Signaling pathways regulating neuron-glia interaction and their implications in Alzheimer’s disease

Abstract

Astrocytes are the most abundant cells in the central nervous system. They play critical roles in neuronal homeostasis through their physical properties and neuron-glia signaling pathways. Astrocytes become reactive in response to neuronal injury and this process, referred to as reactive astrogliosis, is a common feature accompanying neurodegenerative conditions, particularly Alzheimer’s disease. Reactive astrogliosis represents a continuum of pathobiological processes and is associated with morphological, functional and gene expression changes of varying degrees. There has been a substantial growth of knowledge regarding the signaling pathways regulating glial biology and pathophysiology in recent years. Here we attempt to provide an unbiased review of some of the well-known players, namely calcium, proteoglycan, TGFβ, NFκB, and complement, in mediating neuron-glia interaction under physiological conditions as well as in Alzheimer’s disease.

Introduction

From invertebrate animals to mammals, the central nervous system (CNS) exhibits tremendous cell type diversity. Among these, cells capable of conducting rapid electrical signals are designated as neurons, while all other cell types are collectively categorized as glia. The classical view of neuroscience considers neurons to be the central players with glia playing a passive and supportive role. However, there has been a growing interest in glia, and recent work has challenged this neurocentric view. It is now increasingly appreciated that glia play dynamic structural and signaling roles, both during development and in adult and aging brains, and impairment of these properties dictate many of the neurological conditions.

There are three major glial cell types in the CNS: microglia, astrocytes (also known as astroglia), and oligodendrocytes; the latter are highly specified cells that function to insulate axons via the formation of myelin sheath around axons. Microglia are considered the resident macrophages in the brain where they play prominent roles in immune surveillance, phagocytosis and neuroinflammatory processes. Astroglia are the most abundant cell type in the CNS which exert diverse physiological functions through their close association and communication with neurons and other brain structures. Like microglia, astrocytes possess immune and inflammatory properties when activated. This process, referred to as reactive astrogliosis, is associated with morphology, gene expression and functional changes, and is a common feature accompanying many neuronal injury and neurodegenerative conditions, particularly Alzheimer’s disease (AD).

Alzheimer’s disease is the most common form of dementia in the aged population. The pathological hallmarks of AD include extracellular amyloid plaque deposition, intraneuronal neurofibrillary tangle accumulation, synaptic dysfunction and neurodegeneration (Stancu et al. 2014). In addition, the diseased brains are associated with profound neuroinflammation characterized by reactive astrogliosis, microgliosis and increased release of inflammatory cytokines (Heneka et al. 2015, Hensley 2010, Orsini et al. 2014, Lemere 2013). Mutations in APP and presenilins (PSEN) cause a subset of early onset familial AD, establishing their central roles in disease pathogenesis (Guo et al. 2012, Wang et al. 2014b). The identification of multiple genetic risk factors with immune and inflammatory properties, such as TREM2, CR1, CLU (ApoJ), CD33, ABCA7, etc., strongly implicates innate immunity plays a contributing role in AD (Heneka et al. 2015, Guerreiro et al. 2013). Both positive and negative effects of neuroinflammation have been proposed, and these differing outcomes are likely attributed by differences in the timing, duration, and/or strength of the responses, triggered by a wide range of upstream stimulators and the many downstream effectors, either unique to the specific cells in the CNS or through the crosstalk among these cell types (Lucin & Wyss-Coray 2009, Garden & La Spada 2012).

The role of microglia in neuronal health and disease has been extensively studied and discussed (Wyss-Coray & Rogers 2012, Landel et al. 2014, Ransohoff & Brown 2012, Ulrich et al. 2014). The primary objective of this review is to provide an overview of the signaling pathways pertinent to astrocytes and reactive astrocytes, specifically calcium, proteoglycans, TGF-β, NFκB and complement, and describe their impact on neuronal function in physiological conditions as well as in Alzheimer’s disease. Understandably, it is not possible to cover all the potential pathways in a single review. Other signaling molecules that exhibit widespread functions across cell types, such as lipids (Walter & van Echten-Deckert 2013, Wang et al. 2015), oxidative species (Moncada & Bolanos 2006, Landel et al. 2014, Kukreja et al. 2014) and metabolic intermediates (Hertz 2004, Lovatt et al. 2007, Hoos et al. 2014, Kang et al. 2014), will not be discussed in detail but only mentioned as necessary. While we attempt to focus this review on astrocytes, microglia and neurons express many of the common factors and regulatory pathways that act as integral components of the neuron-glia signaling network. Indeed it would be difficult to pinpoint an astrocyte “only” effect, particularly under pathological conditions. Accordingly, we will often discuss the effects of these pathways in this border context.

