Amyloid-β1-42 oligomers enhance mGlu5R-dependent synaptic weakening via NMDAR activation and complement C5aR1 signaling

Summary Synaptic weakening and loss are well-correlated with the pathology of Alzheimer’s disease (AD). Oligomeric amyloid beta (oAβ) is considered a major synaptotoxic trigger for AD. Recent studies have implicated hyperactivation of the complement cascade as the driving force for loss of synapses caused by oAβ. However, the initial synaptic cues that trigger pathological complement activity remain elusive. Here, we examined a form of synaptic long-term depression (LTD) mediated by metabotropic glutamate receptors (mGluRs) that is disrupted in rodent models of AD. Exogenous application of oAβ (1–42) to mouse hippocampal slices enhanced the magnitude of mGlu subtype 5 receptor (mGlu5R)-dependent LTD. We found that the enhanced synaptic weakening occurred via both N-methyl-D-aspartate receptors (NMDARs) and complement C5aR1 signaling. Our findings reveal a mechanistic interaction between mGlu5R, NMDARs, and the complement system in aberrant synaptic weakening induced by oAβ, which could represent an early trigger of synaptic loss and degeneration in AD.


INTRODUCTION
Alzheimer's disease (AD) is a progressive neurodegenerative disease characterized by memory loss, cognitive deficits, and changes to personality and behavior. 1Although it is the most common cause of dementia affecting the aging population worldwide, therapeutic options are limited.3][4] The development of improved treatment options requires a deeper understanding of how synapses are initially impaired and how this leads to overt neuronal degeneration.
A key pathological feature of AD is plaques, which are made up primarily of insoluble aggregates of amyloid beta (Ab) protein derived from the amyloid precursor protein (APP).The original amyloid cascade hypothesis posited that these Ab plaques are the initial trigger of AD pathology. 5However, subsequent research found that levels of soluble oligomers of Ab (oAb) better correlate with the degree of cognitive impairment in AD patients. 6,7[10] Normally, synaptic strength is modulated through long-term potentiation (LTP) and long-term depression (LTD), which are thought to provide the cellular and molecular basis of learning and memory. 113][14][15][16][17][18][19][20] Therefore, the study of oAb-dependent changes to LTD provides a means to investigate the initial synaptic changes relevant to the earliest stages of AD.
Synapse damage and loss caused by oAb does not occur in a neuron-autonomous manner.The complement cascade is an innate immune pathway which has emerged as a central driver of synapse loss in AD. 21,22 Complement cascade activation converges on the cleavage of complement component 3 (C3), generating C3a and C3b, the latter of which covalently attaches to target structures. 23Subsequently, C3b can either be degraded to iC3b to mediate phagocytosis via the CR3 receptor expressed on phagocytes or can form the C5 convertase along with C4b and C2a.Analogous to C3, the C5 convertase cleaves C5 to generate the fragments C5a and C5b.C5a is a diffusible protein fragment that promotes chemotaxis and inflammation primarily through binding to the receptor C5aR1.Upregulation of C5aR1 surrounding plaques in AD mouse models has been observed. 246][27][28] However, the mechanism by which C5aR1 signaling generates neuronal damage in AD is poorly understood.
36][37][38][39] Synaptic plasticity involving the mGluR subtype 5 (mGlu 5 R) is particularly pertinent to the understanding of AD for multiple reasons.For example, mGlu 5 R was found to be the only co-receptor of cellular prion protein (PrP c ) necessary for oAb to activate intracellular signaling in neurons. 40Further, both knockout of mGlu 5 R and pharmacological antagonism is protective in oAb-based and genetic models of AD. 39,[41][42][43] Interestingly, the complement cascade has recently been implicated in oAb-dependent synapse loss mediated by mGluR activation. 44A silent allosteric modulator of mGlu 5 R has also been found to restore synapse density through reduced tagging by C1q and microglia engulfment in a genetic mouse model of AD. 45 Previously, we identified a simple method to study mGluR-mediated LTD, via the brief application of the group I mGluR agonist S-3,5dihydroxyphenylglycine (DHPG), which causes lasting depression of AMPAR-mediated synaptic transmission. 46This DHPG-induced LTD (DHPG-LTD) has since been used extensively to uncover many aspects of mGluR physiological and pathological function, including the regulation of mGluR function in AD mouse models. 13,15,16,36,47,48Here, we have used DHPG-LTD to investigate whether C5aR1 signaling is involved in mGluR-driven synaptic plasticity.We applied oAb as a standard way to induce synaptotoxicity that is relevant to AD pathology.We found that oAb enhanced DHPG-LTD via a mechanism that involves the activation of mGlu 5 Rs, NMDARs and C5aR1 of the complement cascade.These findings reveal a signaling axis between glutamate receptors and the complement cascade that triggers early synaptic dysfunction that may underlie the initial stages of dementia.

