Synapse elimination is a normal developmental process occurring throughout the central and peripheral nervous systems. Meanwhile, gradual and early loss of synapses is a characteristic that is common to several neurodegenerative disease states. Recent evidence has emerged implicating molecules canonically involved in the immune system and inflammation accompanying neurodegeneration (e.g. classical complement cascade) as important players in the normal elimination of synapses in the developing nervous system. As a result, a question has emerged as to whether mechanisms underlying elimination of synapses during normal development are recapitulated and contribute to early synapse loss and nervous system dysfunction during neurodegenerative disease. The present review explores this possibility and provides a description of many neuroimmune proteins that may participate in the elimination of synapses and synaptic dysfunction in the developing and diseased brain.
- neuroimmune mediator
- retinal ganglion cell (RGC)
- synaptic pruning
In order to form the precise finely tuned neural circuitry of the mature nervous system, a specific and ordered developmental programme must occur. A characteristic series of developmental events required for proper nervous system development is synapse elimination (i.e. synaptic pruning). Early in development, neurons extend exuberant connections to neighbouring cells. Synapse elimination involves the selective activity-dependent removal of these excess inputs [1–4]. Whereas synapse elimination occurs during normal nervous system development, loss of synapses has emerged as a hallmark of several neurodegenerative diseases, as well as normal aging [5–8]. As a result, an intriguing possibility emerges that a common mechanism may underlie elimination of synapses during development and disease. If true, these mechanisms offer potential targets for therapeutic intervention, particularly since synapse loss is often associated with the earliest stages of disease. Supporting this idea, recent findings from our laboratory and others demonstrate that molecules normally involved in immune responses and pathogenesis during early stages of CNS (central nervous system) disease (e.g. complement pathway components, members of the MHCI family) mediate synapse refinement and elimination in the developing brain [9–11]. In the present review, we discuss mechanisms of synapse removal during normal development and relate these findings to synapse pathology associated with neurodegenerative disease. In particular, we highlight the role of glial cells and neuroinflammatory molecules.
Models of developmental synapse elimination
During development, excess numbers of synapses are formed followed by a selective elimination or pruning of those which are inappropriate. Whereas it is clear that electrical activity plays a key role, the cellular and molecular mechanisms underlying synapse elimination remain elusive. Studies in the PNS (peripheral nervous system) at the mammalian NMJ (neuromuscular junction) have helped our understanding . In early postnatal development, postsynaptic muscle cells are innervated by multiple motor axons. However, by the second postnatal week, the majority of these inputs are permanently eliminated through a process of activity-dependent competition in which one input is maintained and strengthened. The activity-dependent molecular determinants driving the initiation and progression of this phenomenon remain unclear.
In the CNS (central nervous system), the complexity and heterogeneity of neural networks makes synaptic pruning difficult to study at individual synapses, but growing evidence suggests that postsynaptic neurons initially receive widespread imprecise inputs that are later refined and sharpened during the first several weeks of postnatal development [1–3]. A classic example of local activity-dependent CNS synapse elimination occurs in the retinogeniculate system [12–15]. In binocular animals, dLGN (dorsal lateral geniculate nucleus) relay neurons are initially innervated by multiple (up to ten) RGC (retinal ganglion cell) axons. Furthermore, during the first postnatal week, RGC axons from each eye terminate in overlapping regions within the dLGN. By the end of the third postnatal week, the pruning process is complete: RGCs terminate in non-overlapping eye-specific domains and each dLGN neuron receives stable inputs from one or two RGC axons [13,14]. Together, eye-specific segregation and elimination of multiple innervations represent elimination of thousands of inappropriate synapses and stabilization of the strongest synaptic terminals [12,13,15].
Over a decade ago, a model was proposed describing a possible mechanism for activity-dependent synapse elimination [3,16]. This model proposed that strong synapses facilitate elimination of weaker synapses by releasing two types of signals: (i) a ‘protective’ signal(s) for the stronger synapse, resulting in stabilization and strengthening; and (ii) a ‘punishment’ signal(s) for weaker synapses, resulting in their elimination. Within the last decade, work towards the identification of molecular signals involved in activity-dependent elimination of CNS synapses has revealed a surprising role for immune-related molecules. Included among these molecules are C1q and the classical complement cascade , MHC I molecules [9,17,18] and NPs (neuronal pentraxins) . An exciting possibility is that immune molecules may be ‘punishment’ signals, resulting in the elimination of weaker synapses. In the present review, we highlight the role of these molecules in developmental pruning and discuss the implications for removal of synapses during early stages of neurodegenerative disease.
