. Complement C3 deficiency leads to accelerated amyloid beta plaque deposition and neurodegeneration and modulation of the microglia/macrophage phenotype in amyloid precursor protein transgenic mice. J Neurosci. 2008 Jun 18;28(25):6333-41. PubMed.


Please login to recommend the paper.


Make a Comment

To make a comment you must login or register.

Comments on Primary Papers and News

  1. This important study by Cynthia Lemere of Brigham and Women’s Hospital and collaborators at Harvard Medical School further supports a beneficial role of intracerebral activation of the complement system, possibly by promoting phagocytosis of Aβ. First authors Marcel Maier and Ying Peng, and colleagues have generated a C3-deficient amyloid precursor protein (APP) transgenic mouse model of Alzheimer’s disease (AD) to specifically investigate the role of the complement system central component C3. APP transgenic mice lacking C3 were found to exhibit increased amyloid plaque burden, enhanced neurodegeneration, and shifted microglia activation toward a phenotype often found to be associated with tissue repair processes. These observations further support a beneficial role for complement component C3 in plaque clearance in APP mice. This is in line with a previous report describing enhanced pathology in APP transgenic mice with brain-targeted expression of a soluble form of complement receptor-related protein y (sCrry), a potent inhibitor of C3 convertases (Wyss-Coray et al., 2002).

    Brain cells can produce and mount a fully functional complement system. Together with a wealth of histopathological reports describing the presence of complement proteins in senile amyloid plaques and neurofibrillary tangles, intracerebral activation of the complement system has long been associated with inflammatory processes thought to drive AD pathology. Furthermore, neurons, because they express low levels of complement regulators, have the propensity to spontaneously activate the complement system in vitro in an antibody-independent manner. As a consequence, neurons are believed to be especially susceptible to complement-mediated lysis (Singhrao et al., 2000).

    Overall, the ability of the complement system to serve as a defense system has probably been overlooked. This view is now emerging and this paper by Marcel Maier, Ying Peng, and collaborators further supports the hypothesis that the complement system may actively participate in helping clear the plaques. However, given the complexity of the complement system, it is expected that different arms of the complement cascade are involved in different - beneficial or harmful – processes. Several examples can illustrate this view: 1) APP transgenic mice lacking C1q, the first component of the classical cascade, exhibit less neuropathology, suggesting a detrimental effect of C1q on neuronal integrity (Fonseca et al., 2004). Interestingly, increased activation of the alternative pathway in these mice may also contribute to the observed decreased pathology; 2) As mentioned above and pointed out by Pat McGeer in his comments to the paper by Lemere and colleagues, intracerebral activation of the terminal lytic membrane attack complex is likely to lead to severe damaging effect; and 3) On the other hand, proinflammatory anaphylatoxins C3a and C5a, generated upon complement activation, have previously been reported to be neuroprotective in vitro (van Beek et al., 2003). Overall, the lack of brain-penetrant pharmacological tools to dampen the complement system at different level of activation is a clear limitation to the further understanding of the complex involvement of complement in experimental disease models.

    Data with APP transgenic mice have been overall very informative. However, as suggested by Pat McGeer, these models are incomplete models of AD, partly because APP transgenic mice display weaker inflammation and complement activation than what is observed in AD patients. In light of this, it is difficult to predict how the observations by Marcel Maier, Ying Peng, and colleagues will translate in AD patients. We cannot rule out the possibility that the balance required in experimental models of AD for the complement system to promote plaque clearance may be shifted in AD patients.

    Targeting Aβ by active or passive immunization have consistently been reported to be effective in APP-overexpressing mouse models. Translation in humans proved problematic when AN1792 immunotherapy vaccine trial initiated by Elan Pharmaceuticals was halted because approximately 6% of the volunteers developed encephalitis. Although interrupted, this trial indicated that Aβ immunotherapy still holds promises for the treatment of AD. Long-term follow up of patients immunized with AN1792 has unveiled reduced functional decline in antibody responders. Additionally, postmortem histological analysis of brains of AD patients that had been immunized with AN1792 is suggestive of patchy plaque clearance from affected brain tissue.

