|Home | About | Journals | Submit | Contact Us | Français|
Complement factor C3 is the central component of the complement system and a key inflammatory protein activated in Alzheimer's disease (AD). Previous studies demonstrated that inhibition of C3 by overexpression of sCrry in an AD mouse model led to reduced microgliosis, increased Aβ plaque burden and neurodegeneration. To further address the role of C3 in AD pathology, we generated a complement C3-deficient APP transgenic AD mouse model (APP;C3−/−). Brains were analyzed at 8, 12 and 17 months of age by immunohistochemical and biochemical methods and compared with age-matched APP transgenic mice. At younger ages (8–12 months), no significant neuropathological differences were observed between the two transgenic lines. In contrast, at 17 months of age, APP;C3−/− mice showed significant changes of up to two-fold increased total amyloid-beta (Aβ) and fibrillar amyloid plaque burden in mid-frontal cortex and hippocampus which correlated with: a) significantly increased TBS-insoluble Aβ42 levels and reduced TBS-soluble Aβ42 and Aβ40 levels in brain homogenates, b) a trend for increased Aβ levels in the plasma, c) a significant loss of NeuN-positive neurons in the hippocampus, and d) differential activation of microglia towards a more alternative phenotype (e.g., significantly increased CD45-positive microglia, increased brain levels of IL-4 and IL-10, and reduced levels of CD68, F4/80, iNOS and TNF). Our results suggest a beneficial role for complement C3 in plaque clearance and neuronal health as well as in modulation of the microglia phenotype.
The complement system is essential for immune-mediated defense against pathogens. It may be activated by three pathways: first, via the “classical” activation route through activation of the C1q complex by immunoglobulin/antigen immune complexes or non-immune molecules; second, via the immune complex-independent alternative activation pathway leading to deposition of C3 fragments on target cells; and, third, via the lectin route by binding mannose-binding lectin to pathogen-associated molecular patterns (PAMPs) (reviewed by van Beek et al., 2003). Activation leads to the formation of multimolecular enzymes that cleave and activate the central component C3. This may lead to the formation of the cytolytic membrane attack complex (MAC) resulting in direct lysis of pathogens. Alternatively, coating of complement activating structures (e.g. immune complexes or PAMPs) with complement C1q activates C3 by-products C3b and iC3b which promote phagocytosis by cells (such as microglia) expressing the corresponding complement receptors (CR) such as CR1 (CD35), CR3 (CD11b/CD18) or CR4 (CD11c/CD18). Furthermore, the complement system is involved in the production of anaphylatoxins (C3a and C5a), modulating inflammation and neuroprotective pathways, clearance of immune complexes, and in the regulation of adaptive immunity (van Beek et al., 2003).
In mammals, the liver is the major source of complement proteins, but glial cells as well as neurons express complement proteins upon stimulation by inflammatory cytokines (Levi-Strauss and Mallat, 1987; Gasque et al., 1995; Thomas et al., 2000). Complement activation plays an essential role in the inflammatory reactions in the nervous system, including in chronic neurodegenerative disorders such as Alzheimer's disease (AD). Complement is fully activated and complement components, with their corresponding mRNA levels, are upregulated in AD brains (Eikelenboom et al., 1989; McGeer et al., 1989; Yasojima et al., 1999). Complement components are observed in pyramidal neurons (Shen et al., 1997) and in neurofibrillary tangles and senile plaques in AD brain (Terai et al., 1997; Stoltzner et al., 2000). Dystrophic neurites are immuno-positive for MAC (McGeer and McGeer, 2002) suggesting that MAC may contribute to neuritic dystrophy and neuronal loss in AD.
Complement may have also a protective role in AD and normal brains, but its role remains controversial. For example, the C5-derived anaphylatoxin C5a was reported to protect against excitotoxicity in vitro and in vivo (Pasinetti et al., 1996; Osaka et al., 1999). Overproduction of TGF-β1 in APP transgenic mice results in elevated C3 brain levels, increased microglial activation, and reduced Aβ accumulation (Wyss-Coray et al., 2001). Inhibition of complement activation by transgenic overexpression of the soluble complement receptor-related protein y (sCrry) in APP mice resulted in increased Aβ accumulation and neuronal degeneration, and reduced microglial activation, but did not alter total C3 levels in the brain (Wyss-Coray et al., 2002). However, C1q was found to co-localize with fibrillar Aβ in a PS1/APP mouse model (Matsuoka et al., 2001). APP transgenic mice lacking C1q had no change in Aβ plaque burden compared to C1q-sufficient controls but showed decreased glial activation surrounding plaques and a slowing of neuronal pathogenesis suggesting a detrimental effect of C1q on neuronal integrity (Fonseca et al., 2004).
