Peripheral complement interactions with amyloid β peptide: Erythrocyte clearance mechanisms

Although amyloid β peptide (Aβ) is cleared from the brain to cerebrospinal fluid and the peripheral circulation, mechanisms for its removal from blood remain unresolved. Primates have uniquely evolved a highly effective peripheral clearance mechanism for pathogens, immune adherence, in which erythrocyte complement receptor 1 (CR1) plays a major role.


Introduction
Multiple studies have made clear that amyloid b peptide (Ab) can move from the brain to the peripheral circulation [1][2][3] and from the peripheral circulation to the brain [4,5]. As such, the disposition of circulating Ab may be pathophysiologically important. For example, failure to clear Ab from blood could lead to an unfavorable concentration gradient for the movement of Ab out of the brain [3]. Moreover, the propensity of fluid-phase Ab to form insoluble fibrils and to activate complement and other inflammatory mediators could well play a role in the colocalization of inflammatory mediators with the vascular abnormalities that are observed in Alzheimer's disease (AD), as reviewed by Grammas et al. [6,7]. Mackic et al. [8,9] have provided critical data on serum and organ levels of Ab after its intravenous (IV) inoculation into the bloodstream of nonhuman primates (NHPs). However, the mechanisms by which Ab is purged from the circulation in primates still remain unclear.
Originally elucidated by Nelson et al. [10] in 1953, complement-dependent, erythrocyte-mediated clearance of circulating immune complexes ("immune adherence") has been investigated in detail for more than 60 years and is now considered a primary mechanism for pathogen removal in humans, as reviewed by Birmingham et al. [11,12]. Fig. 1 illustrates some of the major steps in this pathway, several points of which may be worth emphasizing.
First, in humans, immune adherence hinges on the expression of complement receptor 1 (CR1) by erythrocytes, a phenomenon that is unique to primates. Subprimate erythrocytes do express complement receptors (e.g., Crry), but not CR1, so that their capacity to capture complement-opsonized immune complexes appears to be significantly limited compared with human immune adherence mechanisms [19].
Second, polymorphisms in the CR1 gene have been consistently shown to be among the top genetic risk factors for AD [20][21][22][23][24], and .80% of human CR1 is devoted to the erythrocyte compartment [11,12]. Taken together with the unique expression of CR1 by primate erythrocytes, these findings make erythrocytes perhaps the most parsimonious site for CR1 to impact AD risk.
Third, although human erythrocytes only express some 200 to 1500 CR1 molecules per red cell [11], the sheer number of erythrocytes (2-3 ! 10 13 ) in the bloodstream, compared with circulating and fixed macrophages, makes this an extremely powerful and efficient pathway for pathogen clearance. For example, pathogens experimentally infused into NHPs that have been immunized against the pathogen are typically eliminated by immune adherence mechanisms in 10 to 20 minutes [25].
Fourth, immune adherence research has focused on the clearance of immune complexes [11,12]. However, both our research [15,16] and that of others [14,17] have shown that Ab, like certain bacterial and viral antigens [13], does not require immune complex formation to activate complement or to be bound by complement opsonins that serve as ligands for immune adherence. Thus, Ab (and other antibodyindependent complement activators) may have been overlooked as a substrate for immune adherence pathways.
Our laboratory first suggested that immune adherence might play a role in peripheral Ab clearance and that Fig. 1. Simplified schematic of classical pathway complement activation and immune adherence. (A) An epitope on pathogens (gray tubes) is bound by circulating antibodies (YY) specific to it. C1, the first component of the classical complement pathway, then binds to closely apposed antibodies, forming an immune complex. Notably, like certain bacterial and tumor antigens [13], Ab has been shown to bind C1 [14] and to induce activation of the C1r and C1s proteases without antibody mediation [14][15][16][17]. (B) C1s-mediated activation of the classical complement pathway ensues, including generation of C4b, C3b, and iC3b, which become covalently fixed to the antigen (black bars). C1q also remains bound to the antigen. The antigen and/or immune complex is therefore said to be "opsonized" by complement. (C) Primate (but not subprimate) erythrocytes (RBCs) express cell-surface CR1, which has C4b, C3b, and C1q as ligands. Antigen/complement complexes thus become bound to erythrocytes. (D) Erythrocytes then ferry the complex through the bloodstream until they reach specialized macrophages, Kupffer cells, lining the hepatic sinusoids. Kupffer cells recognize the complement tag via cell-surface CRIg receptors and strip off and degrade the opsonized antigen [11,12,18]. Abbreviations: Ab, amyloid b peptide; CR1, complement receptor 1.
erythrocyte Ab levels were significantly deficient in a small sample of AD and mild cognitive impairment (MCI) patients compared with nondemented elderly (NDE) controls [26]. In the present study, we provide new and more definitive, multidisciplinary evidence of immune adherence reactions with Ab, as well as confirmation, in a 140-patient cohort, of AD and MCI deficits in erythrocyte Ab capture.