Astrocytes and reactive astrocytes

Astroglia are the most abundant cell type in the CNS that exert diverse physiological functions through their close association and communication with neuron and with other brain structures. Astrocytes extend many fine branching processes, putting them in direct contact with neuronal cell bodies, dendrites and synaptic terminals. This physical proximity allows astrocytes to sense and respond to neuronal activities. Astroglia express and secrete many signaling molecules that mediate synapse formation and synaptic transmission (Barres 2008, Allen & Barres 2009). Astrocytes also express neurotransmitter receptors, through which they potently regulate neurotransmitter recycling at synaptic sites through the formation of “tripartite” synapses consisting of astroglial projections and neuronal pre- and postsynaptic terminals (Halassa et al. 2007, Perea et al. 2009). Astrocytic end-feet are an essential constituent of the blood brain barrier. Through these widespread contact properties, astroglia have the ability to adjust blood flow for oxygen, ATP and glucose supplies, in addition to maintaining ionic concentrations in the extracellular matrix, to name but a few (Reviewed in: Garden & La Spada 2012, Sofroniew & Vinters 2010). Through the expression of aquaporin 4 water channels in the astrocytic vascular end-feet, astroglia play a vital role in the newly discovered glymphatic system, which is a brain drainage system implicated in the clearance of both Aβ and tau (Iliff et al. 2013, Iliff et al. 2012, Xie et al. 2013, Nedergaard 2013).

Astrocytes become reactive in response to various triggers, and this process is associated with morphological, molecular and functional changes. Reactive astrogliosis, marked by GFAP (glial fibrillary acidic protein) immunoreactivity, is a common feature associated with both acute brain injury and chronic neurological conditions. While it is generally believed that reactive gliosis contributes to disease pathogenesis, it is often overlooked that this process covers a continuum of changes with varying degrees, ranging from subtle and reversible alterations of physiological processes in mild forms to long-lasting scar formation in the most severe cases (Sofroniew & Vinters 2010, Anderson et al. 2014, Verkhratsky et al. 2014). Thus, the signaling pathways described below and their effects on neuronal function and dysfunction are expected to vary depending on the state of the astrocytes and the degree of reactive gliosis, although this can be difficult to define precisely.

Calcium

Astroglial calcium plays pivotal roles in intracellular signaling and intercellular communication. The cytosolic free calcium concentration is critically controlled by extracellular influx and intracellular storage release. Intracellular calcium transients and intercellular calcium waves are the basic forms of astrocytes which, through intimate connections, communicate among themselves and with neurons (Scemes & Giaume 2006). In particular, the production of inositol 1,4,5-triphosphate (IP₃) leads to the activation of IP₃ receptors (IP₃R) and calcium release from internal storages, especially the endoplasmic reticulum, and the rise of calcium transients promotes astroglial release of stimulating substances such as purines (ATP and adenosine), GABA, D-serine, and glutamate, collectively termed as gliotransmitters. The calcium-mediated release of glutamate from astrocytes promotes neurotransmission at the tripartite synaptic sites. The spread of ATP and glutamate can regenerate calcium mobilization in adjacent cells through the activation of purinergic receptors or metabotropic glutamate receptors, which in turn induce IP₃ production inside the cells. As such, calcium transients inside one cell can initiate calcium changes in surrounding cells through the conversion of intracellular calcium transients to intercellular calcium waves, allowing astrocytes to execute both synchronous as well as differential responses through limited messenger diffusion (Reviewed in: Araque et al. 2014, Rusakov 2015).

Substantial evidence supports a significant role of calcium dysregulation in AD pathogenesis (LaFerla 2002, Peterson et al. 1985, Garwood et al. 2013, Bezprozvanny & Hiesinger 2013). Both APP and presenilins were found to regulate calcium homeostasis and calcium-dependent synaptic function (Yang et al. 2007, Yang et al. 2009, Hamid et al. 2007, Leissring et al. 2002, Stutzmann et al. 2006, Tu et al. 2006, Johnston et al. 2006, Lee et al. 2002). Neurotoxicity induced by protein aggregates, such as Aβ, may be conferred by their direct disturbance of calcium signaling in neurons (Kuchibhotla et al. 2008, Choo et al. 2004, Tang et al. 2003, Del Prete et al. 2014, Lopez et al. 2008, Mattson 2007, Zundorf & Reiser 2011, Demuro & Parker 2013, Armato et al. 2012, Ye et al. 1997, Chiarini et al. 2009) or indirectly through affecting calcium homeostasis in glial cells (Alberdi et al. 2013, Chow et al. 2010, Orellana et al. 2011). In mouse models of AD, calcium channels including ATP-gated purinergic receptor cation channels and voltage-gated calcium channels were overexpressed in plaque-surrounding reactive astrocytes and microglia (Willis et al. 2010, Parvathenani et al. 2003). These astrocytes were shown to exhibit escalated basal calcium transients and intercellular calcium waves (Delekate et al. 2014, Kuchibhotla et al. 2009, Riera et al. 2011, Lim et al. 2014), which may trigger the release of toxic factors or dysregulation of neuronal calcium-regulated receptors (Abramov et al. 2003, Hashioka et al. 2012, Kaushal et al. 2007, Orellana et al. 2011). In particular, dysregulation of astroglial calcium has been implicated in extrasynaptic glutamate receptor activation and associated glutamate excitotoxicity (Rudy et al. 2015, Paula-Lima et al. 2013). It may also produce reactive oxidative species leading to neuronal oxidative stress (Ye et al. 2015, Pirttimaki et al. 2013, Abramov et al. 2004, Zhu et al. 2006)(Zhao & Zhao 2013, Axelsen et al. 2011, Pohanka 2014). Consistent with these assessments, calcium channel blockers have been reported to inhibit the neurotoxicity of activated glia (Hashioka et al. 2012, Maezawa et al. 2011). Interestingly, astrocytes are the major cells for antioxidant glutathione (GSH) synthesis where it exhibits distinct responses to different Aβ species (Ye et al. 2015). Whereas monomeric Aβ increases GSH release, aggregated Aβ leads to reduced GSH, consistent with the in vivo observation that GSH levels are higher in pre-plaque AD animals, but are reduced when plaque pathology manifests (Ye et al. 2015).

