Background
Alzheimer's disease is a progressive neurodegenerative dementia of the elderly characterized by a well defined pathology that includes accumulation of β-amyloid in plaques, hyperphosphorylated tau that ultimately forms neurofibrillary tangles, and neuronal loss [
1]. In addition to these hallmarks, a prominent inflammatory reaction, characterized by the presence of reactive glia associated with the fibrillar plaques, upregulation of several complement proteins [
2‐
5] including local synthesis of the components [
6,
7] is observed. C1q is associated with fibrillar plaques as well as tangles [
3,
8], and the presence of C5b-9 associated with dystrophic neurites in plaques and with tangles [
9] indicates that complement is fully activated in AD [
10]. These
in vivo observations, supported by the
in vitro studies demonstrating that fibrillar β-amyloid can activate the classical [
11,
12] and alternative [
13] complement pathways and that the complement activation fragment C5a is chemotactic for microglia [
14], led to the hypothesis that the complement activation triggered by fibrillar ß-amyloid contributes to the inflammatory reaction that can play a detrimental role in the progression of the later stages of Alzheimer's disease [
15].
Both a genetic and a pharmacological approach have been used to investigate this hypothesis. First, a Tg2576 transgenic mouse model of AD was crossed to a C1q-/- mouse to generate the APPQ-/- mouse which lacks C1q (the first component of the classical complement pathways). We observed a decrease in reactive glia associated with fibrillar amyloid plaques in the APPQ-/- compared to the APP mice at all ages analyzed. In addition, the APPQ-/- mice showed greater synaptophysin (SYN) and MAP-2 staining relative to the APP mice indicating a preservation of neuronal integrity [
16]. In a second approach, Tg2576 mice were treated with a specific antagonist for CD88, a receptor for the complement activation fragment C5a, for three months. The treated animals showed a decrease in plaque and glia pathology, an increase in the SYN staining and cognitive improvement [
17]. 3xTg mice, a mouse model of AD that develops neurofibrillar tangles as well as plaques, similarly treated also showed a decrease in plaques, reactive glia and, in addition, a decrease in hyperphosphoryated tau [
18]. These results support the hypothesis that complement activation plays a detrimental role in AD since inhibiting classical complement activation or blocking the downstream pathway by inhibiting C5a/C5aR interaction renders a substantial improvement in pathology and behavior of these animals.
Since it has also been reported that C1q can bind to hyperphosphorylated tau and activate complement
in vitro [
8], the contribution of complement activation on the kinetics of appearance and accumulation of both amyloid plaques and phosphorylated tau, was assessed in the 3xTg and in the 3xTg lacking C1q (3xTgQ-/-) at different ages. In addition, a caveat for the use of standard mouse models for studying the involvement of complement in human AD is the reported weak hemolytic activity of mouse complement [
19]. While the basis for this apparent deficiency seen in
in vitro assays has not been delineated, one possible consequence
in vivo would be a lower or less efficient C5 cleavage and thus a lower generation of both C5a, a proinflammatory peptide, and the membrane attack complex composed of C5b, C6, C7, C8 and C9 (C5b-9). This would result not only in a decrease in bystander damage of cells in the mouse system, but also decreased proinflammatory activity, relative to that in the human system. The BUB/BnJ strain of mice has higher complement hemolytic activity, as measured
in vitro, than that of current transgenic mouse AD models [
15,
20‐
22]. To determine if this parameter translates into differences in pathology in the 3xTg model, the 3xTg was backcrossed (N = 6) to the BUB strain and to BUBQ-/-, and the development of pathology compared with that of the 3xTg on a mixed C57BL/6 background.