    The exact mechanisms by which either active or passive immunization approaches may promote plaque clearance are still unknown. The study by Lemere and collaborators certainly paves the way to a better understanding of mechanisms underlying the brain’s ability to recognize plaques, initiate defense mechanisms, and promote plaque clearance. Further experimental work along that line that will help design better and safer strategies to treat AD is warranted.


    . 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.

    . Spontaneous classical pathway activation and deficiency of membrane regulators render human neurons susceptible to complement lysis. Am J Pathol. 2000 Sep;157(3):905-18. PubMed.

    . Activation of complement in the central nervous system: roles in neurodegeneration and neuroprotection. Ann N Y Acad Sci. 2003 May;992:56-71. PubMed.

    . Prominent neurodegeneration and increased plaque formation in complement-inhibited Alzheimer's mice. Proc Natl Acad Sci U S A. 2002 Aug 6;99(16):10837-42. 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.

  2. The paper by Lemere and colleagues provides further evidence for the role that the complement system plays in inflammation generally, and in Aβ phagocytosis particularly. The group developed a double transgenic APP and complement C3-deficient mouse model (APP;C3-/-). The researchers then found, as one might expect, increased Aβ deposition in 17-month-old, but not 8- and 12-month-old mice, and a shift in microglial phenotype. Their results are in accord with previous results of Wyss-Coray et al. (Wyss-Coray et al., 2002), who used the slightly different strategy of developing transgenic mice overexpressing the soluble complement receptor-related protein y (sCrry) to inhibit complement. Based on such data, one might suppose that complement activation, as an important facilitator of Aβ clearance, should be stimulated to provide benefit in AD. Such stimulation can be provided by vaccination with Aβ. For transgenic mice, this is indeed the case: vaccination with Aβ produces a dramatic reduction in the Aβ load.

    But there are crucial differences between these transgenic models and AD. Human Aβ powerfully activates human complement (Rogers et al., 1992) but not mouse complement. This is due to weaker binding of mouse C1q to human Aβ (Webster et al., 1999). The differences are manifested in weaker complement activation overall in transgenic mice. Mouse Aβ deposits are opsonized, but the membrane attack complex is not assembled (Schwab et al., 2004). In human AD, opsonization of Aβ plaques occurs alongside assembly of the membrane attack complex. These events can be visualized damaging nerve fibers in and around senile plaques (McGeer et al., 1989; Akiyama et al., 2000). Thus, complement, though beneficial in AD mouse models, is a potent neurotoxic agent in AD. As a result, vaccination with Aβ in AD can be expected to stimulate more complement activation and to enhance the damaging effects of the membrane attack complex. That was observed in the clinical trial of Wyeth/Elan’s AN1792 vaccine. The trial was canceled, but many more vaccination strategies have since been developed. These can be expected to produce the same kinds of side effects unless the consequences of enhanced assembly of the membrane attack complex can be overcome (McGeer and McGeer, 2003).

    View all comments by Pat McGeer
  3. The complexity of the role of complement in Alzheimer disease has been further reinforced by this publication from Lemere and colleagues reporting increased Aβ deposition in 17-month APP C3-/- relative to the APP (J20) C3 sufficient mice. As noted by others in this forum, the use of complement-deficient and transgenic rodent models has provided support for both detrimental (Fonseca et al., 2004) and beneficial (Maier et al., 2008; Wyss-Coray et al., 2002) roles of specific complement components in AD models. Certainly the increased expression of complement components and its activation products shown by many labs provide evidence that it is present and activated in mouse models and in the human disease. Others have also seen reduction in pathology using complement inhibitors in APP mouse models (Bergamaschini et al., 2004) and more studies should be forthcoming soon.

    In the APPC3-/- animal model reported here by Maier and colleagues, an increase in plaque load, neuronal loss (albeit quite small), and CD45 staining in the plaque area in the APPC3-/- animals relative to the C3 sufficient APP transgenics were observed. This contrasts with a twofold increase in neuronal markers MAP2 and synaptophysin (i.e., neuronal integrity) and a significant decrease in microglial markers (MAC1, F4/80, and IA/IE) in the absence of C1q in the Tg2576 model. In a recently published manuscript, mentioned by van Beek above, we demonstrate that C3 is deposited on the thioflavine-positive plaques in the Tg2576C1q-/- (Zhou et al., 2008) presumably via direct deposition of C3 (i.e., alternative complement pathway activation). This supports the hypothesis that complement components can mediate protective events as well as detrimental events in this disease. At this time the reason for the difference in outcome due to classical or alternative pathway activation is unknown, since kinetics or absolute level of activation cannot be determined by the immunohistochemical analysis performed. Protective effects have also been reported for C1q (Pisalyaput and Tenner, 2008), C3a (Rahpeymai et al., 2006), and C5a (Osaka et al., 1999), although again at this time, the mechanisms underlying each of these scenarios is not yet understood.

