Glutamate excitotoxicity is one primarily cause of neuronal death by necrosis and apoptosis and a critical player in ALS onset and progression. Indeed the blockade of the glutamate transporter in a SC organotypic slice model from SOD1G93A rats has shown to result in an increased survival of MNs (Yin and Weiss, 2012). Glutamate, by activating the glutamatergic ionotropic receptor N-methyl-D-aspartate (NMDAR), triggers the influx of Ca²⁺ into neurons, increasing its intracellular levels (for review, see Spalloni et al., 2013). In these conditions, changes in mitochondria dynamic properties are produced leading to excessive oxidative phosphorylation and increased generation of ROS and reactive nitrogen species (RNS) that culminate in apoptosis. In parallel, the NMDAR excitotoxicity also results in the release of Ca²⁺ from the ER. In addition, it was shown that ER stress resulting from the accumulation of aggregated mSOD1 and dysfunction of the unfolded protein response (UPR) activation contributes to the apoptotic signaling cascade in ALS (for review, see Sofroniew, 2009).

Besides the synaptic glutamate pool directly implicated in the excitatory neurotransmission it has been lately additionally considered the extra-synaptic glutamate pool that influences cell communication. This pool, mainly derived from astrocyte and microglia release, is greatly increased by pathological stimuli (for review, see Rodriguez et al., 2013). Besides being implicated in many physiological conditions glutamate contribution to neuron–glia alterations of homeostasis and to the pathophysiology of neurodegenerative diseases, such as ALS, needs to be explored in the near future, due to controversial results. In fact, although linkage between glutamate and neuroinflammation was suggested to have a role in potentiating MN death in a model mimicking ALS disease (Tolosa et al., 2011), other Authors have demonstrated that glutamate by itself was not able to induce MN damage in ALS (Tovar et al., 2009b). Indeed, it has been questioned the involvement of glutamate in ALS-induced MN death (Le Verche et al., 2011) and the cytotoxicity of the cerebrospinal fluid (CSF) from patients with ALS evidenced to not be related to glutamate (Gomez-Pinedo et al., 2013).

Fractalkine (FKN) mRNA is dominantly expressed in neuronal cells, particularly in those at the cortex, hippocampus, caudate putamen, thalamus, and olfactory bulb (Harrison et al., 1998). FKN mRNA was also detected in unstimulated astrocytes (more) and microglia (less) (Mizuno et al., 2003). Interestingly, FKN immunoreactivity and mRNA was observed also in the rat SC and dorsal root ganglia neurons, but not in glia, and levels were not enhanced by neuropathic conditions (Verge et al., 2004). FKN, or CX3CL1, exists as a membrane-bound and soluble protein (sFKN) allowing both adhesive and chemoattractive properties (Bazan et al., 1997). In fact, the sFKN has potent chemoattractant activity, recruiting CXCR1-expressing T cells, monocytes, and microglia to the injured neurons, as well as in regulating the phagocytic capacity of microglia (Cardona et al., 2006). sFKN was shown to increase upon stimulation of glutamate (Chapman et al., 2000) and to induce proliferation of human microglia (Hatori et al., 2002). The constitutive expression of FKN and its receptor CX3CR1 in microglia indicates its involvement in fundamental processes of communication between neurons and microglia (Harrison et al., 1998).

Under stimulation, such as an excitotoxic stimulus (Chapman et al., 2000), the membrane-bound form of FKN is rapidly cleaved from cultured neurons and significantly reduces neuronal NMDA-induced apoptosis (Deiva et al., 2004). Indeed attenuated glutamate-induced neuronal cell death was observed after treatment of primary neuron–microglia co-cultures with sFKN (Noda et al., 2011). Proteolytic cleavage of CX3CL1 is mediated by action of metalloproteinase ADAM10 (Gough et al., 2004), ADAM17 (Garton et al., 2001; Tsou et al., 2001), cathepsin S (CatS) (Clark et al., 2009), and MMP-2 (Bourd-Boittin et al., 2009). Thus, cleavage of FKN may be prevented by the inhibition of MMPs (Chapman et al., 2000). The source of CatS is the activated microglial cells that upon stimulation with LPS secrete sFKN to the extracellular media (Clark et al., 2009). Similarly, stromal cell-derived factor-1 (SDF-1) was indicated to stimulate the expression of ADAM17 and to increase sFKN, while up-regulating FKN expression (Cook et al., 2010).

In addition, FKN has a neuroprotective function by inhibiting the nitric oxide (NO) production and the expression of inducible NO synthase (iNOS) mRNA in activated microglia (Mizuno et al., 2003). Interestingly, it was observed that FKN expression was reduced in the brain of aged rats probably accounting for the increase in microglial activation in such condition. In fact, treatment with FKN has attenuated the age-related increase in microglial activation (Lyons et al., 2009).

Taken together, FKN seems to have both intrinsic and anti-inflammatory properties in the CNS and to act by interfering with toxic microglial–neuron interactions (Suzuki et al., 2011). Whether FKN may have a role in the development of ALS is still not known.