The results show that the presence or absence of C1q generated no difference in plaque or tangle pathology or inflammatory response in this 3xTg animal model. Interestingly however, C1q and C4 were not detected associated with the plaques in the 3xTg models, in contrast to prominent deposition in other transgenic models (Tg2576 and Arc48). These data indicate that the classical complement pathway may not contribute to the generation of C5a and inflammation in the 3xTg model which develops pathology more slowly than other models of this disease. However, an increase in pathology was observed in the 3xTg backcrossed to BUB background relative to the initial 3xTg. Since we previously demonstrated a suppression of pathology in this model with a C5a receptor antagonist, these results suggest that activation of the alternative complement pathway or the presence of other C5 cleaving enzymes may be the mechanism by which the proinflammatory peptide C5a is generated in this model.
Discussion
The presence of complement components and indicators of inflammation in AD brain suggests that complement activation may contribute to the progression of AD, and if so, can be a novel therapeutic target.
In vitro data, demonstrating the ability of fibrillar ß-amyloid to activate both the classical and alternative pathway is consistent with complement pathway activation by fibrillar amyloid plaques
in vivo [
12,
13]. In our previous publication, treatment of 3xTg mice with a C5a receptor antagonist [C5a is a proinflammatory peptide that can be generated as a result of complement activation] showed marked decreases in both amyloid plaque and hyperphosphorylated tau pathology [
17]. In the present study the contribution of complement to pathology was defined in the 3xTg model. To eliminate classical complement pathway activation, mice were first bred to generate C1q-deficient 3xTg mice. Conversely, to enhance complement activity, 3xTg were backcrossed 6 generations to the BUB background, a strain with higher levels of serum complement hemolytic activity. No differences in thioflavine plaques, reactive glia, hyperphosphorylated tau, or selected inflammatory markers were detected between the C1q -sufficient and -deficient 3xTg mice on either background, which is in contrast to previous studies with Tg2576 lacking C1q that had shown reduced gliosis and increased neuronal integrity [
16]. Consistent with the lack of C1q-dependent pathology in the 3xTg, there were no early components of the classical complement pathway associated with the plaques, again in sharp contrast to the Tg2576 [
16] and the Arc48 model (Figure
5). Therefore, the positive effect of the specific antagonist for the C5aR (PMX205) in suppressing pathology in the 3xTg animals [
17] suggests that either alternative pathway of complement activation is the predominant and sufficient mechanism of C5a-generation in this animal model (as the alternative pathway is independent of C1q), that the newly described C3-independent enzymatic cleavage of C5 [
35] may be the source of the C5a contributing to pathology in these mice or that the protective effect of PMX205 is independent of C5a effects.
The presence of properdin, an alternative complement pathway component that can stabilize the C3 convertase as well as initiate alternative pathway activation [
36], associated with amyloid plaques in 3xTg as well as 3xTgQ-/- animals supports the possibility that the alternative pathway is activated in this AD model. However, an additional contribution of the C3-independent generation of C5a cannot be ruled out at this point.
A role for the C5a activation fragment of complement in many inflammatory disorders including neurodegeneration has been well documented (reviewed in [
37]). The abrogation of AD pathology in the 3xTg model backcrossed to the C5 deficient FVB strain (D.A.Morrissette, PhD dissertation, 2009, UCI) as well as the delay in amyloid accumulation in the C5-deficent mice containing the human APP gene under its own promoter [
38] supports, though does not prove, a contribution of C5a. While it is possible that PMX205 is inhibiting other receptors, PMX53, the close homolog of PMX205, was screened for inhibition of 44 different receptors including 4 ion channels and two transporter proteins and shown to exert inhibition with only 4 other receptors and only at a minimum of 3 fold higher concentration than that which inhibits CD88/C5aR [
39]. Treatment with PMX205 in mice had no effect on dampening leukocyte migration to the CNS in response to intracranial inoculation with a neurotrophic coronavirus (T.E.Lane, UC, Irvine, personal communication) [
17], emphasizing the variety of selective chemotactic and inflammation inducing mechanisms available that are not inhibited by this antagonist.
