Research ArticleAlzheimer’s Disease

Complement C3 deficiency protects against neurodegeneration in aged plaque-rich APP/PS1 mice

See allHide authors and affiliations

Science Translational Medicine  31 May 2017:
Vol. 9, Issue 392, eaaf6295
DOI: 10.1126/scitranslmed.aaf6295

Avoiding complements

Complement C3 is an immune molecule that protects against pathogens and plays a role in refinement of the developing visual system by removing weak nerve connections (that is, synapses). C3 is up-regulated in Alzheimer’s disease and, therefore, may contribute to the synapse loss that underlies cognitive decline. Shi et al. now report that an aged transgenic mouse model of Alzheimer’s disease that lacks C3 was protected against synapse loss and cognitive decline even in the presence of Aβ plaques, possibly by altering the glial response to Aβ deposition. Thus, modulation of complement signaling may have potential as a new therapeutic strategy for Alzheimer’s disease.

Abstract

The complement cascade not only is an innate immune response that enables removal of pathogens but also plays an important role in microglia-mediated synaptic refinement during brain development. Complement C3 is elevated in Alzheimer’s disease (AD), colocalizing with neuritic plaques, and appears to contribute to clearance of Aβ by microglia in the brain. Previously, we reported that C3-deficient C57BL/6 mice were protected against age-related and region-specific loss of hippocampal synapses and cognitive decline during normal aging. Furthermore, blocking complement and downstream iC3b/CR3 signaling rescued synapses from Aβ-induced loss in young AD mice before amyloid plaques had accumulated. We assessed the effects of C3 deficiency in aged, plaque-rich APPswe/PS1dE9 transgenic mice (APP/PS1;C3 KO). We examined the effects of C3 deficiency on cognition, Aβ plaque deposition, and plaque-related neuropathology at later AD stages in these mice. We found that 16-month-old APP/PS1;C3 KO mice performed better on a learning and memory task than did APP/PS1 mice, despite having more cerebral Aβ plaques. Aged APP/PS1;C3 KO mice also had fewer microglia and astrocytes localized within the center of hippocampal Aβ plaques compared to APP/PS1 mice. Several proinflammatory cytokines in the brain were reduced in APP/PS1;C3 KO mice, consistent with an altered microglial phenotype. C3 deficiency also protected APP/PS1 mice against age-dependent loss of synapses and neurons. Our study suggests that complement C3 or downstream complement activation fragments may play an important role in Aβ plaque pathology, glial responses to plaques, and neuronal dysfunction in the brains of APP/PS1 mice.

INTRODUCTION

The complement cascade, of which C3 is the central molecule, plays a pivotal role in the immune system. Upon activation, classical complement initiating protein C1q binds to and coats dead cells, debris, or pathogens, triggering a protease cascade, leading to the activation of complement protein C3, which opsonizes material for elimination by triggering phagocytosis of macrophages, or leading to the formation of the membrane attack complex, causing cell lysis (1). In addition, complement has been shown to play a role in microglia-mediated synapse elimination in the developing visual system to refine neuronal connections (24). We recently demonstrated that C3 deficiency spared age-dependent synaptic and neuronal loss in a region-specific manner and protected against cognitive impairment in normal aging of C57BL/6 mice (5). Synapse loss occurs early in Alzheimer’s disease (AD) and correlates with cognitive decline (6, 7). Complement cascade components, including C3, are up-regulated and associated with amyloid plaques that typically contain dystrophic neurites in human AD (8, 9); however, the role of C3 in plaque-related synapse loss and neurodegeneration in AD is unknown. Recently, we reported that in early-stage AD in mice, pre-plaque Aβ oligomer–induced hippocampal synapse loss was rescued by both genetic deletion of complement and pharmacological treatment with an anti-C1qa blocking antibody (10). However, it is not known whether blocking the complement cascade also protects against cognitive impairment and neurodegeneration at later stages of AD in mice with accumulation of amyloid plaques in the brain.

Several complement components, including C1q, activated C3 (C3b, C3c, and C3d), and C4 (8, 9, 11, 12), may be produced by glia surrounding plaques in the AD brain. Aβ peptides up-regulate C3 production by microglia and astrocytes in vitro (13, 14) and also increase C3 expression and deposition in AD mouse brain in vivo (10, 1518). Complement can recruit and activate microglia around fibrillar Aβ deposits (19) and mediates, in part, the uptake and clearance of Aβ by the interaction of complement component iC3b with its receptor CR3 (CD11b/CD18) on the surface of microglia (20). We previously reported that aged C3-deficient J20 hAPP transgenic mice had a higher plaque load, slightly reduced neuron number, and altered macrophage polarization compared to age-matched J20 mice (21). Others reported that C3 inhibition by expression of soluble complement receptor–related protein y (sCrry) in J20 hAPP mice resulted in increased Aβ accumulation, neuronal degeneration, and reduced microglia activation (22). In contrast, C1q deficiency resulted in protection of synapses and neuropathology with decreased glial activation surrounding plaques in two other AD amyloidosis models, Tg2576 and APP/PS1 mice (23). Inhibition of C5a, a C3 downstream activation fragment, using a C5aR antagonist resulted in rescue of a hippocampal-dependent memory task (24). However, none of these previous reports addressed whether complement C3 plays a role in cognitive health. Hence, the consequences of C3 in the aging AD brain remain unclear.

Here, we generated APP/PS1;C3 KO to ask whether C3 plays a role in plaque-related neuropathological changes and cognitive decline. We demonstrate that C3 deficiency in APP/PS1 mice protected against age- and plaque-related synapse and neuron loss, decreased glial reactivity, and spared cognitive decline, despite an increased plaque burden in the mouse brain.

RESULTS

Sparing of cognitive decline in plaque-burdened aged APP/PS1;C3 KO mice

Aged APP/PS1 mice (>15 months of age) have been shown to exhibit hyperactivity in the open-field test and impaired learning and spatial memory in the radial arm/Morris water maze test compared with nontransgenic mice (25). To determine whether C3 deficiency affects age-related cognitive changes in mice, we assessed learning and memory in 16-month-old wild-type (WT), APP/PS1, APP/PS1;C3 KO, and C3 KO mice. We used the water T-maze test, which measured spatial learning and memory as the animals learned the location of a hidden platform (acquisition) and cognitive flexibility during reversal learning (26). Although APP/PS1 mice showed significant impairment in finding the platform on acquisition day 2 compared to WT mice (P < 0.05; Fig. 1A), there were no significant differences in the percent of mice that reached criterion (≥80% correct choices on each individual day) between genotypes on day 6, indicating that all groups eventually learned the task (Fig. 1A). Acquisition performance in APP/PS1;C3 KO mice was between that of WT mice and APP/PS1 mice, indicating a nonsignificant trend, suggesting that APP/PS1;C3 KO mice learned the location of the hidden platform more quickly than did APP/PS1 mice. During the reversal trials, in which the platform location was switched, APP/PS1 mice demonstrated cognitive impairment by performing significantly worse than WT mice on days 3 to 5 (Fig. 1A). Reversal learning in APP/PS1;C3 KO mice was similar to that of WT mice and C3 KO mice, and was significantly better than that of APP/PS1 mice on days 4 and 5 (Fig. 1A), suggesting that aged APP/PS1;C3 KO mice had enhanced spatial learning and cognitive flexibility compared to APP/PS1 mice.

Fig. 1. APP/PS1;C3 KO mice show improved cognitive flexibility (reversal) compared to APP/PS1 mice at 16 months of age.

