Complement Activation in Chromosome 13 Dementias

Chromosome 13 dementias, familial British dementia (FBD) and familial Danish dementia (FDD), are associated with neurodegeneration and cerebrovascular amyloidosis, with striking neuropathological similarities to Alzheimer's disease (AD). Despite the structural differences among the amyloid subunits (ABri in FBD, ADan in FDD, and Aβ in AD), these disorders are all characterized by the presence of neurofibrillary tangles and parenchymal and vascular amyloid deposits co-localizing with markers of glial activation, suggestive of local inflammation. Proteins of the complement system and their pro-inflammatory activation products are among the inflammation markers associated with AD lesions. Immunohistochemistry of FBD and FDD brain sections demonstrated the presence of complement activation components of the classical and alternative pathways as well as the neo-epitope of the membrane attack complex. Hemolytic experiments and enzyme-linked immunosorbent assays specific for the activation products iC3b, C4d, Bb, and C5b-9 indicated that ABri and ADan are able to fully activate the complement cascade at levels comparable to those generated by Aβ1–42. ABri and ADan specifically bound C1q with high affinity and formed stable complexes in physiological conditions. Activation proceeds ∼70–75% through the classical pathway while only ∼25–30% seems to occur through the alternative pathway. The data suggest that the chronic inflammatory response generated by the amyloid peptides in vivo might be a contributing factor for the pathogenesis of FBD and FDD and, in more general terms, to other neurodegenerative conditions.

The classic hallmark lesions of Alzheimer's disease (AD), 1 cerebral senile plaques and neurofibrillary tangles (NFTs), have been known for nearly a century. During the past two decades, a wide range of inflammatory markers, typically absent or significantly reduced in the normal elderly population, were reported in AD brains (1), and accumulating evidence suggests that sustained brain inflammation might be an essential cofactor in AD pathogenesis (2,3). In this sense, immunological factors and inflammation mediators, including complement proteins and pro-inflammatory peptides generated at different stages of complement activation (4) as well as various cytokines (5), have been implicated in accelerating the progression of AD.
The complement system is a highly regulated, powerful effector mechanism of the immune system that destroys and clears deleterious substances. It is composed of more than twenty proteins that become sequentially activated in a proteolytic cascade. Originally, activation of the complement system was thought to occur only by binding of immune complexes to C1q, the recognition component of the classical pathway. However, it became then evident that the complement system can directly be activated, in the absence of antibody, by interaction of certain foreign molecules with C3 (alternative activation pathway), C1q (antibody-independent classical activation pathway), or by specific lectins on the surface of certain microorganisms (lectin activation pathway) (3,6,7).
The first step in the classical complement pathway involves the binding of an activator to C1q resulting in the subsequent conversion of the serine proesterases C1r and C1s to their active forms and, in turn, in the activation of C4, C2, and then C3. The alternative pathway differs from the classic pathway in that activation begins at the level of C3 and involves factors B and D and Properdin. Proteolytic modification of C3 by either pathway leads to the cleavage of C5 and the incorporation of C6, C7, C8, and multiple molecules of C9 resulting in the formation of the membrane attack complex (MAC), C5b-9, a transmembrane channel capable to produce cell lysis (8).
Activation-derived proteins of both the classical and alternative pathways have been demonstrated in association with AD lesions by numerous groups (3). We investigated the complement activation cascade in chromosome 13 dementias, two hereditary conditions, familial British dementia (FBD) and familial Danish dementia (FDD), that are also associated with neurodegeneration and amyloid deposition in the central nervous system. FBD has been described in members of three British pedigrees and is characterized clinically by dementia, cerebellar ataxia, and spastic paraparesis with the disease onset typically in the fourth to fifth decade of life and death occurring some ten years later (9). FDD is a disease associated with a single Danish family with the onset of cataracts in patients before the age of thirty. Affected family members subsequently develop sensory hearing loss, cerebellar ataxia, psychosis, and dementia leading to death between the ages of fifty and sixty years (10). The neuropathology in both diseases is remarkably similar to that seen in AD, including cerebral amyloid angiopathy, pre-amyloid lesions, amyloid plaques of various types, and NFTs, ultrastructurally composed of paired helical filaments with an electrophoretic pattern of abnormal hyperphosphorylated tau indistinguishable from that observed in AD. Activated microglia, expressing the major histocompatibility class II antigens that are characteristic of inflammatory processes, can be demonstrated around amyloid plaques and dystrophic neurites but not in pre-amyloid lesions in both FBD and FDD (10,11) in a topographical distribution similar to that found in AD. In these disorders the deposited amyloid proteins, ABri in FBD and ADan in FDD, are proteolytic fragments of a larger precursor molecule BriPP codified by a multiexonic gene BRI2 (also known as ITM2B) located on the long arm of chromosome 13 (12)(13)(14). The amyloid peptides originate as a result of two different genetic defects, namely a Stop-to-Arg mutation in FBD and a ten-nucleotide duplication-insertion immediately before the stop codon in FDD. Regardless of the nucleotide changes, the final outcome is common to both diseases: the ordinarily occurring stop codon is not in-frame causing the genesis of an extended precursor featuring a C-terminal piece that does not exist in normal conditions. These de novo created amyloid peptides, with no sequence identity to any known amyloid protein, are both 34-residues long, share 100% homology of the first 22 residues, and have a completely different 12-amino acid C terminus. When deposited, both ABri and ADan feature pyroglutamate at their N terminus, a post-translational modification also identified in some truncated amyloid species isolated from Alzheimer's disease brains, i.e. A␤pE3 and A␤pE11 (15)(16)(17).