High-mobility group box 1 (HMGB1) protein, also known as amphoterin, is an inflammatory factor that can be released by astrocytes, microglia, and neurons, mainly when cells are dying (for review, see Fang et al., 2012). HMGB1 has dual activities depending on whether is alone (probably promoting inflammation resolution and tissue regeneration) or forming complexes with several proinflammatory mediators (potentiating inflammation and promoting innate immune cell activation) (Bianchi, 2009). Nuclear HMGB1 regulates transcription of different sets of genes, including proinflammatory genes (Bianchi and Manfredi, 2009; Park et al., 2009; Wong, 2013). Beneficial effects were observed in early CNS development but increased levels of HMGB1 were shown to be correlated with apoptosis and degeneration of neurons (Liu et al., 2009; Kawabata et al., 2010). When neurons are injured, secretion of HMBG1 activates microglia through receptor for advanced glycation end products (RAGE), Toll-like receptors (TLRs) 2, 4, and 9, as well as Mac1 receptors (Kim et al., 2006a; Park et al., 2006; Neusch et al., 2007), as depicted in Figure 1. Release of inflammatory mediators by activated microglia further induces neuronal necrosis. HMGB1 was indicated to decrease in neurons and to increase in astrocytes with aging (Enokido et al., 2008). When expression and localization of HMBG1 was evaluated in the lumbar SC of SOD1G93A transgenic mice, although intense reactivity was found, no differences were obtained between controls and the SOD1 mice (Lo Coco et al., 2007). However, since HMBG1 was identified in the cytoplasm of astrocytes and microglia in SC samples from ALS patients (Casula et al., 2011) it may trigger TLR signaling pathways. The finding was observed at the late ALS phase and, thus, it will be important to follow the TLR/RAGE cascade in animal models at different stages of the disease. In contrast, a progressive reduction of HMGB1 immunopositive MNs was found at advanced stages and may reflect the loss of MNs, reduced synthesis or enhanced released of the cytokine (Lo Coco et al., 2007). Additional studies are needed to investigate the causative hypothesis indicated for the decreased HMGB1 immunoreactivity in MNs, inasmuch since it can also have beneficial effects on neuroregeneration (Fang et al., 2012).

CCL21 was shown to be implicated in signaling neuronal injury to microglia through the receptor CXCR3. CCL21 expression was demonstrated to increase in cortical neurons, in vitro, 2 h after excitotoxic stimulus (de Jong et al., 2005). Intriguingly, these authors demonstrated its location within vesicles that are transported along the neuronal process till presynaptic structures. This chemokine is considered a chemotactic agent important to drive microglia to the site of lesion. Astrocytes were also indicated to have the receptor CXCR3, but unless high levels of CCL21 are produced, no relevant changes can be observed, indicating separate functions from microglia (van Weering et al., 2010). CCL21 expression in CNS revealed to induce a massive brain inflammation, but not lymphocytic infiltration in transgenic mice expressing the chemokine (Chen et al., 2002). Interestingly, CCL21/CXCR3 signaling axis was never explored in ALS.

The membrane glycoprotein CD200 is expressed in neurons and in endothelial cells and its receptor CD200R is restricted to cells of myeloid origin including macrophages and microglia. CD200 was also evidenced to be induced by kainic acid in microglia (Yi et al., 2012) and the authors have suggested that microglia are maintained in an activated state with autocrine signaling by interactions between microglial CD200 and CD200R, as well as in a surveillant/quiescent state by interactions between neuronal CD200 and microglial CD200R. In addition, they also demonstrated that IL-4 leads to an increased expression of CD200 what may be the mechanism down-regulating microglia activation whenever IL-4 is produced. In conclusion, deficits in CD200–CD200R system exacerbate microglia activation and the release of proinflammatory cytokines (for review, see Jurgens and Johnson, 2012). Curiously, multivariate analyses of gene expression have previously identified alterations on the CD200R expression in presymptomatic SOD1 mice model (Chen et al., 2004). Thus, more studies on the role of CD200/CD200R in ALS are needed to evaluate whether its disruption is implicated in ALS.

Neuregulins have been showing to be implicated in a wide range of neurological and psychiatric disorders including multiple sclerosis, schizophrenia, and AD. Neuregulin precursors are expressed predominantly in cortical neurons, but also accumulate at the surface of white matter astrocytes. CSF neuregulin was found reduced in ALS and increased in AD (Pankonin et al., 2009). Neuregulin-1 (NRG1) is a growth and differentiation factor that binds to erbB receptors in microglia. NRG1–erbB signaling is activated after peripheral nerve injury and contributes to microgliosis and neuropathic pain (Calvo et al., 2010). Recently, it was suggested that NRG1 released from damaged neurons and other cells in the SC triggers microglial activation leading to progressive MN degeneration in ALS (Song et al., 2012). Indeed, it was observed that in the early stage of ALS there was activation of NRG1 receptors on microglia. Later on, a reduced membrane-bound NRG1 and an increased mRNA expression were noticed in the SC. More studies are, however, needed to sustain the NRG1 contribution to ALS pathogenesis.

Major histocompatibility complex (MHC) class I and class II molecules present fragments of peptide antigens to CD8⁺ or CD4⁺ T cells, respectively, and cells lacking such molecules are unable to interact with immunocompetent T cells (Parnes, 1989). Presentation of antigenic peptides to CD8⁺ cells is mediated by β2-microglobulin, which is non-covalently bound to MHC class I. Despite being recently observed during hippocampal neurodevelopment (Leinster et al., 2013), adult neurons do not constitutively express MHC class I (Arthur-Farraj et al., 2012). In contrast, spinal MNs were shown to display a high constitutive expression of both MHC class I and β2-microglobulin mRNAs (Thams et al., 2009). A latest paper describes a strong up-regulation of β2-microglobulin in MNs during disease progression in the SOD1G93A mice model, which revealed to be important for the ALS mouse survival (Staats et al., 2013). Therefore, β2-microglobulin target driven therapies may be of help in strategies counteracting ALS.