Human subjects
Under institutional review board-approved protocols and consents, human IV blood samples were obtained prospectively from well-annotated, well-matched AD, MCI, and NDE subjects evaluated and diagnosed at a National Institute on Aging Alzheimer's Disease Center, Banner Sun Health Research Institute, using standard National Institute on Aging Alzheimer's Disease Center criteria. The AD group (N 5 59) had a mean age of 80.1 6 1.0 years (range 61-94 years) and consisted of 61% males and 39% females. The MCI group (N 5 19) had a mean age of 80.1 6 1.3 years (range 62-90 years) and consisted of 68% males and 32% females. The NDE group (N 5 62) had a mean age of 80.4 6 1.5 years (range 50-95 years) and consisted of 48% males and 52% females. Routine autopsy samples of AD, Parkinson's disease, and control liver were obtained from the tissue bank at Banner Sun Health Research Institute.

Nonhuman primates
NHP IV blood samples were obtained from two 16-yearold male cynomolgus macaque monkeys and one 19-yearold cynomolgus macaque under Institutional Animal Care and Use Committee-approved protocols.

Initial processing of blood samples
Serum from human and NHP subjects was obtained by drawing blood into Becton-Dickinson (Franklin Lakes, NJ, USA) Serum Vacutainer tubes. After clotting for 30 minutes, the samples were centrifuged at 1100! g for 10 minutes at 4 C and the serum was withdrawn and stored at 280 C. Plasma and erythrocytes were derived from blood drawn into Becton-Dickinson ethylenediaminetetraacetic acid (EDTA) 2K Vacutainer tubes. The plasma/erythrocyte samples were immediately spun at 1100! g for 10 minutes at 4 C. Plasma and buffy coat were removed, and the remaining erythrocytes were washed with five volumes of tris-buffered saline (TBS). Plasma was stored at 280 C. Erythrocytes were used immediately or processed to erythrocyte membranes by lysis in five volumes of doubledistilled (dd) H 2 O with 1! Protease Inhibitor Cocktail (PIC) (Roche, Basel, Switzerland) for 30 minutes at 4 C. Membranes were then pelleted in an ultracentrifuge by spinning for 40 minutes at 40,000! g, washed in five volumes of dd H 2 O with 1! PIC, centrifuged again at 40,000! g, and either used immediately or flash-frozen and stored at 280 C for subsequent experiments.

Ab preparation
Human synthetic Ab40 and Ab42 (Bachem, Torrance, CA, USA or Genscript Biotech Corporation, Piscataway, NJ, USA) were solubilized in 100% dimethyl sulfoxide (DMSO) at 10 mg/mL, gradually diluted in dd H 2 O to 2 mg/mL, and then brought to a 1 mg/mL concentration in 0.1-M Tris buffer (pH 7.4). The 1 mg/mL Ab42 stock solution was aggregated overnight at room temperature with agitation. The 1 mg/mL Ab40 stock solution was aggregated for 3 days at 37 C without agitation.