    However, it is critical to keep in mind that each model of AD/amyloidosis/tauopathy is on different strains, and subtle differences over the lifetime of the animal may account for some/many of the differences. In addition, compensatory or redundant pathways may develop over the lifetime of any knockout mouse. Furthermore, effects may differ between the early and late stages of the disease (which develop at different ages in the different models).

    As Lemere and colleagues note, the increased pathology in the APPC3-/- animals was seen at the “advanced” age of 17 months for the J20 (APP) mouse, and well beyond the appearance of potentially complement-activating thioflavine-positive plaques, as the 12-month data demonstrate. Thus, the lack of differences in the 12-month animals with and without C3 is also puzzling, and does not seem to support the hypothesis proposed based on the C1q-/- model that C5a (or C3a) plays a significant, if partial, role in recruitment of glia to the plaques (Fonseca et al., 2004). As others have pointed out, it is interesting that the C3-/- APP animals had a cytokine profile characteristic of M2 macrophages, i.e., elevated IL-4 and IL-10. Whether this is due to anti-inflammatory signaling of C1q on plaques in the absence of C5a signaling in the C3-/- APP mice and/or a persistent lack of clearance of Aβ in the absence of opsonizing C3b, or to a predisposition to M2 due to the complete lack of C3 during the lifetime of the animal, remains to be determined.

    Nevertheless, each model provides opportunities for assessing the multiplicity of factors that may influence the pathology and loss of cognitive ability in the human disease. There is a complexity and variety of functions of the complement components (and their inhibitors and cell receptors), and a clear importance of achieving balanced responses in order to permit neuroprotection (and repair?) but prevent neurodegeneration. All this necessitates a thorough understanding of the activities mediated by these components and their activation products. In the end, the optimal treatment for AD may very well consist of a therapy combining targeting of complement components and/or inflammation, amyloid production and/or deposition, and neuroprotective strategies.


    . Peripheral treatment with enoxaparin, a low molecular weight heparin, reduces plaques and beta-amyloid accumulation in a mouse model of Alzheimer's disease. J Neurosci. 2004 Apr 28;24(17):4181-6. 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 deficiency leads to accelerated amyloid beta plaque deposition and neurodegeneration and modulation of the microglia/macrophage phenotype in amyloid precursor protein transgenic mice. J Neurosci. 2008 Jun 18;28(25):6333-41. PubMed.

    . Complement-derived anaphylatoxin C5a protects against glutamate-mediated neurotoxicity. J Cell Biochem. 1999 Jun 1;73(3):303-11. PubMed.

    . Complement component C1q inhibits beta-amyloid- and serum amyloid P-induced neurotoxicity via caspase- and calpain-independent mechanisms. J Neurochem. 2008 Feb;104(3):696-707. PubMed.

    . Complement: a novel factor in basal and ischemia-induced neurogenesis. EMBO J. 2006 Mar 22;25(6):1364-74. PubMed.

    . Prominent neurodegeneration and increased plaque formation in complement-inhibited Alzheimer's mice. Proc Natl Acad Sci U S A. 2002 Aug 6;99(16):10837-42. 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.