(A) Percent of mice that reached criterion (≥80% correct choices on each individual day) in the water T-maze test. Compared to WT mice, APP/PS1 mice were impaired in acquisition (day 2) and reversal learning and memory (days 4 and 5) (*P < 0.05, **P < 0.01). APP/PS1;C3 KO mice performed significantly better than did APP/PS1 mice (##P < 0.01), but similar to WT and C3 KO mice, in the reversal test on days 4 and 5, suggesting better cognitive flexibility in APP/PS1;C3 KO mice compared to APP/PS1 mice. (B) In total, fewer APP/PS1 mice reached the reversal criterion (≥80% correct choices over two consecutive days) (*P < 0.05), whereas the percent of WT, C3 KO, and APP/PS1;C3 KO mice that reached criterion in the reversal test was higher compared to APP/PS1 mice (***P < 0.001), indicating that C3 deficiency in APP/PS1 mice had both age-dependent and AD-related effects (WT, n = 13; APP/PS1, n = 11; APP/PS1;C3 KO, n = 10; C3 KO, n = 11). Tests were assessed using one-way analysis of variance (ANOVA) followed by Fisher’s protected least significant difference post hoc test.

The percent of mice that reached reversal criterion across all days (≥80% correct choices on two or more consecutive days) was significantly higher in WT C3 KO and APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.001; Fig. 1B). Fewer APP/PS1 mice reached reversal criterion compared to APP/PS1;C3 KO mice, whereas C3 KO, APP/PS1;C3 KO, and WT mice achieved equivalent reversal criteria, suggesting that C3 deficiency in the APP/PS1 mice protected against both age-related and AD-related cognitive impairment.

Consistent with previous reports (25), we found that 16-month-old APP/PS1 mice traveled a significantly greater distance in the open-field test compared to WT mice (fig. S1A). Overall, C3 deficiency in APP/PS1 mice had no effect on distance traveled in the open-field test (fig. S1A). We also conducted the elevated plus maze test to examine anxiolytic-like behavior in 16-month-old mice and found that APP/PS1;C3 KO mice showed greater anxiolytic-like behavior compared to APP/PS1 mice (fig. S1, B and C) that was not due to differences in locomotor activity (fig. S1D). These behavioral studies suggest that cognitive decline was at least partially spared in APP/PS1;C3 KO mice.

Elevated cerebral Aβ plaque deposition and insoluble Aβ in APP/PS1;C3 KO mice

To determine whether memory improvement in APP/PS1;C3 KO mice was due to decreased Aβ, we examined the Aβ plaque load of 4- and 16-month-old APP/PS1;C3 KO and APP/PS1 mice. No significant differences in Aβ plaque burden were observed between 4-month-old APP/PS1 and APP/PS1;C3 KO mice at the earliest stages of plaque deposition (fig. S2). However, by 16 months of age, Aβ plaque deposition was elevated in both the prefrontal cortex and hippocampus of APP/PS1;C3 KO mice compared to age-matched APP/PS1 mice (Fig. 2, A and B, and fig. S3). In the prefrontal cortex, plaque deposition was significantly increased by 167.37% for Aβ starting at the first N-terminal residue (Aβ1–x; detected by Aβ N-terminal antibody, 3D6) (P < 0.01; Fig. 2A). In the hippocampus, plaque deposition was significantly increased by 63.4% for Aβ ending at residue 42 (the longer, more pathogenic form; Aβx–42) (P < 0.01), by 64.1% for Aβ ending at residue 40 (the shorter, more abundant but less pathogenic form; Aβx–40) (P < 0.05), and by 65.4% for Aβ1–x (P < 0.05; Fig. 2B). Furthermore, we quantified the area of immunoreactivity for different sizes of Aβx–42–labeled plaques. Larger plaques (>50 μm) were markedly increased by 127.0% (P < 0.01), whereas small (<20 μm) and medium (20 to 50 μm) plaques were modestly but significantly increased by 28.7% (P < 0.05) and by 56.0% (P < 0.01), respectively, in the hippocampus of APP/PS1;C3 KO mice compared to APP/PS1 mice (Fig. 2, C and D). Thioflavin S–positive staining of fibrillar amyloid deposits was significantly elevated in the hippocampus of APP/PS1;C3 KO mice (P < 0.05; Fig. 2, C and E).

Fig. 2. Increased Aβ plaque load in 16-month-old APP/PS1;C3 KO mice.

(A and B) Quantification of Aβx–42, Aβx–40, and Aβ1–x immunoreactivities within a designated region of interest (ROI) revealed increased plaque burden in the cortex (A) and hippocampus (B) of 16-month-old APP/PS1;C3 KO mice compared to APP/PS1 mice (*P < 0.05, **P < 0.01 versus APP/PS1 mice, independent unpaired t test per Aβ species; n = 9 mice; six equidistant planes, 150 μm apart). (C) Aβx–42–immunoreactive and thioflavin S–positive plaques were higher in the hippocampus of C3-deficient APP/PS1 mice versus APP/PS1 mice. White circles indicate large plaques (>50 μm); black arrows indicate medium-sized plaques (>20 but <50 μm). Scale bars, 50 μm. (D) Quantification of small, medium, and large hippocampal plaques confirmed an increased plaque load in APP/PS1;C3 KO mice, especially for large plaques (*P < 0.05, **P < 0.01, independent unpaired t tests per plaque size category; n = 10). (E) Quantification of thioflavin S (Thio S) in hippocampus confirmed an increase in fibrillar plaques in APP/PS1;C3 KO mice (*P < 0.05 versus APP/PS1 mice, unpaired t test; n = 6; three equidistant planes 300, μm apart). (F) Increased insoluble cerebral Aβx–40 and Aβx–38 were found in APP/PS1;C3 KO mice compared with APP/PS1 mice (*P < 0.05, independent unpaired t tests per Aβ species; n = 8).

Enzyme-linked immunosorbent assay (ELISA) measuring Aβ in mouse hemibrain homogenates revealed that guanidine-soluble (T-per–insoluble) Aβx–40 and Aβx–38 were significantly higher in APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.05; Fig. 2F). These data are in agreement with the corresponding increase in Aβx–40 and Aβ1–x plaque load. These results indicate that the absence of a key complement factor, C3, or its downstream activation products in APP/PS1 transgenic mice might result in an age-dependent increase in the accumulation of cerebral Aβ plaques and insoluble Aβ in aged APP/PS1 mice.

Altered plaque-associated gliosis in aged APP/PS1;C3 KO mice

Microglia and astrocytes are frequently associated with plaques in AD brain. Therefore, we examined colocalization of these cells with plaques in mouse brain by immunostaining for Iba-1 and CD68 (markers of microglia and macrophages) as well as glial fibrillary acidic protein (GFAP; a marker of astrocytes). Immunoreactivities to Iba-1 and GFAP were similar between APP/PS1;C3 KO and APP/PS1 mice at postnatal day 30 (P30) and 4 months of age (fig. S4). However, 16-month-old APP/PS1 transgenic mice displayed microglia and macrophages with enlarged cell bodies and thicker, shorter processes with less branching compared to Iba-1–positive and CD68-positive cells in APP/PS1;C3 KO mice (Fig. 3A and fig. S5, A and B); this was consistent with morphological changes associated with reactive microgliosis. In addition, GFAP-positive astrocytes were clustered in hippocampus of APP/PS1 mice but were more widely distributed and less clustered around plaques in APP/PS1;C3 KO mice (Fig. 3A). Quantitative image analysis revealed that Iba-1 immunoreactivity (% area) was significantly reduced by 28.8% in the hippocampal CA3 region (P < 0.05) and slightly reduced by 19.3% (nonsignificant trend) in the hippocampal CA1 region in APP/PS1;C3 KO mice; no significant differences were seen in the dentate gyrus (Fig. 3B). In addition, CD68 immunoreactivity, indicative of lysosomal phagosomes, was significantly reduced in hippocampal CA3, CA1, and dentate gyrus in APP/PS1;C3 KO mice (P < 0.05 per region; Fig. 3C). GFAP immunoreactivity of astrocytes in CA3, CA1, and dentate gyrus was also significantly reduced in APP/PS1;C3 KO mice (P < 0.05, CA1; P < 0.01, CA3 and dentate gyrus; Fig. 3D). Further, immunofluorescence intensities for Iba-1 and CD68 in Aβ plaques of similar size were significantly reduced in hippocampal CA3 of 16-month-old APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.01; Fig. 3, E and F). Although glial cell morphology was different, stereological analysis revealed that the number of Iba-1– and GFAP-positive cells in hippocampal CA3, CA1, and dentate gyrus areas was not significantly different between aged APP/PS1;C3 KO mice and APP/PS1 mice (Fig. 3, G and H).