Here, we generated an AD mouse model (APP) deficient for C3 (C3−/−) to study the in vivo role of complement C3 on AD pathogenesis. Our results suggest a beneficial role of complement C3 in AD, particularly with advanced aging and pathogenesis.
Hemizygous APP transgenic mice ((Mucke et al., 2000), line J20, harboring the (PDGF promoter)-human APPsw (K670N, M671L), Ind (V717F) transgene; C57BL/6 background) from our breeding colony were bred with homozygous C3-deficient mice (C3−/−, (Wessels et al., 1995) initially obtained from Jackson Laboratory (line B6.129S4-C3tm1Crr/J; C57BL/6 background) or with C57BL/6 (Jackson Laboratory). To generate APP;C3−/− animals C3−/− mice were bred initially with APP mice, then with APP;C3+/− and later, with APP;C3−/− mice. Animals were genotyped by PCR with the following primers: 5'-CTTgggTggAgAggCTATTC-3' and 5'-ggTTgCAgCAgTCTATgAAgg-3' for C3 wt allele in the same reaction with 5'-AggTgAgATgACAggAgATC-3' and 5'-ATCTTgAgTgCACCAAgCC-3' for mutated C3 allele; 5'-GGTGAGTTTGTAAGTGATGCC-3', 5'-TCTTCTTCTTCCACCTCAGC-3' for hAPP allele together with 5'-GCGCGCTCGTGCACACTTATCACA-3' and 5'-CTGCCCCTGACTTCCTGGAAGCAC-3' for DNA (GFAP) control. Groups at the different ages were gender-balanced, exactly age-matched and the maximal age difference of individual animals within each group was less than 1 month. All animal use was approved by the Harvard Standing Committee for Animal Use and was in compliance with all state and federal regulations.
Mice were sacrificed by CO2 inhalation and blood was collected by cardiac puncture followed by transcardial perfusion with 20–30 ml phosphate buffered saline (PBS) as described (Maier et al., 2005). The brain was removed and divided sagittally. One hemibrain was fixed for 2 h in 10% buffered formalin while the other hemibrain was snap frozen in liquid nitrogen for biochemical analysis. One hemibrain, liver, and kidney were embedded in paraffin as described (Lemere et al., 2003).
For Tris buffered saline (TBS) brain homogenates, frozen hemibrains (without cerebellum and brain stem) were homogenized with a dounce homogenizer in 5 volumes of TBS with a protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN). The samples were centrifuged at 175,000 g for 30 minutes at 4°C. The supernatant (TBS-soluble homogenate) was collected and stored at −20°C. The pellets were resuspended in the same volume of TBS-T (TBS/1% triton X-100 plus protease inhibitor cocktail) buffer, sonicated for 5 min in 4°C water bath, homogenized, and centrifuged at 175,000 g for 30 min at 4°C. The supernatant (TBS-T-soluble homogenate), containing membrane-bound Aβ, was collected and stored at −20°C. The pellets were extracted a third time as previously described (Johnson-Wood et al., 1997) using ice cold guanidine buffer (5 M guanidine-HCl/50 mM Tris, pH 8.0) (herein referred to as TBS-insoluble or guanidine-soluble homogenate). AβX-40 (Aβ40) and AβX-42 (Aβ42) levels were determined in TBS, TBS-T and guanidine brain homogenates and total Aβ in EDTA plasma samples of tail blood. ELISAs specific for human Aβx-40, Aβx-42 and total Aβ were performed (using antibodies kindly supplied by ELAN Pharmaceuticals) according to (Johnson-Wood et al., 1997).
Plates were coated overnight at 4°C with goat anti mouse C3 (1:500, ICN cappel, MPBiomedicals, Solon, OH) in 50mM carbonate buffer (pH 9.6) and then blocked for 2 hours in blocking buffer (PBS/1% BSA). Serum or TBS-brain homogenates were diluted (1:32,000 for serum and straight for brain homogenates) in wash buffer (PBS/0.1% BSA/0.05% tween-20), added to the plate for 2 hours, hand-washed three times, incubated with biotinylated goat anti-mouse C3 antibody (1:1000 ICN cappel) in blocking buffer for 1 hour, washed, incubated with SA-HRP in blocking buffer for 30 minutes, washed and developed with TMB peroxidase substrate for 2–5 minutes. Stop reagent was added and OD was measured at 450nm. Matching antibody pairs, composed of capture and detection antibodies for murine IL-4, IL-10 and TNFα (BD PharMingen), were used according to the manufacturer’s instructions. Serial dilutions of exogenous cytokine were added as standards. TBS and TBS-T brain homogenates were added to the plate without dilution.