The results presented here show (i) the co-localization of complement activation products with amyloid plaques and cerebrovascular amyloid deposits in FBD and FDD and (ii) the activation of both the classical and the alternative pathways of the complement system by synthetic peptides representing the main species deposited in both disorders, activation that proceeds to the terminal stages with the generation of C5b-9. The data suggest that chronic inflammation driven by in vivo complement activation may be a contributing factor to disease progression and pathogenesis in both FBD and FDD.
Complement Reagents-Pooled normal human serum (NHS), C1qdepleted serum, purified C1q protein, polyclonal goat anti-C1q antiserum, and ELISA kits for the quantitation of the activation products C4d, Bb, iC3b, and SC5b-9 were purchased from Quidel, Inc., Mountain View, CA. An EZ diagnostic kit for the assay of total complement activity (CH50) was obtained from Diamedix Corp., Miami, FL.

Immunohistochemical Studies
Sequential paraffin sections from the hippocampus, including the dentate fascia were used for C1q immunostaining in five cases with FBD (mean age at death, 62.4 years; range, 59 -68 years) and three cases with FDD (mean age at death, 53.7 years; range, 43-60 years). For the immunohistochemical detection of C4d, C5b-9, and Bb, sequential frozen sections, fixed in acetone, were used. Sections from four controls (mean age at death, 64.3 years; range, 33-88 years) without a neurological disease and seven cases with Alzheimer's disease (mean age at death, 78.2 years; range, 63-92 years) were also stained in a similar manner. Depending on the antibody, a number of different pretreatments were employed for the paraffin-embedded tissues, which are detailed in Table I. After pretreatment, the tissue sections were incubated with the pertinent primary antibodies, followed by sequential incubations with either biotinylated anti-mouse or anti-rabbit secondary antibodies, as appropriate, and the corresponding ABC complex (Dako, Denmark). Color was developed using diaminobenzidine/H 2 O 2 followed by hematoxylin counterstaining.

Characterization of Amyloid Peptides
Structural Analysis-Secondary structure was assessed by circular dichroism spectrometry in the far-UV range (190 -250 nm) at 24°C using a Jasco J-720 spectropolarimeter (Jasco, Tokyo, Japan) as previously described (19).
Peptide Solubilization and Aggregation-The different ABri and ADan peptides were solubilized in 10 mM NaHCO 3 , pH 9.6, aliquoted, and lyophilized. Before use, each aliquot was dissolved in distilled water at a concentration of 1 mg/ml and immediately used in the complement activation assays. A␤1-42 was dissolved in distilled water, and added of concentrated PBS to reach a final concentration of 150 mM NaCl. Solutions of 1 mg/ml were allowed to aggregate for 7 days at 37°C, as described (19).

Assays for Complement Activation/Consumption
Hemolytic Assays: Total Functional Activity of the Classical Pathway (CH50)-The assay relies on the lysis of sensitized sheep erythrocytes by human serum following activation of the Ca ϩ2 -and Mg ϩ2 -dependent classical pathway. Briefly, to induce complement activation, constant volumes (10 l) of NHS were separately incubated for 45 min at 37°C with 10 l of ABri, ADan, and A␤1-42 solutions containing increasing concentrations (0 -1000 g/ml in Tris-buffered saline, pH 7.4, containing 5 mM CaCl 2 and 2 mM MgCl 2 (TBS)) of the amyloid peptides. The reaction was stopped on ice and immediately analyzed for total complement activity (CH50) employing an EZ diagnostic kit in accordance with the manufacturer's specifications. Results are expressed as percentage of lysis compared with controls of 100% complement activity (no complement consumption) in which NHS was incubated with buffer in the absence of amyloid peptides.
Quantitation of Activation Products by ELISA-NHS was incubated with identical volumes of the different amyloid peptides in increasing concentrations (0 -1000 g/ml in TBS) as described above, and the reaction was stopped by the addition of EDTA to a 5 mM final concentration. The generation of complement activation products was analyzed by capture-ELISA employing C4d, iC3b, Bb, and SC5b-9 kits from Quidel in accordance with the manufacturer's instructions. Samples were diluted of 1:100 for the quantitation of C4d and Bb, 1:300 for iC3b, and 1:200 for SC5b-9 prior to the analysis by ELISA.
To estimate the contribution of the alternative complement pathway to the total activation of the system, in a separated set of experiments and using identical conditions as above, the amyloid peptides were incubated with C1q-depleted serum (reconstituted with Ca ϩ2 and Mg ϩ2 ions to a 10 mM final concentration) instead of NHS, and the generation of SC5b-9 was assessed by ELISA. Confirmation of the contribution of the alternative pathway to the total activation was achieved by quantitating the SC5b-9 generated by incubation of the amyloid peptides with NHS in the presence of Mg ϩ2 /EGTA.
