparent that routine clinical MR imaging did not correlate with any specific histopathologic feature and the MR image was probably a composite that resulted from the contribution of any number of histologic features in a given plaque 109. Thus, demyelination 35, macrophage infiltration 115, vascular permeability 117, edema 125, gliosis 159, could all either individually or in orchestration, produce the images seen on conventional clinical MRI.

MAGNETIC RESONANCE AND MYELIN WATER IMAGING

Water as the dominant source of contrast in MRI

MRI has revolutionized our ability to detect and monitor MS pathology in vivo. The most common type of MR is known as “proton” MR which is sensitive to signal from all of the protons or hydrogen atoms in tissue. The overwhelming majority of the signal measured by proton MRI of the brain and spinal cord originates from hydrogen in water molecules. The properties of the hydrogen govern the three fundamental contrast mechanisms in MRI: [1] proton density (proportional to water content) 165; [2] T ₁ relaxation (influenced heavily by water content as well as the presence of other tissue constituents such as iron and myelin, and factors including field strength, temperature, and MR magnetization exchange processes 40, 47, 138, 162) and [3] T ₂ relaxation (related to water content, the nature of the tissue microstructure, iron, pH, and MR magnetization exchange).

Use and limitations of conventional MRI

Proton density, T₁ and T₂ can provide qualitative contrast weightings that yield superb in vivo visualization of CNS tissue and have proved invaluable as diagnostic and patient management tools in MS 94, 140. Conventional MR techniques play a crucial role in the clinic, and while there is some evidence that certain aspects of image contrast are related to severity of damage (ie, permanent black holes evident on T₁‐weighted imaging are felt to be indicative of parenchymal destruction 169), standard clinical MR methods are not specific to the types of tissue damage they visualize, and they cannot detect subtle abnormalities in tissue. Thus, more quantitative approaches have evolved that focus on measuring specific tissue properties 165. For example, magnetic resonance spectroscopy (MRS) offered some histopathologic correlative specificity as it demonstrates the presence of molecules that serve as specific markers for various CNS cell types. Of particular note is N‐acetyl aspartate (NAA), a marker of axons and coupling between neurons/oligodendrocytes 12, 120, which correlates with axonal loss in MS plaques 8. With respect to myelin specificity, an important scientific breakthrough was the discovery of the short‐T₂ component, or myelin water fraction.

Myelin water imaging

Given the important role myelin damage and loss plays in MS, there has been much interest in the development, validation, and implementation of MR techniques for imaging myelination. While several quantitative methods have been proposed as being sensitive to myelin 81, in this review we shall focus primarily on one of these techniques—myelin water imaging. The concept of myelin water imaging is based on the fact that, while the entire MR signal is from protons in water molecules, individual water molecules can experience very different microscopic environments, depending on their physical location. If the total MRI signal comes from water in different non‐exchanging environments, the resulting T₂ relaxation decay curve of that signal is a sum of exponential decays with amplitudes proportional to the relative amounts of water in each environment. Conceptually, the physical size of the reservoir is a key factor in determining the T₂ relaxation time of the water within that reservoir—water in tightly confined spaces will have a shorter T₂ than water in less tightly confined spaces. For the case of heterogeneous CNS tissue, the T₂ decay can be separated into signal from water trapped in the restricted water reservoir between myelin bilayers (myelin water, Figure 1, T₂ time between 10 and 20 ms), intra/extracellular water (T₂~80–100 ms), additional longer T₂ components seen in some neurological diseases including MS (T₂ ~200–800 ms), and CSF (T₂ of ~2000 ms) 82, 84, 97, 151, 189. The T₂ decay curve can then be separated into its exponential components and expressed as a plot of signal amplitude vs. T₂ time, also known as a T₂ distribution (Figure 2) 188. From the T₂ distribution the myelin water fraction (MWF) is defined as the ratio of the area in the T₂ distribution due to myelin water (<40 ms for humans in vivo at 1.5T and 3T, <30 ms for formalin‐fixed tissue at 1.5T, and <20 ms at 7T) to the area of the entire T₂ distribution. MWF can be visually presented as a myelin water image (Figure 3, 4, 5, 6, 7).

Figure 1.

