Synapse Dysfunctions in Multiple Sclerosis

Multiple sclerosis (MS) is a chronic neuroinflammatory disease of the central nervous system (CNS) affecting nearly three million humans worldwide. In MS, cells of an auto-reactive immune system invade the brain and cause neuroinflammation. Neuroinflammation triggers a complex, multi-faceted harmful process not only in the white matter but also in the grey matter of the brain. In the grey matter, neuroinflammation causes synapse dysfunctions. Synapse dysfunctions in MS occur early and independent from white matter demyelination and are likely correlates of cognitive and mental symptoms in MS. Disturbed synapse/glia interactions and elevated neuroinflammatory signals play a central role. Glutamatergic excitotoxic synapse damage emerges as a major mechanism. We review synapse/glia communication under normal conditions and summarize how this communication becomes malfunctional during neuroinflammation in MS. We discuss mechanisms of how disturbed glia/synapse communication can lead to synapse dysfunctions, signaling dysbalance, and neurodegeneration in MS.


Primer on Multiple Sclerosis (MS): Multi-Faceted Neuroinflammatory Autoimmune Disease with Pathologies in the White and Grey Matter of the Brain
Multiple sclerosis (MS) is a chronic neuroinflammatory autoimmune disease of the human central nervous system (CNS). Almost three million people worldwide suffer from MS [1]. Among those, young adults (in particular women) are the preferentially affected disease group [2][3][4][5][6][7]. Although the cause of the disease remains elusive, it has become evident that environmental factors and multiple gene loci are risk factors for disease susceptibility [3,[8][9][10][11][12][13][14][15][16]. Viral infections, particularly those with the Epstein-Barr virus, have also been proposed to enhance the likelihood of developing MS [17][18][19]. In a longitudinal study, Bjornevik et al. [20] provided evidence that infection with the Epstein-Barr virus is indeed the main trigger for the development of MS, leading to a 32-fold increase in disease susceptibility. In MS, brain-reactive, encephalitogenic T cells (particularly T H -cells) from the body periphery invade the brain [21] and induce an auto-destructive immune response that leads to alterations both in the white and grey matter. These auto-reactive T-lymphocytes are the key drivers of the disease [22]. However, also abnormally activated glial cells [23,24] and B-lymphocytes play an important role [22,[25][26][27][28][29][30][31]. Long-term depletion of B-lymphocytes by targeting CD20 with monoclonal antibodies can attenuate disease progression in relapsing-remitting but also primary progressive multiple sclerosis [32-35], emphasizing the role of different components of the immune system in establishing and maintaining the disease.
MS patients suffer from a plethora of clinical symptoms that mirror the sites of lesion. The symptoms include, for example, visual impairments/optic neuritis, central motor paresis, sensory dysfunctions (numbness/paresthesia), and sensory ataxia [36][37][38][39][40][41]. These clinical symptoms are considered "white matter" symptoms resulting from demyelination and axonal damage of the respective fiber tracts in the white matter. In the CNS, the axonal myelin sheath is produced by oligodendrocytes (OL). Oligodendrocyte precursor cells can differentiate into OL and play an important role in the disease course [42]. In the most frequent form of MS, classified as relapsing/remitting multiple sclerosis (RRMS), the disease is characterized by acute inflammatory episodes that improve to some extent before symptoms become progressively chronic and worse with no or only incomplete remission (progressive forms of MS) [16,43,44]. Brain auto-reactive, T-lymphocytes (mainly CD4 + T-helper cells) that enter the brain (either via the blood-brain-barrier or passage through the meninges) play a central role in white matter changes and axonal demyelination [16,22,[45][46][47]. Auto-antibodies that cross-react with brain epitopes [48-54] and hyperactivated glial cells [55][56][57] also significantly contribute to the disease. Demyelination and neuro-axonal damage in the white matter have been extensively investigated [16,43,[58][59][60][61][62][63][64][65][66] and represent the most appreciated, best-characterized part of MS.
Neuroinflammation strongly contributes to grey matter changes [69,70,74,85,104,105]. Neuroinflammation involves abnormal glial cell activation and an excessive release of inflammatory cytokines that cause excitotoxic synapse damage [69,[106][107][108]. Excitotoxic synapse damage can cause cortical network dysfunctions, cognitive disabilities, neurodegeneration, and neuronal cell death that could lead to irreversible disease progression. Clearly, the pathogenesis of grey matter is complex and not completely understood. However, a general picture of MS-related changes in the grey matter is gradually emerging, which we attempt to summarize in the present review.