Controls of Complement Activation-For aggregated IgG, the procedure was as follows: As a positive control for classical pathway activation in both CH50 and ELISA tests, NHS was separately added of identical volumes of aggregated IgG solutions of increasing concentrations (0 -1000 g/ml) and incubated at 37°C for 45 min. The NHS was subsequently analyzed for total remaining complement hemolytic activity (CH50) and generation of complement activation products by ELISA as described above. Aggregated IgG was prepared by incubating a 5 mg/ml solution of human polyclonal IgG (Cohn fraction II, Miles Laboratories, Inc., Kankakee, IL) at 63°C for 15 min as described (20,21). Following centrifugation at 2000 ϫ g for 10 min, the pelleted larger aggregates were discarded, and the soluble IgG aggregates were aliquoted and stored at Ϫ70°C.
The procedure for cobra venom factor was as follows: positive control for the generation of activation products by the alternative pathway consisted of NHS incubated for 1 h at 37°C with cobra venom factor (CVF, naja naja kaouthia, Sigma Chemical Co., St. Louis, MO) at a ratio of 2 g/100 l serum (22). The reaction was stopped by cooling on ice to 4°C, and the serum was immediately analyzed for the presence of complement activation products by ELISA, as described above.

ABri⅐C1q and ADan⅐C1q Complex Formation
The formation of the ABri/ADan⅐C1q complexes was assessed via amino acid sequence analysis. 10 g of C1q in TBS was incubated with 20 g of either ABri or ADan peptides for 1 h at 37°C. The resulting complexes were separated by electrophoresis on 1% agarose gels in 75 mM veronal buffer, pH 8.6, containing 2 mM sodium lactate and contacttransferred to PVDF membranes. Transferred proteins were stained for 1 min with 0.125% Coomassie Blue R-250 in 47% methanol, membranes were distained and extensively washed with water, and the bands of interest were excised and subjected to N-terminal sequencing on a Procise 494 protein sequencer (Applied Biosystems, Foster City, CA). Fig. 1A shows that antibody 338 specific for the C-terminal region of ABri strongly labels amyloid plaques and amyloid-laden blood vessels, including affected small arteries, arterioles, and capillaries in FBD. Both the vascular and parenchymal amyloid lesions were strongly labeled with anti-C1q (B), anti-C4d (C), anti-C5b-9 (neoepitope) (D), and anti-Bb (E) in a staining pattern similar to that seen for ABri immunohistochemistry. Diffuse deposits, defined as Congo Red and Thioflavin S-negative or weakly positive ABri parenchymal lesions (11), were only faintly stained (not shown). The FDD lesions, mainly vascular amyloid and parenchymal pre-amyloid plaques, were highlighted by antibody 5282 recognizing the C-terminal end of ADan (Fig.  1F). The anti-complement antibodies (anti-C1q (G), anti-C4d (H), anti-C5b-9 (neo-epitope) (I), and anti-Bb (J)) labeled amyloid primarily deposited in small arteries, arterioles, and capillaries. The immunoreactivity with these antibodies was weak or absent in the parenchymal lesions, which have been shown to be composed primarily of protein in pre-amyloid conformation (Congo Red and Thioflavin S-negative or weakly positive deposits) (10). Immunohistochemical analysis of AD cases, which were used as positive controls, showed labeling of both vascular A␤ amyloid deposits and parenchymal A␤-positive plaques by anti-C1q, anti-C4d, and anti-C5b-9 antibodies. Smaller numbers of the A␤-positive lesions were also stained with anti-Bb antibody (not shown). Two of the normal controls were entirely negative for complement proteins, whereas in the two normal control cases with ages of 81 and 88 years, occasional A␤-positive plaques immunoreacted only with the anti-C1q antibody (not shown).

Immunohistochemistry-
Classical Pathway of Complement Activation-In view of the presence of complement activation products in the amyloid lesions of FBD and FDD, we investigated whether their colocalization reflected a secondary phenomenon or a specific interaction between complement proteins and the deposited peptides. As an initial step, we tested in hemolytic assays the ability of ABri and ADan peptides to activate the classical pathway. As shown in Fig. 2A, incubation of NHS with increasing concentrations of ABri1-34 and ADan1-34 resulted in a concomitant decrease in the remaining complement activity (CH50) compared with NHS incubated under the same conditions in the absence of the amyloid peptides. Under our experimental conditions, both peptides consumed complement to approximately the same extent, with remaining complement values that reached a minimum of 18% for ABri1-34 and 23% for ADan1-34 at the maximal concentration tested (final concentration: 500 g/ml, ϳ120 nmol/ml). The consumption of complement induced by the ABri and ADan peptides was also similar to that of 7-day-aggregated A␤1-42 that, under the conditions tested, reduced the complement activity to 23% of the values obtained in the absence of peptide, in agreement with previously published data (23). As a positive control for the classical pathway activation, Fig. 2A also depicts the decrease in complement activity induced by incubating NHS with aggregated IgG, a known activator of the classical pathway. As it can be deduced from the data, aggregated IgG is a more potent activator of the complement system than ABri and ADan, achieving similar levels (ϳ30% of the original complement activity) at a much lower molar ratio (500 g/ml, 3.3 nmol/ml), as described (24). No differences in activation were observed among ABri or ADan peptides bearing different posttranslational modifications, i.e. N-terminal glutamate or pyroglutamate, oxidized cysteines 5 and 22, or peptides containing serine residues replacing cysteines 5 and 22 (not shown).