Glial cells (astrocytes, microglia, oligodendrocytes, and Schwann cells), as well as neurons are known to express different members of the TLR family (Lee et al., 2013b). Increasing evidence indicates that, in the absence of pathogens, TLR signaling can be activated by molecules released by the injured tissue, namely the HMGB1 protein, one of the damage-associated molecular patterns (DAMPs) molecules (Bianchi and Manfredi, 2009). Recent findings have underlined the activation of TLRs and RAGE signaling pathways in ALS (Casula et al., 2011). TLR2, TLR4, and RAGE expression was increased in reactive glial cells in both gray (ventral horn) and white matter in SC. TLR2 was predominantly detected in cells of the microglia/macrophage lineage, whereas TLR4 and RAGE were strongly expressed in astrocytes. Activated macrophages (He et al., 2012) and microglia (personal communication) release HMBG1 that promotes the secretion of IL-1β and IL-18 further inducing the necrosis of neighboring neurons and the amount of extracellular HMBG1, which binds to microglial Mac1 leading to the activation of nuclear factor-κB (NF-κB) pathway (Gao et al., 2011) and inflammasome (Lu et al., 2013), forming a vicious circle that sustains progressive neurodegeneration (Figure 1). Recent data have demonstrated that NF-κB activation in wt microglia causes MN death providing a therapeutic target for ALS (Frakes et al., 2014).

When evaluating whether extracellular mSOD1 caused a direct or indirect injury to MNs, it was observed that neurodegeneration was mediated by microglia and concerted activation of CD14/TLR pathway involving both TLR2 and TLR4 (Zhao et al., 2010). In addition, granulocyte macrophage-colony stimulating factor (GM-CSF), a pleiotropic cytokine predominantly released by astrocytes that up-regulates TLR4 and CD14 expression in microglia, may function to exacerbate TLR signaling in ALS disease (Parajuli et al., 2012). Indeed, when GM-CSF was blocked it was observed a delayed onset and increased life span in the ALS mice (Turner and Talbot, 2008). The activation of these pathways may contribute to the progression of inflammation and to a further injury to MNs.

Oligodendrocytes in CNS, and Schwann cells in the peripheral nervous system (PNS), are responsible for the myelin sheaths surrounding neurons which provide electrical insulation essential for rapid signal conduction. Schwann cells also participate in the clearance of debris and in guiding the axon after neuron damage (Ilieva et al., 2009). In the SC injury, reactive gliosis emerges in the lesion accompanied by the up-regulation of chondroitin sulfate proteoglycans (CSPGs) in oligodendrocytes and Schwann cells.

Considering the regenerative capacity of NG2 cells, one of the first cells responding to any alteration in CNS environment, it will be important to clarify their role in ALS since they can be mobilized to originate cells or to release factors. Nevertheless, NG2 cells were shown to remain committed to an oligodendrocyte lineage in both controls and mutant ALS mice, indicating their scarce participation (Kang et al., 2010). However, increased proliferation rate of NG2 revealed to mediate an elevated number of early-born oligodendrocytes that degenerate, resulting in gray matter demyelination in ALS mice and human CNS (Kang et al., 2013).

Although only a few studies have examined whether oligodendrocytes or myelin sheaths have a role in ALS, the myelin abnormalities consisting in loss of compact myelin and lamellae detachment in the SC of pre-symptomatic SOD1 transgenic rats and aggravated at symptomatic stages (Lasiene and Yamanaka, 2011) suggest that it may be an interesting target for further study. Biochemical examinations of the myelin structural components revealed a decrease in the phospholipid content along disease progression, as well as in cholesterol (already in the early presymptomatic stage) and cerebrosides (in the paralyzed animals) (Niebroj-Dobosz et al., 2007). Degeneration of oligodendrocytes was observed in human patients and mice models of ALS prior to the disease (Kim et al., 2013). Cells that were lost were replaced by newly differentiated oligodendrocyte precursor cells, which, however, evidenced reduced myelin basic protein. This finding is suggestive that they may contribute to MN degeneration in ALS (Philips et al., 2013). Moreover, these cells evidenced a loss of the monocarboxylate transporter 1 (MCT1), thus compromising the supply of lactate to MNs (Table 2).

Little is known about Schwann cell involvement in ALS pathology. Schwann cells are located near the MN axons and are known to bridge between denervated and reinnervated endplates, and to guide axonal sprouts (Magrane et al., 2009). Expression of mSOD1 in perisynaptic Schwann cells was suggested to interfere with the trophic maintenance of normal or regenerating motor axons (Inoue et al., 2003). In vivo evidence suggests that glial fibrillary acidic protein (GFAP) up-regulation in the stressed/proliferating Schwann cells may be the underlying pathological events (Keller et al., 2009). Conversely, different results were obtained in SOD1G73R mice where elimination of mutant SOD1G37R from Schwann cells failed to slow disease progression (Lobsiger et al., 2009). Nevertheless, Wang et al. (2012) found that knockdown of mSOD1 in Schwann cells of SODG85R transgenic mice delayed disease onset and extended survival indicating that SOD1G85R expression is neurotoxic. These results imply that diverse mutations confer different outcomes to cell toxicity and, in the case of Schwann cells, oxidative damage seems to be an important feature in the context of ALS.

