Functional and structural damage of neurons by innate immune mechanisms during neurodegeneration

Over the past decades, our view on neurodegenerative diseases has been mainly centered around neurons and their networks. Only recently it became evident that immunological processes arise alongside degenerating neurons, raising the question whether these represent just meaningless bystander reactions or in turn, contribute to pathogenesis and disease symptoms. When considering any effect of inflammatory events on the CNS one has to consider the site, duration and nature of immune activation. Likewise, one has to distinguish between mechanisms which directly impact the neuronal compartment and indirect mechanisms, which affect cells that are important for neuronal functioning and survival. As discussed in this review, both types of mechanisms may be present at the same time and additively or synergistically lead to neuronal demise. Inflammatory mediators released by the principle innate immune cells of the brain, microglia and astrocytes, can compromise the function and structure of neurons, thereby playing important roles in the pathogenesis of neurodegenerative diseases.


Introduction
Infiltration and activation of immune cells within the central nervous system (CNS) is tightly regulated and therefore the brain has been traditionally regarded as an immune-privileged organ. However, the recent discovery of a brain lymphatic system 1,2 is challenging this traditional view and is leading to the evolvement of new concepts in our perspective of the brain's immune system. Accumulating evidence revealed an involvement of immune cells in the response to acute as well as chronic neurodegeneration in the CNS. The brain's own immune system consists of resident microglia and immunocompetent astrocytes, with external immune cells like, e.g., T-cells and neutrophils usually being separated from the brain by the blood-brain-barrier (BBB) and blood-cerebrospinal fluidbarrier. However, upon injury, these cells may also infiltrate the brain and modulate neuroinflammation in the CNS 3 .
Interestingly, neuroinflammatory processes were shown to be involved in the development and progression of a number of neurodegenerative diseases. In these diseases, immune cells seem to be chronically activated and adopt a special phenotype that is ultimately leading to structural and functional damage of neurons. Several neurodegenerative diseases are characterized by protein mis-folding, leading to the progressive intracellular and extracellular accumulation of different protein aggregates. This usually coincides with neuroinflammation or sometimes is even preceded by the activation of immune cells. For example, in Alzheimer's disease (AD), amyloid-beta (Aβ) is accumulating in extracellular amyloid plaques and hyperphosphorylated tau in intracellular neurofibrillary tangles (NFT). This together with a chronic neuroinflammation characterized by the activation of microglia, which cluster around Aβ plaques (Fig. 1), and astrocytes leads to progressive neuronal loss, ultimately resulting in cognitive decline and memory loss 4 .

Microglia
Microglia are mononuclear phagocytes from the myeloid cell lineage. They derive from macrophages originating from the yolk sac and migrate into the CNS during early embryonic development [5][6][7] . Even though they populate the whole brain, they are not homogenously distributed throughout all regions 8 . Under healthy conditions, microglia show a ramified morphology, but even under resting conditions their processes have been shown to extend and retract, thereby surveying their nearby environment 9 . Their main tasks, which they accomplish by their phagocytic activity and release of compounds, e.g., trophic factors, cytokines, chemokines, nitric oxide (NO) and reactive oxygen species, are immune defense and tissue maintenance as well as pruning and remodeling of synapses. In the past, microglia have been separated into two distinct phenotypes: M1 microglia, which exhibit an activated, proinflammatory phenotype and M2 microglia with a non-inflammatory phenotype. This simplified view, however, has been challenged during the last years with accumulating evidence suggesting a much more diverse and distinct set of different microglia phenotypes being present in the adult brain 10,11 . In general, microglia can be activated by danger-associated molecular patterns or pathogen-associated molecular patterns 12 . Upon activation, microglia can migrate to the site of injury 13 , shift from normal self-renewal to clonal expansion 14 and secrete pro-inflammatory cytokines like interleukin-1β (IL-1β) 15 as well as other immunomodulatory factors (Fig. 2). The involvement of microglia in neurodegeneration has been highlighted in the past decade by the identification of mutations in microglia-expressed genes like Trem2 and Cd33, which have been associated with an increased risk to develop sporadic AD [16][17][18] . Intriguingly, a recent study found AD risk-modifying single-nucleotide polymorphisms enriched in myeloid cells 19 , further strengthening the hypothesis of immune cells as central mediators in disease pathogenesis in sporadic forms of AD. While the initiation of an innate immune response by microglia developmentally is meant to protect the brain, an excessive reaction as seen in chronic neurodegenerative diseases can be detrimental and further promote disease progression.