Overall, overwhelming evidence supports a prominent role of astrocytic calcium in neuronal homeostasis and AD pathogenesis, which are mediated through diverse mechanisms. Although an Aβ triggered astrocytic calcium response has been well documented (Vincent et al. 2010, Mattson & Chan 2003, Bezprozvanny 2009), the astroglial receptors mediating this response has been elusive. Some speculated that the calcium-permeable α7 nicotinic acetylcholine receptors (α7nAChRs) expressed in astrocytes may be the receptor (Pirttimaki et al. 2013, Lee et al. 2014), others implicated the metabotropic glutamate type-5 receptor (mGluR5) as a candidate, which may become clustered on astrocytic membrane due to Aβ –induced biophysical changes (Shrivastava et al. 2013, Hicks et al. 2008, Yang et al. 2010). Further investigation of these receptors will provide better insights into the calcium-dependent neuron-astroglia crosstalk in AD.

Proteoglycans

Proteoglycans are glycosylated proteins composed of a protein core and glycosaminoglycan (GAG) chains (Iozzo & Schaefer 2015). The proteoglycans are mainly detected on cell surfaces and in the extracellular matrix (ECM). The cell surface proteoglycans can bi-directionally regulate signaling through interaction with both ligand and membrane receptor proteins (Clark 2008). Some cell surface proteoglycans themselves are receptors capable of conducting intracellular signaling upon activation (Christianson & Belting 2014). Secreted proteoglycans are major components of the ECM where they maintain ECM homeostasis through binding among themselves and with other molecules, such as collagen (Yanagishita 1993, Kim et al. 2011).

Chondroitin sulfate proteoglycans (CSPGs) and heparin sulfate proteoglycans (HSPGs) are the two main types of proteoglycans in the brain, which are produced in both neurons and glia (Bandtlow & Zimmermann 2000). The functions of neuronal cell surface proteoglycans have been covered in previous reviews (Cui et al. 2013, Schaefer & Schaefer 2010, Maeda 2015). Here we discuss how glial secreted proteoglycans regulate neuronal processes and their possible involvement in AD pathogenesis.

Astrocyte-derived proteoglycans are essential components of the brain ECM that facilitate neurite outgrowth, axon guidance and synaptogenesis, among others (Farhy Tselnicker et al. 2014). Certain HSPGs, such as syndecan 4, glypican 6 and perlecan, are specifically expressed in astrocytes (Giamanco & Matthews 2012, Faissner et al. 2010, Yamada et al. 1997, Allen et al. 2012). Secreted glypican 4 and 6 from astrocytes are required for synaptogenesis and AMPAR recruitment, although whether the glypicans function through specific neuronal receptors or through modulation of other growth receptors remains to be determined (Allen et al. 2012). Elimination of CSPG brevican and neurocan induced astrocyte-dependent, dual-phased effects on synaptic puncta numbers displaying initial promoting, but later suppressing, activities (Pyka et al. 2011, Geissler et al. 2013), which implicates their possible contrasting roles in synaptogenesis and synapse maintenance. Results from axon regeneration studies suggest that CSPGs produced from reactive astrocytes are the main source of glial scars formed after injury and are inhibitory for axonal growth and myelination (Pendleton et al. 2013, Deng et al. 2015, Bradbury et al. 2002, Cafferty et al. 2007, Pizzorusso et al. 2002, Sharma et al. 2012). It was suspected that this may be mediated via specific CSPG receptors or through non-specific repulsion between CSPGs and axons due to charge differences (Sharma et al. 2012). However, a recent study on neural-glial 2 (NG2) proteoglycan revealed that proteoglycans have high affinity for axons and thus, instead of repulsion, the inhibitory effects may be caused by the trapping of axons in the scar area (Filous et al. 2014).

