Множественный склероз: то, что мы узнали и можем узнать из электронной микроскопии.

Introduction

Multiple sclerosis (MS) is a chronic disease of the central nervous system (CNS) in which the fundamental pathophysiology is inflammation, demyelination and dysfunction and ultimate loss of neurons. Clinical symptoms usually start around 20 to 40 years of age and MS is more often diagnosed in women compared to men (ratio around 3:1) [ 1 , 2 ]. Although the disease course and symptomatology are very heterogeneous, several clinical courses can be discerned. The most common type is relapsing–remitting MS (RRMS), which is often diagnosed in 75–85% of the people with MS [ 3 ]. In RRMS, people have periods in which they experience new neurological deficits (relapses) followed by periods of remission. People diagnosed with RRMS have a 50% chance to transit to secondary progressive MS (SPMS) within 20 years after their first diagnosis, and 93% after 30 years [ 4 ]. In SPMS, symptoms gradually worsen over time, with or without occurrence of relapses. A less common type of MS is primary-progressive MS (PPMS), which manifests at a later age and in about 10% of the people with MS. PPMS is characterised by gradual worsening of symptoms over time, without the occurrence of relapses or remissions. Current therapies are mostly based on immunomodulatory approaches acting outside the CNS and there are no treatments available that successfully prevent the onset or halt the progression of MS. A significant number of potential therapies for progressive MS that seemed promising in experimental models of MS, failed the clinical trials [ 5 ]. Although studies on MS tissue have provided some insight in the cellular and molecular factors that contribute to disease pathology, information crucial for developing effective and curative therapies remains elusive.

The underlying mechanisms of MS lesion initiation, development and progression are still unknown, and our knowledge of causal pathways is still very limited. Current postulations on the onset of MS are the outside-in and the inside-out hypotheses. The outside-in hypothesis states that demyelination is driven by autoimmunity and primary activation of peripheral immune cells, while the inside-out hypothesis states that it is driven by a primary cytodegenerative process in the CNS [ 6 ]. Moreover, epidemiological data indicate that both genetic and environmental factors contribute. Genetic epidemiological studies reveal that first-degree relatives have an 10–25 times higher risk developing MS than the general population [ 7 ]. A GWAS study identified 233 MS susceptibility variants that were mainly annotated to immune cell types and an enrichment for MS genes in purified primary human microglia [ 8 ]. Furthermore, several environmental factors increase the risk of developing MS, such as Epstein-Barr virus (EBV), smoking and latitude/vitamin D deficiency [ 7 , 9 ]. Although the primary molecular mechanisms(s) of initiation are unknown, evidence suggests that inflammation, either primary or secondary, triggers a cascade of events including microglia activation, chronic oxidative injury, accumulation of mitochondrial damage in axons and age-related iron accumulation [ 6 ]. These events are augmented by ageing and accumulated disease burden and eventually lead to (secondary) neurodegeneration [ 6 ]. In vitro and ‘omics’ studies provided some insights into the biological and molecular processes involved [ 10 – 15 ] and experimental models that only partly reflect aspects of MS pathology (Box 1 ) have been used to test these hypotheses. However, cell culture and animal models do not fully reflect the biology of human disease [ 16 ]. Moreover, information on the morphology and ultrastructure of individual neurons and glial cells in age and disease is very limited.

Elucidating early changes in MS would have a significant impact on the development of treatment strategies that prevent MS. Furthermore, overcoming remyelination failure in existing demyelinated lesions is essential to reinstate saltatory conduction and restore metabolic support to axons, especially in progressive stages. In addition to relapse-driven worsening, recent findings emphasize the importance of progression independent of relapse activity (PIRA), which is driven by widespread, diffuse changes in normal-appearing matter (NAM) that occur independent of demyelination [ 20 ]. Electron microscopy (EM) may prove to be essential to understand molecular mechanisms underlying MS by providing an overview of human-specific morphological and (sub)cellular features at the nanoscale. This very high-resolution imaging technique is a powerful tool to study ultrastructural features and complements conventional techniques that are used to study cellular and molecular processes, such as cell culture and animal models, but also for example immunohistochemistry, single nucleus RNA sequencing and spatial transcriptomics, lipidomics and proteomics of post-mortem MS brain tissue. The EM data available on MS brain tissue is very limited and largely depends on studies from before the 90’s, or as a confirmation of pathology observed in experimental models for MS. However, the EM field has evolved, and more advanced techniques are now available that can provide more ultrastructural information on a larger spatial scale and are just starting to be used on MS tissue. In this review, we first provide a brief overview on MS pathogenesis, and highlight the current contribution of EM analysis to understand MS pathology by outlining the described ultrastructural features of MS and its (potential) contribution to underlying mechanisms. Secondly, novel EM approaches and their potential on improving our understanding of MS pathology will be discussed.

Animal model Induction and mechanism of action CNS a location Ref. EAE b subcutaneous injection with MOG peptide, peripheral immune cell infiltration followed by myelin breakdown spinal cord [ 17 ] focal EAE b subcutaneous injection with MOG peptide combined with stereotactic injection in the cortex with pro-inflammatory mediators, peripheral immune cell infiltration followed by myelin breakdown cortex [ 18 ] cuprizone dietary administration, disturbs copper homeostasis, OLG c apoptosis mainly investigated in the brain [ 17 , 19 ] lysolecithin local CNS injection, (myelin) membrane disruption, OLG c cell death WM d spinal cord and brain [ 17 , 19 ] ethidium bromide local CNS injection, intercalates in DNA and affects all nuclei in lesion WM d spinal cord and brain [ 19 ] TMEV e intracerebral injection, apoptosis of neurons in acute phase and of OLG d in chronic phase WM d spinal cord [ 17 ] a central nervous system, b experimental autoimmune encephalomyelitis, c oligodendrocyte, d white matter, e Theiler’s murine encephalomyelitis virus

Animal model Induction and mechanism of action CNS a location Ref. EAE b subcutaneous injection with MOG peptide, peripheral immune cell infiltration followed by myelin breakdown spinal cord [ 17 ] focal EAE b subcutaneous injection with MOG peptide combined with stereotactic injection in the cortex with pro-inflammatory mediators, peripheral immune cell infiltration followed by myelin breakdown cortex [ 18 ] cuprizone dietary administration, disturbs copper homeostasis, OLG c apoptosis mainly investigated in the brain [ 17 , 19 ] lysolecithin local CNS injection, (myelin) membrane disruption, OLG c cell death WM d spinal cord and brain [ 17 , 19 ] ethidium bromide local CNS injection, intercalates in DNA and affects all nuclei in lesion WM d spinal cord and brain [ 19 ] TMEV e intracerebral injection, apoptosis of neurons in acute phase and of OLG d in chronic phase WM d spinal cord [ 17 ] a central nervous system, b experimental autoimmune encephalomyelitis, c oligodendrocyte, d white matter, e Theiler’s murine encephalomyelitis virus

a central nervous system, b experimental autoimmune encephalomyelitis, c oligodendrocyte, d white matter, e Theiler’s murine encephalomyelitis virus

MS pathogenesis

Myelin, a specialised membrane structure that enwraps axons, is crucial for efficient nerve impulse conduction and axon maintenance and survival [ 21 ]. In MS, the loss of oligodendrocytes and myelin impairs these functions. The disease course and symptomatology of MS are very heterogeneous, with pathophysiological characteristics that include not only the formation of demyelinated lesions but also inflammation and neurodegeneration [ 22 ]. Upon demyelination, resident microglia and astrocytes are activated by myelin debris and start to produce factors which (i) activate each other, (ii) contribute to the recruitment of monocytes from the blood, and (iii) induce the proliferation and migration of oligodendrocyte progenitor cells (OPCs) [ 23 ]. Microglia and recruited macrophages remove the remyelination-impairing myelin debris, and secrete pro-OPC differentiation factors beneficial for remyelination [ 23 ]. The recruited OPCs differentiate into mature oligodendrocytes that remyelinate denuded axons. While remyelination by newly generated oligodendrocytes is very effective in experimental models of demyelination [ 24 ], in MS lesions, the regeneration of myelin membranes is either incomplete or non-occurring or it could be executed by mature surviving oligodendrocytes rather than matured OPCs [ 25 ]. The signalling events for the remyelination process are closely associated with the inflammatory response and an inappropriate inflammatory response may prevent the more efficient OPC-based remyelination in chronic MS lesions [ 26 , 27 ]. However, areas of remyelination are observed in most MS lesions [ 22 , 28 ], indicating that in addition to heterogeneity between lesions, there is also spatial heterogeneity within the same lesion.