Electron micrograph of myelinated central nervous system (CNS) tissue at low and high magnifications ([low magnification (left), adapted from Figure 4, 5, 6, 7, originally by Dr. W.T. Norton and Dr. C. S. Raine in Morell P, Quarles RH, Myelin formation, structure, and biochemistry. In Siegel GJ, Agranoff BW, Albers RW, Fisher SK, Uhler MD (editors) Basic Neurochemistry. 6th Edition; 1999. Philadelphia: Lippincott‐Raven; ISBN 0‐397‐51820‐X with permission; high magnification (middle) (adapted from Peters A, Palay SL, Webster H deF. Fine Structure of the Nervous System: The Cells and Their Processes. 1^st edition; 1970; page 89, Figure 33, New York: Paul B. Hoeber Inc, with permission from Dr. Alan Peters] depicting the major dense line, which represents the fusion of the cytoplasmic aspects of the oligodendrocyte cell membrane, and the intraperiod line, a potential extracellular space formed by the apposition of the extracellular faces of adjacent oligodendrocyte cell membranes. As shown in the accompanying schematic, the intraperiod line forms a restricted water reservoir, and, thus, is thought to give rise to the short‐T₂ component, the signal of which can be displayed anatomically as the myelin water map (see Figure 4, 5, and 6). The oligodendrocyte cell membrane is a bilayer of lipids in which are embedded the major myelin proteins, which include myelin basic protein (MBP), proteolipid protein (PLP), 2′,3′‐cyclic nucleotide 3′‐phosphodiesterase (CNP), myelin oligodendrocyte protein (MOG), and myelin‐associated glycoprotein (MAG). Note, however, that on the inner aspect of the myelin sheath MAG is restricted to the membrane adjacent to the adaxonal space, which it spans to bind the myelin sheath to its axolemmal ganglioside receptors, GD1a and GT1b. The exact position of some of the components of myelin shown in this schematic has not been determined.

Figure 2.

T₂ distribution from multiple sclerosis (MS) normal‐appearing white matter (NAWM) in formalin at 7T (black), 1.5T (gray) and in vivo (light gray). All three distributions have a similar shape, showing two distinct peaks with the myelin water peak on the left and intra/extracellular (IE) component on the right. However, the IE component is shifted to shorter times for the 1.5T formalin sample, and even shorter for the 7T formalin sample when compared to in vivo. (Reprinted from NeuroImage. 2008;40( 4 ): Laule C, Kozlowski P, Leung E, Li DKB, Mackay AL, Moore GRW. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. pages 1575–1580, Figure  1 , Copyright 2008, with permission from Elsevier).

Figure 3.

58‐year‐old male with a 34‐year history of secondary progressive multiple sclerosis (MS) with clinical evidence of optic, cerebellar and spinal involvement. Note the large, irregular lesion in the periventricular occipital white matter, which appears as an area of increased signal on the proton density scan, an area of absent signal on the myelin water/short‐T₂ component distribution, and an area of gray discoloration of the white matter in the gross photograph. A band of reduced signal is seen in the lesion on both scans and correlates with the gross appearance (long arrows). More rostrally, several smaller lesions are evident (short arrows), which appear as areas of reduced intensity on the short‐T₂ component image. The Luxol fast blue and 2′,3′‐cyclic nucleotide 3′‐phosphohydrolase (CNPase) stains show absence of myelin in most regions of the large periventricular occipital lesion. The Bielschowsky stain for axons is reduced in the lesions but not to the degree of the myelin stains. The faint band detected by the short‐T₂ distribution component and the proton density scan is particularly evident on the CNPase stain (long arrows). (Moore, G.R.W., Leung, E., MacKay, A.L., Vavasour, I.M., Whittall, K.P., Cover, K.S., Li, D.K., Hashimoto, S.A., Oger, J., Sprinkle, T.J., Paty, D.W. A pathology‐MRI study of the short‐T2 component in formalin‐fixed multiple sclerosis brain. Neurology 2000;55( 10 ):1506–1510. Figure  1 . Published by The American Academy of Neurology, with permission. http://n.neurology.org/content/55/10/1506.long)

Figure 4.