A second major finding of this study is that thioflavine positive plaques and glial accumulation was detected earlier (by approximately 2 months) in the 3xTgBUB mice relative to the 3xTg (Figure
1 and
2), consistent with the possible greater generation of detrimental C5a in the BUB background. Since there was no evidence of C1q and C4b associated with plaques in the 3xTg BUB, similar to the mixed background 3xTg, greater alternative pathway complement activation or greater C3-independent C5 cleavage (both of which can lead to higher C5a generation) in BUB may be the basis for this accelerated pathology. Although C3b stably bound to plaques was not detected in the 3xTg, it is possible that low levels of alternative pathway activation could occur, but that amplification is highly regulated (such as by the complement regulatory protein Crry previously shown to be present at high levels in mice [
32]) resulting in very low surface bound C3b that may be difficult to detect by any of antibodies used.
These data, and other results from studies by us and others, suggest that the role of complement in AD is complex, with evidence for both detrimental and beneficial functions [
16,
17,
40‐
42]. For example, transgenic over expression of the murine complement inhibitor of C3 (Crry) or generation of a C3 deficient APP mice resulted in enhanced pathology in these mouse models suggesting a protective contribution of complement [
41,
43] (possibly due to the opsonic effect of C3b for amyloid or cellular debris that is missing when classical and alternative pathways are blocked by C3 inhibition or deletion). This protective role is also consistent with the recent report demonstrating a correlation between induction of early components of complement and suppression of Aß deposition in the TgCRND8 AD mouse model [
44]. However, deletion of C1q in the Tg2576 and APPPS1 models of AD supported a detrimental role for complement activation since the Tg2576C1q-/- and APPPS1C1q-/- mice showed less reactive glia surrounding plaques and increased synaptophysin than the Tg2576 or APPPS1 [
16]. The protection given by the lack of C1q (and thus lack of the classical pathway for complement activation) was substantial but not complete, suggesting that the alternative pathway and/or other non complement mediated events contribute to the inflammatory reaction around the plaques. The deposition of C3b on the plaques of Tg2576C1q-/- in the absence of C1q and C4 demonstrated that the alternative pathway is activated in the Tg2576C1q-/- mice [
33]. The presence of properdin on plaques of all models assessed here indicates that either or both complement pathways are activated in AD mice.
The 3xTg is the first transgenic model of AD in which early components of classical (C1q and C4) complement pathway have not been detected associated with thioflavine positive plaques. Robust deposition of C1q and C4 in the Tg2576 mice [
33] and the APP23 mice [
32] or C1q in the APP/PS1 [
31] on plaques has been demonstrated. In addition, the mouse model for cerebral microvascular amyloid showed increases in C1q, C3 and C4 in areas with fibrillar amyloid deposits [
45]. We cannot rule out the possibility that there is transit binding of C1q in C1 (and thus limited activation of the classical complement pathway) on the plaques of the 3xTg or increased plaque binding of complement regulators such as C4BP (known to bind to plaques in AD brain) [
46]. In any case, the possibility that the complement pathways are more robust in the Tg2576, APP23, (and APP/PS1) provides a plausible explanation as to why the progression of the pathology is faster in those strains than in the 3xTg. Similarly, the accelerated progression of pathology in the 3xTg on BUB background is consistent with a role for complement in determining the rate of progression of the disease. The different C3 and C5 convertases involved in the classical and alternative pathways and/or the balance of both pathways might contribute differently to the extent of downstream activation (C5a generation) and to the resulting pathology in these models. The potential polymorphism in CR1 associated with human AD also suggests a point of control of complement activation, since CR1 is a critical regulator of C3 convertase activity in humans, and remains to be further investigated.
