Complement activation and plaque formation are mutually promoting mechanisms. Aggregated Aβ efficiently binds C1q, activating the classical complement pathway [10
], and this process further enhances Aβ aggregation and fibril formation [30
]. Plaque-associated complement activation has been suggested to be neurotoxic via bystander lysis [31
] and also, indirectly, by chemotactic attraction of microglia [32
], which when activated can secrete toxic reactive oxygen species (ROS) [33
]. However, because aggregated Aβ alone can also activate microglia [34
] and generate ROS [35
], the contribution of complement activation to Aβ's neurotoxicity (and to cognitive decline, to the extent that it is associated with plaques) is unclear. We performed regression analysis in an effort to determine the relationship between plaque-associated complement activation and cognition, but the analysis was not informative because of multicollinearity between Bielschowsky-stained and complement-stained plaques. This problem would be present regardless of the numbers of specimens in this study, because statistical power has no bearing on multicollinearity; when two variables are strongly correlated, the strength of this association is independent of the group sizes for the variables [29
]. A post-hoc power analysis indicated that this study was highly powered: the range of the power for detecting between-group differences with a 0.05 significance level, based upon the standard deviations within each of the three groups, was 78.4% to 98.3% for iC3b staining, 72.7% to 99.98% for C9 staining, and 98.8% to 99.997% for Bielschowsky staining (data not shown).
The aggregated, fibrillar Aβ in neuritic plaques is considered to be the major factor through which plaques activate complement [10
], although nonfibrillar Aβ in diffuse plaques may also be able to do so ([37
]; however, see Nybo et al. [38
]). Previous studies have detected early complement activation proteins on both diffuse and neuritic plaques [1
], but late-stage complement activation on plaques has been demonstrated primarily, or exclusively, on neuritic plaques [2
]. Little or no late-stage complement activation has been reported on plaques in non-demented subjects [3
]. Our results differ somewhat from these studies in that, in each of the three groups of specimens, plaque staining was detected for C9 (late-stage activation) as well as for iC3b (early-stage activation); in addition, C9 staining was occasionally observed on diffuse plaques, as well as more frequently on neuritic plaques. The slight increases in C9-stained vs. iC3b-stained plaques in MCI and AD specimens are likely to be due to differences in antibody sensitivity (i.e., the polyclonal anti-C9 may detect more plaques than monoclonal anti-iC3b). Detection of C9 on some diffuse plaques suggests that complete complement activation may, in some cases, occur on plaques even in the absence of fibrillar Aβ; whether this contributes to plaque "progression" from diffuse to neuritic is unknown. The similar numbers of Bielschowsky-stained and complement-stained plaques in AN and MCI specimens suggest that neither increased plaque formation nor increased plaque-associated complement activation is required for the development of early cognitive deficits. Our finding of similar plaque counts between AN and MCI subjects is in agreement with a recent study by Petersen et al. [42
In addition to plaques, complement can be activated in the brain by NFTs [43
]. NFT density and distribution are more strongly correlated than plaques with dementia severity [44
]. There were few or no NFTs in most AN and MCI specimens in this study, in agreement with reports that only in severe dementia are large numbers of neocortical NFTs typically present [46
]. To determine the relationship between NFT complement activation and cognition, examination of limbic structures such as the hippocampus and entorhinal cortex, in which NFT counts increase early in the disease process [47
], would be required. The resulting analysis would still, most likely, be confounded by multicollinearity. Zanjani et al. [4
] found little complement staining on NFTs in the entorhinal cortex except in late-stage AD, suggesting that complement activation on NFTs during the development of AD may be minimal.
Neuronal complement staining was present in many of the specimens in this study, but there were no differences between groups for this immunoreactivity and it was poorly correlated with cognitive measures. It is unclear whether this staining indicates complement activation on neurons, neuronal synthesis of native complement proteins, or both. The monoclonal anti-iC3b antibody used in this study is stated by its manufacturer, Quidel, to be iC3b-specific, and to not recognize native C3. This suggests that neuronal iC3b staining is likely to represent complement activation rather than native complement protein synthesis. If neuronal C9 immunoreactivity also represents complement activation (i.e., C5b-9 deposition), then it is unclear why the MAC should be present on normal-appearing neurons. This staining could represent sublytic concentrations of C5b-9, which have been suggested to be neuroprotective [8
]. Our findings do not point to complement activation as a direct cause of neuronal injury in AD, although, as stated above, complement-mediated chemotactic attraction of microglia and activation of these cells could indirectly result in neuronal damage due to microglial ROS secretion. Our detection of complement immunoreactivity on normal-appearing neurons differs from an earlier study [48
] in which C1q and MAC staining was reported in the AD brain on degenerating neurons. The reasons for these conflicting results are unknown, but may be related to differences in the antibodies employed for complement staining and/or procedures for processing the brain specimens.
Our results demonstrate the limitations of studies with human brain specimens for determining the role of complement activation in AD. Although the extent of this process in the brain correlates strongly with cognitive loss, differentiating its effects from the effects of AD neuropathology on cognition is difficult. Studies in animal models of AD may eventually be of value in resolving this issue if MAC deposition can be demonstrated to the same extent in these models as in AD. A trial in AD patients with an inhibitor of late-stage complement activation such as Eculizumab (Soliris, Alexion Pharmaceuticals) [49
], a humanized anti-C5 monoclonal antibody which prevents formation of C5b-9, might also help to resolve this issue.