MS lesion heterogeneity

MS lesions in white matter (WM) are histologically classified as active, mixed active/inactive, inactive, and remyelinated (shadow plaques), based on the degree of inflammation and demyelination observed in end-stage post-mortem tissue (Fig.  1 a) [ 22 , 29 ]. Active lesions are characterised by inflammatory activity of macrophages/microglia and hypertrophic astrocytes throughout the lesion [ 22 , 29 ]. Mixed active/inactive (chronic inactive) lesions are typified by inflammatory activity of macrophages/microglia that is limited to the lesion border and astroglial scars [ 22 , 29 ]. Inactive lesions are almost completely lacking inflammatory activity of macrophages/microglia, and characterised by astroglial scars in lesion cores [ 22 , 29 ]. In most demyelinated lesions OPCs are present, while surviving mature oligodendrocytes are only observed in active MS lesions [ 30 , 31 ]. Pre-demyelinating areas can also be discerned and are referred to as reactive sites (pre-active), which are small  clusters of microglia/macrophages. Shadow plaques are post-demyelinating areas with a fainter myelin stain [ 22 , 29 ]. Lesions in grey matter (GM) are named based on their location and classified into leukocortical, intracortical and subpial lesions (Fig.  1 b) [ 22 , 29 ]. Leukocortical lesions are mixed GM-WM lesions and can also be referred to as type I. Intracortical lesions or type II lesions are GM lesions that do not extend into WM or reach the surface of the brain. Subpial lesions, also described as type III lesions, extend from the pial surface of the brain and do not extend into WM [ 22 , 29 ]. Fig. 1 White and grey matter MS lesions: classification and key differences.
a. White matter (WM) MS lesions are classified based on demyelination and inflammatory activity into the following types: pre-active, active, mixed active/inactive, inactive and remyelinated. The black cells indicate inflammatory cells. b. Grey matter (GM) MS lesions are classified based on location into the following types: leukocortical, intracortical and subpial. c. Key differences between WM and GM lesions include gliosis, more inflammation, less remyelination and lower OPC numbers in WM

White and grey matter MS lesions: classification and key differences.
a. White matter (WM) MS lesions are classified based on demyelination and inflammatory activity into the following types: pre-active, active, mixed active/inactive, inactive and remyelinated. The black cells indicate inflammatory cells. b. Grey matter (GM) MS lesions are classified based on location into the following types: leukocortical, intracortical and subpial. c. Key differences between WM and GM lesions include gliosis, more inflammation, less remyelination and lower OPC numbers in WM

A clear difference in pathology is apparent between WM and GM MS lesions (Fig.  1 c). While WM lesions are characterised by immune infiltration, these cells are barely present in GM lesions [ 32 ]. In leukocortical lesions, neighbouring demyelinated GM and WM of the same lesion have a distinct gene expression profile and cellular composition [ 33 , 34 ]. In contrast to WM lesions, OPC numbers do not decrease in GM lesions and in remyelinated GM areas compared to adjacent normal-appearing cortex or control grey matter [ 34 ]. Furthermore, there is hardly astrogliosis in GM lesions, and remyelination is more extensive in the cortex [ 35 – 37 ]. The heterogeneity between WM and GM lesions may at least be partly attributed to the number and regional diversity of glial cells [ 33 , 38 ]. In human WM, microglia numbers are higher and the cells are more prone to become proinflammatory than GM microglia [ 33 , 39 – 41 ]. Furthermore, comparative transcriptomic analysis of human MS microglia reveals that in WM MS microglia genes related to lipid metabolism predominate, while genes associated with glycolysis and iron homeostasis are enriched in GM MS microglia [ 42 ]. Primary rat derived WM microglia are also more elastic and viscous compared to GM microglia [ 43 ]. Additionally, treatment with a TLR4 agonist decreases cell elasticity and viscosity in WM primary rat microglia, whereas it induces an increase in GM microglia [ 43 ]. Hence, the biomechanical response to inflammation differs between GM and WM microglia.

Besides differences between WM and GM microglia, regional diversity in OPCs and astrocytes may have implications for remyelination in WM and GM. WM and GM rodent OPCs display distinct electrophysical properties, gene expression, morphology and proliferation and differentiation rates [ 44 – 46 ]. OPCs acquire neurotransmitter receptors, like α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPAR), kainate (KAR) and N-methyl-D-aspartate receptors (NMDAR), during murine development [ 44 ]. In the first postnatal week, WM and GM OPCs have a similar sensitivity to neuronal activity. However, after the first post-natal week, WM OPCs exhibit a lower KAR density and a higher NMDAR density than GM OPCs. Upon ageing, WM OPCs display a more gradual reduction in NMDAR density. In addition, neonatal rat OPCs in GM seem to be better equipped for remyelination, being less mature on morphological and gene expression level and less sensitive to lesion-derived factors that impair remyelination, such as IFNγ and myelin debris compared to OPCs in WM [ 46 ]. GM primary rat OPCs also proliferate more and differentiate slower compared to WM OPCs [ 46 ]. Similarly, astrocyte diversity may contribute to lesion development and remyelination efficiency. GM and WM astrocytes differ in morphology and function; astrocytes in GM are more protoplasmic and mostly support the synapses, while astrocytes in WM are more fibrous and give support to axons at the nodes of Ranvier [ 38 , 47 , 48 ]. In addition, adult WM rat astrocytes are less supportive for in vitro myelination than adult GM rat astrocytes [ 49 ]. Furthermore, within each WM MS lesion type heterogeneity is observed [ 50 ], including local remyelinated areas [ 22 ], indicating that in addition to glial cell diversity also locally generated environmental cues are important determinants for successful remyelination. This emphasises the importance to analyse MS lesions at the nanoscale with spatial resolution, including analysis at the ultrastructural level.

Affected biological processes in non-demyelinated MS tissue

While demyelination is evident in MS lesions, subtle changes in biological processes are already present in non-demyelinated areas of people with MS, i.e., normal appearing white (NAWM) and grey matter (NAGM) (Fig.  2 ). Several MRI studies revealed abnormalities in NAGM and NAWM, with more extensive changes in NAWM [ 51 , 52 ], including disruption of blood–brain barrier (BBB) permeability [ 53 , 54 ]. Interestingly, the number of functional inhibitory synapses is reduced in NAGM, indicating that inhibitory neurons are more vulnerable to damage compared to excitatory neurons [ 55 – 57 ]. Furthermore, normal appearing brain tissue of people with MS exhibit increased microglia activation compared to brain tissue of non-demented donors, evidenced in the NAWM by altered phenotype, increased HLA-DR positivity, and increased binding of 11 C-PK11195(TSPO)-PET and in the NAGM by [F-18]PBR06(TSPO)-PET [ 58 – 60 ]. One study found a decrease in the number of homeostatic microglia marker P2RY12 in NAWM [ 58 ], while another study reported no difference [ 42 ]. Another macroscopically observable inflammatory change in NAWM is the proliferation of astrocytes [ 61 ]. Furthermore, an increase in protein carbonyl content in both NAWM and NAGM indicate increased ROS-mediated protein oxidation [ 62 ]. The deregulation of metals has also been implicated in MS pathogenesis. Sodium 23 ( 23 Na) magnetic resonance (MR) imaging revealed that total Na + -levels both in NAWM and NAGM are increased in advanced stages of RRMS [ 63 ]. Fig. 2 Abnormalities in non-demyelinated MS brain tissue . BBB blood–brain barrier, CGM control grey matter, CWM control white matter, MS multiple sclerosis, NAGM normal appearing grey matter, NAWM normal appearing white matter, ROS reactive oxygen species