Fig. 3. Morphological changes associated with glial activation were reduced in 16-month-old APP/PS1;C3 KO mice.

(A) Iba-1– and CD68-positive immunostaining showed less activation of microglia and macrophages (that is, smaller cells with thinner processes) and less clustering of GFAP-positive astrocytes in the hippocampal CA3 region of 16-month-old APP/PS1;C3 KO mice compared to APP/PS1 mice. Scale bar, 50 μm. (B to D) Quantification of Iba-1, CD68, and GFAP immunostaining in hippocampal CA3, CA1, and dentate gyrus (DG) showed reduced glial immunoreactivity (IR) in APP/PS1;C3 KO mice compared to APP/PS1 mice (*P < 0.05, **P < 0.01, independent unpaired t tests per region; n = 6 to 8; three equidistant planes, 300 μm apart). (E) High-resolution confocal images of Aβ plaques immunostained with 6E10 antibody show microglia/macrophages (immunoreactive for Iba-1) and phagocytic cells (immunoreactive for CD68) in the hippocampal CA3 region. These findings suggested reduced phagocytosis in APP/PS1;C3 KO mice compared to APP/PS1 mice. Scale bar, 10 μm. DAPI, 4′,6-diamidino-2-phenylindole. (F) Iba-1– and CD68-positive immunofluorescence intensities were lower in APP/PS1;C3 KO mice compared to APP/PS1 mice (**P < 0.01, independent unpaired t tests per marker; n = 5). (G and H) Stereological counts of Iba-1–immunoreactive cells (G) and GFAP-immunoreactive cells (H) counterstained with 3,3′-diaminobenzidine were performed in hippocampal CA3, CA1, and dentate gyrus tissue. The number of Iba-1–positive microglia/macrophages was increased in CA3 and dentate gyrus in APP/PS1 and APP/PS1;C3 KO mice versus WT and C3 KO mice (G). The number of GFAP-positive astrocytes was increased only in the hippocampal CA3 region (H); no differences were observed in glial cell numbers between APP/PS1 and APP/PS1;C3 KO mice (**P < 0.01, one-way ANOVA with Bonferroni post hoc test per region; n = 6 to 8; three equidistant planes, 300 μm apart).

Next, glial cell association within and surrounding Aβ plaques was examined to further elucidate a possible mechanism underlying the C3 deficiency–mediated increase in hippocampal Aβ plaque burden in APP/PS1;C3 KO mice. We quantified the immunofluorescence intensity and the number of microglia/macrophages (Iba-1–positive) and astrocytes (GFAP-positive) surrounding 6E10 antibody–positive Aβ plaques of similar size in the hippocampal CA3 region of 16-month-old APP/PS1;C3 KO mice and APP/PS1 mice using high-resolution Z-stack imaging by confocal microscopy (Fig. 4, A to L). Orthogonal analysis revealed that Iba-1–positive cells and Aβ plaques colocalized (fig. S5A). Three-dimensional (3D) reconstructed images were analyzed for intensity and cell number in the proximal area (within the center of large plaques) and distal area (surrounding large plaques). We observed more Iba-1– and GFAP-positive cells in the center of plaques in APP/PS1 mice, whereas Iba-1– and GFAP-positive cells tended to surround the plaques in APP/PS1;C3 KO mice (Fig. 4, A and G) and were more widely distributed. Upon quantification, we found that the intensity of Iba-1–positive cells in APP/PS1;C3 KO mice was significantly reduced proximal to plaques (P < 0.01) and significantly increased distal to plaques (P < 0.01) compared to APP/PS1 mice (Fig. 4, C and D). Similarly, the intensity of GFAP-positive astrocytes in APP/PS1;C3 KO mice was significantly reduced in the proximal area of plaques (P < 0.05) and significantly increased distal to plaques (P < 0.01) compared to APP/PS1 mice (Fig. 4, I and J). Although the number of glial cells located within plaques (proximal) was decreased (Fig. 4, E and K) and those surrounding plaques (distal) were increased (Fig. 4, F and I) in APP/PS1;C3 KO mice, there was no difference in the total number of glial cells in hippocampal CA3 in APP/PS1;C3 KO versus APP/PS1 mice (Fig. 3, G and H). Thus, although the APP/PS1;C3 KO mice had enhanced plaque deposition, they also had reduced immunofluorescence intensity and localization of glia within the center of plaques, suggesting that C3 deficiency may alter the glial response to plaques.

Fig. 4. Plaque-associated microglia and astrocytes and brain cytokines were altered in APP/PS1;C3 KO mice compared to APP/PS1 mice.

(A, B, G, and H) High-resolution confocal images of Iba-1 (red)/6E10 (Aβ antibody; green)/DAPI (blue) or GFAP (red)/6E10 (green)/DAPI (blue) in APP/PS1 and APP/PS1;C3 KO mice. The inner ring indicates the proximal region of a plaque (that is, the center), whereas the outer ring indicates the distal region. Scale bars, 10 μm. (C, D, I, and J) Immunofluorescence intensities of Iba-1 and GFAP were lower in the Aβ plaque proximal area (C and I) and higher in the Aβ plaque distal area (D and J) in APP/PS1;C3 KO mice compared to APP/PS1 mice (*P < 0.05, **P < 0.01, unpaired t test; n = 6). (E, F, K, and L) The number of Iba-1– and GFAP-positive cells was reduced in the proximal plaque area in APP/PS1;C3 KO mice compared to APP/PS1 mice (E and K) (*P < 0.05) and increased in the distal plaque area (F and L) (*P < 0.05, unpaired t test; n = 6). (M) Assay of cytokines by ELISA in mouse brain homogenates revealed reductions in tumor necrosis factor–α (TNF-α), interferon-γ (IFN-γ), and interleukin-12 (IL-12) and an increase in the IL-10/IL-12 ratio in 16-month-old APP/PS1;C3 KO mice compared to APP/PS1 mice (*P < 0.05, independent unpaired t test per marker followed by Bonferroni correction for multiple comparisons; n = 8). KC-GRO, keratinocyte chemoattractant (KC) chemokines CXCL1/2, mouse homologues of human growth-regulated oncogenes (GRO).

In a previous study, we found that CD45 immunoreactivity was increased in aged J20;C3 KO mice compared to J20 hAPP transgenic mice. However, Iba-1 immunoreactivity and the amount of CD68 by Western blot were not significantly different between J20;C3 KO and J20 mice (15). Here, we performed immunostaining of Iba-1 and CD68 on brains from 18-month-old APP/PS1, APP/PS1;C3 KO, J20, and J20;C3 KO mice. In aged J20 mice, both Iba-1 and CD68 immunoreactivities within Aβ plaques were lower than those in aged APP/PS1 mice (fig. S6, A and D). We observed decreases in Iba-1 and CD68 immunoreactivities within Aβ plaques in APP/PS1;C3 KO mice compared to APP/PS1 mice (fig. S6, E and G) but no difference between J20 mice and J20;C3 KO mice (fig. S6, F and G), consistent with our previous study (16).