Elimination of CNS synapses: the role of glia and the classical complement cascade
It is becoming increasingly clear that glial cells play an active role in synapse development and plasticity. For example, growing evidence suggests that astrocytes provide instructive signals that control the formation and development of synapses [19–21]. In efforts to identify synaptogenic signals secreted by astrocytes, purified rodent RGC–astrocyte co-cultures have proven a useful model system. Immunopurified RGCs cultured alone survive for several weeks, but form few synapses [22,23]. However, when purified RGCs are cultured with a feeding layer of astrocytes, or ACM (astrocyte-conditioned medium), robust synaptogenesis occurs . Using this system, a gene chip screen was performed to identify neuronal genes influenced by astrocyte-derived signals during synaptogenesis. In particular, one gene that was significantly up-regulated 10–30-fold in RGCs in response to astrocyte-derived signals (ACM) was C1q, an immune system protein thought not to be normally expressed within the brain .
C1q is the initiating protein of the classical complement cascade, a component of the innate immune system involved in elimination of dead cells, debris or pathogens (Figure 1). Canonically, C1q binds to and coats (i.e. opsonizes) the surface of dead cells, debris or pathogens, thereby triggering a protease cascade, leading to the deposition of the downstream complement protein, C3. Opsonization with activated C3 fragments (C3b and iC3b) leads to elimination of material in one of two ways. Deposited C3 can directly activate C3 receptors on macrophages or microglia, thereby triggering elimination by phagocytosis, or activated C3 can trigger the terminal activation of the complement cascade, leading to cell lysis through the formation of a lytic membrane-attack complex (Figure 1, bottom). Given that C1q is up-regulated in developing neurons, we hypothesized that C1q could be similarly opsonizing or ‘tagging’ synapses for elimination in the brain .
Consistent with this hypothesis, mice deficient in classical complement pathway components, C1q and C3, have deficits in eye-specific segregation and elimination of excess synapses on dLGN neurons . Involvement of the classical complement cascade in developmental pruning was supported further by the developmental expression and localization of C1q. C1q is specifically and highly expressed by RGCs at time points consistent with pruning. Furthermore, C1q localizes to pre- and post-synaptic terminals on a similar timescale, and work in our laboratory suggests that activated C3 (iC3b/C3b) is also localized to developing synapses (D.P. Schafer and B. Stevens, unpublished work). These data provide a model in which, similar to the peripheral immune system, C1q opsonizes weaker synapses for elimination followed by deposition of iC3b/C3b (Figure 2). Following iC3b/C3b deposition, two possibilities exist: (i) phagocytosis of opsonized targets by microglia; or (ii) cell lysis via membrane-attack complex formation. These possibilities are currently being explored in our laboratory.
A common feature of the complement cascade is that activated components have very limited diffusion and thus result in very little ‘innocent bystander’ damage. As a result, C1q and downstream iC3b/C3b may be acting as a highly efficient local ‘punishment’ signal for weak synapses. Furthermore, it is an intriguing possibility that inhibitors of the complement cascade (e.g. Factor H, C1 inhibitor) may be acting as local ‘protective’ signals for stronger synapses. As a result, important questions for future investigation include the following. (i) Which synapses are ‘tagged’ by complement cascade components and/or inhibitors? (ii) Is electrical activity required for expression and/or activation of complement cascade components and/or inhibitors? (iii) How does neural activity effect complement component and/or inhibitor function?
Other neuroimmune mediators involved in developmental elimination of CNS synapses
Whereas complement plays a key role in the elimination of synapses during normal CNS development, there are other immune-related molecules that have recently been identified as mediators of synaptic refinement and plasticity in both the developing and adult nervous system [24,25]. These molecules include: (i) components of the innate immune system (e.g. complement pathway and pentraxins); (ii) components of the adaptive immune system (e.g. MHC I family of proteins and receptors); and (iii) inflammatory cytokines [e.g. TNFα (tumour necrosis factor α)].