Astrocytes
Astrocytes are the most abundant glial cell type in the brain. Despite not being a classical immune cell, they take on immunomodulatory functions in the CNS. They possess a very complex morphology with an elaborate network of branches terminating in either fine structures, which ensheath and directly interact with synapses or processes that enwrap the surface of blood vessels. Work over the past years showed that astrocytes represent not only morphologically 20 but also functionally a very heterogenous cell population (see 21 for a detailed review). They respond to neuronal activity with calcium waves 22,23 , regulate ion homeostasis, provide energy and precursors for the antioxidant glutathione (GSH) to neurons and modulate synaptic transmission by controlling neurotransmitter removal 24,25 . Furthermore, astrocytes interact with neurons to modulate their activity in a subtype-specific and synapse-specific manner [26][27][28] , and play an active role in synapse formation 29,30 and elimination 31 . Upon acute injury as well as during the development of progressive, chronic diseases in the CNS, astrocytes shift to a reactive phenotype which is characterized by transcriptional, morphological and functional changes 32 . To date, reactive astrocytes have been implicated in a number of neurodegenerative diseases including AD, Parkinson's disease (PD), Huntington's disease (HD) and amyotrophic lateral sclerosis (ALS) 32 .

Others
Other cells of the innate and adaptive immune system are located in the meninges, perivascular spaces and choroid plexus, patrolling the outer border of the brain. The innate immune system in the periphery is comprised of a number of different cell types. For example, natural killer (NK) cells perform cytotoxic functions and neutrophils are the first line of defense against bacteria and other invading microorganisms. Dendritic cells are classical antigen-presenting cells, while macrophages are phagocytic cells that also exert antigen-presenting functions. Cytokine secretion and antigen-presentation by the innate immune system lead to activation of cells in the adaptive arm of the immune system. Here, CD8-positive T-cells exert direct cytotoxic functions while CD4-positive T-cells are so-called "helper" cells, which can stimulate other T-cell effector functions and antibody production by B-cells 33 .
Under normal conditions, these peripheral immune cells can influence the brain by e.g., secretion of different cytokines, which then pass the BBB. However, upon injury or inflammation, these cells may also infiltrate the brain and exert more direct effects.

Functional impairment of neurons
Inflammatory mediators may immediately damage neuronal structures and disrupt neuronal integrity upon release. However, it seems more likely that before causing structural damage, inflammation impairs neuronal function. Alternatively, one may hypothesize that functional or structural damage of neurons depend on the concentration of the respective inflammatory mediator, and thus the same inflammatory event may cause both, functional and structural impairment depending on the distance to the site of release. Furthermore, negative modulation of neuronal activity or damage to neuronal structures may occur at single synapse level, the level of neuronal micro-networks or even affect larger network modules, depending on the nature, site, degree and duration of inflammatory changes. When considering functional impairment of neurons by inflammatory mediators, several mechanisms of action need to be considered: first, microglial cells, which-as already mentioned-are needed to physiologically prune and remodel synapses through their processes, will lose this function simply because any inflammatory activation will cause process retraction 34 , a process mediated via adenosine A 2A receptors in chronic neuroinflammatory states 35 . This most likely terminates this physiological, contact-dependent interaction 36 between neurons and microglia. It