Abnormalities in non-demyelinated MS brain tissue . BBB blood–brain barrier, CGM control grey matter, CWM control white matter, MS multiple sclerosis, NAGM normal appearing grey matter, NAWM normal appearing white matter, ROS reactive oxygen species

Differences in myelin composition and structure in normal appearing matter are also noted. In NAWM, the paranodal domains, which are specialised regions directly adjacent to the nodes of Ranvier, are elongated. This elongation is associated with microglial activation, axon stress and Kv1.2 channel dislocation into these paranodal regions [ 64 – 67 ]. Spectral coherent anti-Stokes Raman Scattering (sCARS) microscopy revealed subtle changes in myelin lipid biochemical signatures and intra-molecular disorder of fatty acid acyl chains in NAWM [ 68 ]. Similarly, subtle changes in lipid polarity of myelin are detected in NAWM using Nile Red spectrometry, which identifies fluorescence from Nile Red dye bound to lipids [ 69 , 70 ]. Lipid composition of lipids extracted from NAWM and NAGM shifts towards a higher proportion of phospholipids and a lower proportion of sphingolipids [ 71 ]. Myelin proteins like myelin basic protein (MBP) also harbour other post-translational modifications in NAM than in control matter, including citrullination and acylation [ 72 , 73 ]. In addition to compositional changes, changes in myelin morphology—such as myelin outfoldings and blister-like swellings formed by myelin detachment from axons—are observed in NAWM and are discussed in Sect. " Ultrastructural features of myelin pathology in MS ". Determining whether changes observed in NAM indicate either early pathology that increases the susceptibility to lesion formation or widespread and diffuse pathology independent of demyelination that drives PIRA is challenging. Therefore, in addtion to a detailed analysis of the ultrastructure in MS lesions, conducting similar analysis in normal appearing MS tissue could offer insights into pathological mechanisms, including local demyelinating or remyelinating processes.

Contribution of EM to our current understanding of MS pathology

Early EM studies in the 70’s on autopsy and biopsy tissue describe that within early MS lesions myelin is absent, only a few or no oligodendrocytes are present, axons appear normal, and most of the space between the axons is occupied by numerous fibrous astrocytic processes [ 74 , 75 ]. In addition, cytoplasmic inclusions could be observed [ 74 ]. Other studies in the 90’s, using brain biopsies from MS lesions, showed similar results [ 76 – 78 ]. More specifically, mature oligodendrocyte numbers are decreased and demyelinated axons are observed among astrocytic processes in areas of chronic demyelination [ 76 – 78 ]. In addition, a variable degree of widening of the extracellular space has been noted in some but not all studies [ 74 – 78 ]. These early EM studies are mostly descriptive and focus on small selected areas.

As biopsy material is scarce in MS research, most analyses on MS tissue are performed on autopsy tissue. Therefore, potential ultrastructural changes due to post-mortem delay should be considered. Immediately after death, ultrastructural changes include a linear increase in vacuolization and a decrease in recognisable synapses for up to 30–35 h, likely resulting from compression due to vacuolation [ 79 ]. Determining a single threshold for an acceptable post-mortem delay is challenging due to the highly variable rates of decomposition, which are influenced by factors such as storage temperature, pH levels, hydration, and oxygen content [ 80 ]. In addition, when using EM to study MS, it is important to consider differences between changes in ultrastructure created by normal ageing and those caused by MS. Using volume EM on mouse optic nerve, ageing results in larger axons, which have thicker myelin and longer and thicker mitochondria (increased mitochondrial volume) [ 81 ]. Furthermore, ageing results in reduced axon-myelin-node-mitochondrion-smooth endoplasmic reticulum (SER) connections, which suggest disruption in calcium homeostasis and the unfolded protein response [ 81 ]. Despite these changes, axons maintain their function and these changes likely do not lead to irreversible functional decline associated with neurodegenerative diseases. Age-related changes do not necessarily lead to the development of MS but may increase the risk of disease progression [ 82 ]. Also, the massive disruption of myelin sheaths is a characteristic feature of MS that does not occur during normal, healthy ageing. Considering factors such as ageing, fixation conditions, and post-mortem delay, EM analyses have uncovered or substantiated cell-type specific and molecular ultrastructural features in both normal appearing matter and MS lesions, which will be discussed next.

Ultrastructural features of myelin pathology in MS

Structurally, myelin is a multi-layered stack of double phospholipid bilayers virtually devoid of cytosol and visible in EM as a periodic structure of alternating electron-dense and electron-light layers, representing respectively the condensed cytoplasm and exoplasm (Fig.  3 a). In fact, EM has been very informative in unravelling myelin architecture. For example, EM clearly visualises non-compacted myelin, such as the paranodal loops, and the inner and outer tongue. In addition, EM demonstrated the presence of diffusion barriers in myelin, including the septate-like junctions at the paranodal loops and the tight junctions or radial component between the different layers of the internode. Clearly visible ultrastructural features of demyelination are the degeneration of inner glial loops and uniform widening of inner myelin lamellae [ 76 , 83 ]. Given its high resolution, EM has proven invaluable not only in confirming demyelination or remyelination, but also in studying the ultrastructural features of these processes. Moreover, EM is instrumental in identifying early changes in myelin architecture that precede extensive demyelination. Fig. 3 Normal and abnormal CNS myelinated axons . a. Schematic overview of normal CNS myelinated axons. CNS central nervous system, IT inner tongue, OT outer tongue, r axon radius, R myelinated axon radius, NF 155 neurofascin-155, Caspr 1 contactin-associated protein 1, PLP proteolipid protein, MBP myelin basic protein. Bar: 100 nm. b. Schematic overview of different types of abnormal CNS myelinated axons in MS lesions; outfolding, vesiculation, blister, vacuolation, and axonal degeneration. Myelin outfoldings, vesiculation, blisters, and vacuolation are thought to represent early events that ultimately lead to myelin degeneration. Axonal degeneration is marked by initial swelling and organelle accumulation, which then likely transitions into disintegration of these organelles reflected by a condensed axoplasm

Normal and abnormal CNS myelinated axons . a. Schematic overview of normal CNS myelinated axons. CNS central nervous system, IT inner tongue, OT outer tongue, r axon radius, R myelinated axon radius, NF 155 neurofascin-155, Caspr 1 contactin-associated protein 1, PLP proteolipid protein, MBP myelin basic protein. Bar: 100 nm. b. Schematic overview of different types of abnormal CNS myelinated axons in MS lesions; outfolding, vesiculation, blister, vacuolation, and axonal degeneration. Myelin outfoldings, vesiculation, blisters, and vacuolation are thought to represent early events that ultimately lead to myelin degeneration. Axonal degeneration is marked by initial swelling and organelle accumulation, which then likely transitions into disintegration of these organelles reflected by a condensed axoplasm

Demyelination (onset)