Altered cytokines in aged APP/PS1;C3 KO mouse brain

ELISA of soluble proteins derived from T-per extraction of brain homogenates revealed that proinflammatory cytokines including TNF-α, IFN-γ, and IL-12p70, which have been reported to be elevated in AD brain (27), were significantly reduced in 16-month-old APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.05; Fig. 4M). The amount of IL-10 did not differ between APP/PS1 and APP/PS1;C3 KO mice, although the ratio of IL-10/IL-12 was significantly increased in APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.05; Fig. 4M). Thus, C3 deficiency resulted in the reduction of several proinflammatory cytokines in aged AD mice.

Partial protection against age-related hippocampal synaptic degeneration in APP/PS1;C3 KO mice

Decreased synaptic number, mRNA, and protein as well as impaired synaptic morphology have been reported in hippocampus and cortex of APP/PS1 mice between 7 and 18 months of age (2830). To determine the effects of C3 on hippocampal synapses in aged APP/PS1 mice, we performed high-resolution confocal microscopy analysis of pre- and postsynaptic markers (that is, synaptic puncta) in hippocampal CA3, CA1, and dentate gyrus areas in 16-month-old mice and compared WT to C3 KO, WT to APP/PS1, APP/PS1 to APP/PS1;C3 KO, and C3 KO to APP/PS1;C3 KO mice (Fig. 5). Consistent with our previous results (5), C3 KO mice had significantly more Vglut2-positive (P < 0.01), GluR1-positive (P < 0.05), and colocalized (P < 0.05) synaptic puncta than WT mice (Fig. 5, A to C). In addition, C3 KO mice had more presynaptic Vglut2-positive puncta and more postsynaptic GluR1-positive puncta than APP/PS1 and APP/PS1;C3 KO mice (P < 0.01), although colocalized synaptic puncta were not significantly different between C3 KO and APP/PS1;C3 KO mice (Fig. 5, A to C). We found that the densities of postsynaptic puncta (GluR1, P < 0.01) and colocalized puncta (Vglut2 and GluR1, P < 0.05) were significantly reduced in hippocampal CA3 in 16-month-old APP/PS1 mice compared to WT mice (Fig. 5, A to C). The density of GluR1-positive and colocalized synaptic puncta was significantly increased in CA3 of APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.01) and normalized to WT, suggesting protection against age-related synapse loss (Fig. 5H). Western blotting of presynaptic markers, synapsin-1 (SYN-1) and synaptophysin (SYP), and postsynaptic markers, GluR1, PSD95, and Homer1, in hippocampal synaptosomes from 16-month-old WT, C3 KO, APP/PS1, and APP/PS1;C3 KO mice revealed that C3 KO mice had elevated GluR1, PSD95, and Homer1 compared to all other groups, and APP/PS1 mice had reduced GluR1, PSD95, Homer1, SYN-1, and SYP compared to WT mice (Fig. 5, D, E, and H). APP/PS1;C3 KO mice showed a significant increase in all five synaptic proteins compared to APP/PS1 mice (P < 0.05, Syn-1; P < 0.01, GluR1, PSD95, Homer1, and SYP), and the amounts were similar to those detected in WT mice (Fig. 5, D and E). In addition, microtubule-associated protein 2 (MAP-2), a marker of the neuronal cell body and dendrites, was more abundant in fibrillar, thioflavin S–positive amyloid plaques of similar size in the hippocampal CA3 region of APP/PS1;C3 KO mice compared to APP/PS1 mice (P < 0.01; fig. S7).

Fig. 5. C3 deficiency resulted in partial preservation of synapse density in APP/PS1 mice despite an increased plaque load.

Comparisons were made between WT and C3 KO, WT and APP/PS1, APP/PS1 and APP/PS1;C3 KO, and C3 KO and APP/PS1;C3 KO mice. (A) Synaptic puncta of pre- and postsynaptic markers Vglut2 and GluR1, respectively, and their colocalization in hippocampal CA3 were analyzed by high-resolution confocal microscopy in 16-month-old mice. Scale bar, 5 μm. (B) C3 KO mice had increased Vglut2 and GluR1 synaptic densities compared to WT, APP/PS1, and APP/PS1;C3 KO mice. APP/PS1 mice had fewer GluR1 synaptic densities than WT mice, whereas APP/PS1;C3 KO mice had more GluR1 densities than APP/PS1 mice and were not significantly different from WT mice (*P < 0.05, **P < 0.01, one-way ANOVA and Bonferroni post hoc test per marker; n = 6 to 8; three equidistant planes, 300 μm apart). (C) Colocalization of pre- and postsynaptic puncta revealed increased puncta in C3 KO versus WT and APP/PS1 but not APP/PS1;C3 KO mice, reduced puncta in APP/PS1 mice versus WT mice, and a rescue of synaptic puncta in APP/PS1;C3 KO mice compared to APP/PS1 mice, suggesting partial protection against synapse loss by deletion of C3 (one-way ANOVA and Bonferroni post hoc test). (D and E) Western blotting of synaptic proteins in hippocampal synaptosomes isolated from aged mice indicated increased postsynaptic proteins GluR1, PSD95, and Homer1 in C3 KO versus WT, APP/PS1, and APP/PS1;C3 KO mice. APP/PS1 mice had lower postsynaptic GluR1, PSD95, and Homer1 and presynaptic SYN-1 and SYP compared to WT mice. APP/PS1;C3 KO mice had more GluR1, PSD95, Homer1, SYN-1, and SYP than APP/PS1 mice and were not significantly different than WT mice, suggesting a sparing of synaptic loss in aged C3-deficient APP/PS1 mice (*P < 0.05, **P < 0.01, one-way ANOVA and Bonferroni post hoc test per marker; n = 6). (F and G) Western blotting and quantification of TrkB, mBDNF, CREB, and pCREB in hippocampal homogenates of 16-month-old mice. C3 KO mice had increased mBDNF and pCREB compared to WT, APP/PS1, and APP/PS1;C3 KO mice. APP/PS1 mice showed reductions in all four markers compared to WT mice. APP/PS1;C3 KO mice had higher mBDNF, CREB, and pCREB than APP/PS1 mice (*P < 0.05, **P < 0.01, one-way ANOVA and Bonferroni post hoc test per marker; n = 6), suggesting a partial rescue of age- and AD-related lowering of BDNF pathway proteins. (H) Table summarizing the effects of C3 deficiency on synaptic and BDNF-related proteins. N.S., not significant.

Mature brain-derived neurotrophic factor (mBDNF) and its upstream (TrkB) and downstream [CREB (cyclic adenosine monophosphate response element–binding protein) and pCREB] signaling partners, key mediators of synaptic plasticity and memory, were examined by Western blot of brain homogenates and quantified (Fig. 5, F to H). C3 KO mice had higher mBDNF and pCREB than WT, APP/PS1, and APP/PS1;C3 KO mice. APP/PS1 mice had reduced TrkB, mBDNF, CREB, and pCREB compared to WT mice. APP/PS1;C3 KO mice had significantly higher mBDNF, CREB, and pCREB compared to APP/PS1 mice (P < 0.01) but lower mBDNF and pCREB than C3 KO mice (Fig. 5, F and G). No significant differences were seen in CREB between C3 KO and APP/PS1;C3 KO mice. This result correlates well with our cognitive and synaptic puncta data and suggests a pro-cognitive phenotype in the APP/PS1;C3 KO mice. Thus, whereas 16-month-old APP/PS1 mice demonstrated hippocampal CA3 synapse loss, APP/PS1;C3 KO mice were at least partially spared synapse loss, which was associated with an increase in BDNF signaling pathway proteins (Fig. 5). We also examined the amounts of human APP and presenilin 1 in APP/PS1 and APP/PS1;C3 KO mice and found no significant differences in either protein between the two genotypes (fig. S8).