Western blot was performed as previously reported (Peng et al., 2006). TBS-brain homogenates were used to detect αAPPs, C3, and iNOS levels, while TBS-T brain homogenates were used to examine total APP, CD68, F4/80 and synaptophysin levels. Briefly, samples were denatured in SDS sample buffer and separated on 10–20% Tricine gels (Invitrogen, Carlsbad, CA). Polyclonal antibody R1736, which recognizes residues 595–611 of APP695 (1:1000, gift of D. Selkoe, CND, Boston, MA) was used to specifically detect the α-secretase-generated ectodomain fragment of APP. Monoclonal antibody 8E5 (1:1000), which recognizes residues 444–591 of human APP, was used to detect the full-length APP (a gift from Elan Pharmaceuticals, San Francisco, CA). Polyclonal anti-C3 (1:500, ICN Cappel, MPBiomedicals), anti-synaptophysin (1:1000, Sigma, Saint Louis, MO), anti-iNOS (1:200, BD Bioscience), anti-CD68 (1:200, Serotec), anti-F4/80 (1:200, Serotec) and anti-β-actin (1:10,000, Sigma, Saint Louis, MO) were used to probe the membrane. Horseradish peroxidase-coupled anti-rabbit or anti-mouse IgG were used as secondary antibodies. The signal was detected with an enhanced chemiluminescence (ECL) kit (Pierce, Rockford, IL). Densitometric evaluation was performed using an imaging system and the corresponding analyzing software (FluorchemTMIS-8800 software, Alpha Innotech, San Leandro, CA). Protein levels were normalized to β-actin.
Serial ten micron sagittal paraffin sections of mouse brain were mounted on glass slides and immunohistochemistry (IHC) was performed as previously reported (Lemere et al., 2002) using Vector ELITE ABC kits (Vector Laboratories, Burlingame, CA). The following antibodies were used for neuropathological analysis: anti-CD45 (1:5000, Serotec, Raleigh, NC), anti-Iba1 (1:500, Wako, Richmond, VA), anti-GFAP (1:500, Dakocytomation, Carpinteria, CA), anti-neuronal nuclei (NeuN; 1:250, Chemicon, Temecula, CA), anti-APP antibody 22C11 (1:200, Chemicon), rabbit polyclonal anti-Aβ R1282 (1:1000, gift of D. Selkoe, CND, Boston, MA), and anti-Aβ40 and anti-Aβ42 (1:500, BioSource, Camarillo CA, USA). Computer-assisted image analysis was used to quantify the percent area of Thioflavin S staining and immunoreactivity as previously described (Maier et al., 2006). Briefly, the percent area occupied by immunoreactivity was calculated for 3–4 equidistant sections per animal of total hippocampus area or for one visual field of the whole cortex dorsal to the hippocampus (mid-frontal cortex, ~1.5mm2). The threshold for detection of immunoreactivity was kept constant for the analysis of an entire series of sections for each antibody. Presynaptic terminals were labeled with anti-synaptophysin (SYN; 1:200, Sigma, Saint Louis, MO), neuronal cell bodies and dendrites with anti-MAP2 (1:200, Sigma), and newly generated neurons with anti-doublecortin (C-18, 1:500, Santa Cruz, CA) followed by Alexa488, Rhodamine Red™-X (Molecular probes, Eugene, OR) or Cy3 (Jackson Lab, Bar Harbor, MI) coupled secondary antibodies. To quantify immunoreactivity, acquisition of images was performed in a single session using a SPOT camera with individual RGB channels set for fluorescence (Sterling Heights, MI, USA), and analyzed by ImageJ software (NIH, Bethesda, MD).
Stereological neuron counts were performed in three 10 µm NeuN stained sagittal brain sections per animal spaced 200 µm apart using the optical dissector technique (Irizarry et al., 1997). The average number of neurons per section in the CA3 region of hippocampus was estimated using approximately 12 optical dissectors and the Bioquant Image Analysis System (Nashville, TN) according to the principle of Cavalieri (Cavalieri, 1966). Each optical dissector included a 50 µm × 50 µm sampling box. Using a 100X oil immersion lens, neurons with a visible NeuN cell body were counted if they were not visible in the initial plane of focus, but came into focus as the optical plane moved through the tissue. The average number of hippocampal CA3 neurons per section was determined by calculating the mean of the neuron counts from three sections per animal.