In MS, the onset of demyelination is still unknown, but ultrastructural analysis of MS lesions revealed a decrease in oligodendrocyte numbers which impairs normal myelinating function [ 83 , 84 ]. In addition, perivascular inflammatory cells, such as lymphocytes, plasma cells, and macrophages are noted, occasionally in contact with myelin [ 76 ]. EM analysis on biopsy tissue revealed that in active demyelinating MS lesions, myelin damage is characterised by the formation of local myelin outfoldings, also called myelinosomes (Fig.  3 b) [ 84 ]. Myelin outfoldings are also observed in biopsy perilesional WM [ 84 ] and post-mortem control white matter (CWM) and NAWM [ 66 ], indicating that myelin damage can be initiated independent of axon damage. The outfoldings are hardly observed in the toxin-induced demyelination experimental models. In contrast, in experimental autoimmune encephalomyelitis (EAE), an inflammatory demyelination model for MS (Box 1 ), myelin outfoldings are induced by anti-myelin antibodies [ 84 ]. This indicates that structural damage is at the surface of the myelin sheaths and then extends to the cell body. Detached, folded and excessive myelin in post-mortem MS lesions and EAE, even at non-inflammatory areas, are also clearly visible with scanning electron microscopy (SEM) [ 85 ], and independent of axon size. SEM analysis in EAE also visualised focal myelin infoldings and areas where myelin is detached from the axon. In addition, immunohistochemical analysis demonstrated that blister-like swellings formed by myelin detachment from axons is more frequently observed in non-inflamed NAWM than in CWM of non-neurological donors and other neuropathologies (Fig.  3 b) [ 67 ]. These swellings showed altered glutamate receptor expression, MAG distribution, lipid biochemical composition and increased citrullination in myelin sheaths [ 67 ]. Detailed ultrastructural analysis of post-mortem optic nerve tissue MS tissue revealed that NAWM myelin is less compact compared to CWM myelin [ 66 ]. The close apposition of the two phospholipid bilayers within one layer is facilitated by MBP. Ultrastructural analysis in a rat model of neuromyelitis optica, also a demyelinating CNS disease, unveiled that the underlying mechanism for the disruption of myelin could be the aberrant transition of MBP molecules from their cohesive state to their non-adhesive state. This results in the loss of interaction between MBP and the membrane, eventually resulting in myelin vesiculation (Fig.  3 b) [ 86 ]. Elevated intracellular calcium levels that disassemble MBP molecular networks are the underlying mechanism of myelin vesiculation [ 86 , 87 ]. Therefore, it could be hypothesised that an unknown mediator damages the oligodendrocyte cell body, leading to an increase in intracellular calcium, which rapidly spreads into the myelin sheath, resulting in myelin vesiculation. Vesiculations at the inner tongue of myelin are observed by EM in EAE and vacuolations in the cuprizone model most frequently at paranodes, indicating a common mechanism of myelin breakdown [ 86 , 88 ]. This loss of paranodal organization precedes demyelination.

At the internodes of myelinated axons, activity-dependent communication occurs between the axon and myelin through specialised chemical synapses, known as axo-myelinic synapses [ 89 ]. 2-Photon microscopy revealed that axonal action potentials result in a rise in calcium levels in optic nerve myelin [ 90 ]. This effect was mediated by myelinic GluN2D and GluN3A-containing NMDA receptors. On the ultrastructural level, no major dysmyelinating phenotype is observed in the single or double knockouts of the GluN2D and GluN3A subunits [ 90 ]. Double knockout mice show increased myelin thickness due to an increased spacing between myelin lamellae, but not an increased number of myelin wraps [ 90 ]. This might indicate that NMDA receptors, which are expressed at the myelin-axon adhesion site, regulate expression of myelin proteins to modulate the myelin nanostructure. This in turn might influence the conduction velocity of propagated action potentials, because of changes in the biophysical properties of myelin [ 90 , 91 ]. Disturbances of this axo-myelinic synapse might play a role in the pathophysiology of MS [ 67 , 89 , 92 , 93 ].

Microglia, the macrophages of the CNS, play an early role in the pathogenesis of MS. Clusters of activated microglia in non-demyelinated NAWM are identified as pre-active lesions [ 22 , 29 ]. Pre-active lesions are thought to either resolve or represent early stages in the formation of lesions [ 94 ]. In active MS lesions, reactive microglia phagocytose myelin, present antigens to T cells and/or release proinflammatory cytokines [ 95 ]. In MS, monocyte-derived macrophages infiltrate the CNS and are difficult to distinguish from microglia because of their similarities in morphology and function. However, using serial block-face scanning electron microscopy (SBF-SEM), Yamasaki et al. showed that, in murine EAE, monocyte-derived macrophages associate with nodes of Ranvier and initiate demyelination, whereas microglia appear to clear debris [ 96 ]. This is also confirmed by their respective gene expression profiles; the gene expression of monocyte-derived macrophages reflects their phagocytic and inflammatory phenotype, while microglia show downregulation of cellular metabolism-related genes [ 96 ]. Accordingly, EM reconstructions and parallel immunohistochemistry of cell-specific markers demonstrated that blood-derived macrophages preferentially engulf myelin outfoldings [ 84 ], and thus are more involved in the stripping of myelin rather than myelin debris clearance. Hence, there seem to be distinct roles for microglia and monocytes in the inflamed CNS and distinguishing these cells will help to better understand inflammatory pathologies [ 96 ]. Furthermore, recent EM studies identified a new microglial phenotype, named dark microglia, which are rarely seen in healthy conditions and become abundant during chronic stress, ageing, fractalkine signalling deficiency (CX3CR1-/- mice), and Alzheimer’s disease (AD) pathology (APP-PS1 mice and post-mortem human AD brain) [ 97 – 100 ]. These cells can be recognised by their condensed, electron-dense cytoplasm and display many signs of cellular stress, including nuclear and chromatin condensation and dilation of their endoplasmic reticulum [ 97 ]. Whether dark microglia are also present in the context of MS remains to be elucidated.

Initially, astrocytes were primarily associated with the formation of a glial scar following the completion of demyelination. However, astrocytes also play an early role in MS pathogenesis by engaging in phagocytosis of myelin debris and facilitating the recruitment of immune cells [ 101 ]. On ultrastructural level in non-demyelinated NAWM sampled from biopsies, astrocytes exhibited an increased number of lysosomes [ 102 ]. These lysosomes, found in the perikaryon and cell processes, often contain material resembling myelin membranes [ 78 , 102 , 103 ], indicating that astrocytes are phagocytosing myelin prior to detectable demyelination. Another EM study, using stereotaxic brain biopsies from MS subjects, demonstrates that astrocytes proliferate away from the demyelinated areas [ 76 ]. In addition, astrocytic processes and collagen deposition could be observed in demyelinated areas, indicating astrocyte-driven collagen production [ 76 ]. Although astrocytes are increasingly recognized as key players in MS, the amount of ultrastructural data pertaining to these cells in the context of demyelination and remyelination in MS is restricted to these studies, along with the observation of filament-rich astrocytic processes within demyelinated MS lesions [ 76 ].

Remyelination (failure)

Remyelination is the spontaneous regeneration of myelin following demyelination resolving functional deficits and preventing axonal degeneration. Determination of the g-ratio by EM is the most reliable method and golden standard for assessing (developmental) myelination and remyelination. The g-ratio is defined as the ratio between the inner axon radius, and the other (myelinated) axon radius (Fig.  3 a), and assumed to be constant across axons [ 104 ]. In experimental models for successful remyelination, remyelinated axons can initially typically be identified by thin myelin resulting in a higher g-ratio, and shorter internodes. As a result, in MS brain tissue remyelinated areas appear fainter in (immuno)histochemistry, and are therefore called ‘shadow plaques’. The appearance of thinner myelin and shorter internodes is also observed during myelin remodelling in the adult rodent CNS, and likely a feature of adult myelination and not exclusive to remyelination [ 105 ]. Notably, in rodents, remyelinated regions eventually resolve in time, closely resembling areas with axons myelinated during development [ 106 ]. Based on this observation, Neumann et al. suggested that NAWM may also harbour remyelinated regions [ 106 ], but whether this is also applicable to humans and whether the content and ultrastructure of myelin in remyelinated axons differ from myelin of axons myelinated during development, has not been extensively investigated yet.

In experimental rodent models of reversible demyelination, myelin is mainly regenerated by newly-formed oligodendrocytes that matured from adult OPCs rather than surviving mature oligodendrocytes [ 107 ], while the remyelination capacity of surviving mature oligodendrocytes is improved following learning [ 108 ]. Data on the ultrastructural level indicated that surviving mature oligodendrocytes can indeed participate in remyelination. Using light and EM analyses, Duncan et al. visualised in two dietary non-rodent large animal models of demyelination oligodendrocytes with cytoplasmic connections to both thick, mature myelin sheaths and thin, likely remyelinated sheaths [ 109 ]. By measuring the integration of 14 C-derived from nuclear testing in genomic DNA, i.e., cell birth dating, oligodendrocytes in MS shadow plaques are identified as ‘old’ oligodendrocytes [ 25 ]. The absence of newly-formed oligodendrocytes in shadow plaques suggests that myelin is either regenerated transiently or not at all, or generated by pre-existing oligodendrocytes. However, surviving oligodendrocytes exhibit limited capacity to form new myelin sheaths [ 110 ]. If myelin sheaths were made by these cells, their growth along axons appeared relatively normal, although immunohistochemical analysis in zebrafish and post-mortem MS lesions revealed that occasional myelin is generated around cell bodies. In contrast, newly-generated oligodendrocytes produce a higher number of sheaths in zebrafish than pre-existing oligodendrocytes [ 110 ]. Detailed large scale and volume EM analysis of shadow plaques and other MS lesion types may provide more insights in remyelination features of lesion-free MS tissue. For instance, EM analysis may discriminate between remyelinating and remyelinated areas, by for example the presence of a few axons enwrapped with non-compact myelin membranes and more ribosomes in oligodendrocytes during active remyelination. Using automated tape-collecting ultramicrotome-SEM (ATUM-SEM) on mouse tissue, an elaborated network of tubular ER was identified within the inner tongue of developing myelin, and contributes to membrane extension [ 111 ]. Hence, the presence of an ER network in the inner tongue, where active myelin growth occurs, may be used to identify remyelinating areas. Additional analysis of these axons, extracellular space and neighbouring cells may provide ultrastructural clues on why for example remyelination fails or occurs.