Complement activation
To assess dose-dependent Ab complement activation, Ab40 or Ab42 was diluted with 0.1-M Tris buffer to various concentrations and then incubated with an equal volume of normal human serum (NHS) (Complement Technology, Tyler, TX, USA) that had been diluted 1:5 in veronalbuffered saline containing calcium and magnesium (pH 7.4) for 1 hour at 37 C. To demonstrate specificity, 10-mM EDTA, which blocks complement activation, was added to Ab/NHS control wells immediately before the incubation. To halt further activation, 10-mM EDTA was also added to all samples after incubation. Complement activation was assessed by the measurement of C3a or sC5b-9 formation using Quidel Human C3a or Human sC5b-9 enzyme-linked immunosorbent assay (ELISA) kits (San Diego, CA, USA). All assays were performed according to the protocols provided by the manufacturer. Conventional assays of complement activation, particularly with antibody-independent activators, typically require very high concentrations of activator under in vitro conditions, as here and in virtually all previous studies of complement interactions with Ab c.f., [14][15][16][17]26,27].

Binding of Ab by complement opsonins
NHS was incubated with Ab42, as mentioned previously, to permit complement activation, generation of complement opsonins, and their covalent binding to Ab. Ab/NHS solutions were then run on conventional, reducing, sodium docecyl sulfate/polyacrylamide gel electrophoresis (SDS/PAGE) Western blots using anti-Ab antibody 6E10 (BioLegend, San Diego, CA, USA) or an antibody directed at C3b or iC3b (Quidel Corporation). In our hands, the iC3b antibody also reacts with purified C3 and its two major chains, C3a and C3b, which are produced under SDS/reducing conditions. As a control to block complement activation and opsonization of Ab, 10-mM EDTA was added to NHS before incubation with Ab.
Opsonization of Ab was also studied in vivo. Here, a 19-year-old male cynomolgus macaque was anesthetized with Telazol (5 mg/kg, intramuscular) and infused intravenously with 61-mg/kg Ab42 diluted in 0.9% sterile saline solution. Femoral artery blood samples were taken 20 minutes later and processed for plasma, as described previously. After a 1:1 dilution with sterile dd H 2 O, the plasma preparations were incubated with either monoclonal anti-C3 antibody (Abcam, Cambridge, MA, USA), which binds C3 and C3 split products (e.g., C3b, iC3b) or monoclonal anti-Ab/ amyloid precursor protein (APP) 4G8 antibody (Covance, Princeton, NJ, USA). Incubations were overnight at 4 C with gentle rocking. The samples were then subjected to immunoprecipitation (IP) using Thermo Scientific spin columns (Rockford, IL, USA). To prepare each column, 50 mL of Protein G Sepharose 4 Fast Flow (GE Healthcare Life Sciences, Pittsburgh, PA, USA) was added and the columns were prewashed with 200 mL of IP wash buffer (Thermo Fisher Scientific Inc.). Samples were loaded into the columns and incubated for 3 hours at 4 C with gentle rocking to harvest immune complexes. The columns were washed three times with 200 mL of 1! IP lysis/wash buffer (Thermo Fisher Scientific Inc.) and were additionally washed with 100 mL of 1! conditioning buffer (Thermo Fisher Scientific Inc.). Ab and C3b protein complexes were eluted with 50 mL of elution buffer (Thermo Fisher Scientific Inc.) and incubated for 5 minutes at room temperature before centrifugation at 1000! g, 4 C. To demonstrate complement opsonization of Ab, the IPbound fractions were boiled 5 minutes with Laemmli buffer containing 5% b-mercaptoethanol and then loaded into 4% to 20% Tris-glycine precast gels (Bio-Rad, Hercules, CA, USA). After electrophoresis, proteins were transferred to a polyvinylidene difluoride (PVDF) membrane using the Trans-blot Turbo Transfer System (Bio-Rad) at 25 V for 30 minutes and blocked with Blotto Blocking Buffer (Thermo Fisher Scientific Inc.) overnight at 4 C. Membranes containing protein complexes that had been immunoprecipitated with the anti-Ab antibody were immunoblotted with a 1:5000 dilution of mouse monoclonal anti-C3b antibody directed at a neoepitope specific to C3b (Quidel Corporation) or with mouse monoclonal anti-Ab/APP antibody 4G8 at a 1:1000 dilution. Conversely, membranes with protein samples that had been immunoprecipitated with the anti-C3 antibody were immunoblotted with a 1:1000 dilution of mouse monoclonal 4G8 or with mouse monoclonal anti-C3b neoepitope-specific antibody at a 1:5000 dilution. After four 5minute washes with 1! phosphate-buffered saline (PBS) containing 0.1% Tween 20, the blots were incubated with a 1:10,000 dilution of anti-mouse Alexa Fluor 680 secondary antibody (Molecular Probes, Life Technologies, Grand Island, NY, USA) and imaged using the Odyssey Imaging System (LI-COR Biosciences, Lincoln, NE, USA). Blots were then stripped and incubated with a mouse monoclonal antibody specific for b-actin (Santa Cruz Biotechnology, Dallas, TX, USA) as a control.