Example of diffusely abnormal white matter (DAWM) at 7T with corresponding myelin water map and histological stains for phospholipids (Luxol fast blue, Weil's), sialic acid groups (Alcian Blue), axons (Bielschowsky), myelin proteins (myelin‐associated glycoprotein (MAG), myelin basic protein (MBP), myelin oligodendrocyte protein (MOG), proteolipid protein (PLP), 2′,3′‐cyclic nucleotide 3′‐phosphohydrolase (CNP)), and astrocytes (GFAP). DAWM, characterized by an area of reduced intensity on the proton density (arrows) and myelin water map, matches a region of reduced staining intensity on the Luxol Fast Blue, Weil's, Alcian Blue, Bielschowsky, and, to a lesser degree, MAG stains. Several small plaques are seen within this region. Note the improvement in resolution in this high‐field strength compared to that at 1.5T shown in Figure 3. (Laule, C., Pavlova, V., Leung, E., Zhao, G., MacKay, A.L., Kozlowski, P., Traboulsee, A.L., Li, D.K., Moore, G.R.W. Diffusely abnormal white matter in multiple sclerosis: further histologic studies provide evidence for a primary lipid abnormality with neurodegeneration. Journal of Neuropathology and Experimental Neurology 2013; 72( 1 ): 42–52, Figure  2 , by permission of Oxford University Press and the American Association of Neuropathologists)

Figure 5.

Examples of the quantitative correlation between myelin water fraction (MWF) and Luxol Fast Blue optical density (LFB OD) for gray matter (GM), lesion, diffusely‐abnormal white matter (DAWM), and normal appearing white matter (NAWM) for 2 multiple sclerosis (MS) cases. (Reprinted from NeuroImage. 2008;40( 4 ): Laule C, Kozlowski P, Leung E, Li DKB, Mackay AL, Moore GRW. Myelin water imaging of multiple sclerosis at 7 T: Correlations with histopathology. pages 1575–1580, Figure  4 , Copyright 2008, with permission from Elsevier).

Figure 6.

Magnetic resonance imaging (MRI) and corresponding histology from a formalin‐fixed multiple sclerosis (MS) spinal cord. Cervical, thoracic, and lumbar regions show anatomical variation in myelin with white matter showing increased myelin water relative to the central gray matter butterfly. MS lesions (arrows) demonstrate myelin water loss. Staining for myelin (Luxol Fast Blue) demonstrates excellent correspondence between MRI and histology. (adapted from Figure  1 a, Laule, C., Yung, A., Pavolva, V., Bohnet, B., Kozlowski, P., Hashimoto, S.A., Yip, S., Li, D.K., Moore, G.R.W. High‐resolution myelin water imaging in post‐mortem multiple sclerosis spinal cord: A case report. Multiple Sclerosis Oct 22 2016, 1485–1489, published by SAGE Publications).

Figure 7.

Heat map of myelin water fraction. Left side: T₂ weighted image of a Multiple Sclerosis (MS) patient. Right side: heat map of a myelin water image (MWI). T₂‐hyperintense MS‐lesions show clear reductions of myelin water fraction (MWF) (white arrows, right side). (Faizy TD, Thaler C, Kumar D, Sedlacik J, Broocks G, Grosser M, Stellmann J‐P, Heesen C, Fiehler J, Siemonsen S (2016) Heterogeneity of Multiple Sclerosis Lesions in Multislice Myelin Water Imaging. PLoS ONE 11( 3 ): e0151496. https://doi.org/10.1371/journal.pone.0151496 , Figure  2).

Myelin water in CNS white matter was first observed in a cat model in 1991 102. The first human in vivo myelin water measurements in the mid‐1990s were slow to acquire and produced only a single brain slice in 25 minutes 97; today it is possible to collect whole brain MWF images in less than 5 minutes 118. At least four different approaches to myelin water imaging have now been explored; for a comprehensive technical overview the reader is pointed to a recent review by Alonso‐Ortiz et al 6. The pioneering, most common, and still considered to be the “gold‐standard” approach to myelin water imaging uses a Carr–Purcell–Meiboom–Gill (CPMG) multi‐echo spin echo data acquisition strategy 97, 130, 193. Variations on the CPMG method in recent years have resulted in significantly faster imaging times 118, 119, 122, 123, 130. Traditional analysis of the T₂ decay used a non‐negative least squares (NNLS) method which makes no a priori assumptions about the number of water environments 129, 188, although other approaches also exist 2, 52, 69, 70, 134, 147, 157. Several groups have obtained myelin water images from gradient echo T ₂ ^* decay curve measurement which examines the echo train derived by magnetic field gradient reversals 32, 89, 113, 142 and measurement of multiple T ₁ relaxation components to isolate myelin water has also been used 72, 124. Finally, the mcDESPOT (multicomponent driven equilibrium single pulse observation of T₁ and T₂) method 29 which uses multiple flip angles to examine signal changes of two fast gradient echo imaging sequences to enable demonstration of the myelin water and intra/extracellular water components in CNS tissue has also been used. mcDESPOT is fundamentally different from myelin water imaging techniques which are derived from T_2, T₂*, or T₁ decay curves.