Ultrastructural features of neuronal pathology in MS

In the CNS, neurons are fundamental functional units responsible for transmitting and processing information. GM areas, particular the cerebral cortex, are populated with neuronal cell bodies and dendrites, while neurons project their axons to WM areas, such as the corpus callosum. Neuronal pathology, and axon degeneration in particular, rather than myelin pathology perse, correlates with irreversible clinical disability in MS [ 112 , 113 ]. Although axonal injury is a consequential outcome of (persistent) demyelination, axonal pathology also occurs independently of inflammation and demyelination. Injured axons often display (focal) swellings and accumulation of organelles (Fig.  3 b) [ 114 ]. In the MS cortex, there is also a reduction in synapses, leading to disruptions in neuronal connectivity and function [ 55 – 57 ]. The mechanisms driving synaptic changes at neuronal axons and/or dendrites, and whether this precedes demyelination, remain to be determined. In the following we delineate how EM contributed to our knowledge on axonal and synapse pathology in MS.

Axonal pathology

Subtle axonal pathology can already be observed in the absence of demyelination. Using large scale 2D scanning electron microscopy (STEM) images (nanotomy), we recently noted a reduced density of myelinated axons in post-mortem MS NAWM brain tissue without a concurrent decrease of their cross-sectional area, suggesting that a reduction in axonal density is compensated by axonal swelling [ 115 ]. Despite this change, the g-ratio remains similar, which led us to hypothesise that the swelling of myelinated axons triggered myelin membrane synthesis to maintain an optimal g-ratio. In line with these findings, EM and advanced microscopy analysis revealed reduced axonal density in perilesional WM [ 116 ] and focal swelling of myelinated axons in MS NAWM [ 67 ]. Similarly, ultrastructural analysis at different stages of EAE demonstrated that swelling of myelinated axons precedes demyelination [ 117 ]. Notably, some of the swollen axons recovers spontaneously, indicating that axonal swelling can be reversible. Serial block-face scanning electron microscopy (volume EM) demonstrated swellings of myelinated axons upon cuprizone-mediated demyelination [ 114 ]. Multiple causes may lead to axonal swelling. For example, a lack of trophic support by oligodendrocytes could become detrimental for axons that are insulated by dysfunctional myelin [ 118 ]. Alternatively, axonal swelling could be initiated independent from oligodendrocytes. For example, Wallerian degeneration, which is the degeneration of the axon distal to an injury prior to demyelination, is also accompanied by axonal swelling [ 119 ].

Axonal swelling accompanied by organelle accumulation is generally considered as a sign of subsequent axonal disintegration or degeneration [ 120 ]. Extensive EM analysis of control CNS tissue revealed that the axoplasm consists of various organelles, such as mitochondria and dense bodies, and cytoskeletal elements, including microtubules and neurofilaments. The axon does not contain rough ER and Golgi, but forms an axoplasmic reticulum, i.e., a neuronal endomembrane system with continuous lengthwise delicate membranous tubules and flattened sacs, acting as a calcium reservoir and is involved in axonal transport. A well-developed axoplasmic reticulum is also noticed in axons in MS tissue [ 85 ]. SEM analysis during the course of EAE, demonstrated that focal development of a well-developed axoplasmic reticulum correlates with abnormal myelin and accumulation of mitochondria, which was not observed in cuprizone-induced demyelinated areas [ 85 ]. However, at early stages of cuprizone-induced demyelination, organelle-filled axons initially accumulate, but decline as demyelination proceeds [ 118 ], indicating reversible axonal injury. Hence, besides alterations in radius, axonal swelling is characterised by a diminished neurofilament spacing and the accumulation and swelling of intracellular organelles forming spheroids.

Axons in EAE display a spectrum of distinct morphologies ranging from focal swellings to fragmentation representing sequential progressive stages of axon damage [ 117 ]. Similarly, in MS lesion biopsy tissue, distinct forms of axonal damage are observed [ 118 ], including myelinated and demyelinated axons with organelle accumulation and myelinated axons with a condensed axoplasm (Fig.  3 b). Dense axons, likely representing decomposed organelle and cytoskeletal elements, are more frequently seen in lesions with less pronounced demyelination and at the lesion rim. The different structural forms of myelinated axons are considered transitional stages, where myelinated axons with condensed axoplasm represent irreversibly damaged axons [ 118 ]. In specific context, such as an acute autoimmune environment, retained myelin can become a risk for axon degeneration [ 118 ]. Thus, while myelin is typically considered protective for axons, its persistent presence on damaged axons in immune-related damage can have adverse effects.

Axonal swellings with organelle accumulation are triggered by mitochondrial pathology and induced by macrophage-derived reactive oxygen (ROS) and nitrogen species [ 117 ]. Swollen axons have swollen mitochondria that are a potential sign of mitochondrial dysfunction. In EAE, myelinated axons that reside in perilesional areas of inflammation also have swollen mitochondria [ 117 ], pointing to ultrastructural alterations of mitochondria as an early sign of axon degeneration and myelin damage. In post-mortem MS NAWM that is not adjacent to the lesion, the cross-sectional surface area of mitochondria was not different [ 66 , 115 ]. However, 3D EM reconstruction is required to appreciate mitochondrial alterations in size. Furthermore, several studies point toward a role of mitochondrial dysfunction as the underlying mechanisms or cause of axonal degeneration following demyelination [ 121 ]. Mitochondria play an important role in calcium handling and ROS production. Indeed, in EAE, calcium influx from the extracellular space leads to increased intra-axonal calcium that drives axonal degeneration [ 122 ]. Demyelinated axons are more susceptible to energy deficits as evidenced by mitochondrial changes and redistribution of Na-channels in axons that lack a healthy myelin sheath [ 123 ]. To adapt to the increased energy demand in the absence of myelin, mitochondrial density and size increase upon demyelination. This increased energy requirement subsequently promotes ROS formation and thereby axonal loss [ 124 – 126 ]. This hypothesis is supported by increased mitochondrial protein expression as well as a significant increase in the number of mitochondria in axons and astrocytes in active and inactive MS lesions [ 127 ]. On the ultrastructural level, the number of mitochondria increase in several rodent models, including Shiverer mice, TMEV-infected mice and Plp1-overexpressing mice [ 128 – 130 ], and upon toxin-induced demyelination, including demyelination induced by lysolecithin and cuprizone [ 131 , 132 ] (Box 1 ). Ultrastructural analysis in different species, including human, revealed that the size of axonal mitochondria is inversely linked to myelin thickness, a correlation that is lost upon demyelination and re-established at robust remyelination [ 133 ]. The homeostatic balance between mitochondrial radius and g-ratio is lost in post-mortem MS NAWM, but not in MS NAGM, control GM and control WM [ 115 ]. The obtained insights in the complex interplay between myelin, axons and immune responses in MS are a good example of the value of ultrastructural analysis to uncover MS pathology.