Primer on Brain Synapses: Communication Nano-Machines with Multiple Adjustment "Screws"
The remarkable capabilities of the human brain are enabled by neuronal networks formed by about 100 billion neurons that are connected by more than 100 trillion chemical synapses (10 14 synapses). Recent evidence revealed that neuronal synapses, the key devices of intercellular communication in the grey matter of the brain, are compromised in MS. Synaptic dysfunction and subsequent neurodegeneration likely account for the cognitive changes and for disease progression in MS. Before reviewing synapse alterations and Figure 1. Glutamatergic brain synapses. It schematically depicts the composition of a healthy, glutamatergic tetrapartite brain synapse consisting of pre-and postsynaptic terminals and perisynaptic processes of astrocytes and microglia. At the presynaptic terminal, synaptic communication occurs via exocytosis of glutamatergic synaptic vesicles at the active zone. The active zone is a protein-rich compartment at which the synaptic vesicle fusion machinery is linked by active zone proteins close to voltage-gated Cav-channels. Influx of Ca 2+ through Cav-channels triggers exocytosis. The postsynaptic membrane contains ionotropic glutamate receptors. AMPA and NMDA receptors are depicted. Perisynaptic processes of astrocytes contain glutamate transporters that remove synaptically released glutamate to prevent spillover of glutamate to neighboring synapses. Perisynaptic processes from astrocytes communicate with microglia processes (magenta arrows). Astrocytes Figure 1. Glutamatergic brain synapses. It schematically depicts the composition of a healthy, glutamatergic tetrapartite brain synapse consisting of pre-and postsynaptic terminals and perisynaptic processes of astrocytes and microglia. At the presynaptic terminal, synaptic communication occurs via exocytosis of glutamatergic synaptic vesicles at the active zone. The active zone is a protein-rich compartment at which the synaptic vesicle fusion machinery is linked by active zone proteins close to voltage-gated Cav-channels. Influx of Ca 2+ through Cav-channels triggers exocytosis. The postsynaptic membrane contains ionotropic glutamate receptors. AMPA and NMDA receptors are depicted. Perisynaptic processes of astrocytes contain glutamate transporters that remove synaptically released glutamate to prevent spillover of glutamate to neighboring synapses. Perisynaptic processes from astrocytes communicate with microglia processes (magenta arrows). Astrocytes sense synaptic activity via various receptors, including metabotropic glutamate receptors, and modulate synaptic activity via release of "gliotransmitters", e.g., glutamate, TNFα and IL1β (magenta arrows). Furthermore, astrocytes control cerebral blood flow and the integrity of the blood-brain barrier. An exemplary capillary is shown with endothelial cell contacts sealed by tight junctions. Arrows in magenta show interactions between components of the multipartite synapse. Abbreviations: CP-AMPA, Ca 2+permeable AMPA receptors, iGluR, ionotropic glutamate receptor, mGluR, metabotropic glutamate receptor, EAAT, excitatory amino acid transporter.
After release into the synaptic cleft, the neurotransmitter (e.g., glutamate) binds to postsynaptic neurotransmitter receptors. The neurotransmitter receptors are enriched in the postsynaptic density (PSD) that is located directly opposite to the active zone. In the PSD, neurotransmitter receptors are immobilized by a dense protein network of scaffold proteins [120,121]. These scaffold proteins include MAGUK family proteins, e.g., PSD-95, SAP-97, SAP-102, in excitatory synapses. Inhibitory synapses contain scaffold proteins such as gephyrin and collybistin [122]. PSD-95 of excitatory synapses is a particularly relevant PSD scaffold protein of excitatory synapses because it links AMPA-and NMDA-type glutamate receptors to each other [120,121]. These glutamate receptors are important for synaptic plasticity and for excitotoxic synapse damage (see below).
The synaptic cleft contains synaptic adhesion protein complexes that align pre-and postsynaptic signaling complexes into functionally connected transsynaptic nanocolumns required for synchronous, efficient information transfer [123][124][125]. An important synaptic adhesion complex consists of presynaptic neurexins and postsynaptic neuroligins and functions as transsynaptic organizers [123,124,126]. Neurexin genes generate large numbers of splice variants relevant for guiding connectivity between distinct, individual neurons. Dysfunctions of neurexin-neuroligin synaptic complexes were correlated with cognitive disturbances and neuropsychiatric diseases [123,124]. Other synaptic adhesion complexes, in part, also interact with the neurexin-neuroligin trans-synaptic adhesion axis [123,126].