VALIDATION OF MYELIN WATER IMAGING

MRI‐histology studies of myelin water imaging have focused almost exclusively on validation of the multi‐echo spin‐echo approach to data acquisition. The myelin water signal is present both shortly after death in situ and upon tissue fixation with formalin, and the shape of the T₂ distribution from formalin fixed CNS tissue is qualitatively similar to that from in vivo, albeit with shorter T₂s, (Figure 2) 77, making MRI‐pathology correlation studies possible. One of the earliest correlation studies conducted at 1.5T in 2000 showed that the anatomic distribution of the short‐T₂ component matched the distribution of myelin and its absence correlated with the absence of myelin in plaques with relative axonal sparing (Figure 3 ) 111. Visual correspondence between MRI and histology has improved significantly with the advent of higher field strength MR systems that can produce MR images from much thinner volumes of tissue (Figure 4, 1mm thick at 7T 80 vs. Figure 3, 5mm thick at 1.5T). Further quantitative brain studies showed a tight relationship between the strength of the short‐T₂ signal and the optical density of myelin staining as indicated by the myelin phospholipid stain Luxol Fast Blue (LFB) 63, 96, 141, 146. (Figure 5) 75, 76. Likewise, comparisons between MWF and myelin staining in human spinal cord also show excellent correspondence between the MR and histology markers for myelin (Figure 6) 88. As a consequence, the distribution of the short‐T₂ component has been referred to as the “myelin water map.”

In addition to the above mentioned human validation studies, a number of animal studies have also demonstrated a strong correlation between myelin water and various myelin histological stains in both peripheral nervous system 121, 132, 158, 168, 185 and CNS animal models 45, 67, 68, 100, 161.

IN VIVO APPLICATIONS OF MYELIN WATER IMAGING IN RESEARCH AND CLINICAL TRIALS

Myelin variation in healthy brain and spinal cord white matter

Initial in vivo studies by MacKay et al of the brain almost 25 years ago demonstrated MWF of white matter to be substantially higher than gray matter, and regional variation of MWF across different white matter structures 97; this observation has been confirmed by numerous studies since 11, 20, 85, 123, 180, 189. Frontal lobe MWF is correlated with age, as well as years of education and reading IQ in healthy adults 44, 73. mcDESPOT‐derived MWF shows a positive correlation between physical activity level and MWF in the right parahippocampal cingulum 16 and one study found regional differences in MWF of the corpus callosum between males and females 91. Reproducibility and reliability of MWF in healthy controls, both at a single site and multiple sites is very good 14, 103, 171, 174.

Myelin water techniques applied in the brain can also be used to study spinal cord myelination. However, spinal cord myelin water imaging studies are far less common, since imaging the spinal cord is difficult for a number of reasons including the small diameter of the cord, cardiac and respiratory motion, magnetic field inhomogeneties, and the presence of flow from CSF. Nevertheless, it is feasible to measure MWF in the spinal cord in vivo. MWF is approximately 50% higher in spinal cord than normal brain white matter and varies along the length of the cord 66, 87, 98, 105, 193. Younger adults (20–30 years old) have a higher cervical cord MWF compared to older (50–75 year) study participants 98. An in‐depth review of myelin water in the cord can be found elsewhere 78.

The aforementioned work, which characterizes MWF in controls and demonstrates sufficiently stable reproducibility, supports the application of myelin water imaging in disease states.