The Mann-Whitney U (MWU) test was used for statistical analysis of the percent area of immunoreactivity in brain sections and Aβ levels in brain homogenates. The critical α level was set to 0.05 for all statistical analyses. All values reported are average ± SEM. A two-tailed Spearman correlation was employed for correlation of plaque load and NeuN counts. SPSS software (V11.5) was used for statistical analysis
To determine the role of central complement component C3 in Aβ accumulation and neurodegeneration we generated APP transgenic mice lacking complement C3 by breeding APP (J20 line; Mucke et al., 2000) to complement C3-deficient mice (Wessels et al., 1995). Complement C3 gene deletion was confirmed by PCR using primers showing the presence of the mutated allele and the absence of the wildtype C3 allele PCR product in APP;C3−/− mice (Fig. 1A). Expression of complement C3 protein was demonstrated by Western blot and C3 ELISA with an anti-C3 antibody in brain homogenates (Fig. 1B, C) and blood plasma (Fig. 1D) in 12 and 17 month-old APP mice. In contrast, C3 protein was undetectable in brain and blood plasma in APP;C3−/− mice at the same ages (Fig. 1). C3 levels in blood as well as brain were slightly higher in 17 month-old compared to 12 month-old APP animals (p=0.14, n=5 mice each at 12 and 17 months) and about four orders of magnitude higher in plasma than in brain tissue.
To study the role of C3 deficiency on the onset of plaque pathology we analyzed total Aβ immunoreactive plaque load in the hippocampi of 8 month-old APP and APP;C3−/− mice (n=5 mice per group). Quantitative image analysis showed no significant difference in plaque burden between the groups (data not shown). Semi-quantitative analysis of CD45-specific immunoreactivity as a marker for activated microglia and GFAP-specific immunoreactivity as a measure to quantify astrogliosis in the hippocampus revealed no major differences between the groups (data not shown).
At 12 months of age, quantitative image analysis of brain sections stained with Aβ42-, Aβ40- or pan-Aβ-specific R1282 antibodies revealed no significant differences in plaque load in hippocampus or mid-frontal cortex in APP;C3−/− mice (n=8) compared to their age-matched APP controls (n=6) (Fig. 2A, B; 11.8±2.0% vs. 10.2±0.9% area of hippocampus for R1282, p>0.05). At this age, Thioflavin S-positive plaques were mainly detected in the hippocampus and again, quantitative image analysis showed no significant differences between the two groups of mice (Fig. 2C). In parallel, no significant differences were found for CD45- and GFAP-specific immunoreactivity (Fig. 2D, E). Quantification of Aβ42 and Aβ40 by ELISA in TBS, TBS/1% Triton (TBS-T) and guanidine HCL brain extracts also showed no significant differences in Aβ levels (Fig. 2F, G). Total Aβ levels in plasma were increased by 31%, but the difference did not reach significance due to variability (Fig. 2H, n=6 per group, p=0.11).
In contrast to 12 month-old mice, total Aβ plaque load at 17 months of age was roughly doubled in both the hippocampus and mid-frontal cortex in APP;C3−/− mice compared to APP mice as shown by Aβ42- and Aβ40-specific immunoreactivity (Fig. 3A, B), general Aβ immunolabeling (Fig. 4A, B, G, H; R1282 antibody), and Thioflavin S-positive staining of fibrillar amyloid deposits (Fig. 3C, Fig. 4M, N,) (p<0.05, n=5 mice per group). Guanidine-soluble (TBS-T-insoluble) Aβ42 and Aβ40 levels were higher in APP;C3−/− mice compared to APP mice (Fig. 3D, E) at 17 months of age, although only the difference in Aβ42 levels reached significance (p<0.01). These data are in agreement with the corresponding Aβ42 and Aβ40 plaque loads. In contrast, TBS-soluble Aβ42 and Aβ40 levels were significantly lower in APP;C3−/− mice compared to APP mice (Fig. 3D, E; p<0.05). TBS-T soluble Aβ42 and Aβ40 levels were non-significantly reduced in the APP;C3−/− mice compared to APP mice. Total Aβ levels in plasma were increased by 51% in APP;C3−/− mice compared to APP mice (Fig. 3F; non-significant trend, p=0.06). These results indicate that a deficiency of the key complement factor C3 results in greater accumulation of insoluble Aβ in the brain and a trend for more soluble Aβ in the plasma with aging in APP tg mice.