Exocytosis of glutamatergic synaptic vesicles at the active zone is the prime mechanism for glutamate release at the synapse. Following presynaptic release, glutamate exerts its action at the postsynapse by binding to different types of glutamate receptors [131][132][133] ( Figure 1). Glutamate receptors are classified into ionotropic and metabotropic glutamate receptors (iGluR, mGluR) [133,141]. iGluRs are sub-divided into AMPA-, Kainate, and NMDA-type receptors based on their molecular composition, physiological properties, and preferential agonists. For fast synaptic transmission, the postsynapse mainly employs iGluRs. All iGluR receptors are composed of several subunits. In the CNS, 2-Amino-3-(3-hydroxy-5-methylisooxazol-4-yl)proprionate (AMPA) receptors are the most abundant glutamate iGluRs. AMPA receptors assemble from four different subunits (GluA1-GluA4). These form homo-or hetero-tetramers [133,142,143]. Glutamate-gated opening of AMPA receptors depolarizes the postsynaptic compartment. AMPA receptors are permeable to Na + , K + and, depending upon subunit composition, also to Ca 2+ . If the GluA2 subunit is absent from AMPA receptors, the resulting AMPA channels are permeable to Ca 2+ [133,142]. AMPA receptors containing GluA2 are not Ca 2+ permeable. N-methyl-D-aspartate (NMDA)-type iGluRs also play an important role for synaptopathy in MS. NMDA receptors are hetero-tetramers that consist of GluN1, GluN2 (GluN2A, GluN2B, GluN2C, GluN2D), and GluN3(A,B) subunits [144,145]. Two GluN1 subunits combine with two GluN2 or GluN3 subunits to form the NMDA channel. AMPA and NMDA receptors functionally interact. Strong depolarization of the postsynaptic compartment obtained by many AMPA-channel openings relieves the Mg 2+ block of NMDA receptors and enables the opening of NMDA receptor channels [141,145,146]. Only strong depolarizations that typically result from multiple simultaneous presynaptic vesicle fusion events lift the block of NMDA glutamate receptors by expulsing Mg 2+ from the channel pore. The influx of Ca 2+ through the NMDA receptor induces early and late phases of LTP (Long-Term Potentiation) through CaMKII (Ca 2+ /calmodulin-dependent protein kinase II) and cAMP/PKA/pCREB-dependent mechanisms [147][148][149][150][151][152][153][154][155][156]. Influx of Ca 2+ also triggers increased surface expression of AMPA receptors via fusion of AMPA receptor-containing subsynaptic vesicles with the postsynaptic plasma membrane [147,157,158]. Elevated surface expression of AMPA receptors increases the efficacy of this individual synapse at which the NMDA receptor was activated and makes it more sensitive to the subsequent release of glutamate by the connected presynaptic terminal. The adjustment of synaptic efficacy based on previous activity is part of a phenomenon called "synaptic plasticity". It is considered the basis for learning, memory, and experience-based behavior [159]. Heavily used synapses become more efficient by this mechanism. Vice versa, the efficacy of less active synapses decreases. These positive feedback mechanisms belong to the Hebbiantype of synaptic plasticity and include short-term and long-term effects [147,157,160]. The Ca 2+ permeability of iGluRs (GluA2-lacking AMPA-and NMDA receptors) is central to this process [142,[161][162][163]. Further Ca 2+ -dependent events also support synaptic plasticity. Ca 2+ regulates the metabolic activity of mitochondria in order to synchronize energy production with synaptic activity [164][165][166]. Ca 2+ also controls intracellular signaling cascades, such as the Ca 2+ /Calmodulin/CaMKII system and the Ras/Raf/MAP-kinase pathway that control the activity of phosphorylation-regulated transcription factors (e.g., CREB, NF-kB). These transcription factors control transcriptional programs needed for the long-term remodeling of synaptic components (e.g., dendritic spines) and/or neuronal survival (e.g., secretion of BDNF [167]). Pre-synaptic mechanisms can also contribute to synaptic plasticity [112,150,168]. Presynaptic mechanisms include modulation of synaptic vesicle release probability, e.g., by modification of active zone components/Cav-channel number/distance/opening properties [112,113]. A second form of synaptic plasticity, homeostatic synaptic plasticity (HSP), prevents over-activation/saturation of active synapses and complete silencing of less frequently used synapses [169][170][171][172][173][174]. HSP is based on negativefeedback mechanisms that lead to counter-acting, compensatory adjustments, and a resetting of the synaptic signaling range. HSP prevents saturation and unresponsiveness of synaptic connections and maintains synaptic activity in a functional range [169,170]. Interestingly, synaptic plasticity and adaptational changes of synapses are strongly influenced by inflammatory cytokines, such as TNFα and IL1β.

A More Extended View on Brain Synapses: Contribution of Glial Cells
In the brain, pro-inflammatory cytokines, such as TNFα and IL1β, are physiologically secreted in small amounts by glial cells, mainly by astrocytes and microglia. These glial cells establish close contacts with synapses and modulate synaptic communication. Gliasynapse interactions are important to adjust synaptic activity under normal conditions but are also relevant for synaptopathy as it occurs in MS.

Primer on Astrocytes: A Network of Guardians of Brain Homeostasis with Strong Impact on Synapses
The human brain contains about 86 billions of neurons and a similar number (85 billions) of glial cells [175]. Within the population of glial cells, astrocytes are the most abundant type of glial cells in the human brain. Astrocytes are multi-branched stellate cells that intensively communicate with each other via gap junctions [176][177][178]. Astrocytes also establish close contact with neurons. At a functional level, astrocytes execute many pivotal homeostatic functions. Some important functions include the maintenance of extracellular ion and fluid balance, provision of metabolites (e.g., glutamine, see below) to neurons, control of blood flow, and maintenance of the blood-brain barrier [176,[179][180][181][182][183][184][185]. Astrocytes are functionally very diverse, as also confirmed by single-cell sequencing [178]. Based on the molecular heterogeneity, a nomenclature has been proposed for astrocytes in which resting astrocytes and various types of "reactive" astrocytes have been discriminated [178].

Figure 2.
