Multiple sclerosis (MS) is a widespread chronic disease of the central nervous system (CNS), which affects the white and gray matter (GM) of the brain and the spinal cord (1, 2). To date, the complex interplay of underlying pathological processes is still insufficiently understood. Besides acute inflammatory events, chronic smoldering processes appear to drive the disease progression relapse independently (3). Although traditionally white matter alterations have been dominating as the core feature of MS, considering also other measures, such as atrophy and GM alterations, has shown to be highly valuable in predicting long-term disability and cognitive impairment (4–9). Histopathological examinations of the GM report neuroaxonal loss, transected dendrites, and synaptic loss (10–14). Notably, GM alterations appear at all disease stages; however, GM lesions are predominantly present in progressive disease phases (15, 16).

To date, insights into the synaptic density of the human CNS of patients with MS are limited to a relatively small number of post-mortem studies. Synaptic loss appears in different areas of the cortex, such as the frontal gyrus, precentral- and postcentral gyrus (13) the frontotemporal cortex and insular cortex (17). Evidence for a reduction has been found also in the hippocampus (11, 12, 18), the thalamus (19), the cerebellum (20), and in the spinal cord (21). A change of in synaptic density has been identified primarily in lesional GM, but in a growing number of studies also in normal appearing gray matter (NAGM), going beyond areas of demyelination and exceeding neuronal loss (17, 20, 21). Thus, it can be hypothesized that the processes driving synaptic loss are surpassing secondary degeneration and do not solely follow neuronal death and demyelination.

Clearly, synaptic pathology includes not only synaptic loss but also more subtle functional disturbances, as well as alterations of electrical synapses. Considering the important physiological role of chemical synapses in neurotransmission and network dynamics, synaptic loss might contribute to functional impairment of MS patients. The aim of this review is to systematically assess the research reporting on quantitative changes of chemical synapses in distinct sites of the human CNS. A comprehensive analysis of available data will be valuable for the interpretation of regional measures in a broader context and to provide a reference for further research. A better understanding of GM pathology in MS, with synaptic loss accounting for a central component, is urgently needed considering the great impact on physical disability and cognition and moreover, the lack of effective treatment, especially for progressive stages of the disease (22).

This systematic review is a survey of the research literature on synaptic density in the human CNS of patients with MS. Our investigation reveals synaptic loss in various areas of the CNS. However, applied methods and results are inconsistent, and most studies are limited on the visualization of presynaptic structures. Conflicting results in synaptic quantification could be a result of varying measurement techniques to detect not only foremost synaptophysin-immunoreactivity but also other synaptic markers. The most abundant methodological approaches to visualize synaptophysin-immunoreactivity were OD measurements and the use of confocal microscopy. Although OD measurements can reach a higher regional resolution compared to other methods, such as western blot, they appear to have considerable limitations in accurate and sensitive quantification of synapses (27). Indeed, the majority of studies using OD to detect synaptic loss in MS report no significant change (46–49). In contrast, confocal microscopy provides a higher spatial precision compared to light microscopy analysis (50), which could be a reason for higher sensitivity in detecting synaptic loss in studies relying on this technique (17–19). Nevertheless, comparable studies also provided evidence for an unaltered overall synaptic density in areas of the cortex and suggested only focal synaptic disappearance in colocation to activated microglia (45). On the other hand, stereological analysis, where the total number of synapses is estimated based on a subgroup of individually counted stained points, has been demonstrated to be both, highly sensitive and accurate in detecting local differences (27). This might account for the observation of pre- and postsynaptic loss in the groundbreaking studies (17, 20, 44). Furthermore, the included studies have been inconsistent in the choice of a reference region. Some studies have compared measurements in demyelinated GM area solely to NAGM in patients with MS (46, 47), whereas other studies compared lesional and NAGM to HCs. In other words, some studies assumed an unaltered histological environment in GM areas without signs of demyelination. Since we know that considerable pathology takes place also in the NAGM (17, 44), this might have led to false conclusions, most likely to an overestimation of the synapse number in demyelinated areas. Importantly, the majority of studies measured solely presynaptic protein levels. Few human studies have taken other anatomical substructures, such as the synaptic gap and the postsynaptic site, into account (11, 17). Additional research is required in order to draw reliable conclusions about the correlation between loss of presynaptic proteins and quantitative changes of whole synaptic connections in MS. Partly, more precision is also needed in the specification of the targeted molecule. For example, Dutta et al. (11) reported no alteration in the density of SV2, a synaptic vesicle protein in the presynapse. However, SV2 comprises the three isoforms SV2A, SV2B and SV2C, out of which only SV2A is ubiquitously expressed throughout the adult brain. The expression of SV2B and SV2C is much more variable (30), a fact which needs to be taken into account for drawing valid conclusions.