Activation of microglia, measured by CD45-specific immunoreactivity, was significantly increased in hippocampus (p<0.01) and cortex (p<0.05) of APP;C3−/− mice (Fig. 4C, D and Fig. 5A). Because CD45 staining was associated mainly with compacted plaques and correlated well with Thioflavin S-positive plaque load (r=0.75, p<0.05, n=5 mice per group), we calculated the CD45/Thioflavin S ratio as an indicator for activation of microglia corrected by the difference in plaque load (as performed by Wilcock et al., 2006). A trend for an increased CD45/Thioflavin S ratio was observed in the hippocampus of APP;C3−/− mice compared to APP mice (p=0.10, Fig. 5B) even after correction for plaque load.
To further characterize the microglial phenotype, we used anti-Iba1 (ionized-calcium binding adapter molecule 1) immunoreactivity and Western blot levels of CD68 and F4/80 as additional markers of microglia/macrophage activity. Western blot was used because immunohistochemistry with anti-CD68 and anti-F4/80 was unsuccessful on paraffin sections. In hippocampus and cortex, Iba1 immunoreactivity was detected mainly in microglia/macrophages associated with compacted plaques but it also stained cells not directly associated with plaques, particularly in the cortex. The number of Iba1-positive cells was somewhat higher in APP;C3−/− compared to APP mice (Fig. 4I, J), however quantitative image analysis revealed no significant differences in the percent area of Iba-1 immunoreactivity between the two groups (Fig. 5C). In contrast to CD45 and Iba1, quantification of CD68 and F4/80 microglia/macrophage markers in TBS-T total brain homogenate by Western blot revealed reduced levels of each in APP;C3−/− compared to APP mice, suggesting a differential regulation of microglia, however the difference was not statistically significant. No differences in CD68 and F4/80 levels were observed in 12 month-old APP;C3−/− mice compared to APP mice (data not shown).
Next, iNOS, TNF, IL-4 and IL-10 levels, as indicators of microglia/macrophage phenotype, were examined. iNOS levels in TBS soluble brain homogenates were significantly lower by Western blot in APP;C3−/− mice compared to APP mice (Fig. 6A, p<0.05). TBS-T soluble, membrane-bound TNF levels were significantly reduced (p<0.05) by ELISA, whereas TBS soluble TNF levels showed a strong trend for reduction (p=0.05; Fig. 6B). A significant increase in IL-4 was observed by ELISA in TBS soluble brain homogenates in APP;C3−/− compared to APP mice (Fig. 6C, p<0.05). IL-10 levels in both TBS soluble and TBS-T soluble brain homogenates were elevated in APP;C3−/− mice compared to APP mice (Fig. 6D), although the differences did not reach statistical differences due to a high level of variability among the mice in each group. Taken together, these results are indicative of a shift of the microglia/macrophage towards a more alternative M2 activation phenotype in APP;C3−/− mice compared to APP mice (Mantovani et al., 2004; Morgan et al., 2005).
To determine if altered plaque pathology is due to altered expression or processing of APP protein, total APP protein levels and αAPPs fragments were quantified in brain homogenates by Western blot and normalized to β-actin levels (Fig. 7, average of two experiments, n=4 mice per group). Total APP levels were slightly increased in APP;C3−/− mice at 12 months of age (Fig. 7A), but in general, APP levels were not significantly different between the two lines of mice at either 12 or 17 months of age. No differences were observed in αAPPs levels between the two groups at either 12 or 17 months, indicating that there was no obvious effect of C3-deficiency on APP processing (Fig. 7C, D).