Glutamatergic retinal ribbon synapses. It schematically depicts a glutamatergic ribbon synapse of the retina. A photoreceptor ribbon synapse is shown. The functional and molecular composition of the presynaptic ribbon terminal is like that of the brain synapse shown in Figure 1. However, the presynaptic terminals of ribbon synapses possess eponymous synaptic ribbons that provide the active zone with additional synaptic vesicles to enable continuous synaptic transmission at this synapse. The active zone is the site at which exocytosis of synaptic vesicles occurs close to voltage-gated L-type Cav-channels. Endocytosis of fused vesicle membrane occurs in the periactive zone, followed by a refilling process of endocytosed vesicles with glutamate. The postsynapse is composed of the dendrites of horizontal and bipolar cells. The postsynaptic dendrites contain ionotropic and metabotropic glutamate receptors (not shown). Müller glial cells form perisynaptic processes that are enriched in glutamate transporters (EAAT1, EAAT2). The presynaptic terminal contains additional glutamate transporters (GLT1v, EAAT5) in the periactive zone. Retinal ramified microglial cells frequently contact synapses and likely communicate with Müller glial cells (magenta arrows). Retinal ribbon synapses are very sensitive to neuroinflammatory changes. Neuroinflammatory changes strongly affect presynaptic terminal functions (for details, see text). Active zone composition, presynaptic Ca 2+ homeostasis and exo-and endocytic vesicle cycling in these synapses are disturbed in EAE. The underlying mechanisms are not yet fully elucidated but might involve simar pathways as described for brain synapses (see text). Arrows in magenta indicate possible interactions between Müller cells, microglia and the photoreceptor synapse. Abbreviations: sv, synaptic vesicles; sr, synaptic ribbon; Cav, voltage-gated Ca 2+ -channel; mGluR, metabotropic glutamate receptor at Müller glial cells; iGluR, ionotropic glutamate receptor at Müller glial cells; h, horizontal cell postsynaptic dendrite; b, bipolar cell postsynaptic dendrite; GLT1v, glutamate transporter 1 splice variant; EAAT1/2, excitatory amino acid transporter 1/2.
Of note, glutamate transporters (GluTs) can also revert the transport direction of glutamate and secrete glutamate. This can occur during pathological conditions/diseases. The Glutamatergic retinal ribbon synapses. It schematically depicts a glutamatergic ribbon synapse of the retina. A photoreceptor ribbon synapse is shown. The functional and molecular composition of the presynaptic ribbon terminal is like that of the brain synapse shown in Figure 1. However, the presynaptic terminals of ribbon synapses possess eponymous synaptic ribbons that provide the active zone with additional synaptic vesicles to enable continuous synaptic transmission at this synapse. The active zone is the site at which exocytosis of synaptic vesicles occurs close to voltage-gated L-type Cav-channels. Endocytosis of fused vesicle membrane occurs in the periactive zone, followed by a refilling process of endocytosed vesicles with glutamate. The postsynapse is composed of the dendrites of horizontal and bipolar cells. The postsynaptic dendrites contain ionotropic and metabotropic glutamate receptors (not shown). Müller glial cells form perisynaptic processes that are enriched in glutamate transporters (EAAT1, EAAT2). The presynaptic terminal contains additional glutamate transporters (GLT1v, EAAT5) in the periactive zone. Retinal ramified microglial cells frequently contact synapses and likely communicate with Müller glial cells (magenta arrows). Retinal ribbon synapses are very sensitive to neuroinflammatory changes. Neuroinflammatory changes strongly affect presynaptic terminal functions (for details, see text). Active zone composition, presynaptic Ca 2+ homeostasis and exo-and endocytic vesicle cycling in these synapses are disturbed in EAE. The underlying mechanisms are not yet fully elucidated but might involve simar pathways as described for brain synapses (see text). Arrows in magenta indicate possible interactions between Müller cells, microglia and the photoreceptor synapse. Abbreviations: sv, synaptic vesicles; sr, synaptic ribbon; Cav, voltage-gated Ca 2+ -channel; mGluR, metabotropic glutamate receptor at Müller glial cells; iGluR, ionotropic glutamate receptor at Müller glial cells; h, horizontal cell postsynaptic dendrite; b, bipolar cell postsynaptic dendrite; GLT1v, glutamate transporter 1 splice variant; EAAT1/2, excitatory amino acid transporter 1/2. Of note, glutamate transporters (GluTs) can also revert the transport direction of glutamate and secrete glutamate. This can occur during pathological conditions/diseases. The physiological basis for this phenomenon is the fact that GluTs are secondary active transporters that depend upon the transcellular Na + gradient [221]. In this electrogenic transport, one glutamate molecule is co-transported with 3 Na + and 1 H + . In the regular "forward" mode, this import is coupled with counter-transport (export) of a K + ion into the extracellular space. Disease conditions that dissipate the electrochemical Na + gradient can drive glutamate transport into the opposite direction, i.e., release of glutamate into the extracellular space (reverse mode of transporter activity [256,257]). Reverse efflux of glutamate through GluTs happens during ischemia and inflammation [257][258][259]. Such a mechanism could also be relevant for glutamate excitotoxicity in MS (see below). Further mechanisms that could contribute to the elevation of extrasynaptic glutamate include glutamate release through the cystine-glutamate antiporter xCT (cystine/glutamate antiporter) [260,261], release through bestrophin-1 anion channel [262][263][264][265], TREK-1 channels [264,265], and volume-regulated anion channels/volume-sensitive organic anion channels (VRACs/VSOACs) [266][267][268]. Moreover, vesicular release of glutamate by astrocytes and microglia could lead to elevated levels of extrasynaptic glutamate and contribute to glutamate excitotoxicity in MS (see below).