NPs are synaptic proteins similar in sequence to short pentraxins of the peripheral immune system. Short pentraxins are traditionally involved in opsonization and phagocytosis of dead cells in the immune system. Similar to complement-deficient mice, mice deficient in NP1 and NP2 and the receptor, NPR, have transient defects in eye-specific segregation in the dLGN . In addition, purified RGCs co-cultured with an astrocyte feeder layer from NP1/NP2-knockout mice exhibited a delay in maturation of glutamatergic synapses [i.e. decreased mEPSC (miniature excitatory postsynaptic current) amplitude and frequency] compared with wild-type RGC-astrocyte co-cultures.
MHC I molecules represent a large family of proteins that mediate removal of foreign antigens (e.g. bacteria, cancer cells) by the immune system. By presenting corresponding polypeptides of foreign (i.e. ‘non-self’) antigens on the surface of cells, MHC I molecules activate T-cells which leads to removal of foreign antigens. The MHC I family of molecules and receptors were the first immune-related molecules found to be involved in developmental synapse elimination [9,17]. MHC I molecules were shown to be up-regulated in neurons by activity, enriched in brain regions undergoing activity-dependent remodelling and enriched in synaptic compartments [9,17,18]. Similar patterns have been detected for MHC I receptors [26–28]. Moreover, animals deficient in MHC I molecules have defects in eye-specific segregation in the retinogeniculate pathway and ocular dominance plasticity, suggesting a role in developmental elimination of CNS synapses [9,26]. In addition to development, MHC I-knockout mice have defects in adult plasticity paradigms such as LTP (long-term potentiation) and LTD (long-term depression) in the hippocampus and cerebellum [9,29].
Similar to MHC I molecules, inflammatory cytokines have been associated with adult plasticity with implications for facilitating developmental plasticity. For example, TNFα is released from astrocytes and promotes an increase in AMPA (α-amino-3-hydroxy-5-methylisoxazole4-propionic acid) receptor surface expression and facilitation of excitatory synaptic transmission [30,31]. Another pro-inflammatory cytokine associated with synaptic plasticity, IL (interleukin) -1β, inhibits LTP induction following tetanic stimulation in the hippocampus [25,32,33]. Although the mechanism is not entirely clear, inhibition of Ca2+ influx through NMDA (N-methyl-D-aspartate) receptors is observed following treatment with IL-1β [32,34]. Besides LTP, synaptic plasticity associated with mechanisms of pain development has been largely attributed to pro-inflammatory cytokines secreted by reactive astrocytes and microglia (IL-1β, TNFα and IL-6). Each of these cytokines has facilitated peripheral and central sensitization by enhancing excitatory and decreasing inhibitory transmission [35,36].
Taken together, traditional regulators of normal immune function have been implicated in developmental and normal adult plasticity of CNS synapses on both a structural and a functional level. It remains an intriguing possibility that components of the complement, pentraxin, MHC I and inflammatory cytokine families may be acting in concert to facilitate synaptic changes in the developing steady-state adult and/or pathogenic CNS. The following section explores the possibility that early synapse dysfunction and loss that accompanies several neurodegenerative disease states may be initiated and/or propagated by immune molecules associated with developmental plasticity. As an example, we focus on the complement pathway.
Synapse loss and neuroimmune mediators during early stages of neurodegenerative disease
Early synapse loss has been described in several neurodegenerative disease and injury states. For example, early synapse dysfunction and loss has been well documented during AD (Alzheimer's disease) . Soluble Aβ (amyloid β-peptide) oligomers, which are present in the earliest stages of AD, inhibit synaptic plasticity such as LTP in the hippocampus and memory of learned behaviour in rodents [37–39]. In mouse models of AD, there is a clear and early decrease in presynaptic terminal and postsynaptic density immunoreactivity before plaque formation, as well as a decrease in spine density as analysed by electron and light microscopy [5,40,41]. This loss of terminals is also reflected in human patients. For example, a decrease in spine density was observed in cortical biopsies from early-stage AD patients . Furthermore, a recent study in mice harbouring the human tau mutation P301S known to induce tauopathy associated with frontotemporal dementia have significant synapse loss and a concomitant increase in activated microglia in the hippocampus . This pathology occurs at 3 months of age before observable neurofibrillary tangle pathology.