Erythrocyte capture of Ab in vitro
To assess the ability of erythrocytes to take up complement-opsonized Ab in blood via CR1-dependent mechanisms, two strategies were used. First, 300 mL of NHS was incubated at 37 C for 1 hour with 300 mL of 20 mg/mL Ab42 to permit complement activation and opsonization, then diluted to various concentrations in 1! TBS. The resulting NHS/Ab42 solutions were incubated with 600 mL of packed erythrocytes (in TBS) at 37 C for 1 hour. After incubation, the mixtures were spun at 100! g in a microfuge for 10 minutes, the supernatant was removed, and the erythrocytes were processed to erythrocyte membranes as described in Section 2.3. For ELISA, 200 mL of the erythrocyte membranes were solubilized in 160 mL of dd H 2 O containing PIC and 40 mL of 1% SDS for 30 minutes at room temperature. The SDS-solubilized membranes were then mixed 1:5 with Wako (Richmond, VA) ELISA sample buffer and assayed using Wako Human Ab40 ELISA kits (#298-62301) or Wako Human Ab42 ELISA kits (#298-62401). The manufacturer's protocols were followed throughout the process.
A second strategy for demonstrating specificity to complement mechanisms used a modified tip plate adhesion assay, previously used to characterize erythrocyte CR1 binding to its ligands [28]. Here, 96-well Costar high-binding microplates (Corning, Corning, NY, USA) were coated for 1 hour with aggregated Ab42, diluted to various concentrations in 10-mM carbonate buffer (pH 9.6), and then blocked using 0.5% polyethylene glycol (PEG) 3350 in 2/3 tris-buffered saline/Tween 20 (TBST) (10-mM Tris pH 7.2, 100-mM NaCl, 0.05% Tween 20). Wells were subsequently exposed for 30 minutes, room temperature, to NHS to permit complement activation and binding (Complement Technology). To demonstrate that Ab binding is mediated by complement opsonization, parallel wells were incubated with heat-inactivated NHS, C1q-depleted NHS (Complement Technology), C4-depleted NHS (Complement Technology), or EDTA 1 NHS, all of which block various stages of complement activation. Heat inactivation was for 30 minutes at 56 C. All sera were diluted 1:32 in veronal-buffered saline (with Mg 11 and Ca 11 ) (Complement Technology).
To evaluate binding, packed erythrocytes were diluted (1:2667) to 375 ppm with adhesion buffer (8-mM Tris pH 7.4, 100-mM NaCl, 140-mM dextrose, 0.45-mM CaCl 2 , 0.17-mM MgCl 2 ). From the diluted erythrocyte sample, 200 mL (w700,000 erythrocytes) was added to each well and incubated for 60 minutes at room temperature. The wells were subjected to gentle, continuous washing/aspiration using 2 mL of adhesion buffer followed by 4 mL of PBS per well.
To further demonstrate that Ab binding to erythrocytes is dependent on CR1, suspensions of erythrocytes were blocked, before exposure to Ab-coated plates, with 2.0 mL of 0.2 mg/mL anti-CR1 antibody J3D3 (Beckman Coulter, Indianapolis, IN, USA) or with 25 mL of 1 mg/mL recombinant C3b (Complement Technology) for 30 minutes, room temperature. Erythrocytes that remained bound to the plate were imaged with bright-field illumination at 100! using an inverted Olympus IX71 microscope (Olympus, Center Valley, PA, USA) and quantified using investigator-independent Im-ageJ software. Erythrocyte counts were normalized to wells coated with anti-CR1 antibody to adjust for any minor differences in the number of erythrocytes/samples.