Activation of astrocytic neurotransmitter receptors leads to changes in intracellular Ca 2+ in astrocytes and to the release of "gliotransmitters" [195,208,272,[281][282][283][284][285][286]. The source of these intracellular Ca 2+ changes is controversially discussed [287,288]. Gliotransmitters are neuroactive molecules, such as glutamate, ATP/adenosine, GABA, NPY, D-serine (a coagonist of NMDA-type of glutamate receptors), IL1β, and TNF-α [286]. Astrocytes release synapse-active components that affect synaptic performance [208,218,[283][284][285]289]. Gliotransmitters released by astrocytes influence pre-and postsynaptic functions [193,195,276,278,290,291]. Important signaling cascades have been identified. Glutamatergic synapses that show only little activity secrete less glutamate. Decreased levels of synaptically released glutamate are sensed by neurotransmitter receptors of perisynaptic astrocytes in hippocampal synapses and induce secretion of TNFα (by astrocytes and microglia via astrocyte/microglia communication). These elevated, non-toxic levels of TNFα (≈100 picomolar TNFα) enhance pre-and postsynaptic glutamatergic signaling [162,163,173,278,[292][293][294]. TNFα induces glutamate release by astrocytes that, in turn, stimulate presynaptic glutamate release at hippocampal synapses (i.e., entorhinal cortex/dentate gyrus synapses) via binding to presynaptic NMDA receptors [162,163,269,275,277,278,290,295]. Of note, the effect of astrocytic glutamate on presynaptic release depends on the expression of distinct neurotransmitter receptors. The binding of glutamate to metabotropic glutamate receptors (mGluR2/3) at the presynapse was reported to inhibit glutamate release from the presynaptic terminal [296][297][298]. At the postsynapse, glial TNFα leads to an increased surface expression of AMPA receptors to scale up synaptic activity and synaptic responsiveness [162,163]. In conclusion, the pro-inflammatory cytokine TNFα at physiological concentrations is an important positive regulator of synaptic activity in the healthy brain. Modulation of TNFα secretion in this physiological range serves homeostatic scaling of synapse activity (HSP, see above). Typically, astrocyte-mediated modulation of synaptic transmission occurs at a slower timescale as fast synaptic communication between the pre-and postsynaptic compartment due to the signal integration in the glial compartment [197,272,278]. At the systems level, physiologically elevated levels of proinflammatory cytokines have been shown to be important for memory formation in freely moving animals reflecting their effect on synapses also in-situ [299].
Astrocytes establish intimate contacts with microglia at synapses and exchange important mutual signals with microglia. Astrocytes provide microglia with cues on synaptic activity. Microglia vice versa, provide signals for the differentiation of astrocytes into distinct sub-types, i.e., either into a beneficial, homeostatic, and activity-maintaining sub-type (initially denoted as "A2" astrocytes [217,284,300]) or into a neurotoxic, neuroinflammatory subtype ("A1" astrocytes [208]) whose activity is detrimental to the CNS [208,217,284,300]. C1q, IL-1α, and TNFα promote differentiation of resting astrocytes into the A1 neurodestructive state [208], whereas co-stimulation with TNFα and IL1β promotes differentiation towards the neuro-supportive A2 phenotype in astrocytes [217,284,300,301].
As mentioned, the A1/A2 dichotomy of astrocytes is too simplified based on singlecell sequencing data but still useful as a simplified working model. Transcript analyses revealed markers common to all reactive astrocytes (e.g., GFAP [178,302]). The expression of complement protein C3 is a marker for inflammatory "A1" astrocytes [178,208,284,303,304]. Complement proteins were found to be highly elevated in the brain of MS patients, particularly in cortical grey matter lesions [305][306][307][308]. The complement system is likely involved in synapse dysfunctions in the MS brain based on its well-known function of synapse removal during brain development [309,310]. Astrocyte-microglia interactions are highly relevant for neuroinflammatory disease in multiple sclerosis and the resulting synaptic changes (see below).

Primer on Microglia: Never-Resting Brain "Police" with the Mission to Survey and to Take (Strong) Action
Microglia are the main resident immune cells of the brain, comprising ≈10% of total CNS brain cells with some regional differences [311][312][313][314][315]. Microglial cells develop from monocyte-like precursor cells of the bone marrow and yolk sac [316,317]. During embryonic development, they invade the brain, in which they mature and proliferate [216,318,319]. Cell surface markers, such as TMEM119, allow to discriminate microglia from bloodborne macrophages invading from capillaries [320,321]. A few years ago, microglia were believed to be active only during disease conditions. Recent investigations demonstrated that microglia perform important functions already in the healthy, non-injured brain. The range of functions performed by microglia, both in the healthy and injured brain, is broad [216,285,[322][323][324].