Acute oAb exposure enhances mGlu 5 R-dependent LTD
To investigate the impact of oAb on synaptic function, we incubated acute hippocampus slices for 2 h with oAb (500 nM).We subsequently measured excitatory synaptic transmission and plasticity with field potential recordings at hippocampal CA3-CA1 synapses.As a positive control for the bioactivity of our oAb preparation, we confirmed previous findings 15,49 that oAb was able to impair LTP in slices pretreated with oAb (+oAb) versus interleaved non-treated slices (-oAb; Figures S1A and S1B).Although LTP was impaired by acute application of oAb, basal synaptic transmission was unaltered, indicated by the overlapping input-output fEPSP relationship in +oAb compared to -oAb slices (Figures S1C and S1D).As such, this oAb application paradigm allowed us to study early disruptions to plasticity that precedes the loss of synapses.
To investigate how oAb alters mGluR-dependent synaptic plasticity mechanisms, we induced LTD using S-DHPG (referred to hereafter as DHPG), a selective agonist for group I mGluRs (mGlu 1 R and mGlu 5 R), that is widely used to study mGluR function.A short application of DHPG (10 min) induced a stable LTD (DHPG-LTD) in -oAb slices from wild-type mice (Figures 1A and 1B).We observed that the magnitude of DHPG-LTD was significantly enhanced in interleaved +oAb slices (Figures 1A and 1B).We then sought to determine the group I mGluR subtype responsible for mediating the enhancement of DHPG-LTD by oAb.We co-applied the mGlu 1 R and mGlu 5 R antagonists, YM298,198 (2 mM) and MTEP (1 mM), respectively.Under these conditions, the induction of DHPG-LTD was blocked in both -oAb and +oAb slices (Figures 1C and 1D), indicating that both control and oAb-enhanced DHPG-LTD shared a group I mGluR-dependency.Next, to determine whether the LTD phenotype exhibited a group I mGluR subtype-specificity, we applied each antagonist independently and measured DHPG-LTD in the presence and absence of oAb.YM298,198 failed to prevent the enhancement of DHPG-LTD by oAb, indicated by the significantly greater LTD magnitude in +oAb compared to -oAb slices (Figures 1E and 1F).Conversely, in the presence of MTEP, the level of DHPG-LTD was not significantly different between control and oAb-treated slices (Figures 1G and 1H).This suggested that specifically the mGlu 5 R subtype was responsible for mediating the effects of oAb on DHPG-LTD.To examine this using an orthogonal approach to interfere with mGlu 5 R activity, we examined DHPG-LTD in mice genetically lacking mGlu 5 R. 50 We observed that in mGlu 5 R À/À mice, DHPG-LTD was absent in both -oAb and +oAb slices (Figures 1I and 1J).Together, our pharmacological and genetic data indicate that the enhancement of DHPG-LTD by oAb requires the activation specifically of mGlu 5 Rs (referred to hereafter as mGlu 5 R-LTD).

NMDARs are required for oAb-mediated enhancement of mGlu 5 R-LTD
There are two distinct forms of DHPG-LTD that can be distinguished by their dependence upon the activation of NMDARs. 46,51To determine whether NMDARs are required for mGlu 5 R-LTD either under control conditions or the enhanced mGlu 5 R-LTD observed in the presence of oAb, we used a selective NMDAR antagonist, L689,560 (5 mM).L689,560 was chosen as it acts at the glycine site of the NMDAR and so is independent of L-glutamate concentration, which might be altered by oAb treatment. 16In these experiments, we interleaved a new set of controls and again observed greater mGlu 5 R-LTD magnitude in +oAb compared to -oAb slices (Figures 2A and 2B).We found that NMDAR activation was not required for mGlu 5 R-LTD in control (-oAb) conditions.However, the enhancement of mGlu 5 R-LTD by oAb was completely prevented by L689,560 (Figures 2C and 2D) and therefore requires NMDAR signaling.