Absence of age-dependent hippocampal CA3 neuron loss in APP/PS1;C3 KO mice

As previously reported (5), there was a small (28%) but statistically significant age-dependent reduction in the number of NeuN-positive neurons in the hippocampal CA3 region from P30 to 16 months of age in WT mice that was rescued in C3 KO mice (P < 0.05; Fig. 6F). A stronger (40%) reduction in CA3 neurons was seen from P30 to 16 months of age in APP/PS1 mice (P < 0.01; Fig. 6F). APP/PS1;C3 KO and C3 KO mice showed no significant age- or plaque-related loss of hippocampal CA3 neurons (Fig. 6F). Instead, 16-month-old APP/PS1;C3 KO mice had significantly more hippocampal CA3 neurons compared to age-matched APP/PS1 mice (P < 0.05; Fig. 6, A and B). No differences in neuron numbers were found in CA1, dentate gyrus, and prefrontal cortex areas in APP/PS1;C3 KO mice compared to APP/PS1 mice (Fig. 6, A and C to E). In addition, CA3 neurons were disorganized in 16-month-old APP/PS1 mice but not in APP/PS1;C3 KO mice (Fig. 6A). These data suggest that hippocampal CA3 neurons are more vulnerable to aging than neurons in other brain regions in both WT and APP/PS1 mice and that APP/PS1;C3 KO mice may be protected against this age-related neuronal vulnerability even in the presence of abundant plaques.

Fig. 6. C3 deficiency resulted in partial sparing of neuron loss in hippocampal CA3 in 16-month-old APP/PS1 mice.

(A) NeuN immunoreactivity in hippocampal CA3, CA1, and dentate gyrus (DG) and prefrontal cortex (PFC) regions in APP/PS1 and APP/PS1;C3 KO mice. Scale bars, 50 μm. (B) APP/PS1;C3 KO mice had more neurons in hippocampal CA3 compared to APP/PS1 mice (*P < 0.05, unpaired t test). (C to E) No significant differences were observed in neuron numbers in CA1, dentate gyrus, and PFC between APP/PS1 and APP/PS1;C3 KO mice. (F) Age-dependent neuron loss was observed between P30 and 16 months of age in WT and APP/PS1 mice and also between 4 and 16 months of age in APP/PS1 mice. However, neuron loss was not observed in C3 KO or APP/PS1;C3 KO mice (*P < 0.05, **P < 0.01, two-way ANOVA and Bonferroni post hoc test; n = 6 to 8; six equidistant planes, 150 μm apart).

DISCUSSION

Complement C3 is an innate immune system molecule that is important for removing pathogens and eliminating synapses during brain development and aging (25). Complement activation is elevated in human AD brain, especially in the presence of fibrillar amyloid plaques containing clusters of reactive microglia (8, 9, 11, 12) and in the brains of AD amyloidosis mouse models (10, 17, 31). To address the role of complement C3 in age-dependent, plaque-related neuropathology, we generated APP/PS1;C3 KO mice and analyzed the role of C3 in brain biochemistry, neuropathology, and behavior in 16-month-old mice. We demonstrated that C3 deficiency protected against cognitive impairment, loss of hippocampal CA3 synapses and neurons, and glial alterations even in the context of increased Aβ plaque deposition. Our results suggest that, despite extensive plaque deposition, C3 deficiency protects against both age- and AD-related synapse loss and cognitive decline in aged APP/PS1 mice by altering the glial response to plaques. These data indicate that plaque load is less critical than the reaction of glia to the plaques, and implicate complement activation as a possible link between fibrillar amyloid and accompanying microglial clustering in late-stage AD pathogenesis in mice, similar to that observed in human AD brain.

Progressive cognitive impairment in water maze tests is observed in aged APPswe/PS1dE9 mice (an AD-like mouse model of amyloidosis) and is correlated with progressive Aβ deposition, neuroinflammation, and degeneration of synapses and neurons (25, 32). Here, we found that APP/PS1;C3 KO mice were protected against cognitive deficits, especially during water T-maze test reversal learning, a task that is hippocampal-dependent (20). This result agrees with previous reports, including our own, demonstrating that C3 KO mice are protected against synapse and neuron loss and are spared from cognitive impairment during normal aging (5, 33). Many of our results in the current study point to age-dependent protection by C3 deficiency in aged APP/PS1 mice. However, C3 deficiency also protected against some AD-related changes because there were no significant differences between APP/PS1;C3 KO, WT, and C3 KO mice in the percent of mice that reached the reversal criterion in the water T-maze test, indicating better cognitive flexibility. Similarly, C3 deficiency protected against AD-related decline in hippocampal SYP and CREB expression in APP/PS1 mice. In addition, inhibition of complement C5aR, downstream of C3 activation, with an antagonist has been reported to reduce Aβ pathology and improve cognitive function in transgenic 2576 AD mice (24), raising the possibility that the protective cognitive effects we observed in aged C3-deficient APP/PS1 mice might be due to the lack of downstream C5a/C5aR signaling. Synaptic degeneration has been correlated with cognitive decline in AD (6, 7); therefore, our observations of cognitive sparing in aged APP/PS1;C3 KO mice may be due to the preservation of hippocampal synapses and neurons, even in the presence of increased plaque load. Whether the protective effects of C3 deficiency overpower the toxic effects of Aβ or cause sequestration of Aβ into plaques remains to be determined. Neuroinflammation (for example, glial clustering and activation within plaques) was reduced in APP/PS1;C3 KO mice, suggesting that C3 may play a critical role in driving neuroinflammation, which facilitates synaptic decline.

In the brain environment undergoing injury or disease, chronic activation of microglia may contribute to synaptic degeneration and memory decline (34). Elimination of microglia and inhibition of microglial proliferation by inhibiting colony-stimulating factor 1 receptor have been reported to prevent spine and neuronal loss and improve memory performance in AD mouse models, without significant modulation of Aβ pathology (35, 36). In addition, we recently demonstrated that complement and microglia mediate synapse loss in AD mouse models at early AD stages before plaque deposition (10). Our data here demonstrate that complement plays an additional role in the glial response to amyloid deposition and synaptic health at later AD stages as well. Furthermore, early C1q-mediated microglial activation and neurodegenerative changes in retina observed in a mouse model of retinal ischemia and reperfusion were abrogated in C1qa KO mice (37). Similarly, the loss of retinal ganglion cell synapses and dendrites (which precede axon and soma loss) in a genetic mouse model (DBA/2J) and an inducible rat model (microbead) of glaucoma were rescued by genetic deletion of C1qa in the mouse model and C1 inhibition in the rat model (38). Together, these findings suggest that inhibition of complement and its interaction with microglia may serve as a target to prevent the progression of synapse and neuron loss in AD.