Multiple sclerosis plaques

Much of the pioneering in vivo work in myelin water imaging has been studies of MS. MS plaques, or lesions, show variably decreased MWF (Figure 7) 38, 58, 64, 85, 93, 97, 122, 166, 171, 180, averaging approximately half that of normal‐appearing white matter (NAWM) 85. This variability in MWF is probably reflective of the myelin content or pathology in different lesions. Indeed, MWF can vary between lesions observed on T₂‐weighted imaging, black holes evident on T₁‐weighted imaging and lesions with contrast enhancement 38, 85, 166. MWF can also be used to distinguish plaques based on their age, with older lesions showing a larger reduction in myelin water 178. The difference in MWF between new and old lesions suggests there is less advanced demyelination in new lesions or possibly ongoing remyelination which eventually fails in older lesions. Longitudinal study of MWF can provide insights into demyelination and remyelination in plaques, where a reduction in MWF in some lesions can be followed by MWF increase, suggesting remyelination over time 90, 171, 176.

Multiple sclerosis normal‐appearing white matter

It has become increasingly apparent that what on routine MRI and casual histopathologic examination appears to be “normal‐appearing white matter” is far from normal when more sophisticated tools in either of these spheres are used for interrogation of this region 110. Furthermore, these changes in MS NAWM are clinically relevant as they present very early in the course of the disease at the time of the first clinical presentation, and they correlate with disability, cognitive impairment, and the degree of brain atrophy 104.

The literature prior to the MRI era showed conflicting results with respect to MS NAWM neurochemistry and myelin damage 108. Some studies showed reductions in total phospholipid 23, 49, 136, phosphatidylserine 116, phosphatidlylinositol 116, fatty acids particularly linoleic acid 9, 49, cerebroside 1, 135, 136, sulfatide 5, the gangliosides GM4, GM1, GD1b, GQ1b 195. Other studies showed increased levels of cholesterol esters in NAWM 1, 23, 195. However, others reported normal levels of cholesterol esters 194, total cholesterol 194, total phospholipids 194, ethanolamine phospholipids 24, cerebroside 24, and sulfatide 24. In addition, while a variety of enzymes were found to be increased in NAWM, these results were also inconsistent 108. At the time, it was felt that these discrepant results were due the inadvertent inclusion of small macroscopically invisible plaques in the material assayed as NAWM and it was thought that biochemically NAWM was truly normal 163. Nevertheless, the very few neuropathologic studies of MS NAWM showed subtle abnormalities that could not be necessarily considered plaques. These included perivascular inflammation, perivascular lipofuscin deposition, cells with increased numbers of lysosomes, and occasional demyelination 3. Subsequent studies showed microglial activation 4, upregulation of factors involved in Class II Major Histocompatibility Complex (MHC) expression 50, expression of peripheral benzodiazepine binding sites 50, upregulation of osteopontin and αB‐crystallin 150, extracellular matrix enzymes, and modification of extracellular matrix components 154. Further studies showed blood–brain barrier breakdown in MS NAWM 61, 128.

Very consistently, however, is the axonal loss that is evident in NAWM 36 and this loss correlates with plaque volume, consistent with the notion that it is due to Wallerian degeneration as a consequence of axons transected in plaques 37. Furthermore, this axonal loss appears to involve small‐diameter axons predominately 28. Evidence for axonal degeneration is also apparent in the upregulation of ephrin A1 and receptors to ephrin‐A3, ‐A4, and ‐A7 153 and axonal amyloid precursor protein, dephosphorylated neurofilament, and neuropeptide Y receptor Y1 in periplaque white matter 33. An important driver of neurodegeneration in MS NAWM may be the bystander effect on the axon by the products of inflammatory infiltrates, which, while mild in degree, are scattered throughout the NAWM 71, and may be sequestered behind the blood–brain barrier 74.

There is compelling evidence from unconventional MRI techniques for abnormalities in NAWM. Reinforcing the neuropathologic findings of axonal degeneration and loss in NAWM is the finding of reduced NAA by MRS 27. In general, MRI‐demonstrable axonal degeneration does not correlate with plaque load, suggesting that factors in addition to Wallerian degeneration may contribute to neurodegeneration in NAWM 104. Numerous studies support widespread and varying abnormalities in MS NAWM including increases in creatine 57, myo‐inositol 42, choline 57, and lipid peaks 114, a higher apparent diffusion coefficient 187, reduced fractional anisotropy 51, reduced magnetization transfer ratio 43, prolonged T₁ 170, and increased total water content 85.