Proteoglycans may contribute to AD pathophysiology through multiple pathways. In neurons it is known that APP is processed during sorting and trafficking (Wang et al. 2014b, Del Prete et al. 2014). APP splicing variants lacking exon 15 are modified by GAG chains and are rapidly transported to the cell surface through indirect endosomal sorting (Mihov et al. 2015). This GAG-directed redistribution of APP could result in differential APP processing and Aβ generation (Cui et al. 2012, von Einem et al. 2015, Zheng & Koo 2011, Vassar et al. 1999). Proteoglycans have been shown to interact with APP (Narindrasorasak et al. 1991, Buée et al. 1993), and disruption of the interaction may inhibit its physiological function, such as soluble APP (sAPP)-induced neurite outgrowth (Small et al. 1994, Clarris et al. 1994). Proteoglycans are known to associate with both amyloid plaques and neurofibrillary tangles, and their expressions are dysregulated in AD (Snow et al. 1988, DeWitt et al. 1993, Perry et al. 1991, Leveugle & Fillit 1994, Bellucci 2007). These may be caused by a direct induction of astrocytic proteoglycan overexpression by Aβ (Canning et al. 1993), or as a consequence of astroglial reactivation (Karimi-Abdolrezaee & Billakanti 2012, Properzi et al. 2003, Dow & Wang 1998). Aberrantly expressed proteoglycans may promote Aβ aggregation by facilitating its structural conversion from non-amyloidogenic random coil to amyloidogenic β-sheet through hydrophobic and ionic interactions between Aβ and proteoglycans (Snow et al. 1994b, Geneste et al. 2014). It may also attenuate Aβ clearance by inhibiting Aβ proteolysis due to the steric shielding of Aβ peptide by naturally protease-resistant GAG chains (Gupta-Bansal et al. 1995) and by altering the ApoE-HSPG complex formation and compromising the ability of ApoE to transport Aβ (Libeu et al. 2001, O’Callaghan et al. 2014, Kanekiyo & Bu 2009). Overall the studies combined support a model whereby neuronal secreted Aβ and astroglial produced proteoglycans form a feed forward loop to promote amyloid pathology through increased Aβ production and aggregation and reduced Aβ clearance. Consistent with this model, degradation of proteoglycans through enzymatic digestion of the GAG chains were shown to alleviate disease pathology in AD animal models (Jendresen et al. 2015). However, in light of the essential roles of proteoglycans in the CNS, further studies are needed to assess the therapeutic potential of targeting this pathway.

TGF-β

Transforming growth factor β (TGF-β) is a family of pleiotropic cytokines consisting of three members, TGF-β1, TGF-β2, and TGF-β3. They are secreted as a latent complex composed of a TGF-β dimer and a pro-peptide called latency associated peptide (LAP), which in most cases is associated with latent TGF-β binding protein (LTBP) to form a large latent complex in the extracellular matrix. Recognition of LTBP by cell-surface proteins and subsequent cleavage of LAP generates active and mature TGF-β. This process is tightly regulated to ensure proper activation of TGF-β signaling. On the cell surface, TGF-β type II receptor (TβR-II) binds to the activated TGF-β dimer and recruits type I receptor (TβR-I) to form a four-receptor complex (Figure 1). The TβR-II kinase domain, activated upon ligand binding, phosphorylates TβR-I which subsequently phosphorylates the transcriptional factor complex, the Smad2/3 complex. Phosphorylated Smad2/3 then binds to the common Smad4, translocates to nucleus and activates target gene expression. TβR-I also activates kinases, such as PI3K and MAPK, to trigger downstream pathways in a Smad independent manner (Zhang 2009) (Figure 1). The diverse outcomes of TGF-β activation are possibly determined by differences in the stimulators, the extracellular factors that TGF-β interacts with which in turn modifies its receptor binding affinities and the recruitment of different Smad proteins as well as other transcription factors.

Figure 1.

Schematic diagram of TGF-β signaling. LTBP: latent TGF-β binding protein; LAP: latency associated peptide; TβR-I and TβR-II: TGF-β receptor type I and II respectively.

In the CNS, neuron, glia and brain vasculature all express and respond to TGF-β (Flanders et al. 1991, Unsicker et al. 1991, Flanders et al. 1998). During embryonic development, the temporal and spatial expression of TGF-β tightly regulates neuronal survival, neurogenesis, synaptogenesis and gliogenesis (Bottner et al. 2000, Gomes et al. 2005) (Stipursky et al. 2014, Stipursky et al. 2012). These effects may depend on secreted astrocytic TGFβ and through both canonical and non-canonical signaling pathways (Yu et al. 2014, Diniz et al. 2014, Diniz et al. 2012). In particular, TGF-β1 deficiency causes neuronal death and microgliosis; these phenotypes can be reversed by astrocytic TGF-β1 expression (Brionne et al. 2003). Furthermore, astroglial expression of TGF-β mediates synaptic refinement during visual system maturation and, interestingly, this process was shown to be dependent on neuronal C1q (see “Complement” section) (Bialas & Stevens 2013). Finally, Schachtrup et al., reported that astrocytic TGF-β signaling promotes glial scar formation (Schachtrup et al. 2010). Interestingly, recent studies by the same group revealed that this pathway is regulated by γ-secretase cleavage of p75 neurotrophin receptor, raising the intriguing possibility that astroglial presenilin/γ-secretase activity may play a role in TGF-β-dependent neuron-glia communication (Schachtrup et al. 2015).