Complement C3 plays a pivotal role in plaque deposition and clearance (20); however, it is not known whether C3 mediates the glial response in plaque-enriched brain. Here, we show that C3 deficiency in aged APP/PS1 mice decreased microglial CD68 activity, maintained microglial branching, and increased plaque load, while preserving cognition and hippocampal CA3 synapses and neurons. Both C3 deficiency and inhibition of C3 convertase by overexpression of sCrry in J20 hAPP transgenic mice led to an age-dependent increase in plaque burden (21, 22). In agreement with these studies in J20 mice, we found that aged APP/PS1;C3 KO mice had enhanced plaque deposition and increased insoluble Aβ in brain homogenates. Microglial uptake and clearance of Aβ (that is, by phagocytosis) may be mediated by the interaction of iC3b with its receptor, CR3 (CD11b/CD18), on the surface of microglia (20, 39). Previous studies have reported the reduction of microglial marker F4/80, I-A/I-E, or CD68 in C1q-deficient transgenic 2576 APP mice (23), J20 hAPP/sCrry mice (22), and J20 hAPP;C3 KO mice (21). Here, we also found that microglia/macrophage and astrocyte markers were reduced in aged APP/PS1;C3 KO mice compared to APP/PS1 mice, although there was no difference in glial cell number. Moreover, we found fewer microglia and astrocytes located within plaques, suggesting that C3 deficiency alters the glial response to plaques in aged, plaque-rich APP/PS1;C3 KO mice. Thus, these findings suggest that the increased fibrillar Aβ plaque load in the aged APP/PS1;C3 KO mice is due to a muted glial response and reduced phagocytosis of Aβ.

Senile plaques have been proposed to act as a potential reservoir of soluble oligomeric Aβ. When nonsequestered, Aβ colocalizes with the postsynaptic density, is associated with dendritic spine collapse, and is the major synaptotoxic Aβ species in the AD brain (40). Therefore, in theory, it is possible that although C3 deficiency reduced microglia-mediated phagocytosis of Aβ and increased plaque deposition in APP/PS1 mice, it also may have resulted in more sequestration of Aβ oligomers into plaques. This might reduce the availability of Aβ oligomers to bind to synapses, thereby blocking microglia activation–induced synapse loss in the hippocampus in aged APP/PS1;C3 KO mice. This possibility would be consistent with our previous findings that genetic deletion or antibody blocking of C1qa protected hippocampal synapses from Aβ oligomer–induced damage (8). Although purely speculative at this stage, future studies are under way to further clarify this possibility.

Complement C3 may contribute to region-specific and age-dependent synapse and neuron loss in AD mice via C3-dependent chronic microglial activation. Complement is involved in synaptic elimination in brain development and aging. Upon activation, complement mediates synapse elimination in the developing visual system by complement tagging of synapses that are then phagocytosed by microglia (2, 3). Complement C3 contributes to age-dependent and region-specific synapse and neuron loss during normal aging (5). Synapse loss is an early and important change in AD brain (41), and hippocampal synapse loss is observed in APP/PS1 mice (28, 29). Although both plaque-associated and plaque-independent hippocampal synaptic degeneration have been observed in aged APP/PS1 mice (42), we found reduced synaptic degeneration despite increased plaque load in aged APP/PS1;C3 KO mice compared to APP/PS1 mice, indicating that the presence of plaques alone did not induce neurodegenerative changes. Given that microglia and astrocytes, both of which express complement, are highly phagocytic and participate in synapse pruning (10, 43), our findings that C3 deficiency protected against hippocampal synaptic decline in two mouse models (C3 KO and APP/PS1;C3 KO) suggest that complement C3 or its downstream activation fragments, such as C3a, C5a, and C5b-9, may play an important role in synapse loss and neurodegeneration. Other reports support this hypothesis. For example, a recent report suggests that complement C3a secretion by astrocytes and binding to C3aR on microglia modulate Aβ in vitro and amyloid plaques in vivo in AD mouse models (44). The same group previously demonstrated that Aβ-induced nuclear factor κB–mediated release of C3a from astrocytes bound to C3aR on neurons and induced neurodegeneration and cognitive impairment, whereas treatment with a C3aR antagonist rescued these effects (14). Others have reported that low-dose treatment with the same C3aR antagonist in an ischemia/reperfusion stroke model in C57BL/6 mice resulted in enhanced neurogenesis, histologic and functional neuroprotection, and an anti-inflammatory effect (for example, the absence of T lymphocytes expressing C3aR migrating into the ischemic region) 7 days after injury (45). However, several caveats of the C3aR antagonist including potential off-target and agonist activity have been reported (46), leaving the role of C3a/C3aR signaling in neurodegeneration unclear.

Previously, we reported an age-dependent loss of CA3 neurons in 16-month-old C57BL/6 mice (5). Age-dependent neuron loss in APPswe/PS1dE9 mice has been found in striatum (47) but not cortex (48) or hippocampal CA1 (30). Here, we report age-dependent loss of hippocampal CA3 neurons in the same APP/PS1 mouse model, which was reduced in the absence of complement C3. Age-dependent neuron loss was not observed in CA1, dentate gyrus, and prefrontal cortex in aged APP/PS1;C3 KO mice, indicating that certain brain regions are more vulnerable to neuron loss with aging than others. It is possible that the selective vulnerability of CA3 neurons during aging in APP/PS1 mice is due to early CA3 synapse loss preceding late neuron loss induced by C3-mediated synaptic elimination, similar to what we found in aged WT mice (5). CA3 neurons receive input from other CA3 neurons through recurrent connections as well as medial septum and the diagonal band of Broca, both of which show degenerative changes during aging (49, 50). C1qa-deficient transgenic 2576 hAPP mice show increased iC3b deposition in brain, reduced gliosis and neuron loss in several brain regions, and partial protection of synapses in the hippocampal CA3 region, suggesting selective regional vulnerability of synapses and differential effects of various complement pathways (23).

It has been shown that microglia engulf presynaptic inputs via the C3/CR3-dependent microglia phagocytic signaling pathway in the early developing postnatal rodent brain (3). Recently, Hong et al. reported Aβ oligomer–induced, complement-mediated, and microglia-dependent synaptic loss in young, pre-plaque AD mouse brain that was rescued by the genetic deletion of C1q or C3, or by treatment with a C1qa blocking antibody (10). This study suggests a role for complement in oligomeric Aβ–induced synaptic pathology at early AD stages before plaque deposition. Here, we demonstrate a role for complement C3 signaling in Aβ fibril–induced clustering and activation of microglia within plaques at a much later disease stage in aged APP/PS1 mice. Aβ-induced complement activation is strongly determined by the degree of Aβ fibrillarity (51, 52). In humans, early, nonfibrillar diffuse plaques are associated with little or no microglial clustering (53) and accumulate little or no complement proteins (8, 9). Complement has been shown to promote the nucleation phase of Aβ aggregation (54), suggesting that, once activated, complement may perpetuate AD pathogenesis. At later stages of AD pathogenesis in humans, various complement proteins and clusters of activated microglia have been found to colocalize with fibrillar amyloid plaques, which often contain dystrophic neurites (8, 9). Our current study supports the role of complement in the link between Aβ fibrillarity and clustering of activated microglia and shows that lifelong C3 deficiency suppresses the glial response to plaques, reduces the production of some proinflammatory cytokines, rescues hippocampal synapses and neurons, and spares cognition in aged APP/PS1 mice despite increased plaque deposition.