Astrocytes are a potential source of both pro- and anti-inflammatory cytokines and are ideally placed in close proximity to BBB and BSCB, thus translating signals from the periphery to the CNS. Although not being immune cells, they can also contribute to the immune response. Astroglial activation, or astrogliosis, is characterized by hyperplasia, hypertrophy of cell bodies and cytoplasmic processes, up-regulation of intermediate filament proteins, namely GFAP and vimentin, mediating a histologically apparent glial scar at the lesion site in the damaged SC (Sofroniew, 2009). Reactive astrogliosis in ALS was first revealed in 1990s by increased GFAP staining in the subcortical white matter (Kushner et al., 1991), and later on similarly observed in the ventral and dorsal horns of the SC, as well as in the transition between gray matter and anterior and lateral funiculi, where the dystrophy of neuritis exists (Schiffer et al., 1996). SC astrocytes were shown to assume a neurotoxic phenotype in response to extracellular ATP. In SOD1G93A astrocytes this activation is mediated through P2X7 receptor signaling (Gandelman et al., 2010). Reactive astrocytes surround degenerating MNs in patients and transgenic animal models of ALS, and in particular those localized in the ventral SC of SOD1G93A mice, are a source of nerve growth factor (NGF) and NO, which are both required for MN apoptosis (Table 2) (Pehar et al., 2004). Interestingly, the NO-induced MN death was shown to be mediated by astrocytes expressing SOD1 and TDP-43 mutations (Rojas et al., 2014). Ferraiuolo et al. (2011) have shown that the increase in total NGF is due to the secreted pro-NGF fraction before cleavage to the mature form, achieving a twofold increased ratio in the SOD1G93 mice as compared to control conditions. Such finding may derive from the increased ALS-associated MMP-9 (Niebroj-Dobosz et al., 2010), one of the enzymes that degrade mature NGF in the extracellular space. In this scenario, neuroprotection of MNs could be achieved by promoting pro-NGF cleavage.

Recently, it was demonstrated that loss of GFAP did not affect disease onset and marginally shorten the SOD1H46R mice survival, indicating that GFAP only plays modulatory effects (Yoshii et al., 2011), at least in this mice. On the contrary, astrogliosis was observed in the SOD1G93A mice in either symptomatic or presymptomatic phase, preceding microglial activation (Yang et al., 2011). Indeed, depending on the ALS animal model used, astrocyte activation was observed to occur earlier or later and to be rather complex, increasing and decreasing in waves at different times throughout the disease (for review, see Evans et al., 2013). Despite no differences in the astrocyte number, increased GFAP labeling in the ventral horn of lumbar SC from SOD1G93A mice was observed before MN loss (Gerber et al., 2012). Astrocytes derived from such mice evidenced to uptake glutamate less efficiently and showed a reduced trophic response to activation, deficiently protecting MNs (Benkler et al., 2013). Attesting this low astrocyte efficiency, the mainly expressed astroglial S100B protein was found to be decreased in CSF (Sussmuth et al., 2003) and serum (Otto et al., 1998) from ALS patients. Interestingly, exposure of primary cultures of astrocytes from ALS patients to CSF evidenced to enhance GFAP and S100B expression (Shobha et al., 2010). Similarly to neurons, astrocytes have demonstrated constitutive and regulated expression of FKN, which may control the migration and function of the microglia (Sunnemark et al., 2005). Additional contribution of astrocytes to MN death involves the release of ROS mediated by mitochondrial dysfunction and of D-serine that is a co-activator of the NMDARs, thus exacerbating glutamate excitotoxicity, among other factors as indicated in Table 2 and reviewed in Valori et al. (2014).

A recent publication has characterized a specific astrocytic phenotype (aberrant astrocytes) obtained from primary SC cultures of SOD1G93A symptomatic rats (Diaz-Amarilla et al., 2011). These aberrant astrocytes isolated on the basis of marked proliferative capacity and lack of replicative senescence, lack GLT-1 and the NG2 marker and were shown to release toxic factors accounting for a MN hostile environment. This specific astrocyte phenotype expressing increased S100B and connexin-43 (Cx-43) is abundant in the symptomatic phase of the disease and seems to be located close to MNs, representing a new potential target for delaying ALS progression.

Data in postmortem tissues of ALS patients revealed changes in the morphology of astrocytes together with elevated GFAP and aldehyde dehydrogenase family 1, member L1 (ALDH1L1) (Philips and Robberecht, 2011), corroborating findings in animal models of ALS. Indeed, mSOD1 gene excision from microglia and selective reduction in astrocytes significantly slowed disease progression (Yamanaka et al., 2008b). However, there is still some controversy on whether astrogliosis is detrimental or beneficial, what surely will depend from spectrum intensity and toxicity potential of the aberrant phenotype.