oAb enhances mGlu 5 R-LTD via C5aR1 signaling
Early studies on the effects of oAb on synaptic plasticity found that inhibition of LTP was prevented by application of minocycline, an inhibitor of microglia activation. 52Further, it has recently been found that agonism of group I mGluRs in vivo leads to complement cascade activation and eventual synapse loss. 44We, therefore, hypothesized that complement cascade activity is necessary for the enhancement of mGlu 5 R-LTD by oAb.Within the complement cascade, we chose to examine the role of C5aR1 signaling, which has been implicated in the pathology of genetic mouse models of AD. [25][26][27][28]53,54 For these experiments, we also interleaved a new set of controls and again observed that mGlu 5 R-LTD magnitude was significantly higher in +oAb compared to -oAb slices (Figures 3A and 3B). To instigate the role of C5aR1 in this phenotype, we bath-applied PMX205 (0.3 mM), a peptide-based C5aR1 antagonist.55 In control (-oAb) slices, DHPG application still induced a stable LTD in the presence of PMX205.However, there was no significant difference in mGlu 5 R-LTD magnitude between -oAb and +oAb slices (Figures 3C and 3D).These data suggested that C5aR1 signaling had a specific role only in the oAb-enhanced portion of mGlu 5 R-LTD.To provide further evidence for this, we also used a nonpeptide based C5aR1 antagonist, W54011.56 As with PMX205, bath-application of W54011 (0.3 mM) had no effect on the level of mGlu 5 R-LTD induced under control conditions (-oAb) but prevented the enhancement induced by oAb (Figures 3E and 3F).Together, these data indicate that the oAb-mediated enhancement of mGlu 5 R-LTD at hippocampal CA3-CA1 synapses occurs via activation of complement C5aR1 signaling.

DISCUSSION
The complement cascade is emerging as a key innate immune pathway for shaping synapses during brain development, as well as driving synapse loss in various disease states. 22,57,58However, the synaptic signals that lead to complement activation remain poorly understood.
In the present study, we uncover a mechanistic link between innate immune signaling and the amplification of glutamatergic synapse depression induced by oAb.We found that mGlu 5 R-dependent LTD induced by DHPG application was enhanced in slices treated with oAb (Figure 1).Unlike vehicle-treated slices, the oAb-mediated enhancement of mGlu 5 R-LTD was dependent on NMDAR signaling (Figure 2).Similarly, antagonism of the complement cascade receptor C5aR1 prevented oAb from enhancing mGlu 5 R-LTD (Figure 3), without affecting the underlying mGlu 5 R-LTD per se.Given that basal synaptic transmission was unaltered by oAb (Figure S1), our findings reveal a signaling axis between mGlu 5 Rs, NMDARs, and the complement cascade that mediates early synaptic plasticity alterations, upstream of overt synapse loss.