Anti-NeuN is a neuronal marker that labels most differentiated neurons in the adult neocortex (Mullen et al., 1992) but degenerating neurons seem to lose this marker (Larsson et al., 2001). Consequently, the amount of NeuN-positive neurons was used as a marker to determine the role of complement C3 on neuronal survival. The average number of NeuN-positive neurons per section (based on three 10 µm sections) in the CA3 region of hippocampus, determined by stereological counting, was similar between 12 month-old APP and APP;C3−/− mice (Fig. 8C, 221±8 vs 225±5; n=6 mice per group). In contrast, at 17 months of age, the average number of NeuN-positive neurons per section in the CA3 region of hippocampus was significantly reduced in APP;C3−/− mice (201α8) compared to APP mice (225α8; p<0.05) (Fig. 8 A–C). Furthermore, the average number of NeuN-positive neurons per section was significantly reduced in 17 month-old versus 12 month-old APP;C3−/− mice (p=0.015) but not in C3-sufficient APP mice. The average number of NeuN-positive neurons per section within individual mice was inversely correlated with the Thioflavin S plaque load (Fig. 8D, r=−0.636, p<0.05, n=5 mice per group), There was a more pronounced loss of neurons in the CA3 compared to the CA1 region of the hippocampus (data not shown). This may correlate with the observation of an increased expression of the hAPP protein (Fig. 4 K, L) and increased Aβ plaque load (Fig. 4 A, B, G, H) in CA3 compared to CA1 in the hippocampus of this mouse model.
MAP-2 immunoreactivity, a marker for neuronal dendrites and cell bodies, was slightly reduced in APP;C3−/− mice compared to APP mice, however the difference was not significant (Fig. 8E). Synaptophysin is used as a marker to assess presynaptic terminals and neuronal integrity. Similar to MAP-2, no significant difference was observed in synaptophysin staining intensity in the CA1 and CA3 region of hippocampus in brain sections between the two groups (data not shown). Synaptophysin levels, quantified by Western blot of TBS-T brain homogenates, were similar between APP and APP;C3−/− mice at 12 months of age (data not shown), but showed a tendency to be lower in APP;C3−/− compared to APP mice at 17 months of age, however this difference was not significant (Fig. 8F).
A reduction of neuronal numbers might possibly be due to a reduction in neurogenesis, especially because it was recently reported that complement C3 and its receptor C3aR play a role in basal and ischemia induced neurogenesis (Rahpeymai et al., 2006). Neurogenesis usually takes place in the subventricular zone (SVZ) and in the granular cell layer (GCL) of the dentate gyrus (reviewed by Lledo et al., 2006). Therefore, we used Doublecortin (DCX) as a marker for newly generated neurons and quantified DCX staining in the SVZ and GCL. Quantification of DCX positive cells in the GCL, the potential source of new neurons in the hippocampus, was not possible as the labeling of DCX positive cells in the GCL of 12 and 17 month-old APP or APP;C3−/− mice was below the limit of detection unlike the clusters of DCX positive cells in the SVZ. This finding is in agreement with earlier studies characterizing neurogenesis in this animal model (Jin et al., 2004) showing very low levels of neurogenesis in the GCL.
To experimentally address the role of complement C3 in the pathogenesis of AD, we generated C3-deficient APP transgenic mice and analyzed them neuropathologically and biochemically at 8, 12 and 17 months of age, comparing them to age-matched APP transgenic mice. At younger ages (8 and 12 months), no significant differences were observed between APP and APP;C3−/− mice in plaque load, biochemical levels of Aβ, or in any of the neuronal or glial markers examined. In contrast, at 17 months of age, the APP;C3−/− mice showed a significant ~two-fold increase in total Aβ and fibrillar amyloid plaque burden that correlated with significantly increased guanidine soluble Aβ42 levels and reduced TBS soluble Aβ42 and Aβ40 in brain homogenates, a non-significant trend for increased plasma Aβ levels, a significant reduction of NeuN positive neurons in the CA3 region of the hippocampus, and a shift of microglia/macrophage towards a more alternative activation/M2 phenotype (according to Mantovani et al., 2004; Morgan et al., 2005). Our findings demonstrate that the complete absence of the central complement component, C3, accelerates AD-like plaque pathology with aging once plaque pathogenesis is underway. Although many of the differences between the APP:C3−/− and APP mice reported here were statistically significant, many comparisons resulted in a non-significant trend due to the relatively small number of mice studied and the high variability observed in APP transgenic mice, in general. However, these trends provide additional support to the overall findings.
The rather late effect of C3-deficiency may be related to an attempt of APP mice to protect the brain by upregulating complement C3 once AD-like pathogenesis is underway. Similar to a previous report by Wyss-Coray and colleagues (2002), we observed an increase in C3 protein levels with aging and AD pathogenesis in APP mice. In their study, Wyss-Coray and colleagues demonstrated that C3 mRNA levels increased with age in APP mice, and that C3 protein levels were elevated when TGFβ was overexpressed in APP mice (Wyss-Coray et al., 2002). Inhibition of complement C3 convertase (to block activation of C3) by overexpression of sCrry in APP mice resulted in increased plaque burden and neurodegeneration, even in the presence of C3 protein (Wyss-Coray, 2002).