In healthy adult CNS, TGF-β2 and 3 are stably expressed. In contrast, TGF-β1 is absent or expressed at very low levels (Flanders et al. 1991, Koefer et al. 1995). Abnormal TGF-β1 hyperactivation has been detected in pathological conditions such as AD, Parkinson’s disease, amyotrophic lateral sclerosis and brain injury; astrocytes and microglia are the main source of TGF-β1 production (Phatnani et al. 2013, Endo et al. 2015, Finch et al. 1993, Morgan et al. 1993). Specific to AD, microglia and astrocytes secrete TGF-β upon Aβ stimulation (Tu et al. 2015, Tichauer & von Bernhardi 2012), consistent with increased levels of TGF-β1 and TGF-β2 in postmortem brain samples of AD (Grammas & Ovase 2002, Tarkowski et al. 2002, Malaguarnera et al. 2006, Swardfager et al. 2010, Flanders et al. 1995, Noguchi et al. 2010) and age-dependent TGF-β1 upregulation in AD mouse models (Wirths et al. 2010, Salins et al. 2008). Secreted TGF-β may directly engage neuronal response such as NFκB activation to promote cell survival as supplementing TGF-β protects neurons from Aβ toxicity (Ma et al. 2012, Chacon & Rodriguez-Tebar 2012), and this protective activity was shown to be antagonized by Aβ (Huang et al. 1998). Besides Aβ, neurofibrillary tangles may also interfere with neuronal response to TGF-β by sequestrating Smad proteins in the cytoplasm, eventually resulting in the death of tangle-bearing neurons (Chalmers & Love 2007). The expression of TGF-β type II receptor (TβR-II), which is mainly expressed by neurons, is reduced in AD brains (Tesseur et al. 2006). Supporting a beneficial role of TGF-β and TGF-β receptor signaling pathway, overexpression of a dominant negative TβR-II mutant promotes Aβ accumulation, dendritic loss and neurodegeneration in an AD mouse model (Tesseur et al. 2006). Conversely, supplementation of TGF-β1 rescued neurodegeneration in rats injected with Aβ42 (Chen et al. 2015). Besides a direct interaction with neurons, secreted TGF-β may also modulate neuronal physiology by participating microglial-mediated inflammatory responses (Huang et al. 2010, Cekanaviciute et al. 2014, Norden et al. 2014, Makwana et al. 2007). For instance, TGF-β enhances microglial Aβ uptake via Smad3-dependent modification of scavenger receptor expression (Tichauer & von Bernhardi 2012, Wyss-Coray et al. 2001). This could be the underlying mechanism by which astrocytic TGF-β1 expression reduces plaque pathology in APP transgenic mice (Wyss-Coray et al. 2001).

Nevertheless, negative association between TGF-β signaling and AD pathogenesis has also been reported (Hayes et al. 2013, Medeiros et al. 2013). In particular, TGF-β1-mediated neuronal signaling has been shown to potentiate APP transcription and Aβ production (Lesne et al. 2003, Docagne et al. 2004, Lahiri et al. 2003). Additionally TGF-β could contribute to AD through interacting with the cerebrovascular system (Wyss-Coray et al. 2000). Cerebral amyloid angiopathy (CAA) is a common feature in AD (Li et al. 2014). Transgenic mice expressing TGF-β1 in astrocytes develop CAA, and this phenotype may be attributed by the overexpression and accumulation of basement membrane proteins such as perlecan to facilitate amyloid deposition (Wyss-Coray et al. 1997, Wyss-Coray et al. 2000, Snow et al. 1994a). Lastly, TGF-β binds to ApoE in an isoform-dependent manner and this may indirectly influence Aβ clearance (Tesseur et al. 2009). Contrasting to the protective effects described above and supporting a detrimental role of heightened TGF-β signaling in AD pathogenesis, pharmacological inhibition of brain TGF-β was associated with alleviated AD pathology in mouse models (Sachdeva & Chopra 2015, Li et al. 2015, Zhang et al. 2014, Liu et al. 2014). Interestingly, blocking TGF-β signaling in peripheral macrophages was also reported to be efficacious in mitigating AD pathology (Town et al. 2008). Taken together, TGF-β signaling could play rather complicated and context-dependent activities during AD progression. Targeting the TGF-β pathway for potential therapeutic intervention will require further mechanistic and functional understanding.