Mechanisms leading to synaptic loss are likely multifactorial. Several mechanisms have been proposed; nevertheless, a comprehensive understanding is still missing. Consistent with our findings, demyelination has a notable impact on the synapse number. Demyelination leads to an altered gene expression; it is disrupting the physiological microenvironment and contributes to a decrease of essential synaptic proteins (11). Closely related is the excitotoxic hypothesis, which states that an environment of enhanced glutamate and reduced GABA is having a negative impact on synapses. This is supported by the observation that the selective inhibition of AMPA receptors in experimental autoimmune encephalomyelitis (an animal model for MS) can prevent clinical disability and dendritic spine loss (51, 52). Further involved processes might be neuron-autonomous synaptic elimination (20) and trans-synaptic anterograde and retrograde synaptic degeneration as a consequence of lesional areas (20, 53). Additionally, histopathological studies have pointed toward a complement-assisted (C1q-C3) elimination of synapses by microglia in a mildly inflamed environment (18). This process is of special interest because it has been discovered also in NAGM (18). Reports of lesion-independent synaptic loss, which can surpass axonal loss, give rise to the question of a primary process leading to a reduction of synapses (17). Complement-assisted elimination of synapses by microglia appears to have similarities to the physiological pruning of synapses throughout life, an essential mechanism to ensure functional neuronal circuits (54, 55).

The quantification of synapses in post-mortem studies has clearly its limitations. Subjects are for the most part in advanced stages of the disease, and dynamic processes are difficult to unravel. Animal models used to study MS, such as the experimental autoimmune encephalomyelitis (EAE), are therefore an important tool. Indeed, EAE studies have demonstrated alterations in the synaptic density to be a dynamic process. Synapse numbers are suggested to partly normalize completely after the recovery from EAE (56). Evidence for regeneration is also visible in human MS studies, such as an increased ratio of GAP-43/SYN in leucocortical lesions (GAP-43 is expressed in growth clones and axons). A relative increase of GAP-43 might indicate local remodeling processes (13). Also when compared to traumatic injury, leucocortical lesions of early MS appear to present a higher regeneration potential (57). To which extent synaptogenesis is part of a regeneration process in MS is still elusive. However, mechanisms involved in synaptic elimination and regeneration might be a suitable target for therapeutic interventions. Therefore, new methods to study the synaptic density directly in humans in the living organism are urgently needed. A promising approach is a recently introduced PET imaging technique using the synaptic vesicle glycoprotein 2A (SV2A)-binding radioligand ¹¹C-UCB-J, which can be used to visualize the synaptic density in vivo (58). With this approach, changes in synaptic density have been reported from several neurological and psychiatric diseases, such as Parkinson's, dementia with Lewy bodies (59, 60), Alzheimer's disease (61–63), temporal lobe epilepsy (58, 64), major depressive disorder (65), post-traumatic stress disorder (65), and schizophrenia (66). Comparable in vivo studies still need to be carried out for MS.

We have found evidence for widespread synaptic loss throughout the MS brain and in the MS spinal cord, partly exceeding lesional areas. However, results are inconsistent, probably due to differences in the applied methods. We have found a tendency toward an increased methodological effort to tackle the question of synaptic loss in MS. In order to gain a better understanding of the underlying pathomechanisms and dynamic processes, further research and in vivo studies are needed.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

EM did the study design, data extraction, drafting, and revision of the manuscript, as well as the screening and selection of potential publications. EH contributed to the screening and selection of potential publications. LA carried out the study supervision and participated in the revision of the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by Sigrid Juselius Foundation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.