oAb-induced glutamatergic synapse dysfunction
Although AD is an extremely slowly progressing disease, it is likely that at the level of a single synapse there is acute damage, which in many cases may be orchestrated by toxic oAb species.This would presumably occur when the processes that ordinarily prevent the accumulation of oAb are sufficiently compromised such that the local concentration increases to a level that can impair synaptic function and structure.Therefore, the transient application of oAb has been widely employed to model this early, critical stage of the disease.Consistent with the validity of this model, mechanisms that are engaged by the acute application of oAb to impact synaptic plasticity, such as the involvement of GSK-3, caspase-3, 17 tau, 19,59 and microglia 52 are all directly relevant to human AD pathology.7]36,49,60,61 Altered activity of both group I mGluRs and NMDARs by oAb is central to this glutamatergic synaptic failure.mGlu 5 R was identified as a co-receptor of cellular prion protein (PrP c ) for binding oAb and is necessary for oAb-induced intracellular signal transduction and synapse loss. 37,40,62,631,66 Despite these important advances, whether NMDARs and mGluRs are synergistically involved in oAb-induced synaptic weakening and loss is poorly understood.Of relevance to the present study, at least two distinct forms of DHPG-LTD exist, which can be distinguished by the dependence on NMDARs. 46,51In our current study, DHPG-LTD induction under control conditions was not affected by NMDAR antagonism and was completely dependent on group I mGluRs.The finding that full inhibition of DHPG-LTD pharmacologically required antagonism of both mGlu 1 Rs and mGlu 5 Rs but was also absent in mGlu 5 R À/À mice alone, is most readily explained by the receptor being an mGlu 1 /mGlu 5 heterodimer. 67,68In mGlu 5 R À/À mice, these mGlu 1 /mGlu 5 heterodimers would fail to assemble thereby completely preventing the induction of DHPG-LTD.In contrast, the oAb-enhanced portion of the LTD fully depended on the activation of NMDARs and specifically required mGlu 5 Rs but not mGlu 1 Rs.Therefore, oAb may preferentially induce and/or bias toward mGlu 5 R-dependent signaling, which is blocked by MTEP but not YM298,198.
Our observations of enhanced mGlu 5 R-LTD in the presence of exogenously applied oAb are in line with previous studies using acute oAb exposure. 15,16,47Conversely, studies utilizing amyloidogenic AD models with overexpression of APP have observed an inhibition of mGluR-LTD. 13,48One explanation for these disparate findings is the duration of oAb exposure between AD models.In models of acute oAb exposure (including this study), few if any synapses would be expected to undergo endogenous mGlu 5 R-LTD in the presence of oAb prior to experimental LTD induction utilizing DHPG.Conversely, in models utilizing APP overexpression, oAb exposure is chronic, providing a longer time window for endogenous mGlu 5 R-LTD to occur in the presence of oAb.Thus, the impairment of mGlu 5 R-LTD in these models may be the result of occlusion.A non-mutually exclusive alternative explanation is that synaptotoxic effects of oAb require accumulation in intracellular compartments. 69,70In such a scenario, exogenously applied oAb would need to enter neurons via uptake by surface receptors which could include mGlu 5 R and/or PrP c , a known binding partner of mGlu 5 R and oAb.Conversely, in AD models using overexpression, oAb is already present intracellularly and would not require an active uptake mechanism.
Both mGlu 5 R and NMDARs are also implicated in AD pathogenesis in genetic mouse models and humans.Antagonists of mGlu 5 R and genetic deletion have both been found to reduce synaptic and behavioral deficits in AD mouse models. 39,41,42,45,71Indeed, the mGlu 5 R silent allosteric modulator BMS-984923, 71 is currently in phase I clinical trials.Evidence that altered NMDAR signaling is causally related to the human condition is the beneficial action of memantine, a non-competitive NMDAR antagonist, for the treatment of AD.3][74] Our experiments reveal a mechanism by which NMDARs and mGlu 5 Rs may act synergistically to control synaptic integrity.

C5aR1 signaling in AD
In both postmortem human samples as well as numerous AD mouse models, many components of the complement cascade have been found to be upregulated. 22,58Importantly, pharmacological and genetic manipulations have provided evidence that C5aR1 signaling is necessary for synaptic dysfunction and cognitive impairments in AD genetic mouse models.In particular, one of the C5aR1 antagonists used in the current study, PMX205, has been found to reduce pathology in genetic mouse models of AD. 26,27 Furthermore, C5aR1 genetic ablation is protective in AD mouse models. 25,28However, the role of oAb in detrimental C5aR1 signaling, and its intersection with glutamatergic synaptic signaling to cause this dysfunction was previously not understood.Our study provides insight into this communication, by demonstrating that oAb enhances synaptic weakening, via an NMDAR and mGlu 5 R synergistic mechanism that requires C5aR1.
An important avenue for future studies to investigate is the mechanism by which C5aR1 activation leads to increased LTD.Transcriptomic data indicate that C5aR1 expression is restricted to microglia in the brain 75 and upregulated C5aR1 protein surrounding Ab plaques has been found to co-localize with microglia. 24Thus, it is likely that a paracrine signal would need to be released by microglia to act on synapses to translate C5aR1 activation into synapse strength changes.The canonical inflammatory function of C5a-C5aR1 includes cytokine production. 23,76Interestingly, neurons are known to express cytokine receptors, the activation of which can induce synaptic plasticity. 77,78Additionally, activation of a different microglia receptor, CR3, during hypoxic and neuroinflammatory conditions leads to the production of reactive oxygen species, which acted in a paracrine fashion to induce AMPAR endocytosis at synapses. 79Therefore, cytokines and/or reactive oxygen species are candidate paracrine signals that could mediate the synaptic effects of C5aR1 activation in the presence of oAb.
A second interesting area for future work is to determine whether the enhanced mGlu 5 R-LTD in the presence of oAb is associated with differential structural outcomes compared to physiological mGlu 5 R-LTD.C3b, generated by the cleavage of C3, can either form part of the C5 convertase or can be degraded to iC3b, which induces phagocytosis upon binding to its receptor CR3.In vivo oAb administration has been found to induce synapse elimination by microglia via CR3. 80Further, complement-dependent synapse pruning by microglia under physiological conditions has been found to be preferentially targeted toward weaker synapses. 81,82Therefore, an intriguing possibility is that upon oAb exposure, the NMDAR-and C5aR1-mediated enhancement of mGlu 5 R-LTD identified in this study serves as the trigger for subsequent elimination of synapses by microglia.