Morphologically, microglia are very diverse and dynamic [325]. In the healthy brain, microglial cells often display a highly ramified morphology with many processes. Ramified microglia were previously considered "resting", i.e., inactive. This old view is not correct. Novel technologies, particularly live imaging analyses with genetically engineered fluorescent microglia, demonstrated that "resting" ramified microglial cells are indeed highly active already under physiological conditions in the healthy brain [322,[326][327][328][329][330][331]. Live imaging experiments with genetically tagged microglia revealed that microglial processes are highly mobile and frequently expand and retract [322,326]. These processes continuously scan and monitor the extracellular environment of the CNS for a broad range of signals [216,285,322,323]. Thus, the term "resting" microglia should be replaced by the term "surveying" microglia [323].
Live-Imaging with these genetically engineered mice possessing fluorescent microglia revealed that ramified microglia use their processes to contact synapses [322,326]. The soma of the microglia typically stays stationary in this process, whereas the processes expand and retract continuously. A sub-type of ramified microglia that is positive for a 5D4 keratan-epitope and rich in IL1β is particularly active in this process [332]. The close relationship between microglial processes to both pre-and postsynaptic compartments, as well as to perisynaptic astrocytes has been referred to as tetra-partite synapse [285,333].
The interaction between ramified microglia processes and synapses is complex and depends upon synaptic activity [327,328]. Synapses and microglial synaptic processes influence each other in a mutual manner. On the one hand, synaptic activity regulates the contact properties between microglial processes and synapses (e.g., process motility, duration, and frequency of synaptic contacts) in a differential manner [334]. On the other hand, ramified microglia signal back to synapses and lead to changes in glutamatergic synaptic transmission [276,334]. The contact between ramified microglial processes and synapses was reported to increase synaptic activity [335]. This feedback from microglia to synapses is predominantly indirect and mediated via astrocytes that secrete gliotransmitters [276,335]. Similarly, synaptic activity also influences the motility of microglial processes in the retina [336]. In the retina, glutamatergic neurotransmission enhances microglia process motility, whereas GABAergic transmission inhibits microglial process motility [336].
The precise mechanisms of synapse-microglia interactions are complex and not fully understood. As mentioned, the effects of synaptically released glutamate on microglia are likely indirect because ramified surveying microglia do not express glutamate receptors (in contrast to "activated" microglia, see below). In the retina, the neurotransmitter effects of glutamate on microglia motility were reported to be mediated via ATP release from Müller glial cells/astrocytes [336]. Thus, synaptic neurotransmitter release likely does not signal directly to ramified microglial cells but indirectly via signals from perisynaptic astrocytes [326,[335][336][337][338]. ATP binds to microglial ionotropic P2X7 receptors to induce the release of IL1β [339]. IL1β is a powerful modulator of synapse function (see below). ATP can also bind to microglial metabotropic purinergic receptors (P2Y12/13) that are relevant for chemotactic guiding of microglia [340][341][342]. Released ATP can also directly affect the postsynaptic terminal by activation of postsynaptic P2X receptors [343,344].

Multivalent Microglia: Potentiator of Inflammatory Signals with Strong Impact on Synapses
Any kind of homeostatic disturbance in the brain can activate microglia and lead to a transformation into a reactive, particularly alerted state [216,285,318,319,323]. Microglia activation is often associated with de-ramification or even loss of microglial processes. A process-lacking amoeboid shape promotes the movement of the activated microglia towards the area in which a potential threat has been detected. The character and degree of microglia activation differ broadly [216,285,323,[326][327][328]336,350]. Two extreme forms of activated microglia sub-types have been previously discriminated and denoted as M1-and M2-microglia [284,323,[351][352][353]. "M1"-type microglia are pro-inflammatory and neurotoxic with low phagocytic activity. They show surface expression of MHC-II proteins, CD11b, CD16, CD68, TREM2, and release large quantities of glutamate (and glutamate-like toxic kynurenines) as well as pro-inflammatory cytokines, such as IL1β, TNFα, IL-6, IL-12, and IFNγ [216,284,323,342,354,355]. "M2"-type microglia are anti-inflammatory, neuroprotective with high phagocytic activity, express distinct surface markers such as CD163 and CD206, and secrete IGF-1 and TGF-β [216,284,323,342,354,355]. Single-cell sequencing revealed that the classification of activated microglia in only two classes is simplified. Activated microglia are functionally more diverse [302,356,357]. The dichotomic M1/M2 category still serves as a simplified model.
Microglial cells are equipped with various receptors for detecting homeostasis-threatening signals [358][359][360][361][362][363][364]. Large amounts of ATP released from damaged or dead cells are a strong attracting and activation signal for microglia [285,326,365,366]. Complement proteins (e.g., complement proteins C1, C3) are also detected by microglial cells via cell surface complement receptors [177,185,285,309,329]. During development, complement proteins tag excess or dysfunctional synapses and mark these for subsequent elimination by microglia, a process called "synapse stripping". Synapse stripping is important for the development and refinement of functional neuronal circuits during development [216,309,329]. In the mature brain, less active synapses also contain increased levels of C1q that promote their removal by microglial cells, thus allowing activity-dependent refinement of neuronal circuits in the healthy, postnatal brain [327][328][329]345,[367][368][369]. Astrocytes are also involved in synapse stripping of silenced or dysfunctional synapses together with the microglia [177,370].