can therefore be speculated that sustained activation of microglia impairs the efficacy of the synaptic compartment by failing to control the necessary synapse turnover, which is required to guarantee normal and  Under healthy conditions, microglia promote neuronal survival by secretion of e.g., BDNF, but also play an active role in synapse pruning. Once the microglia are immune activated by binding of e.g., amyloid-beta, they secrete a multitude of immunomodulatory factors like nitric oxid (NO) and IL-1β, which ultimately can lead to negative effects on neuronal function, structure and survival failure-free neuronal functioning. Second, microglial cells generate and secrete a plethora of neurotrophic factors of which brain-derived neurotrophic factor (BDNF) was shown to be mandatory for distinct forms of learning and memory in rodents 37 . Sustained immune activation, however, strongly reduces the generation of these factors and it does not seem too far-fetched to conclude that reduced levels of neurotrophins contribute to neuronal failure and loss during neurodegeneration. A third potentially damaging component arises from the release of inflammatory mediators itself (Table 1). Here, either the respective immune molecule may negatively affect neurons, or alternatively alter the toxicity of known executors of neurodegeneration. Thus, it has been shown that the generation and persistent release of nitric oxide, due to induction of nitric oxide synthase type 2 (NOS2) leads to the suppression of synaptic plasticity, measured as hippocampal long-term potentiation (LTP) 38 . In these experiments, exposure of murine hippocampal brain sections to fibrillar Aβ suppressed LTP, depending on the presence of NOS2, as this effect was observed in wildtype, but not brain sections of NOS2-knockout mice or was prevented by addition of a NOS2 inhibitor. This suppressive effect may depend on the direct modification of neurophysiological processes, e.g., through the inhibition of mitochondrial respiration or through posttranslational changes of synaptic proteins 39 . Additionally, NO may alter the toxicity of Aβ itself, since the nitration of Aβ at its tyrosine residue at position 10 increased its potency to suppress hippocampal LTP 40 .
While several of the immune mediators described in neurodegenerative diseases 12 may account for the induction of NOS2 in microglia and astroglia, some of these factors have been shown to act as suppressors of LTP or to compromise trophic factor functions themselves. Thus, IL-1β has been shown to suppress LTP in hippocampal brain slices and also to counteract the function of BDNF 41 . Similarly, tumor necrosis factor alpha (TNF) and interleukin-6 (IL-6) are further cytokines for which suppression of synaptic plasticity has been demonstrated [42][43][44] . Suppression of LTP may not be restricted to cytokines, because further immune molecules such as astroglialderived complement factor c3 has been equally shown to suppress LTP 45 . Since most of these experiments, however, have been performed in slice cultures and not yet been replicated by recording in vivo LTP, they have to be interpreted with caution. Next to the missing control of arising neuronal projections from other brain areas, it remains unclear whether the experimentally applied concentrations of immune mediators are reflecting those reached under pathophysiological conditions in vivo. Another yet understudied mechanism is the impact of astroglial changes on neuronal dysfunction and death independent of the release of immune factors. While astroglia equally provide the neuronal compartment with neurotrophic factors, it is still less clear if this action is compromised by the presence of immune factors. Nevertheless, given the importance of the astrocyte compartment for molding synapses and participating in synaptic transmission, it seems likely that any sustained changes of the astroglial compartment will negatively feed back on neuronal functioning and survival.