Consistent with an important role of complement in late stage AD-like pathogenesis is a study (Matsuoka et al., 2001) showing the colocalization of complement-activating C1q and an upregulation of C1q with increased formation of fibrillar Aβ plaques in a PS1/APP transgenic mouse model. In addition, we previously reported prominent C1q and C3 immunoreactivity in highly compacted Aβ42-positive neuritic plaques associated with microgliosis in the cortex of middle-aged and older individuals with Down syndrome (Stoltzner et al., 2000). Whether the elevation of complement proteins in AD brain is an attempt to protect the brain or a consequence of neuronal damage is unclear but our results, along with those of Wyss-Coray (2002), suggest that complement C3 may play a protective role in the brain.
Amyloid fibrils have been detected in microglia in human AD brain suggesting that microglia are involved in the clearing of Aβ protein deposits (Wegiel and Wisniewski, 1990). Furthermore, it has also been shown in multiple studies that different forms of aggregated Aβ activate and become bound by complement opsonins such as C3b (Webster et al., 1997; Bradt et al., 1998) which facilitates CR3-mediated phagocytosis of Aβ (Ehlers, 2000). Recent mouse studies report a reduction of the microglial markers F4/80 and I-A/I-E (marker of MHCII alloantigen) in old C1q-deficient Tg2576 (APP) mice (Fonseca et al., 2004). In addition, F4/80-positive microglia were reduced in APP/sCrry mice (Wyss-Coray et al., 2002). By Western blot, we, too, found a non-significant trend for reduced F4/80 as well as CD68 levels in our 17 month-old APP;C3−/− mice. Therefore, it is possible that the increased fibrillar plaque load observed in our C3-deficient APP tg mice may be due to less efficient phagocytosis of Aβ fibrils in the absence of C3 due to lack of C3b or iC3b- mediated opsonization.
Interestingly, although F4/80 and CD68 levels were reduced in our 17 month-old APP;C3−/− mice compared to age-matched APP mice, we found increases in other microglia/macrophage markers of activation including CD45 (significant) and Iba1 (non-significant trend), correlating with increased Aβ plaque burden and neurodegeneration. C3-deficiency in 17 month-old APP mice resulted in increased IL-4 (p<0.05 in TBS soluble fraction) and IL-10 (non-significant trend) and, reduced iNOS (p<0.05) and TNF (p<0.05 in TBS-T membrane-bound fraction) brain levels. Others have demonstrated that anti-inflammatory cytokines, such as IL-4 and IL-10, increased fibrillar Aβ phagocytosis by murine primary microglia in vitro (Koenigsknecht-Talboo et al., 2005). Although anti-inflammatory cytokine levels were elevated in the APP;C3−/− mice in our study, Aβ deposition was increased as well, indicating that Aβ phagocytosis by microglia was ineffective at removing Aβ deposits. Thus, it is possible that the presence of complement C3 may be necessary for anti-inflammatory cytokines to stimulate microglial phagocytosis of aggregated Aβ.
Taken together, our findings indicate a shift of the microglia/macrophage response towards an alternative M2 activation phenotype (Manotovani et al., 2004; Morgan et al., 2005) in the absence of complement C3. M2-type microglia/macrophage are often found in association with apoptotic cells to scavenge debris and promote tissue repair which is in agreement with the increased neurodegeneration we observed in the 17 month-old APP;C3−/− mice. Upregulation of CD45, which is expressed at high levels in infiltrating microglia/macrophage (Ford et al., 1995), could represent enhanced infiltration of peripheral macrophages due to increased fibrillar Aβ deposits and a higher incidence of dying cells in brain. It should be noted that in the study of Wyss-Coray et al. (2002) inhibition of C3 convertase by sCrry overexpression may have only affected certain microglial functions such as Aβ phagocytosis, while the complete absence of C3 may have additional affects on microglial function and/or molecules that normally suppress microglial activation along the phagocytic and cytotoxic pathways. Ours is the first study to fully examine the role of complement C3 on the microglia/macrophage phenotype and cytokines in APP mice.