  4. This is a very timely and important paper that supports a more sophisticated view of the role of inflammation in amyloid deposition. Ten years ago, most in the Alzheimer research community believed that inflammation was part of the pathogenic mechanism in AD. However, increasingly, literature from the APP mice argues that the classical, M1 form of inflammation with IL-1 and TNFα expression can motivate microglia/macrophages to clear amyloid plaques. Studies ranging from LPS injections to complement inhibition (as in Maier et al.) to IL-1 overexpression demonstrate Aβ reductions associated with microglial activation (DiCarlo et al., 2001; Shaftel et al., 2007). Instead, it appears that it is the alternative, or M2 activation state of microglia, that is associated with toxicity. A key proponent of this perspective has been Carol Colton, who demonstrated increased expression of type 2 markers in AD and APP mouse brains, and enhanced toxicity when iNOS, a traditional M1 protein, was knocked out in APP mice (Colton et al., 2006a; Colton et al., 2006b). It appears that anti-Aβ immunotherapy can shift the microglial activation state from the M1 phenotype to the M2 phenotype, possibly explaining some of its actions.

    Although most data supporting a beneficial role for microglial activation come from mouse models of amyloid deposition, the recent failure of the ADAPT trial of AD prevention with NSAIDs can also be consistent with this perspective (ADAPT Research Group et al., 2007). One of the key links in the argument for inflammation being pathogenic in AD is the reduced rate of dementia in those using NSAIDs. The assumption is this reflects protection due to use of the drug. However, it may equally be the case that the drug use signals individuals who are prone to inflammation. If this were true, drug use may identify these folks, but if anything, to the extent they entered brain, the NSAIDs might thwart the protective effects of the proinflammatory trait. When administered randomly, rather than by self-selection, ADAPT found that NSAIDs increased the risk of AD (in those who did not already have the disease).

    Another intriguing feature of this manuscript is the appearance of neuron loss. APP mice have generally been regarded as having modest neuron loss at best. However, using more sophisticated methods, and genetic manipulations in mice, amyloid-dependent neuron toxicity is emerging in this and other models (O'Neil et al., 2007; Wilcock et al., 2008).

    The real question is the extent to which this will translate to the human circumstance. AD is more than simply amyloid deposition, and the role of inflammation in the other pathologies associated with the disease will be important to determine. For this reason, the tauopathy mice that have pronounced neuronal loss will be important new models in which to evaluate these same manipulations for their potential role in preventing or exacerbating the tauopathy and neurotoxicity.

    View all comments by Dave Morgan
  5. It would be nice to see Andrea Tenner weigh in on this discussion. She has created a mouse with "humanized" C1q. Contrary to expectations, that study indicated there are no important differences in how Aβ interacts with C1q in humans and rodents (Li et al., 2008). She also showed that addition of C1q to cultured neurons could protect against Aβ toxicity (Pisalyaput and Tenner, 2008). The latter, if I may say so, complements the papers discussed above.

    View all comments by Steve Barger
  6. The paper by Maier, Lemere, and colleagues (2008) provides an extension of previous findings by Wyss-Coray, Masliah, and coworkers (2002) showing that inhibiting the complement cascade in aged AD model mice (in the former case by knocking out C3; in the latter by expressing the complement inhibitor, soluble complement receptor-related protein y) promotes cerebral amyloidosis, as judged by Aβ plaque load and biochemical analysis of insoluble Aβ. Maier and colleagues further noted a trend toward increased Aβ levels in blood plasma from cross-bred (APP;C3-/-) mice, reduced NeuN-positivity in crossed mouse hippocampal pyramidal neurons, and an increase in more “anti-inflammatory” microglia. These results add to the emerging complex picture of brain inflammation in the context of AD-like pathology, and offer additional insight into the beneficial role of the complement cascade in these transgenic AD model mice.

    This interesting work by Maier et al. raises a number of questions regarding the interplay between innate immune cells (i.e., microglia and macrophages) and AD-like pathology. It is somewhat unfortunate that multiplex immunofluorescence and/or FACS analysis could not be performed on the CD45-positive microglia/macrophages that the authors show to be increased in brains of their crossed mice. Due to poor antigenicity of paraffin-embedded brain tissue, the authors instead performed Western blots for F4/80 Ag and macrosialin (CD68), and found a trend toward reduction of these proteins in crossed mice. However, it is unclear whether these reductions correspond to the same CD45-positive microglial population or another population of CNS-resident microglia or pericytes, or perhaps invading monocytes/macrophages. Did the authors note vascular cuffing in these crossed mice or otherwise detect presence of blood-borne macrophages?