Whether microglia and complement work together to mediate synapse loss in plaque-rich AD brain is unclear, but this hypothesis is supported by our data in APP/PS1 mice. Others have suggested that C3-dependent microglia overactivation is responsible for neuroinflammation, which accelerates pathogenesis in neurodegenerative diseases (55). For instance, lipopolysaccharide-induced loss of dopaminergic neurons was inhibited by C3 deficiency (56). In experimental autoimmune encephalomyelitis, a mouse model of human multiple sclerosis, C3 overactivation in Crry-deficient mice increased microglial overactivation and exacerbated neurodegeneration (57). Mice deficient for complement receptor 2 (Cr2 KO), which lack the CR1 and CR2 receptors that bind to C3 activation fragments, had reduced secondary brain damage after closed head injury including decreased neuron death, astrocytosis and microglial activation, and improved neurological outcomes (58). In addition, genetic deletion of C3 (but not C1q) protected against synapse loss and promoted faster recovery 1 week after axotomy of spinal cord motor neurons by sciatic nerve transection (59). Furthermore, mice deficient in BDNF in microglia have deficits in learning and hippocampal synapse formation, indicating that microglial BDNF signaling plays an important role in hippocampal function (60). Together, these results provide evidence for a deleterious effect of complement activation and its downstream consequences in brain in the context of aging, trauma, and neurodegenerative diseases.

Seemingly contradictory results of complement have been observed among different AD mouse models. C3 inhibition enhanced plaque deposition and promoted neurodegeneration in J20 hAPP mice either by overexpression of sCrry (22) or by genetic deletion of C3 (21). The addition of a mutant human PS1 transgene is one possible explanation for the difference in C3 deficiency between J20 mice in the previous studies and APP/PS1 mice in this study. Previous reports demonstrated up-regulation of C1q in the hippocampus and GFAP and cathepsin S in the cortex and hippocampus in conditional PS1/2 double-knockout mice and increased expression of complement cascade components C1q, C3aR, and C4b in the hippocampi of PS1/2 double-knockout mice (61, 62). Thus, it is possible that the presence of the mutant human PS1 gene in the APP/PS1 mice may explain, in part, the differential effect of C3 deficiency on neurodegeneration compared to J20 hAPP mice. However, we found no significant difference in the amount of APP or PS1 in the APP/PS1 versus APP/PS1;C3 KO mice used here. Other studies demonstrated that lifelong C1q deficiency rescued SYP and MAP-2 expression in hippocampal CA3 in aged transgenic 2576 APP and APP/PS1 mice (23), similar to our findings in 16-month-old APP/PS1;C3 KO mice. However, an additional study suggests that C1q does not appear to play a role in synapse loss during aging but instead contributes to a decline in functional brain wiring (31). Here, we observed less reactive microgliosis and reduced astrocytosis in aged APP/PS1;C3 KO mice, compared to no change in Iba-1 and CD68 markers in J20;C3 KO mice, similar to our previous study. In addition, our current findings showing hippocampal region–specific and glia/synapse marker–specific effects described in APP/PS1;C3 KO mice may have been missed in our previous J20;C3 KO study that used fewer markers and examined the entire hippocampus, not individual subregions. Together, the differences in baseline glial responses to plaques and effects of C3 deletion on microglia in aged J20 versus aged APP/PS1 mice may explain the differences in neurodegeneration that we have observed in the two models.

Last, we acknowledge the limitations of our study, including the overexpression of human mutant APP and PS1 in mice, which likely generates other APP fragments that could affect behavior. Our findings are also limited by the differences in the immune system and life span between mice and humans. In addition, our APP/PS1 mouse model lacked tau pathology (which better correlates with cognitive decline in human AD), thus limiting the translation of our findings to the full human disease. Future studies in APP knockin mice or mice that have both amyloid and tau pathologies (for example, 3xTg-AD mice) would be useful to confirm the protective effects of C3 deficiency that we observed in our aged APP/PS1 mice, and these other models might better reflect human AD.

In summary, we found that C3 deficiency in APP/PS1 mice confers neuroprotection against age- and AD-related neurodegeneration and cognitive decline despite enhancing plaque burden. Our results suggest that C3 may play a detrimental role in synaptic and neuronal function in plaque-rich, aged brain and suggest that inhibition of C3 signaling may be a potential therapeutic target for AD treatment. Our future studies using inducible C3 conditional KO mice will determine whether depleting C3 after brain development or after the onset of AD pathology will protect synapses and spare cognitive decline.

MATERIALS AND METHODS

Study design

The objective of our study was to determine whether complement C3 plays a role in late-stage AD pathogenesis and cognitive decline. We crossbred an AD-like mouse model with C3 KO mice and aged them to 16 months. The APP/PS1;C3 KO mice were compared to APP/PS1, WT, and C3 KO mice in terms of behavioral, neuropathological, and biochemical end points. Operators were blinded to mouse genotype for all outcome measures. We used power analysis for the water T-maze test, the outcome measure with the highest variability, to compare cognitive behavior between 16-month-old APP/PS1;C3 KO mice and APP/PS1 littermates and determined that we needed a minimum of 10 mice per group to achieve statistical significance of P < 0.05 (mean % correct APP/PS1;C3 KO = 90.0; mean % correct APP/PS1 = 45.5; SD = 10.0; α = 0.05; power = 0.80; www.statisticalsolutions.net). To improve the rigor of the study, we included WT (n = 13), APP/PS1 (n = 10 to 11), APP/PS1;C3 KO (n = 11 to 12), and C3 KO (n = 10 to 11) for behavioral tests and n = 6 to 9 per group for biochemical and immunohistochemical tests. We excluded the data from one APP/PS1 mouse from the elevated plus maze test because it was an outlier of approximated 3 SDs from the mean.

Mice

C57BL/6J mice, APPswe/PS1dE9 mice, and homozygous C3–deficient breeder mice (C3−/−; line B6.129S4-C3tm1Crr/J) (63) were obtained from The Jackson Laboratory. C3 KO mice were crossed with APP/PS1 mice to generate APP/PS1 (heterozygotes);C3 KO (heterozygotes) mice, which were backcrossed with C3 KO mice to generate APP/PS1 (heterozygotes);C3 KO (homozygotes). These mice were then bred with C3 KO mice to generate APP/PS1;C3 KO mice and littermate C3 KO mice. However, a separate cohort of APP/PS1 mice was bred with C57BL/6J mice to generate WT littermates and heterozygote APP/PS1 mice. Mice were genotyped by polymerase chain reaction using the following primers: APP/PS1, 5′-GACTGACCACTCGACCAGCTT-3′ and 5′-CTTGTAAGTTGGATTCTCATAT-3′; C3 WT, 5′-ATCTTGAGTGCACCAAGCC-3′ and 5′-GGTTGCAGCAGTCTATGAAGG-3′; C3 mutant, 5′-CTTGGGTGGAGAGGCTATTC-3′ and 5′-AGGTGAGATGACAGGAGATC-3′. Mice were aged to P30 or 4 or 16 months of age. Only males were used for this study to reduce gender-specific variability in the behavioral tests. At the end of the study, mice were anesthetized, blood was collected, and the brain was perfused with saline before harvest.

All animal protocols were approved by the Harvard Medical Area Standing Committee on Animals, and studies were performed in accordance with all state and federal regulations. The Harvard Medical School animal management program is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International and meets all National Institutes of Health standards as demonstrated by an approved Assurance of Compliance (A3431-01) filed at the Office of Laboratory Animal Welfare.

Water T-maze

The water T-maze behavioral paradigm assesses spatial learning and memory by training mice to use the spatial cues in a room to navigate to a hidden platform to escape water. Testing was performed as previously reported (5). Note that the 16-month-old WT and C3 KO mice used for behavioral testing in this large study were the same mice that we reported in (5), although they were tested concurrently with the APP/PS1 and APP/PS1;C3 KO mice in the present study. Here, all analyses were performed on these same mice.