Relevance to other neurodegenerative diseases
Pathological upregulation of complement cascade proteins and activity is not exclusive to AD. 83 For example, studies from in vitro and mouse models of amyotrophic lateral sclerosis (ALS) have also identified the role for C5a/C5aR1 in driving pathology. 84,85Interestingly, mGlu 5 R has separately been identified as an emerging therapeutic target in ALS as well. 86,87][90][91][92][93][94][95][96][97][98][99][100][101] Therefore, it will be critical for future studies to elucidate whether the pathological mGlu 5 R-NMDAR-C5aR1 signaling axis identified in our study underlies synapse deterioration in multiple brain disorders.

Limitations of the study
Our chosen experimental paradigm (acute oAb application to brain slices) permitted the precise control of the exposure timing and species of Ab.However, this paradigm does not capture chronic aspects of AD or non-amyloidogenic signaling pathways.Therefore, it will be essential for future work to determine whether the mGlu 5 R-NMDAR-C5aR1 axis identified in our study plays a key pathogenic role in various genetic mouse models of AD.Further, experiments in such mouse models will permit the investigation of whether enhanced mGlu 5 R-mediated LTD through NMDAR/C5aR1 activation has a causal role in behavioral impairments observed in AD mouse models.

Concluding remarks
mGluRs, NMDARs, and C5aR1 have been independently studied for their role in mediating the deleterious effects of oAb.Our study has found a direct link which places glutamate receptor-innate immune interactions at the center of early synaptic plasticity dysfunction.As oAb is thought to be a key trigger in AD, a deeper understanding of the cellular and molecular mechanisms of these interactions has the potential to elucidate therapeutic targets which halt synaptic dysfunction and AD progression.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Figure 2 .
Figure 2. Enhancement of mGlu 5 R-LTD by oAb requires NMDARs (A and B) In a separate cohort, DHPG-LTD was enhanced in slices treated with oAb (34 G 2%, n = 22) compared to the interleaved -oAb slices (21 G 2%, n = 22; p = 0.00031 [***]).Quantification of LTD magnitude shown in B. (C and D) In the presence of the NMDAR antagonist L689,560 (5 mM), the magnitude of DHPG-LTD was not significantly different between +oAb slices (26 G 4%, n = 9) and -oAb slices (22 G 2%, n = 9).Quantification of LTD magnitude shown in D. Sample fEPSP traces in all panels are the mean of 4 consecutive responses at the indicated time points (1-4).For each experiment, sample traces before and after DHPG treatment are superimposed.A two-tailed t test was used in B and D. In all graphs, data are presented as mean G SEM.

Figure 3 .
Figure 3. Complement C5aR1 is necessary for oAb-driven enhancement of mGlu 5 R-LTD (A and B) In a third cohort, oAb-treated slices had significantly greater DHPG-LTD magnitude (43 G 3%, n = 16) compared to the interleaved -oAb slices (26 G 3%, n = 16; p = 0.00064 [***]).Quantification of LTD magnitude shown in B. (C and D) In the presence of the C5aR1 antagonist PMX205 (0.3 mM), the level of DHPG-LTD was not significantly different between +oAb slices (35 G 4%, n = 8) and -oAb slices (29 G 6%, n = 8).Quantification of LTD magnitude shown in D. (E and F) A separate C5aR1 antagonist, W54011 (0.3 mM), also prevented the enhancement of DHPG-LTD magnitude in +oAb slices (32 G 5%, n = 6) compared to -oAb slices (29 G 4%, n = 6).Quantification of LTD magnitude shown in F. Sample fEPSP traces in all panels are the mean of 4 consecutive responses at the indicated time points.For each experiment, sample traces before and after DHPG treatment are superimposed.Two-tailed t tests were used in B, D, and F. In all graphs, data are presented as mean G SEM.
TABLE d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS B Animals d METHOD DETAILS B Oligomeric amyloid-b 1-42 (oAb) preparation B Hippocampal slice preparation B Electrophysiology d QUANTIFICATION AND STATISTICAL ANALYSIS