In view of the values obtained in the CH50 hemolytic assay, we quantitated the in vitro formation of the activation products C4d, iC3b, and SC5b-9 via specific capture ELISAs employing specific monoclonal antibodies directed against neo-epitopes originated in the activation-derived fragments (8). The C4d levels generated by incubation of NHS with ABri and ADan peptides are shown in Fig. 2B. C4d, together with C4c, are the physiological degradation products of C4b as a result of proteolytic cleavage by the complement regulatory protein Factor I in the presence of either C4-binding protein or complement receptor 1 (CR1) (25,26). The ability of both amyloid peptides to generate C4d in a dose-response manner indicates that activation of the complement system occurred through C1 activation, because the conversion of the proenzyme C1s is solely needed to produce the C4 cleavage. Proteolytic fragments of C3 (Fig. 2C) and the soluble terminal complex SC5b-9 (Fig. 2D), on the

FIG. 1. Immunohistochemical identification of complement activation products in FBD (A-E) and FDD (F-J).
In the FBD case used for illustration, the age at death was 68 years preceded by an 11-year-long period of cognitive decline. The FDD patient, whose case is demonstrated, died at the age of 43 years and had dementia in the final 3 years of life. Deposition of ABri (antibody 338) and ADan (antibody 5282) in blood vessels and parenchymal lesions in the hippocampus of FBD and FDD cases is shown in A and F, respectively. In both FBD (B-E) and FDD (G-J) the anti-C1q, anti-C4d, anti-C5b-9 neo-epitope, and anti-Bb antibodies prominently label amyloid depositing in blood vessels, including small arteries and arterioles (arrow) as well as capillaries (double arrows) with a staining pattern similar to that seen in ABri and ADan immunohistochemistry. Although in FBD the parenchymal amyloid lesions are also strongly labeled with the anti-complement antibodies, in FDD the primarily pre-amyloid deposits (arrow- other hand, may originate by activation of both the classical and the alternative pathways. Quantitation of iC3b was used to estimate C3b generation, because, once produced, C3b is rapidly inactivated by Factor I in conjunction with either Factor H or CR1 as cofactors (26,27). As indicated in Fig. 2C, both ABri and ADan were able to generate iC3b in a dose-dependent manner. The assembling of SC5b-9 shown in Fig. 2D, determined by a widely used standard method to assess complement activation (24,28), confirmed the ability of ABri and ADan to in vitro trigger the complement cascade to full completion, including the terminal stages. The terminal cytolytic C5b-9 complex generated by the assembly of C5b, C6, C7, C8, and multiple C9 molecules in the absence of a target membrane (as in the case of these experiments) binds to the naturally occurring serum S protein (vitronectin) resulting in the formation of the soluble, non-lytic form of the MAC, SC5b-9. The data in Fig. 2 also indicate that incubation of NHS with ABri and ADan peptides results in the production of activation-generated fragments of the complement proteins C4 and C3 as well as the complex SC5b-9 to levels comparable to those induced by incubation with aggregated A␤1-42, in agreement with the CH50 findings ( Fig. 2A). In addition, the levels of SC5b-9 produced by incubation of A␤1-42 with NHS confirm previously reported data acquired under similar experimental conditions (29). Similarly to the results obtained using the hemolytic assay, aggregated IgG produced a comparable level of activation-generated fragments at a lower molar ratio. In the presence of EDTA, a chelator of both Ca ϩ2 and Mg ϩ2 ions essential for the activation of the complement cascade, none of the activation products C4d, iC3b, and SC5b-9 were generated, as expected (not shown).
Alternative Pathway of Complement Activation-The ability of the ABri and ADan peptides to trigger the alternative pathway was assessed by measuring the generation of Bb by ELISA, a method that provides a direct, specific indication of alternative pathway activation (30). As shown in Fig. 3A both ABri and ADan are able to induce the production of Bb to similar levels as A␤1-42 following incubation with NHS. These values are significantly different from those originated spontaneously by incubation of NHS with buffer in the absence of amyloid peptides, which are also shown in the figure for comparison purposes. Although the Bb levels that result from incubation with the amyloid peptides are similar to those reported in certain infectious pathological conditions that take place with activation of the alternative pathway (31), they are significantly lower that those induced by incubation of NHS with CVF, a potent activator of the pathway. CVF, a functional analog of human C3b, forms stable C3/C5 convertases that are not inactivated by the plasma regulatory proteins factor H and factor I. It binds factor B rendering it available for cleavage by factor D and initiating in this way a potent alternative pathway activation.