Immunohistochemistry

Hemibrains were fixed in 4% paraformaldehyde for 24 hours, and immunohistochemistry was performed as described previously (5). Fixed, frozen sections were incubated with anti–Aβx–42 (1:200, Covance), 3D6 (Aβ1–5) (1:1000, Elan), 6E10 (1:1000, Covance), MAP-2 (1:500, Millipore), and NeuN (1:200, Serotec) mouse monoclonal antibodies; Aβ40 (1:200, Covance), Iba-1 (1:200, Wako), and GFAP (1:1000, Dako) rabbit polyclonal antibodies; or CD68 (1:250, Serotec) rat polyclonal antibody overnight at 4°C. After washing with tris-buffered saline, sections were incubated with biotinylated secondary antibodies and developed using Vector ELITE ABC kits (Vector Laboratories) and 3,3′-diaminobenzidine (Sigma-Aldrich) or immunofluorescence-labeled secondary antibodies and coverslipped with mounting medium (Vector Laboratories). Thioflavin S staining of amyloid fibrils was performed as previously described (21).

Aβ load analysis

Images of Aβ immunoreactivity were captured in a single session under a Nikon Eclipse E400 microscope. Quantification of Aβ plaque load and analysis of Aβ plaque size were measured within an ROI using the Bioquant image analysis system.

Three of six equidistant planes of 10-μm-thick sections from each mouse were analyzed. The operator was blinded to mouse genotype and age when analyzing.

Aβ and cytokine ELISAs

MSD Aβ Triplex ELISA was performed on T-per–insoluble, guanidine hydrochloride–extracted brain homogenates from mouse hemibrain, including cortex and hippocampus. The V-Plex Proinflammatory Cytokine ELISA kit (Meso Scale Discovery, catalog no. K15048) was used for detecting cytokine levels by MSD plate reader QuickPlex SQ 120 (Meso Scale Discovery). Brain homogenization and ELISAs were performed as previously described (5).

Confocal analysis of glia morphology and association with Aβ plaques

Immunofluorescence for Aβ plaques (6E10), microglia/macrophages (Iba-1), and astrocytes (GFAP) was detected by confocal microscopy (LSM 710, Carl Zeiss) using a methodology described previously (64). Images were collected using the same exposure settings. Confocal Z-stack images (optical slices of 0.2 μm) of plaques and surrounding glia were collected using a 63× objective. Three images were acquired from three equidistant planes, 500 μm apart, per mouse. Immunofluorescence intensity analysis and 3D reconstruction of Z-stack images were performed with confocal image analysis software (ZEN Black, Carl Zeiss). Glia counts were performed using stereological methods after collecting images.

Synaptic puncta staining and analysis

Synaptic puncta staining and analysis were performed as previously described (5, 65). Confocal imaging was performed using a Zeiss LSM710 confocal microscope and a 63× oil objective. Images were acquired using a 1 airy unit (AU) pinhole while holding constant the gain and offset parameters for all sections and mice per experiment.

Preparation of synaptosome fractions

Synaptosome fractions were prepared as described previously (66).

Western blotting of synapse markers and BDNF signaling proteins

Brain hippocampal and cortical tissues were homogenized, and Western blotting was performed as previously described (5) using rabbit polyclonal antibodies [SYN-1 (Millipore, 1:200), BDNF (Abcam, 1:200), and pCREB and CREB (Chemicon, 1:1000)] and mouse monoclonal antibodies [PSD95 (1:200; Millipore) and glyceraldehyde-3-phosphate dehydrogenase (1:200; Millipore)]. Blots were scanned using the LI-COR Odyssey Infrared Imaging System. Intensity of bands was measured by LI-COR Odyssey software.

Stereological quantification of neuron and glia counts

Immunohistochemistry for NeuN, Iba-1, and GFAP was performed as described previously (21). Stereological counting of neurons and glia was performed on 10-μm sagittal, immunostained sections at each of six planes, 250 μm apart, per mouse using the optical dissector method. Cells were counted in an area of ~300 μm within a 10-μm depth per brain region examined. The number of neurons and glia per section in each brain region was estimated using ~15 optical dissectors and the Bioquant image analysis system according to the principles of Cavalieri as described by West and Gundersen (67).

Statistical analysis

Data are means ± SEM. Comparisons were made between WT and C3 KO, WT and APP/PS1, APP/PS1 and APP/PS1;C3 KO, and C3 KO and APP/PS1;C3 KO mice. Significance for all behavioral tests was assessed by one-way ANOVA followed by Fisher’s protected least significant difference using StatView version 5.0 software. All other data were analyzed by one- or two-way ANOVA followed by Bonferroni’s post hoc test or Student’s t test using Prism version 6.0 (GraphPad Software). P < 0.05 was considered significant.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/9/392/eaaf6295/DC1

Materials and Methods

Fig. S1. Behavioral testing of locomotion and anxiety.

Fig. S2. C3 deficiency had no effect on Aβ deposition in young APP/PS1 mice at the earliest stage of plaque deposition.

Fig. S3. Immunoreactivity of Aβx–42, Aβx–40, and Aβ1–x in aged mice.

Fig. S4. Glial immunoreactivity was similar in young WT, APP/PS1, and APP/PS1;C3 KO mice at the earliest stage of plaque deposition.

Fig. S5. C3 deficiency resulted in morphological differences in Iba-1–immunoreactive cells in APP/PS1 aged mice.

Fig. S6. C3 deficiency resulted in reduced activation of microglia within Aβ plaques in aged APP/PS1 mice but not in aged J20 mice.

Fig. S7. MAP-2–positive dendrites were better preserved within Aβ plaques in 16-month-old C3-deficient APP/PS1 mice versus APP/PS1 mice.

Fig. S8. The amounts of APP and PS1 were not significantly different between APP/PS1 and APP/PS1;C3 KO mice.

References (68, 69)

REFERENCES AND NOTES

Acknowledgments: We thank J. Frost, J. Kenison, and Y. Hu [Ann Romney Center for Neurologic Diseases (ARCND)] as well as K. Merry [Boston Children’s Hospital (BCH)] for technical assistance. We thank C. White (ARCND) for statistics advice and B. Liu (ARCND) for data discussions. We are grateful to S. Matousek (ARCND) and K. Colodner (BCH) for assistance with mouse breeding. We thank D. Selkoe (ARCND) for providing the 3D6 antibody for immunohistochemistry and O. Butovsky (ARCND) for critical reading of the manuscript. Funding: This work was funded by Fidelity Biosciences Research Initiative (F-PRIME) (C.A.L. and B.S.), NIH/National Institute on Aging (R21 AG044713 to C.A.L.), BrightFocus Foundation Fellowship (Q.S.), and Edward R. and Anne G. Lefler Fellowship (S.H.). Author contributions: C.A.L. and Q.S. co-designed the project and oversaw all experimentation and data analysis. Q.S. and B.J.C. performed and analyzed the behavior studies. Q.S., S.C., and R.M. performed immunohistochemistry and image analysis for quantification, stereological counts, and ELISAs. K.X.L. performed cryosectioning and immunohistochemistry. Q.S. performed confocal imaging and Western blotting. B.S. and S.H. contributed to the discussion of the data. Q.S. and C.A.L. wrote the manuscript with assistance from all authors. Competing interests: C.A.L. serves on the scientific advisory boards of Probiodrug AG, Amgen, and Cognition Therapeutics. B.S. serves on the scientific advisory board and is a minor shareholder of Annexon LLC. The following patents related to this project have been granted or applied for: PCT/2015/010288 (S.H. and B.S.): Biomarkers for dementia and dementia-related neurological disorders; US14/988387 and EP14822330 (S.H. and B.S.): Methods of treatment for Alzheimer’s disease and Huntington’s disease; US8148330, US9149444, US20150368324, US20150368325, US20150368326, and US20120328601 (B.S.): Modulation of synaptic maintenance. The other authors declare that they have no competing interests.
View Abstract

Navigate This Article