Structural damage of neurons
Over the past years, microglia as well as astrocytes have emerged as potential candidates in mediating non-cell autonomous neurodegeneration. Mutations in superoxide dismutase 1 (SOD1) lead to substantial spinal motor neuron loss, a hallmark pathology of ALS 46 . In an elegant study, Boillée et al. showed that reducing mutant SOD1 accumulation specifically in microglia without changing expression in other cell types slowed disease progression 47 . Additionally, astrocytes expressing mutant SOD1-without expression in any other cell type-were shown to trigger non-cell autonomous degeneration of motor neurons [48][49][50] , an effect at least partially mediated via a complex composed of α-adducin and the ion pump α-NA/K-ATPase 51 . Furthermore, Bergmann glia, a specialized astrocyte subtype in the cerebellum, were shown to be involved in Purkinje cell degeneration in the cerebellar cortex, leading to atrophy and ataxia. Elimination of Bergmann glia as well as cell-specific expression of mutant ataxin-7, a protein involved in dominant spinocerebellar ataxia type 7, resulted in downregulation of glutamate transporters. Therefore, excitotoxic stress due to the decreased reuptake of glutamate has been postulated as the mechanism for the resulting Purkinje cell loss in both models 52,53 . In line with this, selective activation of the IKK2/NF-κB signaling pathway in mouse astrocytes was shown to lead to Purkinje cell loss due to NF-κB-induced downregulation of glutamate transporters in Bergmann glia 54 . Structural damage by neuroinflammatory processes has also been implicated in the development and progression of AD. In amyloid precursor protein (APP)-transgenic mice, a model for pathological Aβ deposition in AD, reactive microglia and astrocytes were detected already before amyloid deposition 55 , a finding that was at least partially confirmed for both cell types in human patients later on as well 56,57 . Mechanistically, it has recently been shown that microglia in APP-transgenic mice, which have been exposed to soluble Aβ-oligomers, mediate synapse loss already before amyloid plaque deposition in a complement-dependent manner 58 . Interestingly, microglia depletion for 2-4 weeks in three different mouse models with amyloid plaques (APPPS1, APP23 and 5× FAD) did not result in reduced plaque burden 59,60 . Further analysis of neuronal structures revealed no benefit for plaque-associated neuritic dystrophy in APPPS1 or APP23 mice 59 , but prevented neuronal loss in 5× FAD mice 60 . Work from Condello et al. suggests that the degree of neuronal dystrophy in 5× FAD mice depends on coverage of the respective area by microglia, most likely due to the higher accumulation of Aβ42 when microglia are absent 61 . Therefore, microglia might have direct as well as indirect effects on neuronal structures. In addition to microglia, astrocyte function influences neuronal survival in AD. Glutamate transporter expression and function in astrocytes in AD patients was shown to be reduced, an effect that correlated with disease progression and therefore led to the hypothesis of glutamate excitotoxicity as a potential inducer of neuronal loss [62][63][64] .
One important pathway that modulates microglia activation involves signaling via the fraktalkine receptor CX3CR1 on microglia. CX3CR1 deficiency in wildtype mice resulted in activated microglia and promoted increased neuronal cell loss after a systemic lipopolysaccharide (LPS) challenge 65 . Ablation of CX3CR1 in APPtransgenic mice led to reduced amyloid plaque burden most likely due to an increased phagocytic activity of CX3CR1-negative microglia in these mice 66,67 . However, the exacerbated inflammatory response still resulted in neuronal damage including increased intracellular accumulation of pathological hyperphosphorylated tau species 68 . These differential effects on Aβ and tau pathology are also underlined further by work done in tauopathy mouse models. In these models, expression of mutant forms of human tau (htau) lead to aggregation of tau in intracellular NFT, which models neurodegenerative diseases like e.g., frontotemporal dementia. The deposition of tau in these mice is preceded by microglia activation and ultimately results in synaptic loss and neuronal death 69,70 . In contrast to the effects seen on amyloid plaques in APPtransgenic mice, CX3CR1-deficiency in htau-transgenic mice accelerated the development of tau pathology 71 , an effect most likely mediated by cytokines of the interleukin-1 family secreted by microglia 71,72 . In line with this, microglia-deficiency decreased tau pathology in a rapid tau propagation mouse model 73 .
In addition to the neuronal damage seen in AD models, loss of CX3CR1 also promoted neuronal death in PD and ALS models 65 . Interestingly, different neuronal subtypes showed a differential vulnerability to neuroinflammation. Intranigral injection of LPS led to microglia activation and pro-inflammatory cytokine-mediated selective loss of neurons in the nigrostriatal dopaminergic system, while neurons in the serotoninergic system were only transiently affected 74,75 . In addition to different susceptibilities of neuronal subtypes to neurotoxic stimuli, the abundance of microglia seems to be important. Injection of the same concentration of LPS into the hippocampus, cortex or substantia nigra of rats resulted in neurodegeneration only in the substantia nigra, the region with the highest density of microglia 76 .