To estimate the contribution of the alternative pathway to the total activation of the complement system, we quantitated the levels of SC5b generated by incubation of the peptides with NHS under conditions in which both pathways are activated (presence of Ca ϩ2 and Mg ϩ2 ions) and compared them with those originated by activation of the AP only (presence of Mg ϩ2 / EGTA or substitution of NHS by C1q-depleted serum). As shown in Fig. 3B, whereas addition of EDTA resulted in complete absence of SC5b-9 corroborating that activation of the system is required to produce SC5b-9, the incubation with Mg ϩ2 /EGTA reduced the levels to an average of 25%, 32%, and 31% of the total values of SC5b-9 for ABri, ADan, and A␤1-42, respectively. Corroborating these results, incubation with C1qdepleted serum reduced SC5b-9 levels to similar values (an average of 35% for ABri, 24% for ADan, and 28% for A␤1-42). Therefore, the classical pathway appears to be the major route of activation of the complement system for all the amyloid peptides tested here, representing 70 -75% of the total activation, whereas the alternative pathway accounts for the remaining ϳ25-30%.
Complex Formation with C1q-All the above data strongly suggested that the trigger of the complement cascade by ABri and ADan mainly proceeds through the classical activation pathway, most likely through a direct binding interaction of the peptides to C1q. In a set of solid-phase binding experiments, incubation of either ABri-, ADan-, or A␤1-42-coated microtiter wells with increasing concentrations of C1q (0 -20 nM) resulted in a dose-response relationship that reached saturation (Fig.   FIG. 3. Activation of the alternative pathway by ABri and ADan peptides. A, generation of Bb. To determine activation of the alternative pathway, NHS was incubated for 45 min at 37°C with ABri, ADan, and A␤1-42 peptides at a concentration that resulted in maximum classical pathway activation (final concentration, 500 g/ml). The concentration of Bb was subsequently assessed by ELISA employing Bb quantitation kit (Quidel, Inc.) as described under "Experimental Procedures." For comparison purposes the levels of Bb, produced by incubation of NHS with CVF and in the absence of any activators, are also indicated. Results show mean Ϯ S.D. of three independent experiments. B, generation of SC5b-9. To estimate the contribution of the classical and alternative pathways to the total complement activation, ABri, ADan, and A␤1-42 peptides (500 g/ml) were incubated with NHS: (i) under conditions in which both pathways are active (presence of Ca ϩ2 and Mg ϩ2 ions), (ii) under conditions in which only the alternative pathway may be activated (presence of Mg ϩ2 ions/EGTA), and (iii) in the presence of EDTA in which no complement activation takes place. Additionally the peptides were incubated in the presence Ca ϩ2 and Mg ϩ2 ions but substituting NHS with C1q-depleted serum to corroborate the activation induced by the alternative pathway. In all cases the concentration of SC5b-9 produced was analyzed by ELISA. Results represent mean Ϯ S.D. of two independent experiments. Open bar, EDTA; dotted bar, Ca ϩ2 /Mg ϩ2 ; checked bar, Mg ϩ2 /EGTA; black bar, C1q-depleted serum. 4A). Non-linear regression analysis of the binding data estimated the dissociation constants for ABri and ADan as 1.9 Ϯ 0.4 and 1.2 Ϯ 0.3 nM, respectively, within the same range to that of A␤1-42 (K d : 2.4 Ϯ 0.5 nM). These high affinity interactions correlated well with the pronounced capability of the amyloid peptides to activate the classical complement cascade in hemolytic assays ( Fig. 2A). In addition, these high affinity values suggested that the ABri or ADan peptides likely form complexes with C1q. In vitro complex formation was performed in a 50-fold molar excess of the peptides, to assure that most (if not all) the C1q was part of the complex. To visualize the final product, we took advantage of the characteristic cathodic electrophoretic mobility of C1q in agarose gels and the mostly anodic migration of the peptides in the same system. As indicated in Fig. 4B, Coomassie blue staining of the PVDF-transferred material shows C1q (lane 2) with its characteristic cathodic migration, matching the very end of the gamma region in the human serum profile shown for comparison purposes (lane 1). When complexed with ABri or ADan, the electrophoretic mobility shifted noticeable toward more anodic positions (lanes 3 and 4, respectively). To corroborate the formation of the complexes, the bands were excised and subjected to limited N-terminal sequence analysis. Three identifiable sequences were recovered in each case (see Fig. 4B for details); two sequences corresponded to the A and C chains of C1q while the other matched the N terminus of the ABri or the ADan peptides. No sequence was retrieved for the B chain of C1q known to contain a blocked N-terminal glutamine residue (pyrrolidone carboxylic acid) (protein ID no. P02746, Swiss-Prot data base of the Swiss Institute of Bioinformatics, available at www.expasy.org).