There is also strong in vivo MR evidence of myelin damage in MS NAWM. When compared to healthy controls, MWF is reduced in brain NAWM by 6%–37% 38, 64, 85, 90, 122 and in spinal cord by 11%–25% 87, 193. NAWM MWF can differentiate between different subtypes of MS, with greater myelin loss found in more progressive forms of the disease 62, and reduction of MWF is related to increased clinical disability 62, 64. Changes in NAWM MWF can also be discerned over time 60; for example, in untreated relapsing‐remitting MS patients there was an 8% reduction in brain MWF over 5 years 175, suggesting that chronic, progressive myelin damage is an evolving process occurring over many years. Longitudinal assessment of brain myelin water in non‐lesional tissue has also been successfully used in clinical trials, with a recent study demonstrating NAWM MWF stability after 24 months on MS disease‐modifying therapy 179. Changes in NAWM MWF can also be reliably detected in the spinal cord, with one study showing a 10% myelin loss in primary progressive MS cervical cord over 2 years, while controls remained stable, suggesting ongoing demyelination may be contributing to the disease process in this subgroup of patients 87.

The histopathologic correlate of the NAWM MWF abnormality has not yet been determined. Based on the discussion of the origin of the MWF above, it could represent a change in the periodicity of the spacing of myelin lamellae in the myelin which by the usual histologic stains appears normal. Another alternative is that it may simply be a reflection of the concomitant widespread loss of axons in NAWM 36, 37, 149. Supporting these notions, one study showed that NAWM, as defined by magnetization transfer imaging, showed histopathologic correlates that were spatially dependent 106. NAWM near a plaque shows correlation with microglial and axonal pathology, that latter presumably being secondary to axonal damage within the plaque. Whereas, NAWM remote from the plaque correlates with microglial activation but not axonal damage, suggesting again a factor in addition to Wallerian degeneration is operative in these regions 27, 106.

Multiple sclerosis diffusely‐abnormal white matter

In 2000, Zhao, Li, and colleagues first described “dirty‐appearing white matter” in routine MRI in MS 197. This abnormality, which has subsequently been referred to as “diffusely‐abnormal white matter” (DAWM), has a signal intensity intermediate between that of NAWM and that of plaque, similar to gray matter on proton density and T₂ weighted imaging. It is evident in approximately 20%–25% of MS patients, who tend to have a more rapidly progressive clinical course 196. DAWM shows ill‐defined boundaries and is sometimes adjacent to a plaque, particularly in the periventricular occipital white matter.

Pathologic studies have shown reduced myelin on the LFB stain and reduced numbers of axons in DAWM 145. There is also evidence of blood–brain barrier breakdown in DAWM 182. When the DAWM myelin abnormality is interrogated with a variety of stains, it is apparent that while LFB 96, 141, 146 and another phospholipid stain, the Weil's stain 186, are reduced in DAWM, immunohistochemical staining for various myelin proteins is relatively preserved (Figure 4), suggesting that there is a selective lipid abnormality in DAWM myelin 80, 83, 107. There is also a reduction of staining for sialic acid groups (on the Alcian blue stain) 80. Since the major source of sialic acid groups in the CNS is gangliosides, this finding suggests that in DAWM there is a perturbation of gangliosides, which are located particularly in the axolemma rather than the myelin sheath (Figure 1). Interestingly, the only myelin protein that is occasionally reduced in DAWM is myelin‐associated glycoprotein (MAG) 80, which is located adjacent to the adaxonal space and serves as the ligand that binds the myelin sheath to the axon by interaction with its axolemmal ganglioside receptors GD1a and GT1b 48, 101. Axonal loss, as evident on the modified Bielschowsky stain, is often but not always evident in DAWM, indicating neurodegeneration may occur in DAWM and this might possibly be a result of the MAG‐axolemmal ganglioside perturbation.