Synapses

Synapses, abundantly present in GM, are essential for CNS function enabling neuronal communication, information processing, learning, memory and the coordination of neural activity. Excitatory synapses are predominantly located on dendrites and dendritic spines, while inhibitory synapses are mainly found on the cell soma and axonal initial segment, with sparse distribution along dendrites. In demyelinated and non-demyelinated post-mortem GM MS tissue, synapse density decreases [ 134 – 136 ], especially for inhibitory synapses [ 55 – 57 ], altering the balance between excitatory and inhibitory synaptic activity. The extent of synapse loss likely varies across GM brain regions, for example hippocampus, thalamus, cerebellum, spinal cord, and cortex, between cortical layers [ 137 ]. Synapse loss is likely attributed to inflammation [ 135 , 138 ], specifically from C3-complement-activated microglia that engulf and strip synapses [ 139 , 140 ]. In fact, as analysed in post-mortem tissue, in the MS cortex, a particular microglia population may initially remove pre-synapses from neuronal soma as a protective measure to prevent neuronal loss [ 141 ]. Synapse elimination has also been demonstrated in experimental models [ 139 , 142 – 144 ]. In EAE, synapse loss in the spinal cord is most pronounced at the disease peak and recovers during remission [ 142 ]. Ultrastructural analysis demonstrated a significant elimination of both inhibitory and excitatory synapses on neuronal cell bodies during peak disease, with partial recovery during remission stages. However, the recovery of inhibitory synapses is particularly incomplete during the second remission, leading to an increased proportion of excitatory synapses. This is likely driven by a combination of pro-inflammatory cytokines and axonal damage. In addition, synaptic changes in the visual thalamus, independent of neuronal loss, are observed in the cuprizone model [ 144 ].

Synaptopathy in MS GM occurs early and independently from demyelination and axonal loss, and leads to excitotoxic damage. Therefore, further delineation of abnormal synapses may provide therapeutic opportunities to target synaptopathy, preserving synaptic function and potentially preventing widespread synapse loss and neuronal death in later stages of the disease. To achieve this, gaining deeper insight in changes in synaptic structure and function is crucial. Most studies used immunofluorescent staining and confocal analysis to measure synaptic loss, assuming that staining of a single synapse marker or the colocalization of presynaptic and postsynaptic markers represent mature functional synapses [ 145 ]. Although still static, ultrastructural analyses of synaptic features allow morphological observation of synapses and their surrounding cells. For example, in an EM study from the 70’s of biopsy MS NAGM tissue, ultrastructural features are observed in post-synapses, such as twisted neurotubules, while the pre-synapse appears normal [ 146 ]. In EM, the postsynaptic density of excitatory synapses appears as a prominent electron-dense thickening underneath the synaptic cleft, enabling analysis of the proportions excitatory and inhibitory synapses and separate characterisation of ultrastructural characteristics. For example, due to their small size, EM can visualise synapses and their fine structural features, elucidating alterations in (i) the width and length of the synaptic cleft, (ii) number, shape, aggregation and localisation of synaptic vesicles, and (iii) the presence and morphology of presynaptic and postsynaptic mitochondria. Alterations in these ultrastructural synapse characteristics may have functional implications [ 147 ]. For instance, a reduction in synapse width increases the probability of neurotransmitter release [ 147 ]. EM reveals not only synaptic morphology but also surrounding cells, which can elucidate roles of microglia in synapse pruning and astrocytes in synapse plasticity and maintenance, thus enhancing our understanding of synaptopathy. Ultrastructural evidence from the cerebellar dentate nucleus of a single MS case suggests at least two distinct mechanisms of synaptic loss in MS, namely by a glia‐mediated and a neuron‐autonomous process [ 148 ]. Lysosomal degradation of synaptic components, indicative of synaptic autophagy, is observed in dendrites but not in microglia, along with aggregation of remaining synaptic vesicles and their dislocation from the synaptic cleft [ 148 ]. Moreover, presynapses and postsynapses are separated from neuronal cell bodies by filamentous-rich astrocytic process containing glycogen, indicating a role for astrocytes in (axosomatic) synaptic dysfunction in MS [ 148 ]. Hence, further delineating the ultrastructural organisation of (aberrant) synapses and the surrounding glial cells in both non-demyelinated and demyelinated areas is crucial for understanding the mechanisms underlying (primary) synapse pathology in MS.

Ultrastructural features of BBB pathology in MS

The BBB consists of endothelial cells that line the surface of blood vessels forming an interface between the circulating blood and the brain extracellular fluid to preserve brain homeostasis (Fig.  4 a). Pericytes partially surround the endothelium and are embedded in the basal lamina. The endothelial cells, pericytes and basal lamina are enclosed by astrocytic perivascular endfeet, which provide the cellular link to neurons. Microglia are also closely associated with the endothelium [ 149 ]. Pathology of the BBB plays a substantial role in the pathogenesis of MS. In the early stages of MS, i.e., in active MS lesions, the BBB is disturbed and peripheral immune cells migrate into the CNS. As the disease progresses, the BBB becomes ‘closed’, i.e., decreased BBB permeability reflecting compartmentalization of inflammation [ 150 ]. In addition, fibrillar collagens are strongly induced in active lesions and to an even greater extent in inactive lesions [ 151 ]. Fig. 4 Ultrastructural features of BBB pathology in MS . a Schematic overview of the blood–brain barrier (BBB) illustrating endothelial cells with tight junctions lining the blood vessels, partially surrounded by pericytes enclosed by the basal lamina, and further encased by astrocytic endfeet. b Electron microscopic image of the BBB in post-mortem NAWM. c. Electron microscopic image of the BBB in a post-mortem active MS lesion, showing swollen astrocyte end feet and collagen deposition in longitudinal (top inset) and cross-sectional view (bottom inset). CL capillary lumen, RBC red blood cell, BL basal lamina, EC endothelial cell, TJ tight junction, AC astrocyte (endfeet), COL collagen. Bars are 2 µm (overview) and 500 nm (insets)

Ultrastructural features of BBB pathology in MS . a Schematic overview of the blood–brain barrier (BBB) illustrating endothelial cells with tight junctions lining the blood vessels, partially surrounded by pericytes enclosed by the basal lamina, and further encased by astrocytic endfeet. b Electron microscopic image of the BBB in post-mortem NAWM. c. Electron microscopic image of the BBB in a post-mortem active MS lesion, showing swollen astrocyte end feet and collagen deposition in longitudinal (top inset) and cross-sectional view (bottom inset). CL capillary lumen, RBC red blood cell, BL basal lamina, EC endothelial cell, TJ tight junction, AC astrocyte (endfeet), COL collagen. Bars are 2 µm (overview) and 500 nm (insets)

EM helps in visualising morphological alterations and perivascular inflammatory cells of the neurovascular unit in MS brain tissue. An ultrastructural study on post-mortem tissue of active (post-phagocytic) lesions reveals a marked disruption of astrocyte endfeet, leaving bare stretches of glia limitans. In contrast, the astrocyte endfeet appear normal in chronic lesions [ 152 ]. In another study, using biopsy samples of CNS tissue from chronic progressive MS patients, brain capillaries were analysed morphometrically for alterations in organelle distribution and density [ 153 ]. An increase in the density of pinocytotic vesicles and a decrease in the density of mitochondria in gliotic areas compared to normal appearing matter is noticed. Furthermore, accumulation of perivascular fibrin, perivascular collagen deposits, degenerative changes in pericytes and swelling of astrocytic endfeet are observed [ 153 ]. Another EM study on post-mortem perfusion-fixed WM MS lesions reported thickening of the blood vessels with excessive collagen deposition, and a widened perivascular space, while endothelial cells appear well lined and morphologically normal [ 77 ]. In our own large scale EM datasets, we also observed BBB abnormalities in WM MS lesions, such as swelling of astrocytic endfeet and fibrillar collagen deposition (Fig.  4 b, c). These ultrastructural changes are suggestive of an increased permeability of the BBB in WM MS lesions, evidenced by the presence of peripheral immune cells, including plasma cells.