NFκB

The nuclear factor-kappa B (NFκB) is a family of homo- or heterodimeric transcription factors present in nearly all cell types. NFκB is a master regulator of innate immunity that plays critical controls in various cellular processes such as proliferation, differentiation, survival, and apoptosis. NFκB is normally expressed in an inactive state in the cytoplasm where it is bound by its inhibitor protein family IκB. NFκB activation involves IκB phosphorylation and proteasome-mediated degradation leading to NFκB release. Once released, NFκB translocates to the nucleus and promotes gene expression upon binding to consensus κB DNA sites in promoter regions of target genes (Gilmore 2006). NFκB can be activated by a remarkable number of inducers such as intercellular mediators, bacterial or viral pathogens, reactive oxygen intermediates, and physical damage (Pahl 1999), which in turn regulates the expression of a growing list of target genes including those involved in innate and adaptive immunity, inflammatory responses, growth factors, and many others (http://www.bu.edu/nf-kb/gene-resources/target-genes/). The capability for certain inflammatory mediators, such as TNFα, IL-1β and IL-6, to act both as NFκB stimuli and downstream targets ensures augmentation of the immune responses. On the contrary, the fact that the principal NFκB inhibitor protein IκBα is a bona fide NFκB target allows formation of an elegant auto-inhibitory feedback loop to effectively terminate the pathway (Shim et al. 2011, Lian et al. 2012b, Chiao et al. 1994, Peng et al. 2010, Lian et al. 2015). These combined afford robust activation and precise regulation of NFκB-mediated responses (Figure 2).

Figure 2.

Schematic representation of the canonical NFκB signaling pathway. IKK: IκB kinase; GLT-1: glutamate type I transporter.

In the adult rodent brain, the most common form of the NFκB/IκB complex is a p50/p65 heterodimer bound by IκBα (Bakalkin et al. 1993, Meberg et al. 1996, Suzuki et al. 1997). NFκB can be induced in the CNS by a wide range of signals including neurotransmitters, neuropeptides and cytokines; neuronal NFκB expression has been reported to mediate morphogenesis, synaptogenesis, and learning and memory (Gutierrez & Davies 2011, Kaltschmidt & Kaltschmidt 2009, Maqbool et al. 2013). However, several recent studies raised the concern that some of the proposed neuronal NFκB detected may be caused by the contaminating glia or the non-specific nature of the antibodies (Herkenham et al. 2011, Lian et al. 2012a, Listwak et al. 2013, Massa et al. 2006, Mao et al. 2009). Since astrocytes and microglia possess robust NFκB responses, the precise contribution of neuronal vs. glial NFκB in neuronal development and function warrants further examination using cell-type specific assays and tools.

Substantial evidence supports a prominent role of astroglial NFκB in both physiological and pathological conditions. For instance, astrocytic expression of glutamate type I transporter (GLT-1) is essential in glutamate uptake and recycling (Rothstein et al. 1996, Robinson 1998, Ullensvang et al. 1997, Lehre & Danbolt 1998). Failure of glutamate homeostasis leads to excitotoxicity and neuronal cell death (Petr et al. 2015). NFκB is a positive regulator of GLT-1 (Gupta & Prasad 2014, Karki et al. 2013, Ghosh et al. 2011), and activity-dependent glutamate uptake through GLT-1 requires astroglial NFκB activation (Ghosh et al. 2011). As an NFκB inducer and downstream target, glial TNFα has been shown to mediate homeostatic synaptic scaling, although NFκB per se has not been explicitly implicated in this process (Stellwagen & Malenka 2006).

NFκB activity needs to be stringently controlled in the CNS to ensure normal neuronal development and function (Guerrini et al. 1997, Boersma et al. 2011, Meffert et al. 2003, Sylvie 2006). Aberrant NFκB activation has been observed in various neurodegenerative conditions, including Alzheimer’s disease (Kaltschmidt et al. 1997, Town et al. 2005, Mori et al. 2010), Parkinson’s disease (Hunot et al. 1997), and Huntington’s disease (Hsiao et al. 2013). In AD brains, p65 immunoreactivity is found to be enriched in neurons and glia in the vicinity of amyloid plaques (Kaltschmidt et al. 1997), indicating a possible NFκB activation by Aβ. Indeed, in vitro experiments demonstrated that NFκB activity can be induced by Aβ in neurons, astrocytes, and microglia (Kaltschmidt et al. 1997, Bales et al. 1998, Akama et al. 1998, Heurtaux et al. 2010, Carrero et al. 2012). As an transcription factor, activation of NFκB in reactive astrocytes induces expression changes of a large set of target genes resulting in both morphological ramifications and astrocytic functional changes (Frakes et al. 2014, Jayakumar et al. 2014, Wu et al. 2012, Corneveaux et al. 2010, Vincent et al. 2012, Hsieh et al. 2013, Walker-Caulfield et al. 2015). The astrocytic NFκB targets fall into many categories among which the inflammatory factors are most well understood. Those targets include nitrite oxide, cytokines, chemokines, etc. (Brambilla et al. 2014, Kim et al. 2012, Wang et al. 2013, Hiscott et al. 1993, Sparacio et al. 1992). Compromised astrocyte physiology by NFκB activation is associated with elevated mitochondrial oxidative metabolism, which restricts the supply of pyruvate substrates to neurons (Jiang & Cadenas 2014). The increased production of inflammatory factors also influences neurons directly by inducing neuronal oxidative stress and apoptosis (Akama et al. 1998, Wang et al. 2014a, Akama & Van Eldik 2000, Ye et al. 2013, Friedlander et al. 1996, Downen et al. 1999). Several studies showed that NFκB in astrocytes may regulate the integrity of the blood-brain-barrier essential for neuronal environmental homeostasis and compromised in AD (Park et al. 2014, Coelho-Santos et al. 2015)(Marques et al. 2013). Accordingly, inhibition of NFκB activation in AD models has been shown to ameliorate AD pathology and improve neuronal survival and cognitive function (Xuan et al. 2014, Lai et al. 2014, Medeiros et al. 2013, Shi et al. 2013, He et al. 2011, Brambilla et al. 2014, Jayakumar et al. 2014, Frakes et al. 2014, Brambilla et al. 2005), supporting the notion that NFκB antagonists may be therapeutically beneficial. Nevertheless, like other immune and inflammatory regulators, opposite effects have also been reported (Dvoriantchikova et al. 2009, Bracchi-Ricard et al. 2008), and the contrasting effects are likely attributed by the timing, duration and degree of the responses which require further investigation using spatially and temporally regulable tools and manipulations.