Mapping the Complement Activation Activity of ABri and ADan-To map the complement activating activity to a specific region of the ABri and ADan molecules, we employed synthetic peptides representing different regions of the amyloid subunits in the CH50 hemolytic assay. The peptides tested consisted of the common N-terminal region of the amyloid subunits (ABri/ ADan1-23), as well as the C-terminal fragments from both molecules ABri24 -34 and ADan23-34. Because the molecular mass of the peptides corresponding to different regions of the molecules differ significantly, identical molar concentrations of the peptides were used in the experiments. For comparison purposes, the molar concentrations of the full-length ABri and ADan peptides used were equivalent to the peptide concentrations displayed in Fig. 2 (e.g. 12, 48, and 120 nmol/ml, corresponding to 50, 200, and 500 g/ml, respectively). As indicated FIG. 5. Localization of the complement-activating site within ABri and ADan molecules. To promote complement activation, NHS was incubated with increasing concentrations (0 -120 nmol/ml in TBS) of the full-length ABri and ADan amyloid peptides, the common N-terminal 1-23 peptide, and the C-terminal fragments of both molecules ABri 24 -34 and ADan 24 -34. The remaining total complement lytic activity (CH50) was assessed as described in Fig. 2A. Because the molecular mass of the peptides corresponding to different regions of the molecules differ significantly, identical molar concentrations of the peptides were tested. For comparison purposes, the molar concentrations of the full-length ABri and ADan peptides employed are equivalent to those in Fig. 2A (e.g. 12, 48, and 120 nmol/ml correspond to 50, 200, and 500 g/ml, respectively). Open bar, 0 nmol/ml; light gray patterned bar, 12 nmol/ml; dark gray patterned bar, 48 nmol/ml; black bar, 120 nmol/ml. Results represent the mean Ϯ S.D. of three independent experiments.
FIG. 4. ABri⅐C1q and ADan⅐C1q complexes formation. A, binding isotherm. ELISA microtiter wells were coated with 400 ng of ABri1-34, ADan1-34, and A␤1-42 peptides and incubated for 1 h at room temperature with increasing concentrations (0 -20 nM in TBS) of C1q. Bound C1q was detected with goat polyclonal anti-C1q antiserum (Quidel, Inc., 1/1000 in TBST) followed by alkaline phosphatase-labeled swine anti-goat IgG (BioSource International, 1/5000 in TBST). The reaction was developed with p-nitrophenyl phosphate, and the Absorbance at 405 nm was quantitated. Data represent the means Ϯ S.D. of three independent experiments performed in duplicate. Binding data were analyzed by non-linear regression using GraphPad Prism (Graph-Pad Software, Inc.). f, ABri; E, ADan; , A␤. B, ABri⅐C1q and ADan⅐C1q complex formation. C1q was incubated with either ABri or ADan peptides in the presence of Ca ϩ2 and Mg ϩ2 as described under "Experimental Procedures." The resulting complexes were separated by electrophoresis on 1% agarose gels, transferred to PVDF membranes, and stained with 0.125% Coomassie Blue R-250, and the bands of interest were subjected to N-terminal sequencing. Lane 1, NHS used as a control for the electrophoretic separation on agarose; lane 2, purified C1q; lane 3, ABri⅐C1q complex; lane 4, ADan⅐C1q complex. The amino acid sequences yielded by the different bands are shown in one-letter code. In both ABri⅐C1q and ADan⅐C1q complexes, the N-terminal sequences corresponded to those of the A and C chains of the C1q in addition to the N-terminal of the ABri or ADan peptides, respectively. No sequence information was retrieved for the B chain of C1q that begins with a blocked N-terminal glutamine.
in Fig. 5, both C-terminal fragments lack the ability to activate and consume complement proteins, whereas the common N-terminal region of the amyloid molecules retains the functional activity (27% at a concentration of 120 nmol/ml for ABri/ ADan1-23 compared with 18% for ABri and 23% for ADan). The localization of the complement-activating site to the common N-terminal fragment correlates with the similar behavior of both ABri and ADan peptides in inducing almost identical levels of complement activation. In addition, the mapping of the complement-activating activity to the ABri/ADan1-23 region correlates with the capacity of the N-terminal peptide to bind C1q with almost identical affinity as the full-length peptides (ABri1-23 K d , 1.45 Ϯ 0.3 nM).
Peptide Oligomerization and Complement Activation-An important element in the activation of the complement system by amyloid peptides seems to be directly related to their degree of oligomerization. In our experimental conditions, the activation of the classical pathway takes place with freshly solubilized peptides, reaching complement activation values similar to those obtained with 7-day-aggregated A␤1-42. To clarify this issue, we examined the secondary structure and the oligomerization state of ABri and ADan peptides immediately after solubilization and after 45-min incubation at 37°C, the incubation time required for the hemolytic assays and the ELISA experiments. As shown in Fig. 6A, full-length ABri and ADan rendered circular dichroism spectra compatible with high ␤-sheet content, whereas the respective C-terminal fragments (ABri24 -34 and ADan23-34) showed a random coil configuration suggesting a role for ␤-sheet structure in the complement activating ability of the peptides. Both peptides were already heavily aggregated at the starting conditions (Fig. 6, B and C, lanes 1), although the oligomerization was even more evident after the incubation required for all the complement assays (B and C, lanes 2). Similar results were observed with the N-terminal peptide ABri/ADan1-23 (not shown). These findings indicate that the ABri/ADan peptides have a higher tendency to form oligomers than A␤1-42 and suggest a faster aggregation kinetics. Of note, although both peptides share a similarly high ␤-sheet content, ABri (B) seems to aggregate even faster than ADan (C) as indicated by the presence of higher molecular mass oligomers under the same experimental conditions. Despite this apparently different aggregation kinetics, both full-length peptides trigger complement activation to practically the same extent suggesting that other factors besides the degree of oligomerization play a role in the activation mechanism.