Microglia are the immune resident cells of the CNS with both a supporting role to neurons and astrocytes, and immunological properties with either neuroprotective or neurotoxic potential. Though microglia activation was indicated to precede astrocyte reactivity (Alexianu et al., 2001), reduced neuroprotective behavior of mSOD1 microglia in a resting-state (Sargsyan et al., 2011), and at the disease end-stage (Nikodemova et al., 2013) was documented. If confirmed, boosting of microglia with stimulatory factors may reveal to be clinically useful.

In a recent paper, Roberts et al. (2013) verified that incubation of microglia with aggregated SOD1 first drives its location at the membrane level, and later within the cell. Moreover, supernatants of these SOD1 activated microglia caused a significant decrease in MN viability, which was not related with tumor necrosis factor-α (TNF-α) secretion, NO, or superoxide anion radical. The toxic factors involved are currently unknown, although it was shown that expression of mSOD1 increases TNF-α secretion (Liu et al., 2009), which may induce neurotoxicity by increasing the glutamate release by microglia in an autocrine manner (Takeuchi et al., 2006). Therefore, a very early stage of the disease may rely in this link between protein aggregation and microglial activation. Indeed, neurotoxic potential of mutant microglia in MN degeneration in ALS is a very well documented issue: (1) mSOD1 acting on microglia is required to cause the disease (Clement et al., 2003); (2) limiting mutant damage to microglia slows progression (Boillée et al., 2006b); replacement of mSOD1 microglia for wt microglia delays disease and prolongs mice survival (Lee et al., 2012).

Microgliosis at sites of MN injury is a neuropathological hallmark of ALS and recent therapeutic interventions are looking into factors capable of skewing microglia neurotoxic potential into a neuroprotective phenotype. However, a trophic role for activated microglia has been suggested at early stages of the disease (for review, see Lewis et al., 2012). In the SOD1G93A mice, it was observed a decrease in microglia number in the entire SC at the pre-symptomatic age (Gerber et al., 2012). The authors identified two diverse microglia subpopulations (low and high Iba1 expression cells), suggesting a distinct microglia involvement at pre- and early-symptomatic ALS stages. Interestingly, microglia heterogeneity observed between cervical and lumbar SC regions in ALS mice may derive from different environmental specificities determined by local MN/astrocyte/lymphocyte disposition and activation (Beers et al., 2011b). While cortical microglia appear unaffected by the disease, additional studies evidenced an increased microglial number in the lumbar SC at symptom onset, and neither the typical inflammatory nor the anti-inflammatory phenotypes were identified at end-stage (Nikodemova et al., 2013). In addition, elevated TNF-α gene expression and immunoreactivity were observed in lumbar SC of mSOD1 mice and related with invading microglia (Yoshihara et al., 2002). Early microglia activation in ALS may be further explored in vivo by positron emission tomography (PET). The possibilities of PET suggest its valuable contribution to monitor the progression of the disease and the efficacy of the therapy in use (Corcia et al., 2012).

Context-dependent neuroprotective and neurotoxic properties. An important neuroprotective role of microglia is phagocytosis. Following signaling by neuronal FKN, milk-fat globule EGF factor-8 protein (MFG-E8, SED1) is up-regulated and serve as a bridge via specific integrins between apoptotic neurons and microglia (Figure 1) (Leonardi-Essmann et al., 2005). The authors hypothesize that MFG-E8 is assembled on the surface of exosomes and apoptotic neurons so microglia can recognize their target cells. MFG-E8 seems to be essential for microglia engulfment and removal of the dying neurons. An interesting concept is that inflammatory microglia can also phagocyte viable neurons (Fricker et al., 2012a) through MFG-E8 mediation (Fricker et al., 2012b) accounting for a reduced number of neurons. Phagocytic ability may, however, be lost if the cell is continuously stressed by a neurotoxic stimulus.

Our recent data with the neurotoxic unconjugated bilirubin have evidenced that following the phagocytic ability at eliminating cell debris microglia changes to a more inflammatory phenotype (Silva et al., 2010) and even to a senescent-like cell morphology and death if the duration of exposure is prolonged. However, microglia behavior in the presence of unconjugated bilirubin is modulated by the presence of both astrocytes and neurons (Silva et al., 2011). Also to consider is that during the inflammatory phenotype, microglia may intervene in the glutamate homeostasis, but may also contribute to neurite degeneration through the release of NO (Silva et al., 2012). Interestingly, microglia phagocytic features were shown to occur early in ALS disease. Microglia were revealed to aggregate, proliferate, and phagocyte in the lumbar SC of pre-symptomatic mutant SOD1H46R transgenic mice (Sanagi et al., 2010). However, in other studies microglia have shown to contribute to MN death (Dibaj et al., 2011; Brettschneider et al., 2012) and to decrease in number within disease progression (Butovsky et al., 2012), thus contributing to the disease propagation. By using in vivo imaging by two-photon laser-scanning microscopy and axonal transection, it was observed different phases of microglia-mediated inflammation in the ALS mice model. Indeed, a first phase (preclinical) with highly reactive microglia was followed by another (clinical stage) with morphologically transformed microglia presenting reduced surveillance activity and reactivity (Dibaj et al., 2011). Regional, temporal, and immune environmental differences may contribute to changes in microglia phenotypes and response heterogeneity, thus requiring differentiated immunomodulatory or even combinatory therapeutic approaches along ALS disease progression.