Complement

The complement pathway represents an evolutionarily conserved host defense and immune surveillance system known for its ability to recognize non-self pathogens and dying cells and eliminate them through phagocytosis or cell lysis (Holers 2014, Ricklin & Lambris 2013). Full complement activation involves concerted actions of over 30 proteins that participate in three distinct pathways: classical, alternative and mannose-binding-lection (MBL) (Reviewed by: Veerhuis et al. 2011, Yanamadala & Friedlander 2010, Zipfel & Skerka 2009). While the three pathways differ in their triggers, all converge on the cleavage of the central complement protein C3 by its convertase to C3a and C3b, which enables the conversion of C5 to C5a and C5b. While the C3 and C5 derivatives can mediate downstream signaling or phagocytic pathways through binding to their respective receptors, full complement activation requires the recruitment of other complement factors, including C7, C8, and C9, by C5b to form the terminal pore-forming complex termed the membrane attack complex (MAC). MAC insertion into the cell membrane induces cell death.

The complement pathway was originally assumed to be excluded from immune-privileged CNS and active only when the blood-brain-barrier is compromised. In recent years, however, expression of a full spectrum of complement proteins in the CNS has been detected (Shen et al. 1997, Singhrao et al. 1999, Walker & McGeer 1992, Thomas et al. 2000), where they have been implicated in both signaling and innate immune functions (Stephan et al. 2012). In particular, the C3a-C3aR interaction has been shown to regulate the migration and differentiation of neural stem cells through which it may influence adult neurogenesis (Shinjyo et al. 2009). A groundbreaking work by Stevens and colleagues revealed that astrocyte induced neuronal expression of proteins in the classical complement pathway, particularly C1q and C3, is required for proper elimination of unnecessary synapses during CNS development, and this is possibly regulated by neuronal activity and CR3-dependent microglial phagocytosis (Schafer et al. 2012, Stevens et al. 2007).

The complement pathway executes primarily inflammatory functions in the CNS. Astrocytes and microglia both express a large variety of complement factors as well as complement receptors (Bénard et al. 2008, Pisalyaput & Tenner 2008, Sayah et al. 2003, Veerhuis et al. 2011). Upon activation of complement receptors, glial cells change their profile to secrete proinflammatory cytokines, oxidative products, and also alter their phagocytosis ability (Farber et al. 2009, Fu et al. 2012, Sayah et al. 1999, Davoust et al. 1999, Griffiths et al. 2010, Sayah et al. 2003). In addition, cell lysis through the MAC has been proposed to mediate neuronal cell death associated with AD and other neurodegenerative conditions (Itagaki et al. 1994, Shen et al. 1998, Xiong et al. 2003).

Elevated complement expression at the RNA and protein levels have been consistently observed in brains of AD patients and mouse models (Fischer et al. 1995, Stoltzner et al. 2000, Walker & McGeer 1992, Zhou et al. 2008, Loeffler et al. 2008). The identification of clusterin (CLU or ApoJ), a potent regulator of complement activation, and complement receptor 1 (CR1) as genetic risk factors for AD lend support for a contributing rather than an associative role of complement activation in AD pathogenesis (Lambert et al. 2009, Harold et al. 2009). In vitro experiments demonstrated that Aβ can induce complement expression in neurons and glia (Fu et al. 2012, Haga et al. 1993, Carrero et al. 2012, Lian et al. 2015). However, the consequences of complement activation and the functional effects of complement blockade in vivo remain enigmatic. For example, Fonseca et al., reported that C1q deletion attenuated plaque load and synaptic protein loss in an AD mouse model (Fonseca et al. 2004); the same group also showed that application of C1q proved to be protective against Aβ toxicity (Pisalyaput & Tenner 2008). Likewise, both inhibition and activation of C5aR presented similar beneficial effects in AD mouse models: One study revealed that blocking C5aR by its antagonist reduced plaque pathology and glia activation in addition to improving behavioral performances (Fonseca et al. 2009), and another report documented that immunoglobulin treatment which induces C5aR activity resulted in improved synaptic plasticity and cognitive function (Gong et al. 2013). Moreover, inhibition of C3 by overexpressing the soluble complement receptor-related protein y (sCrry) or by C3 genetic ablation resulted in exacerbated Aβ pathology and late-age neurodegeneration (Maier et al. 2008, Fu et al. 2012, Wyss-Coray et al. 2002), implicating a detrimental effect of C3 inactivation. However, follow up studies by Lemere and colleagues revealed that, despite worsened amyloid pathology, C3 inhibition was associated with functional improvement and that the contrasting protective and neurotoxic effects appear to be age-dependent (Lemere and Zheng, personal communication). While the reasons for these apparently conflicting results remain to be resolved, it is clear that the complement-mediated responses are complex and context-dependent.