DISCUSSION
Although viewed for many years as an immune-privileged organ, the CNS contains many immune system components among them proteins of the complement system that are synthesized by astrocytes, microglia, and neurons. Whereas the pathogenic role of complement is well recognized in systemic disorders, its contribution to neurodegenerative diseases has only recently emerged. One of these pathologic entities in which the activation of the complement system was studied in more detail is Alzheimer's disease, the most common form of human dementia. Proteins of the classical pathway and their activation fragments, namely C1q, C3b, and C4b, as well as the terminal MAC have all been identified in senile plaques, cerebrovascular amyloid deposits, and in association with dystrophic neurites and neurofibrillary tangles, indicating that the complete cascade can be fully triggered in vivo (4,5,23,(32)(33)(34)(35). Of note, the presence of C5b-9 was demonstrated in AD but not in non-demented elderly control cortices (36), suggesting that complement-induced injury and the chronic inflammation resulting from the system activation may be at least partially responsible for the progression of the disease. Although originally not described, both mRNAs and proteins of the alternative pathway have more recently been demonstrated in AD brains together with their specific activation fragments (37). Our data demonstrate that in other unrelated neurodegenerative disorders resulting in dementia, namely FBD and FDD, complement proteins of both the classic and alternative pathways co-localize with parenchymal plaques and cerebrovascular amyloid deposits, closely resembling the findings in AD. Complement immunoreactivity in FBD and FDD was mainly associated with Congo Red/Thioflavin-positive amyloid deposits rather than Congo Red/Thioflavin-negative parenchymal pre-amyloid lesions.
FIG. 6. Peptide oligomerization and complement activation. A, circular dichroism spectrometry in the far-UV (190 -250 nm) of ABri and ADan peptides. Both full-length peptides exhibit similar ␤-sheetrich structure, whereas the C-terminal fragments ABri24 -34 and ADan23-34 show random coil configuration. Dotted lines, C-terminal fragments; solid lines, full-length peptides. Red, ABri; blue, ADan. B and C: The degree of aggregation of ABri (B) and ADan (C) peptides employed in the complement activation assays was assessed by Western blot analysis after SDS-PAGE. The synthetic peptides either immediately after solubilization (lanes 1) or following incubation 45 min at 37°C under the same conditions employed to determine their ability to activate the complement proteins (lanes 2) were loaded on 16% Tris-Tricine gels, transferred to Immobilon-P membranes, and analyzed as described under "Experimental Procedures." The fluorograms showed were developed by ECL.
Activation of the classical complement pathway in an antibody-independent manner was demonstrated for various nonimmune substances such as C-reactive protein, serum amyloid P component, DNA (38 -40), amyloid A␤ (23), and neurofibrillary tangles (24). Aggregated A␤ peptides in vitro are able to directly activate the classical complement system both in fluid phase and immobilized onto solid matrices by binding to the recognition component C1q (23,28,29,41). Our results indicate that both ABri and ADan peptides are also able to induce activation of the classical complement system through a similar mechanism. The data from the hemolytic assays, the quantitation of activation fragments by ELISA, and the immunohistochemical analysis demonstrate that both amyloid molecules can trigger the activation of the classical pathway and proceed to the terminal stages with the in vitro and in vivo generation of the terminal MAC. Their direct high affinity binding to C1q and the corresponding formation of complexes achieved under physiological conditions strongly suggest that both peptides trigger the classical pathway of complement activation mainly through direct interaction with the recognition protein C1q.
Activation of the alternative pathway may be initiated by a variety of elements or cellular surfaces, including pathogenic bacteria, parasites, viruses, and virus-infected cells. Different studies demonstrated some degree of in vitro activation of this pathway by A␤ aggregates leading to the production of the activation fragments C3b, which remained covalently bound to the fibrillar A␤, as well as of the alternative pathway-specific Bb fragment (29,42,43). The results presented here demonstrate that ABri and ADan are also able to trigger the alternative pathway resulting in the production of Bb in comparable levels to A␤1-42. However, the classical pathway appears to be the major route of activation of the complement system accounting for ϳ70 -75% of the total activation as indicated by the concentration of SC5b-9 generated under specific conditions for alternative pathway activation. These findings coincide with previous reports for A␤ peptides in which 70% of the C3 convertase activity formed by incubation of aggregated A␤1-42 with NHS originated from the classical pathway (42).