Ultrastructural changes of the BBB are also visible in animal models. In the cuprizone mouse model, which is a model of de- and remyelination (Box 1 ), tight junctions of adjacent endothelial cells are separated and endothelial and astroglial basement membranes are locally split [ 154 ]. Furthermore, hypertrophic astrocytes and sporadic endothelial cells with atypical ultrastructure are present in cuprizone-demyelinated tissue [ 154 ]. Additionally, cuprizone treatment leads to the swelling of astrocyte endfeet [ 155 ]. Although, in the cuprizone mouse model, there is minimal infiltration of peripheral lymphocytes and monocytes [ 156 ], and only a slight loss of BBB integrity [ 154 ]. In addition, lipopolysaccharide (LPS)-induced neuroinflammation in mice led to ultrastructural remodelling of the BBB. LPS-treated mice exhibited increased brain endothelial cell transcytosis, remodelling of pericytes and activation of brain endothelial cells [ 157 ]. In EAE guinea pigs, the ultrastructural integrity of the endothelial basal lamina is disturbed upon infiltration of mononuclear cells through the endothelia [ 158 ]. These studies indicate that demyelination and/or inflammation can induce ultrastructural changes of the BBB.

Ultrastructural alterations of the BBB have been mostly studied in MS animal models and studies in humans are limited. Expanding research to include larger human cohorts and a variety of lesion types could provide more insights. For example, investigating whether specific morphological alterations in the BBB – such as collagen deposition, swollen astrocyte endfeet, and the presence of peripheral immune cells – correlate with the extent of demyelination and inflammation in MS, and whether these changes are associated with functional differences and disease severity.

Recent advances in EM techniques and its potential to study MS tissue

The previous highlight the importance of EM as a valuable and powerful tool to most precisely assess the morphology of cross sections of single axons and myelin pathology. The EM field is rapidly evolving and more advanced techniques are now widely available that can provide novel insights into MS pathology, including lesion initiation and progression, remyelination failure and diffuse pathology.

The two main types of electron microscopy are transmission electron microscopy (TEM) and scanning electron microscopy (SEM) (Fig.  5 a) [ 159 ]. TEM imaging enables high resolution imaging of the cellular ultrastructure by exposing ultrathin sections to a parallel beam of electrons and projecting the electrons that pass through. In SEM, a focused electron beam is used instead which scans a sample in a raster-like fashion, enabling pixel-by-pixel mapping of various signals. The incident beam can eject electrons from the probed atom, referred to as secondary electrons (SE), which can only escape the sample’s surface due to their relatively low energy thereby enabling topographic imaging. However, contrast formation in SEM can additionally be based on the scattering of electrons, which can be both transmitted (STEM) and electrons that get scattered back (i.e., backscatter electrons (BSE)). While both enable the cellular ultrastructure to be investigated, backscatter imaging is compatible with thicker samples [ 160 ]. To fully capture the complexity of biological structures in tissues, volumetric information is preferred, which can be obtained using both TEM- and SEM-based volume EM approaches (Fig.  5 b) [ 159 ]. TEM-based approaches include classical 3D reconstruction from serial sections (ssTEM) and electron tomography. The axial resolution of 3D reconstruction from serial sections is limited to the section thickness. Electron tomography uses a series of EM images taken from different angles to reconstruct a 3D representation of a sample. SEM-based volume EM approaches include focused ion beam SEM (FIB-SEM), serial block-face SEM (SBF-SEM) and array tomography/serial-section scanning electron microscopy. With FIB-SEM, an additional beam, the focused ion beam, etches the surface of the sample between images taken by the electron beam achieving an axial resolution of only a few nanometres that can be calibrated to the XY resolution for isotropic imaging. With SBF-SEM, the sample’s surface is disposed by cutting the top layer with a diamond knife between images, resulting in a larger Z step size (15–30 nm) but of larger XY areas. Unlike FIB-SEM and SBF-SEM, array tomography is nondestructive, imaging serial sections sequentially. Fig. 5 Overview of different electron microscopy (EM) techniques . a In SEM (scanning EM), secondary electrons (SE) and backscattered electrons (BSE), are primarily used to form an image. In (S)TEM (transmission EM), transmitted electrons,  those that pass through a sample, are detected to create an image. b Overview of volume electron microscopy (EM) techniques: ssTEM (serial sectioning transmission EM), electron tomography, FIB-SEM (focused ion beam scanning EM), SBF-SEM (serial block-face scanning EM), and array tomography. c The elemental composition of samples can be determined in EM using energy dispersive X-ray spectroscopy (EDX), which detects characteristic X-rays emitted from a sample, and electron energy loss spectroscopy (EELS), which measures the energy lost by electrons as they pass through a sample. EDX and EELS are illustrated using nitrogen (N) as an example. N has seven electrons in two orbital shells: K and L. When an incident electron from the electron beam ejects a K-shell electron, the unstable vacancy is filled by an L-shell electron, releasing the energy difference as a characteristic X-ray used to fingerprint N in the collected spectra. The incident electron loses energy roughly equal to the binding energy of the ejected electron, enabling identification of N in the EELS spectrum

Overview of different electron microscopy (EM) techniques . a In SEM (scanning EM), secondary electrons (SE) and backscattered electrons (BSE), are primarily used to form an image. In (S)TEM (transmission EM), transmitted electrons,  those that pass through a sample, are detected to create an image. b Overview of volume electron microscopy (EM) techniques: ssTEM (serial sectioning transmission EM), electron tomography, FIB-SEM (focused ion beam scanning EM), SBF-SEM (serial block-face scanning EM), and array tomography. c The elemental composition of samples can be determined in EM using energy dispersive X-ray spectroscopy (EDX), which detects characteristic X-rays emitted from a sample, and electron energy loss spectroscopy (EELS), which measures the energy lost by electrons as they pass through a sample. EDX and EELS are illustrated using nitrogen (N) as an example. N has seven electrons in two orbital shells: K and L. When an incident electron from the electron beam ejects a K-shell electron, the unstable vacancy is filled by an L-shell electron, releasing the energy difference as a characteristic X-ray used to fingerprint N in the collected spectra. The incident electron loses energy roughly equal to the binding energy of the ejected electron, enabling identification of N in the EELS spectrum

By using advanced volume EM methods it is possible to study the structure of interest as a whole and thereby providing more information compared to a 2D cross-section of a structure. For example, upon inducible ablation of MBP in adult mice, FIB-SEM uncovered internode shortening, membrane tubules at the inner tongue, and the presence of myelinoid bodies at the abaxonal myelin and protruding into the inner tongue [ 161 ]. In addition, FIB-SEM revealed outfoldings in Plp1- or Mag-null mutant myelin, with axonal pathology underneath, including axonal sprouting [ 162 ]. Furthermore, 3D reconstructions of oligodendrocyte morphology using SBF-SEM revealed that an individual oligodendrocyte enwraps myelin around axons of a limited range of diameters [ 163 ]. ATUM-SEM, a type of array tomography, identified a complex ER network close to sites of myelin membrane biogenesis during development [ 111 ]. These examples illustrate the advantages of volume EM compared to 2D EM.

In addition to ultrastructural mapping, the elemental composition of a sample can simultaneously be measured using electron energy loss spectroscopy (EELS) or energy-dispersive X-ray spectroscopy (EDX) (Fig.  5 c) [ 164 ]. The physical principles of both detection methods is based on the incidental removal of an electron from one of the inner shells of an element by the electron beam. To fill the unstable vacancy, an electron from an outer shell fills its place and releases an characteristic x-ray which is specific for the element. This principle makes it possible to determine which elements are present by detecting either the energy loss of the incident electron with EELS or the characteristic x-ray with EDX. The obtained elemental map can be shown in grey-scale or multiple elemental maps can be combined by overlaying single color images (‘ColorEM’). In addition to endogenous elements, exogenous elements are also introduced during sample preparation. For example, heavy metals such as osmium are used to fix membranes and further enhance contrast, thereby improving visualisation. Information about the spatial elemental composition of a sample can contribute to better understanding of the processes involved in MS. For example, calcium, magnesium, zinc and copper are present in purified human myelin [ 165 ] and MBP binds to some of these elements, i.e., magnesium, zinc and copper [ 166 ]. Despite the documented posttranslational modifications of MBP in MS, including citrullination, methylation, deimination and phosphorylation [ 167 , 168 ], it remains unclear whether these modifications affect MBP’s metal-binding capacity. In addition, the distribution of iron and zinc is heterogeneous in MS lesions [ 169 ]. Iron tends to decrease with increasing age and disease duration and is safely stored in ferritin by reactive astrocytes in a subset of inactive lesions [ 169 ]. A subset of mixed active/inactive lesions are bordered by a rim of iron-loaded macrophages/microglia [ 169 ], which are associated with a more severe disease course [ 170 ]. Such lesions are detectable on MRI as paramagnetic rim lesions and are very specific for MS. Zinc in MS lesions is generally decreased [ 169 ]. Hence, visualisation of these elements in MS brain tissue at the nanoscale could contribute to a better understanding of myelin and lesion pathology.

