Neurons respond awkwardly when astrocytes shower them with too many complements. According to a study in the December 18 Neuron online, astrocytes pump out the complement protein C3 in response to Aβ, and neurons react by shriveling some synapses and ramping up the activity of others. The astrocytes’ toxic overture is turned on by the transcription factor NFκB, well known for its role in inflammation. Led by Hui Zheng at Baylor College of Medicine in Houston, the researchers reported that blocking this pathway eliminated memory problems in an AD mouse model. The work adds to a growing body of evidence implicating neuroinflammation as a key player in the neurodegenerative process.

The complement system consists of roughly 30 proteins. Throughout the body, it targets microbes, sickly cells, and biological flotsam for disposal, and ramps up inflammatory responses to ensure this happens. In the central nervous system (CNS), complement expression surges during development, and again in the context of neurodegenerative disease or brain injury. In early life, complement plays a role in pruning synapses, a process that optimizes neural transmission (see Stevens et al., 2007). Effects on the adult brain are less clear, but complement proteins are expressed by astrocytes in response to inflammatory signals (see Zamanian 2012). The complement protein C3 triggers the clearance of Aβ in AD mouse models; however, some research suggests that ridding the brain of complement protects such mice from cognitive decline even in the face of mounting Aβ (see Fu et al., 2012, and Aug 2013 conference coverage).

First author Hong Lian and colleagues wanted to clarify the role of complement activation and NFκB in neurodegeneration. The transcription factor activates complement expression along with many other inflammatory genes, such those that make cytokines. When silent, NFκB is held in check in the cytoplasm by its inhibitor, IκB. Kinases that phosphorylate the inhibitor target it for destruction, freeing NFκB to move into the nucleus and switch on genes. High levels of activated NFκB have been observed in brain tissue from people with Alzheimer’s, Parkinson’s, and Huntington’s diseases.

Lian started by generating transgenic mice without IκB, leaving any NFκB permanently active. They created mice lacking the inhibitor in neurons only, in the whole brain, or in astrocytes only. In the latter two, the researchers detected about sixfold higher levels of C3 mRNA expression in the brain, resulting in nearly double the amount of C3 protein as in control mice. Mice lacking IκB only in neurons had a normal amount of the protein, suggesting that astrocytes are the cells that make C3 in the brain, and that this is regulated by NFκB.

To study what astrocytic C3 does to neurons, the researchers co-cultured mouse hippocampal neurons with normal astrocytes or with those lacking IκB. After sharing a dish with IκB-deficient astrocytes for two weeks, neurons started losing dendritic spines. As these shrank or disappeared, the dendritic arbors lost some of their branches. The number of synapses declined, as well, as measured by a loss of the synaptic proteins synaptophysin and MAP2, and of the excitatory synaptic protein VGluT1. None of this occurred when C3 was depleted from the co-culture medium, while adding C3 to pure neuron cultures recapitulated the effects.

Complement Overload.

In neurons treated with the complement protein C3 (right), dendrites (red) were adorned with fewer synapses (green). [Image courtesy of Lian et al., 2014, Neuron.]

To see how this played out in vivo, the researchers used a viral system to label neurons and their dendrites with green fluorescent protein. Fluorescent microscopy revealed that in mice lacking IκB in their astrocytes, dendritic spines of all shapes and sizes—from stubby, to mushroom, to delicate long ones—took a hit.  Long-term potentiation, a measure of synaptic plasticity, weakened. The mice were forgetful. In contextual fear testing of learning and memory, they froze less often than normal mice when placed in an environment where they had previously experienced a foot-shock.

If astrocytes are sending out toxic C3 signals, how are the neurons receiving them? The researchers suspected the C3a receptor, and indeed, adding a C3aR antagonist to the co-cultures prevented spine loss and C3aR-deficient neurons thrived among IκB-negative astrocytes. Treating the IκB-negative mice with a C3aR antagonist restored dendritic spines, long-term potentiation, and memory. Together, these results suggested that through its receptor on neurons, C3 secreted from astrocytes dampens synaptic plasticity and compromises learning in mice.