Microglia phenotypes. Activation of microglia may be observed through the up-regulation of CD11b, Iba1, and CD68 markers. Primary microglia differ from other blood macrophages in the expression of CD11b/CD45^low/high and CD68^low/high (for review, see Hinze and Stolzing, 2011), although not specific for microglia in a pathologic brain (Matsumoto et al., 2007). These cells show a round morphology with an enlarged cell body and smaller and thicker processes in the resting/quiescent state (M0; Figure 2). When activated in response to an insult or injury, microglia are capable of acquiring diverse and complex phenotypes, allowing them to participate in the cytotoxic response, immune regulation, and injury resolution. Microglia may then favor the entrance of inflammatory T cells with which the cells seem to interact. Recent work have classified and characterized M1 and M2 phenotypes (Table 2) (for review, see Evans et al., 2013). The first is cytotoxic, characterized by the release of proinflammatory cytokines and influenced by T helper cell type 1 (Th1) that release GM-CSF and interferon-γ (IFN-γ), triggering M1 proliferation. Cytotoxic M1 markers include IL-1β, IL-6, TNF-α, iNOS, COX2, and CX3CL1, and known inducers are the TLR4 agonist LPS and IFN-γ (Chhor et al., 2013).

The M2, promoted by the cytokines IL-4 and IL-13 released by Th2, contribute to neuroprotection once they also secrete anti-inflammatory cytokines, such as IL-4 and IL-10, and growth factors such as insulin-like growth factor 1 (IGF-1). Indeed, IL-4 was shown to protect MNs from the injury produced by LPS-activated microglia (Zhao et al., 2006). Thus, M2 polarization may be desirable, although excessive or prolonged M2 polarization may become prejudicial in allowing unwanted fibrotic responses and scarring, not facilitating axonal growth. However, there are three M2 phenotypes: the M2a or alternate activation repair/regeneration/remodeling phenotype, the M2b immunoregulatory and the M2c acquired-deactivating (Figure 2). Reliable markers for M2a stimulated by IL-4 and IL-13 are high IL-1 receptor antagonist (IL-1Ra) and high arginase (Arg1). M2b is stimulated by immune complexes, TLR agonists and IL-1R ligands. Useful markers are IL-1Ra and SOCS3. Characterization of M2c stimulated by IL-10, transforming growth factor-β (TGF-β) and glucocorticoids is obtained through the increased levels of anti-inflammatory cytokines (IL-10, TGF-β), low levels of pro-inflammatory cytokines and enhanced IL-4Rα, Arg1, SOCS3, and CD206 (David and Kroner, 2011; Wilcock, 2012; Chhor et al., 2013). While LPS, IL-1β, TNF-α, and IFN-γ lead to cytotoxic M1 and immunomodulatory M2b activation states, IL-4 mainly triggers M2a phenotype (Chhor et al., 2013).

Therefore, balance between M1 and M2 phenotypes may be a desirable therapeutic goal. Interestingly, mSOD1 microglia isolated from ALS mice at disease onset showed higher levels of Ym1, CD163, and brain-derived neurotrophic factor (BDNF) (M2 markers) and lower levels of Nox2 mRNA (M1 marker) as compared to the end-stage disease (Liao et al., 2012). Interestingly, when co-cultured with wt MNs the first microglia phenotype exerted neuroprotection while the M1 phenotype was neurotoxic, supporting the pathoprogression-related changes in microglia. We may probably consider that microglia display the M2 phenotype at an early stage of the disease switching to the M1 phenotype during the late rapid phase (reviewed in Zhao et al., 2013). However, we should also consider that microglia may become functionally impaired at the end-stage, as recently observed in the SOD1G93A SC microglia (Nikodemova et al., 2013). Different roles of microglia depending on the neurological disease context was recently suggested by Chiu et al. (2013) based on the observation by FACS-transcriptome comparisons that SOD1G93A microglia show a unique phenotype that differs from M1 or M2 macrophages, and from activation with LPS. Robust up-regulation of MMP-12, IGF-1 and osteopontin were pointed as hallmarks. Transcriptomic technology has been carried out to examine the gene expression changes of ALS tissue as compared to controls (reviewed in Heath et al., 2013) and in the future may provide insights into the microglia phenotype profiling prior to disease onset and along ALS progression to assist in the development of new treatments for ALS at different stages.

Inflammatory microRNA profiling. MicroRNAs (miRNAs), small, non-coding RNAs, have been recently pointed to mediate cell-to-cell communication (Xu et al., 2013). Inflammatory phenotypes are miR-155, miR-21, miR-146a/b, and miR-124 (Quinn and O’Neill, 2011). Recent studies demonstrated the miR-124 involvement in promoting microglia quiescence by skewing their polarization from an M1 to an M2 phenotype (Willemen et al., 2012) and that miR-155 together with miR-124 are likely to be directly related to M1 and M2 phenotypes, respectively (Cardoso et al., 2012; Ponomarev et al., 2013). Up-regulation of miR-155 is induced by TNF-α and Il-1β (Pottier et al., 2009), as well as by HMGB1 through the activation of the TLR2/MyD88/miR-155 pathway (Wen et al., 2013), while it is down-regulated by TGF-β (Pottier et al., 2009). Increase of miR-146a expression was shown to occur in the aged mice (Rodier and Campisi, 2011), as well as after LPS stimulation (Jiang et al., 2012), and is related with the microglia phagocytic potential (Saba et al., 2012). Moreover, up-regulation of miR-21 in murine models of SC injury was pointed as a modulator of the pro-reactive effects of inflammatory signaling cascades (Nieto-Diaz et al., 2014). Dysregulation of miRNAs expression was lately found in the SC of ALS patients and in microglia isolated from the hSOD1G93A mice. Promising candidates that were found to be altered in patients were miR-146a^*, miR-524-5p, and miR-582-3p (Campos-Melo et al., 2013), while those up-regulated in the mice model were miR-155, miR-146b, miR-22, miR-365, and miR-125b (Parisi et al., 2013), which need to be further investigated for their relevance as ALS biomarkers and therapeutic targets. Actually, it was recently discovered that inhibition of miR-155 prolongs survival in the mSOD1 mice (Koval et al., 2013).