As mentioned, the activation of microglia can be strong, particularly if the inflammatory event stays unresolved and remains to continuously activate the microglia. Under these conditions, i.e., when microglia are strongly and permanently activated, microglia become harmful to the host brain and promote disease progression/aggravation by secreting large amounts of reactive oxygen species and biologically active cytokines that cause neuroinflammation [216,284,285,348,353,371,372]. Inflammatory over-activation of microglia, as it occurs in active, progressive MS, can lead to significantly elevated release of the pro-inflammatory cytokines, e.g., TNFα, IL1β [70,84,373], that adversely affect synapse functions (see below). As mentioned, microglia also communicate with astrocytes and influence the differentiation of astrocytes. Pathologically activated microglia of the "M1" subtype activate astrocytes towards a harmful "A1" sub-type by the secretion of IL1α, TNFα, and C1q [162,177,208,216,276]. A1 astrocytes, together with the activated M1 microglia, represent the major source of the pro-inflammatory cytokines IL1β and TNFα in the brain [69,162,177,216,276,284]. The large amount of secreted IL1β and TNFα can lead to synapse damage in MS by multiple mechanisms and produce the observed grey matter dysfunctions in MS (see below).

Neuroinflammation-Induced Synapse Dysfunctions in MS
The levels of the pro-inflammatory cytokines TNFα and IL1β can severely increase in MS, particularly in the active, progressive stages of MS [69,70,78,84,85,128,[373][374][375][376][377]. During persistent strong inflammation, activated microglia potentiate inflammatory signals leading to excessive pathological TNFα values up to the mM range [275,278]. A growing body of evidence indicates that the highly elevated levels of these pro-inflammatory cytokines (TNFα, IL1β), as it can occur in MS [378,379], are the main reason for early dysfunctions of synaptic transmission. These lead to glutamatergic excitotoxicity, neurodegeneration, and ultimately, neuronal cell death. Several mechanisms contribute to neuroinflammationinduced glutamatergic synapse dysfunction and glutamatergic excitotoxicity.
The highly elevated levels of inflammatory cytokines released by activated microglia, astrocytes, and inflammatory CD3 + T-cells in MS, inhibit the expression of glial glutamate transporters (EAAT1/2), resulting in a decreased clearance of glutamate from the synaptic cleft [85,128,129,[380][381][382][383][384][385] (Figure 3). The decreased glutamate clearance results in increased levels of extrasynaptic glutamate. Extrasynaptic glutamate binds to extrasynaptic glutamate receptors, including Ca 2+ -permeable NMDA receptors and Ca 2+ -permeable AMPA receptors. Stimulation of these extrasynaptic glutamate receptors is considered the central mechanism causing glutamate excitotoxicity, neurodegeneration, and neuronal cell death [386][387][388][389]. Many of these mechanisms involve elevated levels of Ca 2+ . Extrasynaptic NMDA receptor activation will trigger a deleterious signaling cascade that includes structural degeneration of the synapse, mitochondrial damage, and transcriptional shut-off of neuroprotective pathways [386][387][388][389]. Paradoxically, increased extrasynaptic glutamate can further inhibit the expression of astrocytic glutamate transporters [390], thus fostering a vicious cycle that leads to glutamate excitotoxicity. Figure 3. schematically depicts a brain synapse, as shown in Figure 1, but under neuroinflammatory conditions as in multiple sclerosis (MS). Perisynaptic astrocytes and microglia are strongly activated (encircled in red) and secrete large amounts of inflammatory cytokines. Inflammatory cytokines produce glutamatergic excitotoxicity (see text). Neuroinflammation leads to decreased expression of astrocytic glutamate transporters and to elevated levels of extracellular glutamate (colored in red). Increased extracellular glutamate binds to extrasynaptic glutamate receptors. This initiates a series of deleterious events at the synapse consisting of structural synapse damage, mitochondrial dysfunctions and transcriptional down-regulation of neuroprotective pathways. Synaptic excitotoxicity is further aggravated by increased incorporation of AMPA and NMDA receptors into the postsynaptic membrane, together with a concomitant decrease of inhibitory synaptic transmission (not shown). During neuroinflammation, the integrity of the blood-brain-barrier is compromised and inter-endothelial cell contacts become leaky (indicated by red arrows in the depicted capillary) allowing the entry of blood-borne immune cells into the CNS. Arrows in magenta show interactions between components of the multipartite synapse. Arrows in red denote pathologically activated signaling events during neuroinflammation. Abbreviations: CP-AMPA, Ca 2+ -permeable AMPA receptors; iGluR, ionotropic glutamate receptor; mGluR, metabotropic glutamate receptor; EAAT, excitatory amino acid transporter.