    While the biological role of F4/80 Ag (a member of the epidermal growth factor transmembrane 7 family) remains unclear (a controversial role has been proposed in promoting T cell tolerance; see van den Berg and Kraal, 2005), macrosialin’s location in late endosomes of innate immune cells, together with its belonging to the LAMP/scavenger receptor family, suggests that it plays a role in phagocytosis (Wong et al., 2005). This agrees well with the findings of Maier and colleagues, if we assume that microglia are at least partially efficient at phagocytosing/clearing Aβ in mouse models of the disease, and that CD68 plays a role in promoting microglial Aβ uptake/clearance. The authors also show significant reduction in iNOS and TNFα, and a significant increase in IL-4 in these crossed mice (IL-10 trended toward increased in crossed mice as well). Based on these alterations in innate immune molecules, the authors conclude that microglial/macrophage activation is shifted toward a more alternate (anti-inflammatory or M2) form. I’m just not sure that we know enough yet about the immunology of these rather enigmatic innate immune cells to begin to classify them into distinct, well-defined, biologically meaningful subgroups. We have previously suggested that microglia/macrophage “activation” is better defined in terms of a continuum of responses, ranging from pro-inflammatory/anti-phagocytic to anti-inflammatory/pro-phagocytic, including intermediate phenotypes that have properties of both ends of the spectrum (Town et al., 2005). Of course, such taxonomy is further complicated by multiple functionally distinct subpopulations of microglia/macrophages, defined by markers such as CD11b and Ly-6C (à la Geissmann and colleagues, 2003).

    I wanted to comment on a few points raised by Dave Morgan regarding LPS administration to AD model mice and also the ADAPT trial. It is interesting to note that an acute (single intra-hippocampal injection) of LPS increased reactive microglia (as defined by MHC II immunoreactivity) in AD model mice and promoted reduction of Aβ deposits (DiCarlo et al., 2001). However, in AD model mice a chronic treatment regimen with LPS (weekly for 12 weeks) actually increased both microgliosis (by F4/80 Ag) and astrocytosis (by GFAP) and resulted in increases in cerebral Aβ1-40 and Aβ1-42, intraneuronal Aβ, and F4/80 Ag-positive microglia surrounding neurons containing Aβ deposits (Sheng et al., 2003). It is unclear how AD patients would react to intrathecal/intracranial injections of LPS, but certainly, given the aseptic meningoencephalitis that occurred in just over 5 percent of AD patients given the AN1792 vaccine, the safety of such an approach would need to be thoroughly evaluated in non-human primates before consideration in humans.

    It deserves mentioning that the ADAPT trial was prematurely halted in late 2004 for a number of reasons, including an investigation launched by the public watchdog group “Public Citizen” and the “coxib” (Vioxx [naproxen]) cardiovascular side effects scare, which has made interpretation of this incomplete RCT complex (because treatment was suspended just after a few years following initiation; ADAPT Research Group, 2007). Interestingly, a recent epidemiologic study has carefully investigated “confounding by indication” (here, the idea that the inverse risk relationship between NSAIDs and AD was owed to arthritis, a presumed surrogate for NSAID use), and the authors found that, even after correcting for arthritis incidence, the inverse NSAID-AD risk relationship holds up (Szekely et al., 2008).

    Finally, I wanted to mention our recent work targeting TGFβ signaling on peripheral macrophages in an AD mouse model, where we found dramatic (>90 percent by some methods) reduction of cerebral amyloidosis accompanied by 1) increased infiltration of blood-borne macrophage-like cells, 2) increased CD45, CD68, CD11b, and CD11c immunoreactivity (but cells were mostly negative for Ly-6C), and 3) reduced TNFα and increased IL-10 brain levels (Town et al., 2008). Our results suggest that it is possible to target a subpopulation of macrophages to productively clear Aβ deposits without provoking potentially damaging brain inflammation. Of course, we will have to wait and see if/when this therapeutic modality makes it to human clinical trials before we know whether we can “have our cake and eat it, too.”

    In summary, Maier and colleagues should be credited for contributing to our understanding of the potentially protective role of C3 in AD-like pathology. As we learn more about the biological roles that these various “activated” microglial products play, we will be better equipped to target the correct phenotype for eventual AD therapy.

    View all comments by Terrence Town