Autotaxin. Autotaxin (ATX) is a secreted lysophospholipase D that converts lysophosphatidylcholine (LPC) to lysophosphatidic acid (LPA), a phospholipid growth factor that activates several transduction pathways, and is involved in migration, proliferation, and survival of various cells (Mori et al., 2011; Furukawa et al., 2013). LPA acts as an autocrine and/or paracrine signaling molecule mediating a broad range of intracellular signaling cascades, especially the RHOA pathway (reviewed in Willier et al., 2011). LPA receptors are expressed by astrocytes, oligodendrocytes, and microglia (Lorenzl et al., 2006; Shi et al., 2010a; He et al., 2013).

Data suggest that ATX levels increase within reactive astrocytes following neurotrauma (Savaskan et al., 2007), but other studies have obtained a down-regulation instead (Shi et al., 2010b). Most fascinating, it was observed that overexpression of ATX inhibits microglial activation and protects the cell against oxidative stress (Awada et al., 2012). Thus, reduced expression of ATX in microglia may contribute to a sustained neuroinflammation condition. Since ATX is also a motility driver that stimulates cell invasion (Sau et al., 2007; Hoelzinger et al., 2008; Magrane and Manfredi, 2009) it can be involved in microglia migration, a property not exploited till now. Although ATX is considered an inflammatory mediator we were not able to find studies demonstrating its relevance or not in ALS.

The neuroprotective/neurotoxic role of CX3CL1/CX3CR1 signaling is still a matter of debate once it seems to depend upon the CNS insult (for review, see Desforges et al., 2012). Deletion of CX3CR1 in a transgenic model of ALS mice was shown to extend neuronal cell loss, suggesting that CX3CL1/CX3CR1 signaling limits microglial toxicity in ALS (Cardona et al., 2006). Interestingly, it was also shown that treatment with LPS down-regulated the expression of CX3CR1, thus suppressing the functional response to FKN (Boddeke et al., 1999) and potentiating LPS neurotoxic effects (Zujovic et al., 2000). Accordingly, in ALS, the disruption of CX3CL1/CX3CR1 signaling evidenced to promote neurodegeneration following LPS administration (Cardona et al., 2006).

Although it has been proposed that cathepsins compensate for each other because of their overlapping substrate specificities, there is increasing evidence that disturbance of the normal balance and extralysosomal localization of cathepsins contribute to age-related diseases (for review, see Nakanishi, 2003). CatS that is a lysosomal/endosomal cysteine protease that degrades extracellular matrix proteins, even at neutral pH, was shown to be increasingly expressed by microglia upon LPS surcharge (Petanceska et al., 1996). However, some authors indicate that a further challenge with ATP is required to observe CatS release (Clark and Malcangio, 2012). This protease is preferentially expressed by antigen-presenting cells, including microglia.

Microglial CatS seems to be responsible for the liberation of neuronal FKN (Figure 1), which via the specific receptor CX3CR1 in microglia activate p38 MAPK pathway leading to the release of mediators that interact with neurons (Clark et al., 2007). Several inhibitors have been tried to regulate the immune modulatory effects of CatS (Clark and Malcangio, 2012) and a shift from a Th1/Th17 type response to a Th2 type of response was obtained (Baugh et al., 2011).

The up-regulation of cathepsins appears as a general transcriptional event in ALS (Gonzalez de Aguilar et al., 2008; Offen et al., 2009; Boutahar et al., 2011). However, reference to CatS and microglia activation was not until now indicated in the context of ALS. Therefore, if CX3CL1/CX3CR1 demonstrates to have a role in any stage of ALS progression, inhibition of CatS may constitute a therapeutic approach for ALS.

Exosomes are secretory vesicles deriving from late endosomes and multivesicular bodies that mediate neuron–glia communication, have 50–100 nm in size and carry specific protein and RNA cargo (Fitzner et al., 2011; Fruhbeis et al., 2012; El Andaloussi et al., 2013). Exosomes contain both miRNAs and mRNAs that may be delivered and be functional in another cell (Valadi et al., 2007). Selectivity for miRNA incorporation into exosomes is proposed based on the fact that some are exclusive in exosomes derived from immature dendritic cells while others only exist in those from mature dendritic cells (Stoorvogel, 2012). Indeed, exosomes and miRNAs have been found to participate in cellular senescence and contribute to aging (Xu and Tahara, 2013). In addition senescent cells produce high levels of exosomes thus interacting and inducing the senescence of neighboring cells. As so, exosomal miRNAs might become useful biomarkers of disease. Recently, it was suggested that exosomes may be important candidates to deliver siRNA (El Andaloussi et al., 2013) and specific drugs, based on the perceived advantages of nanoparticle size and non-cytotoxicity (Sun et al., 2013), thus constituting a therapeutic platform.