The data described above raise the question: are mechanisms of extranumerary synapse removal during development recapitulated in early stages of neurodegenerative disease (Figure 2)? If so, could these mechanisms be initiating factors determining disease onset and/or progression? Consistent with this idea, neuroinflammatory mediators known to underlie normal developmental synapse elimination have long been associated with neurodegenerative disease and injury. Components of the complement cascade are among such molecules [44,45]. For example, several studies implicate the initiating molecule in the classical complement cascade, C1q, in the pathophysiology of AD. C1q is up-regulated in CNS neurons during early stages of AD and is highly concentrated in amyloid plaques . Deficiency in C1q decreases neuronal cell loss in mouse models of AD  and elevation of C1q enhances AD pathology . In addition to AD, C1q is also up-regulated in other human neurodegenerative diseases, such as glaucoma and ALS (amyotrophic lateral sclerosis) . Consistent with these studies, we have assessed a mouse model of glaucoma, the DBA/2J mouse. During early stages of disease, before any significant neurodegeneration, C1q is up-regulated in the retina and localizes to the inner plexiform layer, a region enriched in postsynaptic terminals from RGCs . Furthermore, C1q localization to synapse-enriched sites increased with disease progression and was temporally correlated with decreases in synapse density. These data suggest that, similar to development, C1q may be opsonizing or ‘tagging’ synapses for removal in the retina during early stages of glaucoma. We are currently investigating the role of C1q in neurodegenerative disease initiation and propagation in the glaucoma model, as well as in models of other neurodegenerative disease such as AD. In addition, we are investigating downstream effectors such as C3 activation and recruitment and activation of phagocytic cells (i.e. microglia).
Summary and conclusions
Studies in the developing CNS suggest that neuroimmune molecules, including classical complement cascade components, MHC I molecules and NPs, play a central role in the elimination of excess synapses to create a finely tuned mature circuitry. Work in our laboratory suggests that components of the classical complement cascade localize to and opsonize developing synapses. We propose a model in the developing CNS in which complement components, in concert with MHC I molecules and pentraxins, may be opsonizing or tagging excess synapses for removal (Figure 2). We suggest that clearance of these synapses may be mediated by resident phagocytic microglia or the lytic membrane-attack complex. Similar to the developing system, during early stages of many neurodegenerative diseases, there is a clear loss of synapses. However, the underlying mechanism(s) of synapse loss remains to be elucidated. We hypothesize that similar neuroimmune pathways underlying development are reactivated during pathological conditions by an as yet to be determined mechanism (Figure 2). One plausible mechanism may be that synaptic dysfunction triggers complement up-regulation/activation. Similar to development, we propose that activation of the complement cascade results in complement deposition at dysfunctional synapses, which are subsequently eliminated by phagocytic glia and/or membrane-attack complex formation. Thus elucidating mechanisms underlying development and refinement of nervous system circuitry may, in turn, contribute to our understanding of neurodegenerative disease onset and progression. If these mechanisms are confirmed, they offer new targets of therapeutic intervention in the progression of debilitating nervous system disease.
Our work is supported in part by the Smith Family Award for Excellence in Biomedical Research (to B.S.) and a National Institutes of Health Training Grant [grant number T32 NS007473 (to D.P.S.)].
Synaptopathies: Dysfunction of Synaptic Function: A Biochemical Society Focused Meeting held at The Hotel Victoria, Newquay, U.K., 2–4 September 2009. Organized and Edited by Nils Brose (Max Planck Institute for Experimental Medicine, Göttingen, Germany), Vincent O'Connor (Southampton, U.K.) and Paul Skehel (Centre For Integrative Physiology, Edinburgh, U.K.)
Abbreviations: ACM, astrocyte-conditioned medium; AD, Alzheimer's disease; CNS, central nervous system; dLGN, dorsal lateral geniculate nucleus; IL, interleukin; LTD, long-term depression; LTP, long-term potentiation; NP, neuronal pentraxin; RGC, retinal ganglion cell; TNFα, tumour necrosis factor α
- © The Authors Journal compilation © 2010 Biochemical Society