Collectively, early synapse changes in MS (and mouse models of MS) appear to result from increased levels of inflammatory cytokines. Synapse dysfunctions are likely correlates of the known cognitive disabilities and memory dysfunctions in MS patients [68,71,76,92,279,280,[420][421][422][423][424][425][426][427][428][429]. Signals from glial cells, particularly microglia, play a prominent role in synaptic pathology. In support of this suggestion, paralysis of microglia ameliorates EAE [430]. Clearly, the role of microglia in this process is complex. As mentioned, pro-inflammatory microglia ("M1" microglia) can generate pathologically elevated levels of inflammatory cytokines in active MS. M2 microglia can counteract these events and disease pathology. Thus, influencing the differentiation behavior of activated microglia toward the anti-inflammatory M2 microglia subtype will likely provide potential for the development of novel therapeutic strategies in MS. Interestingly, MS susceptibility genes are more frequently associated with microglia functions than with neuronal or astrocyte functions emphasizing the central role of microglia for MS [431,432].
Many of the described synaptic alterations in MS/mouse models of MS can be assigned to the postsynaptic compartment. In glutamatergic ribbon synapses of the retina, also strong alterations of presynaptic events have been observed in EAE [239][240][241]. These alterations included changes in the molecular composition of components of the active Figure 3. Schematically depicts a brain synapse, as shown in Figure 1, but under neuroinflammatory conditions as in multiple sclerosis (MS). Perisynaptic astrocytes and microglia are strongly activated (encircled in red) and secrete large amounts of inflammatory cytokines. Inflammatory cytokines produce glutamatergic excitotoxicity (see text). Neuroinflammation leads to decreased expression of astrocytic glutamate transporters and to elevated levels of extracellular glutamate (colored in red). Increased extracellular glutamate binds to extrasynaptic glutamate receptors. This initiates a series of deleterious events at the synapse consisting of structural synapse damage, mitochondrial dysfunctions and transcriptional down-regulation of neuroprotective pathways. Synaptic excitotoxicity is further aggravated by increased incorporation of AMPA and NMDA receptors into the postsynaptic membrane, together with a concomitant decrease of inhibitory synaptic transmission (not shown). During neuroinflammation, the integrity of the blood-brain-barrier is compromised and inter-endothelial cell contacts become leaky (indicated by red arrows in the depicted capillary) allowing the entry of blood-borne immune cells into the CNS. Arrows in magenta show interactions between components of the multipartite synapse. Arrows in red denote pathologically activated signaling events during neuroinflammation. Abbreviations: CP-AMPA, Ca 2+ -permeable AMPA receptors; iGluR, ionotropic glutamate receptor; mGluR, metabotropic glutamate receptor; EAAT, excitatory amino acid transporter.
Collectively, early synapse changes in MS (and mouse models of MS) appear to result from increased levels of inflammatory cytokines. Synapse dysfunctions are likely correlates of the known cognitive disabilities and memory dysfunctions in MS patients [68,71,76,92,279,280,[420][421][422][423][424][425][426][427][428][429]. Signals from glial cells, particularly microglia, play a prominent role in synaptic pathology. In support of this suggestion, paralysis of microglia ameliorates EAE [430]. Clearly, the role of microglia in this process is complex. As mentioned, proinflammatory microglia ("M1" microglia) can generate pathologically elevated levels of inflammatory cytokines in active MS. M2 microglia can counteract these events and disease pathology. Thus, influencing the differentiation behavior of activated microglia toward the anti-inflammatory M2 microglia subtype will likely provide potential for the development of novel therapeutic strategies in MS. Interestingly, MS susceptibility genes are more frequently associated with microglia functions than with neuronal or astrocyte functions emphasizing the central role of microglia for MS [431,432].
Many of the described synaptic alterations in MS/mouse models of MS can be assigned to the postsynaptic compartment. In glutamatergic ribbon synapses of the retina, also strong alterations of presynaptic events have been observed in EAE [239][240][241]. These alterations included changes in the molecular composition of components of the active zone, presynaptic Ca 2+ homeostasis, and decreased exocytic and endocytic synaptic vesicle cycling. The detailed underlying mechanisms for these presynaptic changes remain to be elucidated but might possibly also involve glutamatergic excitotoxicity. As mentioned, glutamate excitotoxicity can also affect presynaptic events via presynaptic glutamate receptors [275,276]. Gliotransmitter like IL1β and TNFα could also exert effects on such presynaptic events based on the presence of receptors for IL1β and TNFα at the presynaptic terminal [342]. Disorders of the visual system are frequent symptoms of multiple sclerosis, and the early dysfunctions of retinal ribbon synapses observed in the EAE mouse model of MS [239][240][241] could contribute to these symptoms.

Conclusions and Outlook
Glial cells play an important role in modulating synaptic activity under normal, healthy conditions. Glial cells are targets and amplifiers of neuroinflammatory signals that if secreted in excessive amounts, damage brain synapses and lead to progressive neurodegeneration and brain dysfunctions. The underlying mechanisms and signaling cascades are not fully understood despite enormous recent scientific advancement. Further analyses will likely provide not only an improved understanding of synaptopathy in MS but will also help in the development of novel therapeutic strategies. Acknowledgments: The work of the authors was supported by Dr. Rolf M. Schwiete Foundation (2021-022) and Deutsche Forschungsgemeinschaft DFG (FOR2289). We apologize to our colleagues that not all relevant studies/original work could be mentioned in the review due to limitations of space.

Conflicts of Interest:
The authors declare no conflict of interest.