With EM, cellular context is imaged in high resolution but specific molecules are hard to define. To localize specific molecules, it is possible to label them using immunolabelling with electron-dense particles, such as gold (immunoEM) [ 171 ]. Correlated light and electron microscopy (CLEM) can also be used, which combines fluorescence microscopy with EM [ 171 ]. Fluorescence can be used to preselect specific areas of interest in post-mortem tissue, followed by more detailed EM analysis. For example, myelin was labelled in acute murine EAE lesions to find myelin outfoldings that were subsequently imaged for EM [ 84 ]. However, labelling in post-mortem tissue is challenging because genetically encoded molecular labels can not be applied and antigen masking from EM fixation complicates immunolabelling. Despite these challenges, advancements in immunolabelling are improving the application of CLEM to post-mortem brain tissue [ 172 ]. Additionally, testing over 30 different antibodies, different CNS cells could be labelled using CLEM in post-mortem brain tissue from multiple system atrophy donors [ 173 ]. Applying CLEM to post-mortem MS brain tissue could allow for yet unexplored investigation of cellular interactions in MS disease pathology, such as detailed examination of direct physical interactions between cell types in lesions including the possibility to assess organelle dysfunction in each cell type. CLEM offers the advantage of not obscuring ultrastructural details, unlike labelling methods such as diaminobenzidine (DAB), immunogold, or quantum dots, which can mask ultrastructure by adding label intensity to the original EM contrast. Recent developments, such as large-scale EM and super-resolution fluorescence, decreased the gap between EM and fluorescence microscopy [ 172 ].

The above mentioned EM techniques can also be combined. Additionally, imaging speed can be increased by using multiple electron beams (Delmic’s FAST-EM, Zeiss’ multi-SEM) instead of a single beam [ 174 , 175 ] and multiple parallel cameras (TEMCA) to enlarge the field of view of the microscope [ 176 ]. For example, the acquisition speed of large-scale volume EM can be increased by using FAST-EM. Furthermore, adjacent tissue sections could be used for different techniques, as was recently demonstrated using spatial transcriptomics (MERFISH) and large-scale EM [ 177 ]. Combining these approaches facilitates a more effective integration of transcriptional, translational, elemental, and ultrastructural data in a spatial context which will help us learn more about MS pathology.

Concluding remarks

EM studies on post-mortem MS tissue have benefitted insights in MS pathology and will continue to evolve with the advancement in EM microscopy modalities, including large scale imaging. Table 1 provides a summary of the outlined contributions of EM in uncovering MS pathology, its future potential, and clinical relevance. The exact molecular mechanisms underlying MS pathology, such as lesion initiation and progression, remyelination failure and diffuse pathology, are still unknown. Advanced EM techniques could contribute to unveil these mechanisms by providing detailed insights into the distinct pathological features of (de)myelinated WM and GM. Post-mortem MS brain tissue offers only static information and lacks temporal information and is often limited to end-stage disease due to the rarity of early-stage autopsy and biopsy material. Nonetheless, it offers an accurate depiction of MS pathology at the nanoscale. Although multiple factors that contribute to MS have been identified, crucial pieces to the puzzle are still missing and advances in EM research could lead to a better understanding of the neuropathology of MS in both GM and WM. Table 1 Summary the contributions of EM in uncovering MS pathology, future potential, and clinical relevance Disease process EM contribution Future of EM Clinical relevance Myelin pathology (onset of) Demyelination Outfoldings, infoldings, vesiculation, blister, vacuolation, axonal degeneration, decompaction, phagocytosis Uncover the onset of demyelination and ultra-structural features of local glial cells in NAM (FIB-SEM) Prevent demyelination, identify new targets for myelin imaging Diffuse pathology n.a Uncover myelin heterogeneity in NAM (EDX, multilabel CLEM) PIRA modulation, biomarker discovery Remyelination (failure) Thickness (g-ratio), ER in developing myelin Compare ultrastructural features of myelin and glial cells in remyelinating / remyelinated areas to areas of remyelination failure in MS lesions (STEM, EDX, multilabel CLEM, volume EM) Remyelination therapy, biomarker discovery Neuronal pathology Axon degeneration Density, local swellings, organelle accumulation (spheroids), neurofilament spacing, condensed cytoplasm n.a Neuroprotection strategies, identify imaging correlates, treatment decision making Diffuse pathology Ultrastructural alterations in myelinated axons (axonal mitochondria size, compaction) Detailed mapping of diffuse changes (mitochondria, axon, volume EM), uncover myelin heterogeneity in NAM (EDX, multilabel CLEM) PIRA modulation, neuroprotection strategies, identify imaging correlates, biomarker discovery Synaptopathy Twisted neurotubules in post-synapse, synaptic autophagy, separation by filamentous-rich astrocytic process Characterise number and ultrastructural features of synapses (width, length, numbers, synaptic vesicles, mitochondria) and surrounding glial cells in NAGM and GM lesions (STEM) Neurodegeneration prevention therapy, identify imaging correlates BBB pathology BBB integrity Disruption and swelling of astrocyte endfeet, collagen deposition and increase of pinocytotic vesicles and mitochondria Detailed characterisation of alterations in distinct MS lesion types Relevance of BBB modulation: better open or closed?

Table 1 Summary the contributions of EM in uncovering MS pathology, future potential, and clinical relevance Disease process EM contribution Future of EM Clinical relevance Myelin pathology (onset of) Demyelination Outfoldings, infoldings, vesiculation, blister, vacuolation, axonal degeneration, decompaction, phagocytosis Uncover the onset of demyelination and ultra-structural features of local glial cells in NAM (FIB-SEM) Prevent demyelination, identify new targets for myelin imaging Diffuse pathology n.a Uncover myelin heterogeneity in NAM (EDX, multilabel CLEM) PIRA modulation, biomarker discovery Remyelination (failure) Thickness (g-ratio), ER in developing myelin Compare ultrastructural features of myelin and glial cells in remyelinating / remyelinated areas to areas of remyelination failure in MS lesions (STEM, EDX, multilabel CLEM, volume EM) Remyelination therapy, biomarker discovery Neuronal pathology Axon degeneration Density, local swellings, organelle accumulation (spheroids), neurofilament spacing, condensed cytoplasm n.a Neuroprotection strategies, identify imaging correlates, treatment decision making Diffuse pathology Ultrastructural alterations in myelinated axons (axonal mitochondria size, compaction) Detailed mapping of diffuse changes (mitochondria, axon, volume EM), uncover myelin heterogeneity in NAM (EDX, multilabel CLEM) PIRA modulation, neuroprotection strategies, identify imaging correlates, biomarker discovery Synaptopathy Twisted neurotubules in post-synapse, synaptic autophagy, separation by filamentous-rich astrocytic process Characterise number and ultrastructural features of synapses (width, length, numbers, synaptic vesicles, mitochondria) and surrounding glial cells in NAGM and GM lesions (STEM) Neurodegeneration prevention therapy, identify imaging correlates BBB pathology BBB integrity Disruption and swelling of astrocyte endfeet, collagen deposition and increase of pinocytotic vesicles and mitochondria Detailed characterisation of alterations in distinct MS lesion types Relevance of BBB modulation: better open or closed?

Summary the contributions of EM in uncovering MS pathology, future potential, and clinical relevance

BBB blood–brain barrier, CLEM correlative light and electron microscopy, EDX energy dispersive X-ray, EM electron microscopy, ER endoplasmic reticulum, FIB - SEM focused ion beam scanning EM, GM grey matter, NA(G)M normal appearing (grey) matter, PIRA progression independent of relapse activity, STEM scanning transmission EM.