In addition to exerting separate effects, microglia and astrocyte function in neurodegenerative diseases have also been shown to be dependent on each other. Chronic expression of interleukin-3 by astrocytes led to the development of a progressive motor disorder due to recruitment of microglia to the white matter. Here, microglia activation promoted primary demyelination, thereby recapitulating features of demyelinating diseases like e.g., multiple sclerosis (MS) and HIV leukoencephalopathy 77 . Another study identified three factors (Interleukin-1α, TNFα and C1q), secreted by activated microglia as inducers of a special neurotoxic phenotype in astrocytes 78 , characterized by a unique transcriptional profile 79 . Aside from causing neuronal cell death, these neurotoxic astrocytes release complement factors, which the authors speculate could be involved in synapse loss in neurodegenerative diseases. Interestingly, this special reactive astrocyte subtype was detectable in post-mortem tissue from patients with AD, PD, HD, ALS and MS, but their involvement in the progression of all of these diseases has not yet been proven 78 .
To date, little is known about the role of immune cells other than microglia and astrocytes in neurodegeneration. While peripheral immune cells usually cannot access the CNS, a special scenario can be observed in MS patients, since it is a neuroinflammatory, neurodegenerative autoimmune disorder characterized by the infiltration of mainly T-cells and macrophages 80,81 , which directly attack neurons and myelin-forming oligodendrocytes 82 . In addition, several studies describe infiltration of immune cells into the brain parenchyma in other neurodegenerative diseases. For example, CD4-positive and CD8-positive Tcells have been detected in the brain parenchyma of a PD mouse model as well as post-mortem samples from human PD patients 83,84 . In the brains of APP-transgenic mice, NK cells and macrophages were detected. The production of interferon-gamma by NK cells, a potent activator of microglia, has been postulated to influence glia activation in the diseased brain 85 . Furthermore, neutrophils were shown to be present close to Aβ plaques in transgenic mice as well as in the brain parenchyma of AD patients 86 . Additionally, macrophages, B-cells 87 and CD4-positive as well as CD8-positive T-cells are detectable in the brains of AD patients [88][89][90] . Recent work revealed infiltration of CD8-positive T-cells into the brain parenchyma of frontotemporal dementia patients and a tauopathy mouse model. In this model, T-cell depletion did not affect tau pathology, but had positive effects on the expression of proteins involved in synaptic plasticity. However, it remains unclear if the infiltrating T-cells have direct effects on neuronal structures in this particular model 91 . Dependent on the context, infiltration of immune cells from the periphery can also have neuroprotective effects. For example, infiltrating CD4-positive T-cells have been shown to promote motor neuron survival after axotomy 92 . In addition, CD4-positive T-cells were shown to be important in restricting damage from secondary injuryinduced neuronal loss in MS mouse models 93 and led to increased clearance of myelin debris by microglia, thereby possibly promoting regeneration 94 .

Conclusion/Perspective
Microglia and astrocyte function modulate neurodegenerative diseases at different levels of disease progression. Microglia, for example, not only secrete immunomodulatory factors, but also directly interact with neurons. This in the end does lead to functional as well as structural damage of neurons (Fig. 3). In most cases it still remains unclear whether neuroinflammation is a cause or consequence of the disease and which mechanisms are underlying neuronal dysfunction. Further work is necessary to shed light on neurotoxic events and the sequential steps in the progression of neurodegeneration. For prospective, more specific treatment strategies, it will be important to understand if and how functional and structural damage of neurons by microglia, astrocytes and other immune cells are linked to each other. Also, disease-specific as well as common neuroinflammatory events have to be analyzed in more detail to develop clear treatment paradigms for different diseases. For example, nitric oxide (NO) has emerged as one common toxic insult in many neurodegenerative diseases including AD, PD, MS and ALS 95 , and its reduction could be a potential treatment strategy to alleviate neuronal damage. But the optimal timing for a treatment with an antiinflammatory agent needs to be taken into consideration, which could be target and/or disease specific. So far, little is known about potential benefits of certain aspects of a neuroinflammatory response in early or even late stages of neurodegeneration, but have to be taken into account when new therapies are designed. This clearly needs further investigation within the next years to enable us to better predict as well as treat neurodegenerative diseases.