Whereas NFκB and complement can mediate neuronal function and innate immunity independently, our group recently revealed that they can also collaborate to regulate neuronal homeostasis and AD pathogenesis through an astroglia-neuron signaling pathway (Lian et al. 2015). Specifically, we found that C3 is an astroglial target of NFκB that is upregulated upon NFκB activation. C3 release from astrocytes acts through the neuronal C3a receptor (C3aR) to trigger aberrant intraneuronal calcium, leading to disrupted AMPA receptor trafficking, dendritic morphology and network function (Figure 3). Directly relevant to AD, we found that exposure to Aβ activates astroglial NFκB and C3 release in vitro and that C3aR antagonist treatment rescues cognitive impairment in APP transgenic mice (Lian et al. 2015). Our results support the therapeutic potential of C3aR antagonists for treating chronic neuroinflammation conditions accompanying AD and other neurodegenerative diseases. However, it is important to note that, although we demonstrated a direct astroglial C3 and neuronal C3aR interaction, the in vivo effects of the C3aR blockade are likely attributed by inhibiting the C3aR in all cell types, particularly microglia, where C3aR is highly expressed (Davoust et al. 1999, Martin et al. 2007) and where complement signaling has been shown to modulate Aβ phagocytosis (Fu et al. 2012). Taken together, we propose a complement-mediated intercellular crosstalk model in which neuronal overproduction of Aβ activates astroglial NFκB to elicit extracellular release of C3. This promotes a pathogenic cycle by which C3, in turn, interacts with neuronal and microglial C3aR to impair synaptic structure and function and Aβ phagocytosis, respectively, resulting in network dysfunction and AD pathogenesis (Figure 3).

Figure 3.

An NFκB- and C3-mediated astrocyte-neuron and astrocyte-microglia signaling network that mediates neuronal homeostasis and Alzheimer’s disease.

Conclusions

The last decade has witnessed an impressive growth of knowledge regarding the signaling pathways regulating glial biology and pathophysiology. We focus on some of the important examples in this review, namely the calcium, proteoglycan, TGFβ, NFκB, and complement pathways. It is clear that these factors play essential and multifaceted roles in maintaining neuronal homeostasis, and they impact neurodegenerative processes in a complex and context-dependent manner which may be dictated by their unique neuron-glia signaling profiles or the crosstalk among these pathways. While we attempt to provide a comprehensive and unbiased summary of the wealth of information contained in the published literature, it is important to note that interpretation of the findings should take system and technical limitations into consideration. In particular, many of the studies discussed in this review use in vitro systems or mixed cell-type analysis, both of which do not offer the spatial and temporal resolution needed to decipher the cell-type specific contribution to network function and dysfunction. Many fundamental questions remain within the realm of astrocytes. For example, how many types of astrocytes are there in the CNS? What are their morphological and molecular signatures, and how do these signatures alter during aging and in pathological conditions? Do different types of astrocytes exhibit distinct responses to brain insults, and do these in turn convey distinct neuron-glia signaling pathways? With the development of sophisticated cell-type specific targeting and genetic and functional manipulations combined with powerful imaging technology, we are now poised to perform molecular, biochemical, morphological and functional characterizations at subcellular, cellular and network levels as well as under physiological and disease relevant context. These will not only reveal unprecedented insights into the signaling mechanisms underlying glial pathobiology and neuron-glia interaction, but also provide therapeutic understanding on targeting neuron-glia interaction for AD and other neurodegenerative diseases.

Abbreviation list

AD

Alzheimer’s disease

C3aR

C3a receptor

C5aR

C5a receptor

CAA

cerebral amyloid angiopathy

CGSP

Chondroitin sulfate proteoglycan

CLU

clusterin

CNS

central nervous system

CR1

complement receptor 1

ECM

extracellular matrix

GLT-1

glutamate type I transporter

GSG

glycosaminoglycan

GSH

glutathione

HSPG

heparin sulfate proteoglycan

IκB

inhibitor protein κB

LAP

latency associated peptide

LTBP

latent TGF-β binding protein

MAC

membrane attack complex

MBL

mannose-binding-lection

NFκB

nuclear factor-kappa B

sCry

soluble complement receptor-related protein y

TGF-β

transforming growth factor β

TβR

TGF-β receptor