Alternatively, increased Aβ deposition in brain and Aβ levels in plasma in C3-deficient APP tg mice may also be due to reduced peripheral degradation of Aβ. Indeed, a recent study suggested an important role of complement C3 in the peripheral clearance of Aβ by C3b-dependent adherence to complement receptor 1 (CR1) on erythrocytes in blood of humans (Rogers et al., 2006). A recently described complement receptor, CR1g, was found to be important for C3b-dependent clearance of pathogens from the blood (Helmy et al., 2006) and thus, may represent an additional pathway for clearance of complement-opsonized Aβ in the periphery. Such a mechanism should be affected in C3-deficient APP mice. Indeed, the absence of C3 resulted in the accumulation of Aβ in the periphery, thereby increasing the amount of peripheral Aβ available for influx into the brain.
In contrast to guanidine-soluble Aβ42 levels which were increased with the fibrillar Aβ plaque load, TBS-soluble and to some extent also TBS-T soluble Aβ levels were reduced in the brains of C3-deficient APP mice. Therefore, it is possible that complement C3 also plays a role in suppressing aggregation of Aβ in which case, the lack of C3 would result in reduced soluble Aβ levels in brain and increased deposition of aggregated Aβ into plaques. However, this seems unlikely as C3 levels were elevated in the complement-sufficient APP tg mice with age as Aβ became more aggregated.
Increased amounts of fibrillar Aβ deposition in 17 month-old APP;C3−/− mice correlated with a reduction of NeuN positive neurons in hippocampus, suggesting a role for complement C3 in neuronal survival and health. Neuronal loss was highest in CA3 compared to CA1, possibly due to increased hAPP transgene expression and Aβ deposition. Neurons express C3 thus the neurons in C3-deficient APP mice may be more vulnerable to the cytotoxic effects of Aβ and/or microglial activation. However, C3-deficient neurons in 12 month-old APP;C3−/− mice appeared relatively healthy. MAP2 and synaptophysin immunoreactivity were non-significantly reduced in APP;C3−/− mice, in agreement with the findings of Wyss-Coray (2002) in which a similar neuronal phenotype was observed upon inhibition of C3 convertase. By electron microscopy, they reported an accumulation of degenerating neurons in APP/sCrry mice compared to wildtype mice that correlated with the reduction of NeuN-positive neurons. Our findings agree with other studies that have proposed a neuroprotective effect of complement proteins and complement activation products such as C3a and C5a (van Beek et al., 2003).
A recent study attributes an important role to complement C3 and its receptor C3aR in basal- and ischemia-induced neurogenesis in young adult mice (Rahpeymai et al., 2006). The reduction of NeuN positive neurons in APP;C3−/− mice may be due to a reduction of neurogenesis over the lifespan of C3 deficient mice. However, because the average number of hippocampal CA3 NeuN-positive neurons per section was similar in 12 month-old APP;C3−/− versus APP mice this seems unlikely. Quantification of DCX positive cells in the granular cell layer (GCL) of the dentate gyrus, the source for new hippocampal neurons, was not possible as the DCX staining in GCL was below detectable levels in the 12 and 17 month-old APP and APP;C3−/− mouse brain sections (in contrast to DCX positive cell clusters in the SVZ), consistent with earlier reports in APP mice (Jin et al., 2004).
In addition, C1q and C3 were recently reported to play a role in CNS synapse elimination as shown by the failure of C1q-deficient and C3-deficient mice to fully refine retinogeniculate connections during postnatal development (Stevens et al., 2007). Complement-independent mechanisms of synapse elimination may also exist (as suggested by the authors) and may have allowed synapse elimination during development in our APP;C3−/− mice. Furthermore, initial studies in a glaucoma mouse model by the same group suggest that complement-mediated synapse elimination may become aberrantly reactivated under disease conditions via upregulation of C1q at the synapse leading to synaptic and neuronal degeneration. The presence and possible upregulation of complement C1q in our APP;C3−/− mice with aging may have contributed to the synapse reduction and neuronal loss we observed at 17 months of age.
In summary, we demonstrate that C3-deficiency in APP mice resulted in increased cerebral Aβ deposition and neuronal loss as well as a shift to a M2 microglial/macrophage response, thereby supporting a beneficial, neuroprotective role of complement C3 in brain. Further studies are underway to further tease apart the various mechanisms by which complement proteins influence neuronal health at different ages.
The authors would like to thank Guiquan Chen, Admar Verschoor, Diana Li, Eva Luo, Wei Liu and Weiming Xia for help with experiments, Dan Frenkel for providing some APP mice and Lennart Mucke for sharing the APP (J20) transgenic mouse line. This work was supported by NIH grant AG20159 (C.A.L.).