Signaling through C3aR has been reported to boost levels of intraneuronal calcium, and the researchers found that this was the case in neurons co-cultured with astrocytes lacking IκB. Those neurons had more AMPA receptors on their cell surfaces and their mini excitatory post synaptic currents (mEPSCs) crested to higher amplitudes. The C3aR antagonist blocked all these effects. These results suggested that the synapses that remained signaled more vigorously.

How would this relationship between astrocytic C3 and neurons play out in the context of AD? To chip away at this question, the researchers first treated normal astroglial cultures with a fibrillar preparation of Aβ42 peptides (as described in Stine et al., 2003). These boosted translocation of NFκB to the nucleus and ramped up C3 expression. The researchers also detected elevated levels of C3 in APP/TTA transgenic mice, which overexpress human APP under control of the tetracycline promoter (see Jankowsky et al., 2005). Surprisingly, treatment with the C3aR antagonist completely reversed memory deficits in the mice: They performed as well as wild-type mice on the Morris water maze test. Postmortem brain samples from AD patients expressed more nuclear NFκB, as well as C3, than control samples.

Lian proposed a model whereby C3 secreted by astrocytes engages the C3a receptor, which triggers an uptick in intraneuronal calcium. This then leads to the other problems, including loss of dendritic spines. How these effects are connected is still unclear.

Chris Norris of the University of Kentucky in Lexington considers the paper impressive, but found it odd that C3 triggered loss of dendritic spines while at the same time stimulating synaptic activity. Lian said the researchers are currently working to tease out how these seemingly contradictory effects are related.

Mark Mattson of the National Institute on Aging in Bethesda, Maryland, who was not involved in the study, said that the pathway may be more detrimental when neurons are also dealing with toxic proteins, such as complement or Aβ. “When cells are under stress, they may be less able to clear calcium,” he said. “Loss of spines in response to complement may have to do with a toxic calcium overload,” he suggested.

Ben Barres of Stanford University in California thought the study was very interesting, but was skeptical that neurons express C3aR (see full comment below). He wrote that the receptor is primarily expressed on microglia—the immune cells of the CNS (see Schafer et al., 2012). Lian and colleagues did not examine the role of microglial cells in the C3 pathway, but said that finding out how they fit into the picture is a prime focus of their current research.—Jessica Shugart


  1. I think this is a very interesting study! It had already been long known that classical complement cascade is profoundly overactivated in Alzheimer's disease brains. The question had been whether this is driving neurodegeneration or just secondary to it. Most people seemed to have favored the latter, until Beth Stevens and I showed that complement component C1q bound to many synapses in the developing brain and that the classical complement cascade was driving synapse pruning/loss in the normal developing brain (Stevens et al., 2007). Not only did we show that the classical cascade mediated synapse pruning, we showed that the classical complement pathway became reactivated again as the earliest sign of pathology in the neurodegenerative disease glaucoma. In fact Andrea Tenner had shown that C1q deficiency was neuroprotective in mouse models of Alzheimers which had less synapse loss (Fonesca et al., 2004). So Beth and I proposed that C1q and the classical complement cascade would be a universal driver of synapse loss in many neurodegenerative diseases, including Alzheimer's. I became so enthusiastic about this that four years ago I co-founded a company, Annexon Biosciences, that has made the first therapeutic that targets C1q and blocks the classical complement cascade.

    Lian et al. have now provided further direct evidence for a role of the complement cascade component C3 in driving synapse loss in mouse models of Alzheimer's. As others, for example Tenner's group, had done before them, they showed that C3 levels are elevated (see Zhou et al., 2008). This C3 is made by reactive astrocytes. In fact, my lab showed a year or two ago that reactive astrocytes strongly upregulate all the needed classical complement cascade components to run synapse attack, including C1r, C1s, C4, C2, and C3 (Zamanian et al., 2012), with microglia and sick neurons both making C1q. All the pieces seem to be coming together.