Interpretation of the immunohistochemical data assessing the presence of C3 in inflammatory disease models requires an understanding of the various forms of C3 being recognized by the antibodies used. While C3 has been shown to be synthesized by astrocytes and microglia in culture [
47] and in brain tissue by in situ hybridization in neurodegenerative diseases or with injury or inflammation [
7,
48], reports of the association of both native and activated C3 differ among mouse models of neurodegeneration. For example, plaque labeling at different levels, but not astrocyte labeling was observed in the APP23 model with the polyclonal anti C3d antibody [
32,
49], while weak to absent plaque staining in Tg2576 mice was reported with this antibody [
49]. Here, the polyclonal anti human C3/C3d antibody labels astrocytes (probably via epitopes on C3d exposed in intact C3) in the Arc48 model and in the Tg2576 and 3xTg. In addition, the monoclonal anti C3 antibody, clone 11H9 (that recognizes intact C3), labeled only astrocytes in 3xTg and Tg2576 mice, suggesting that this C3 was newly synthesized, uncleaved C3 (consistent with the identification of C3 as an acute phase protein) [
50]. The polyclonal C3/C3c antibody that recognizes the C3c region within C3 (C3c can also be dissociated from the thioester surface bound C3d upon a second cleavage of C3b by Factor I), labeled mainly astrocytes in all three models. The lack of astrocyte labeling by the 2/11 anti C3b/iC3b/C3c that recognizes only cleaved C3, supports the conclusion that the astrocyte labeling in these models represents native, uncleaved C3. Importantly, activated C3 neoepitopes detected by the 2/11 monoclonal antibody were associated with plaques in the Arc48 model but not on the plaques in the 3xTg. Plaques in Arc48 were also stained with the polyclonal anti C3/C3d antibody likely reflecting reactivity with activated C3b cleaved to C3d on those plaques. Overall, these results are consistent with differential activation of C3 in these mouse models. [It should be noted, while we consistently observed astrocytes labeling with all the C3 antibodies tested (except 2/11, which is specific for cleaved C3) in all three mouse models, in another murine model of neurodegeneration C3 was reported exclusively in microglia [
45].]
In conclusion, there is substantial evidence showing C1q, C4 and C3 strongly associated with fibrillar plaques in human AD brain and Down's Syndrome with AD [
3,
5,
51,
52] even in early stages of the disease correlating with the appearance of fibrillar amyloid [
2]. Complement proteins of the alternative pathway are also detected associated with plaques in the human disease [
53]. The production of complement factors by brain cells [
6,
7,
54] and the presence of the terminal complement membranolytic complex C5b-9 detected on plaques and tangles in human AD [
9] provide further evidence that complement is present and fully activated in AD brain and therefore might contribute to the enhancement of neurodegeneration at later stages of the disease when fibrillar plaques are present. Mouse models of AD exhibit to some degree many of the pathological features of AD [
55,
56]. However, the absence of definitive evidence for some late complement factors [
32,
49] and/or differences in the ratios of complement components and complement inhibitors between AD and AD animal models have been reported [
32] suggesting that there might not be quantitatively comparable complement activation in mouse models as in human AD. Here, the presence of properdin in all mouse models of AD provides evidence for the activation of the alternative complement pathway, consistent with the benefit demonstrated in the 3xTg as well as Tg2576 mice of treatment with an antagonist of the proinflammatory C5a receptor [
17]. Nevertheless, the development of new mouse models that more closely mimic the human system in all aspects of the disease will further improve the ability to assess the contribution of complement to AD neuropathology, define the targets most likely to promote beneficial effects and/or prevent detrimental activities [
57‐
59] and aid in developing treatments for this devastating human disease.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MIF carried out part of the IHC experimental work and statistical analysis, prepared the figures and drafted the manuscript. SHC generated and genotyped the animals used in the study, processed the tissue, and performed some of the IHC experiments and statistical analysis. AB processed some of the tissue used and performed some IHC experiments and contributed to the analysis. YK provided the anti properdin antibody, and design of some IHC experiments. MEB designed and analyzed the qPCR experiments and contributed to the manuscript, and DGP performed qPCR experiments and contributed to the analysis. AJT designed the study, contributed to the analysis of all data and to the preparation of the manuscript. All authors read and approved the final manuscript.