The degree of oligomerization of the amyloid peptides represents an important element in the activation of the complement system. In the case of AD, fibrillar or aggregated A␤ species activate complement in vitro, whereas non-aggregated peptides do not (28,43). Although both A␤1-40 and A␤1-42 are able to trigger the activation, on a molar basis A␤1-42 was found to be a more potent activator (41,43), a difference that most likely reflects the ability of A␤1-42 to aggregate more readily and at lower concentrations than A␤1-40 (44). In vivo, complement activation components co-localize with parenchymal and vascular A␤ amyloid deposits and are almost absent in the nonfibrillar (pre-amyloid) lesions (3,45) and in "cotton wool" plaques seen in a variant form of Alzheimer's disease due to a deletion of exon 9 of presenilin 1 (46). In the case of FBD and FDD, components of the complement activation cascade also co-localize with fibrillar but not with non-fibrillar deposits in vivo. In vitro, ABri and ADan peptides form spontaneous ␤-sheet-rich structures that exhibit very fast aggregation kinetics, modifying their degree of oligomerization even after the short incubation time required for the various assays. Under these conditions, both peptides achieve complement activation values similar to those obtained with 7-day-aggregated A␤1-42. Although both ABri and ADan are able to activate complement in vitro to practically the same extent, the in vivo accumulation of activation components in FDD parenchymal lesions is significantly lower than in FBD parenchymal deposits. This most likely reflects the fact that FDD parenchymal lesions are mainly of a pre-amyloid, non-fibrillar nature (Congo Red-negative) and thus unable to achieve high levels of complement activation, as demonstrated by the in vitro studies. Despite their similarities FBD and FDD present striking differences in their respective CNS pathology, including the more severe neocortical involvement in FDD and the nature of most of the hippocampal and neocortical lesions showing features of amyloid in FBD (11) and of pre-amyloid in FDD (10). The difference in the aggregation/fibrillization state between these two types of lesions with the concomitant difference in their ability to activate the complement system most likely accounts for the different topographical distribution of associated complement proteins. Aggregated/fibrillar deposits translate in the presence of activation-derived components in association with vascular amyloid in both diseases and with FBD parenchymal amyloid plaques, and in their absence in non-fibrillar lesions. However, the paucity of complement-derived proteins observed in the parenchymal, pre-amyloid lesions in FDD suggests that activation of the complement system may not be the only critical factor in neurodegeneration. This is also supported by similar observations in the cotton wool variant of familial AD, in which the characteristic morphological feature is the presence of plaques largely composed of Congo Red-negative preamyloid A␤ species unassociated with complement activation (46 -48).
The importance of inflammation in neurodegenerative processes, in particular in Alzheimer's disease, has become clear over recent years. The presence of dementia invariably correlates with the detection of inflammatory markers, activated microglia, and reactive astrocytes and increased levels of cytokines and complement products around amyloid plaques and dystrophic neurites (3). Epidemiological studies have shown that anti-inflammatory drugs reduce the risk of AD (49), whereas animal models of sustained CNS inflammation lose cholinergic nerve cells in the hippocampus and exhibit memory and learning impairments (50,51). These findings, together with the in vitro demonstration that complement activation can lead to cell death in both rat hippocampal and neuronal cell lines (52,53), point to the importance of chronic inflammation in the pathogenesis of AD dementia. Supporting this notion, complement activation products and inflammatory mediators have also been found in Down's syndrome in association with A␤ amyloid deposits (54) and in animal models of AD (55). Immune and inflammatory responses in the CNS are also observed in other chronic and acute neurological conditions, including multiple sclerosis, myasthenia gravis, head trauma and stroke, as well as in animal models of some of these disorders (3,7,56,57). Complement activation proteins have been also demonstrated in association with Lewy and Pick bodies in Parkinson and Pick's disease, respectively (58,59), with dystrophic neurites and early stage extracellular neurofibrillary tangles in the Parkinsonism dementia complex of Guam (60), with clusters of degenerating axons in amyotrophic lateral sclerosis (61) as well as around immunoglobulin light chain (AL) and transthyretin (TTR) amyloid deposits in peripheral nerves in both acquired and hereditary neuropathy (51,62). In Creutzfeldt-Jacob and Gerstmann-Straussler-Scheinker diseases (63), complement activation products have been found associated with amyloid plaques, and it has been recently demonstrated that the complement system plays a role in the early prion pathogenesis in transmissible spongiform encephalopathies (64,65). In this case, follicular dendritic cells participate in the prion replication before the infective agent moves through the nerves into the spinal cord or brain stem, and finally into the brain. Mice bearing deficiencies of either one of the early complement factors or complement receptors present significant delays in both the onset of disease symptoms and the splenic accumulation of the pathological prion protein following injection of infective scrapie strains indicating that activation of complement is most likely involved in the initial trapping of prions in lymphoreticular organs.
The activation of the complement system demonstrated in all these neurological diseases and the concomitant generation of opsonins and anaphylotoxins that drive numerous inflammatory mechanisms, including scavenger cell activation, chemotaxis, frustrated phagocytosis, and the secretion of cytokines, chemokines, and reactive oxygen and nitrogen species (1), may contribute to some of the neuropathological features of the different disorders. Our studies suggest that the chronic inflammatory response generated by the amyloid peptides in vivo might be a contributing (although not the solely responsible) factor to the pathogenesis of FBD and FDD and, in more general terms, to other neurodegenerative disorders. Therapeutics directed toward controlling complement activation to prevent or ameliorate the course of the disease may provide a useful approach in the management of these pathological entities.