It is also of considerable interest, from the point of view of myelin biology and imaging, that the MWF is exquisitely sensitive for the detection of DAWM, showing 23% reduction of MWF in this region in vivo 83 and 30% loss post‐mortem 80. Again, given that it is thought the MWF emanates from restricted water in the tight lamellar compaction of myelin, we postulate that the lipid abnormality in DAWM leads to myelin membrane permeability to water, which would result in widening of the myelin lamellar water reservoir resulting in reduction of the value of the short‐T₂ component and also lead to the observed increase in mean T₂ and total water content seen in vivo 83. This, however, would not necessarily affect the concentration of myelin protein constituents within the myelin lipid bilayers, with the exception of the perturbed ganglioside‐MAG interactions and subsequent axonal degeneration in more advanced DAWM pathology. Other quantitative MRI studies also show abnormalities in DAWM 46, 79, 83, 126, 139 and differences in DAWM in different clinical subtypes of MS, with primary progressive MS showing higher T₁ and lower magnetization transfer ratio than the secondary progressive form of the disease 183. Clearly, further research is necessary to sort out the complex but fascinating changes in DAWM, which may well be important clinically and could represent the early events in propagating the expansion of the MS plaque that it borders.

Neurological applications beyond MS

Beyond MS, myelin water imaging has been used to study many other neurological disease applications include neuromyelitis optica 58, 99, schizophrenia and first episode psychosis 44, 73, phenylketonuria 151, autism 30, stroke 14, neurofibromatosis 10, Niemann–Pick disease 25, primary lateral sclerosis 65, amyotrophic lateral sclerosis 65, concussion 192, and Krabbe disease 86. Other spinal cord applications are also feasible, for example, a recent study of cervical spondylotic myelopathy demonstrated a correlation between MWF in the dorsal columns and functional measures of myelin through somatosensory evoked potential latency times 92.

OTHER MR METHODS SENSITIVE TO MYELIN

Several other MRI techniques have been proposed to be sensitive to changes in myelin. Magnetization transfer (MT) imaging measures decreases in MR signal following off‐resonance excitations 191; the effect is typically quantified by a magnetization transfer ratio, which is sensitive to small differences between groups. There is an extensive literature on using MT to study myelination in MS 7 and several studies have showed correlation between MT parameters and histological measures of myelin 19, 144. A limitation of using MT to monitor myelin is that while a change in myelin will cause a change in MT, a change in MT is not necessarily due to a change in myelin. Changes in other tissue components such as axons and glia, as well as changes in water content due to inflammation or edema will result in changes in MT 112, 177. A newer magnetization transfer related method termed inhomogeneous magnetization transfer (ihMT) shows promise for being more specific to CNS lipids; given that myelin is 70%–80% lipids and ihMT correlates well with MWF, this is a exciting area of ongoing myelin imaging research 34, 172, 173. Several metrics acquired using diffusion tensor imaging (DTI), which examines water movement, have been linked to myelin. Most notably the perpendicular component of the diffusion tensor (often called lambda perp or radial diffusivity) is inversely related to myelination in animal models 155, 184. However, the presence of edema and neuroanatomy such as crossing fibers can confound DTI measurements. More sophisticated diffusion modeling and analysis approaches are now emerging including constrained spherical deconvolution (CSD), Q‐ball imaging (QBI), diffusion orientation transform (DOT), persistent angular structure (PAS), and neurite orientation dispersion and density imaging (NODDI) and diffusion basis spectrum imaging (DBSI) which may provide more specific links to tissue components, including myelin 143, 184. Finally, ultrashort echo time (UTE) measures signal from non‐water sources of hydrogen, including, but not limited to, the lipids and proteins that make up myelin 137. Several studies have used UTE for myelin measurement 15, 55, 190 and, as this method is becoming more commonly available on newer MR systems, it is expected that research on using UTE for myelin imaging will continue to expand 31, 39, 148.

CONCLUSION

There have been numerous substantial advances in myelin imaging and exciting research is ongoing in this area. New techniques or modifications of currently employed techniques to demonstrate myelin in vivo will continue to be developed. All of these new methodologies, however, must pass the scrutiny of histopathologic validation before they can be accepted as appropriate tools to image myelin and its disorders.

CONFLICT OF INTEREST

CL has nothing to declare. GRWM has received a grant‐in‐aid of research from Berlex Canada, has acted as a consultant for Schering, and has received honoraria from Teva for teaching. He is a member of the Medical Advisory Committee of the Multiple Sclerosis Society of Canada.