    Overall I think the main importance of what Lian et al. have done is to show that pharmacological blockade of the complement system is therapeutic in mouse models of Alzheimer's. Because there are not yet any good drugs to block the classical complement cascade, they could only test an already-established C3aR blocking drug. The problem with this approach is that it is very indirect, because C3aR is not required for the classical complement cascade to mediate synapse loss. That requires synaptic C1q binding and then activation of the cascade leading to deposition of C3b on synapses, which are then eliminated by microglial phagocytosis, the latter being mediated by the microglial C3b receptor called CD11b. Microglia make C3aR, and in fact I doubt the authors' assertion that neurons express the receptor. Therefore, when C3 is cleaved upon complement activation, the C3a fragment does not bind to the synapse, rather it is a chemotactic factor that helps recruit microglia to eat the synapse. But because microglia are already everywhere, the synapses tagged by C3b will be eaten regardless of whether C3aR is blocked or not. Most likely this is why Lian et al. only demonstrate relatively small effects on synapse number and protection. I think the critical question now is whether direct pharmacological blockers of C1q and the classical complement cascade will be neuroprotective. 


    . The classical complement cascade mediates CNS synapse elimination. Cell. 2007 Dec 14;131(6):1164-78. PubMed.

    . Absence of C1q leads to less neuropathology in transgenic mouse models of Alzheimer's disease. J Neurosci. 2004 Jul 21;24(29):6457-65. PubMed.

    . Complement C3 and C4 expression in C1q sufficient and deficient mouse models of Alzheimer's disease. J Neurochem. 2008 Sep;106(5):2080-92. PubMed.

    . Genomic analysis of reactive astrogliosis. J Neurosci. 2012 May 2;32(18):6391-410. PubMed.

    View all comments by Ben Barres
  2. Overall, this paper from Hui Zheng's lab is very interesting. The researchers provide further evidence for the role of C3 and complement in neuronal function and dysfunction in Alzheimer's disease as well as the complement pathway as a potential therapeutic target in AD and other neurodegenerative diseases. 

    This work also highlights a novel mechanism by which astrocytes and astrocyte-neuronal signaling contribute to neuronal and synaptic function and dysfunction. Their finding that reactive astrocytes make C3 (along with other complement components) is consistent with work from the Barres lab and others.

    Most of their paper focuses on the relationship between NFκB and C3 under basal conditions. Their hypothesis, which is a bit complicated, is that NFκB-induced astrocytic C3 activates the neuronal C3a receptor, which leads to increased intraneuronal Ca2+ signaling, which ultimately leads to alterations in neuronal structure and function. It is interesting that they see this affecting excitatory but not inhibitory synapses. This raises questions about what makes excitatory synapses vulnerable in this context.

    The authors report that C3aR antagonist treatment in plaque-deposited APP/TTA mice leads to behavioral rescue in spatial working memory. This is an important finding because they are able to rescue cognitive defects at later stages of disease progression; however, the underlying mechanisms for that rescue are not yet clear. It will be important to address whether (and how) C3aR antagonists restores neuronal and synaptic dysfunction in AD models.

    View all comments by Beth Stevens

Make a Comment

To make a comment you must login or register.


News Citations

  1. Curbing Innate Immunity Boosts Synapses, Cognition

Paper Citations

  1. . The classical complement cascade mediates CNS synapse elimination. Cell. 2007 Dec 14;131(6):1164-78. PubMed.
  2. . Genomic analysis of reactive astrogliosis. J Neurosci. 2012 May 2;32(18):6391-410. PubMed.
  3. . Complement component C3 and complement receptor type 3 contribute to the phagocytosis and clearance of fibrillar Aβ by microglia. Glia. 2012 May;60(6):993-1003. PubMed.
  4. . In vitro characterization of conditions for amyloid-beta peptide oligomerization and fibrillogenesis. J Biol Chem. 2003 Mar 28;278(13):11612-22. PubMed.
  5. . Persistent amyloidosis following suppression of Abeta production in a transgenic model of Alzheimer disease. PLoS Med. 2005 Dec;2(12):e355. Epub 2005 Nov 15 PubMed.
  6. . Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron. 2012 May 24;74(4):691-705. PubMed.

Further Reading

Primary Papers

  1. . NFκB-activated astroglial release of complement C3 compromises neuronal morphology and function associated with Alzheimer's disease. Neuron. 2015 Jan 7;85(1):101-15. Epub 2014 Dec 18 PubMed.