Microglia was shown to internalize oligodendroglial exosomes, thus participating in the degradation of oligodendroglial membrane (Fitzner et al., 2011). However, microglia also release exosomes (Potolicchio et al., 2005; Hooper et al., 2012), which generate IL-1β to the extracellular environment propagating inflammation (Turola et al., 2012). Increased exosome discharge by microglia was observed after stimulation with α-synuclein and such activated exosomes revealed an increased membrane content in TNF-α (Chang et al., 2013). Interestingly, it was shown that mouse MN-like NSC-34 cells overexpressing hSOD1G93A secrete SOD1 via exosomes probably accounting to cell–cell-mediated mutant toxicity in ALS pathogenesis (Gomes et al., 2007). Therefore, exosomes from microglia may spread pathogenic factors, such as SOD1 and promote inflammation while influencing neuronal survival (Chang et al., 2013). In line with this, a latest study reports that cell-to-cell transmission of SOD1 misfolding is mediated by two non-exclusive mechanisms: through the release of protein aggregates that are taken up by macropinocytosis or via exosomes secreted from living cells (Grad et al., 2014).

Recent data obtained in the mSOD1 mouse model suggest a dominant neuroinflammatory response in the CNS, with a reactive microglia in preclinical stages that turns into an irresponsive cell during disease progression, and a degenerative process in the PNS (Dibaj et al., 2011). Earlier studies evidenced that mutated SOD1 microglia have an age-dependent cytotoxic potential which reveals upon a stimulatory effect (Weydt et al., 2004). Interestingly, extracellular SOD1G93A mediates the activation of CD14–TLR2 pathway with the consequent release of TNF-α and IL-1β, thus propagating the proinflammatory stimuli (Bowerman et al., 2013). However, since immunosuppressive strategies have not proven consistent efficacy (Appel et al., 1993, 2011; Baugh et al., 2011), one may believe that microglia function may change along disease progression with consequent differential effects on MNs. Late data evidenced that SC microglia proliferate and that TNF-α mRNA expression decreases during disease progression in SOD1G93A rats (Nikodemova et al., 2013). Indeed, microglia revealed to not be polarized to M1 or M2 phenotypes at any disease stage and CNS region evaluated. In addition, SC microglia evidenced to be irresponsive to experimental systemic inflammation at ALS end-stage. Diminished glial neuroprotection by senescent and/or dysfunctional microglia has been suggested to play a role in neurodegenerative diseases, mainly in the late stage (Streit, 2006). In such circumstances the aged microglia evidenced a “dystrophic” morphology with the loss of finely branched cytoplasmic processes, cytoplasmic beading/spheroid formation, and cytoplasmic fragmentation (cytorrhexis; Figure 2) (Streit et al., 2004). Such severe abnormalities in microglia, including cell fusion (multinucleated giant cells) at the symptomatic stage and cytorrhexis at the end stage are indicative of microglial aberrant activation and degeneration, respectively, and were observed in SOD1G93A transgenic rats (Fendrick et al., 2007). Although prevalent in older human subjects this dystrophic cells may also in rare instances be observed in the young brain (Luo and Chen, 2012). Pathogenic miRNAs, such as miR-155, miR-146a, and miR-124 in microglia may be associated with the acquisition of a senescent phenotype (Saba et al., 2012; Ponomarev et al., 2013). We recently observed that aged-cultured microglia exhibit lower phagocytic ability and higher expression levels of miR-146a than younger cells (Caldeira et al., 2013). Whether these findings are related with a dysfunctional microglia at ALS end-stages deserve to be further investigated.

Although new oligodendrocytes were observed in the SC of SOD1G93A mice, they do not fully maturate, resulting in progressive demyelination and accelerated disease (Liu et al., 2002). The authors also found similar myelination defects in postmortem samples taken from SC and motor cortex from ALS patients. Actually, there are many potent inhibitors of axonal regeneration in the injured CNS including myelin-associated proteins, fibrinogen, and axonal guidance molecules, where epidermal growth factor receptor (EGFR) and eukaryotic ribosome biogenesis protein 1 (Erb1) may have a special role in reducing the effects of multiple inhibitors of axonal regeneration (Arthur-Farraj et al., 2012; Leinster et al., 2013). However, the role of EGFR on the protection of MN synapses and survival extension in SOD1G93A mice is still a matter of debate; hence, it was reported that EGFR inhibitors failed to extend ALS mouse survival although influencing disease progression (Le Pichon et al., 2013).

Another important point to be considered is the decrease in the phagocytic clearance by the dysfunctional microglia that results in the accumulation of myelin debris, leading to oligodendrocyte differentiation arrest and decreased recruitment of oligodendrocyte precursor cells (Walter and Neumann, 2009). Activated microglia were also shown to attenuate the proliferation of the oligodendrocyte precursor cells, thus concurring for demyelination progression (Taylor et al., 2010), and reinforcing the recovery of healthy microglia as a potential therapeutic target in ALS. Indeed, studies on the global gene expression during demyelination and remyelination by microarray analysis reinforced that the primary function of microglia is the tolerance induction and support to regeneration (Olah et al., 2012).