Erythrocyte capture of Ab in vivo
After Telazol (5 mg/kg) intramuscular anesthesia, two 16-year-old male cynomolgus macaques received 183 mg/kg or 366 mg/kg Ab40, respectively, through an IV catheter inserted into the saphenous vein. The IV blood samples were taken from the same catheter at various intervals from baseline to 60 minutes after Ab infusion. After Ab infusion and after each withdrawal, the cannula was thoroughly flushed to prevent contamination of subsequent samples. The blood samples were centrifuged at 1100! g for 10 minutes at 4 C to isolate erythrocytes. The erythrocytes were then lysed and their membranes solubilized and assayed using Wako Human Ab40 ELISA kits, as described previously.

Electron microscopy of liver samples
Liver samples from rapid (,4 hours) autopsies of AD, Parkinson's disease, and NDE patients were dissected and processed using standard immunohistochemical and ultrastructural methods, as previously described in detail by our laboratory [27]. Antibodies directed at CD68 (Abcam, Cambridge, UK), a marker for Kupffer cells, and anti-Ab antibody 4G8 (BioLegend) were used. Like most antibodies to Ab, 4G8 also reacts with APP.

Ab capture by AD, MCI, and NDE erythrocytes
Erythrocyte samples from AD, MCI, and NDE participants were processed to erythrocyte membranes, as mentioned previously, solubilized in SDS, and stored at 280 C. Plasma and the solubilized membranes were subsequently assayed for Ab42 using a Covance (Princeton, NJ, USA) BetaMark Ab42 ELISA kit (now marketed by Bio-Legend).

Statistics
Parametric (analysis of variance) and Pearson correlation statistics were used throughout the process. P values are two tailed.

Ab is an antibody-independent activator of complement
To be cleared by erythrocyte/CR1-mediated mechanisms, pathogens must first activate the complement cascade. Here, we show that such activation occurs for Ab and is significantly dose dependent and complete through C5b-9 (R 5 0.93, P 5 .007), the terminal step of the classical and alternative pathways, as well as C3a (Ab40: R 5 0.96, P , .001; Ab42: R 5 0.87, P 5 .002), the step in which C3b/iC3b complement opsonins are generated (Fig. 2). Addition of 10-mM EDTA, a standard inhibitor of complement activation, reduced activation to background (i.e., serum only).

Complement activation by Ab results in its opsonization
Substances that activate complement generate cleavage fragments of C3, including C3b, iC3b, and C4b, which bind back, covalently, to the activator and are said to "opsonize" it. Here, we show that Ab intravenously infused into an NHP can be retrieved by IP with an antibody to C3b, and, conversely, that putative C3 opsonins can be retrieved by IP with an antibody to Ab. In both cases, two major bands were observed for the opsonins and Ab, both of which colocalized at w75 kD and .250 kD ( Fig. 3A and 3B). Importantly, these two bands were not observed with EDTA treatment, which abolishes complement activation and opsonization. These in vivo findings in an NHP extend two previous in vitro studies using human blood wherein colocalized bands for Ab and C3b were also detected at high molecular weights on Western blots [17,29], consistent with the fact that only Ab aggregates, particularly Ab fibrils, activate complement [16]. In addition, colocalization at the same molecular weights after IP and the reducing/denaturing conditions of the Western blot strongly suggest that Ab and C3b were covalently bound, a characteristic feature of complement opsonization.
Binding of Ab to a second complement opsonin, iC3b, was also demonstrated-here, in human blood samples exposed to Ab in vitro. Under the reducing/denaturing conditions of the present experiment, iC3b is cleaved to 39, 63, and 75 kD fragments, and, like the C3b from which it derives, iC3b remains covalently bound to activators through a thioester bond. After incubation of Ab42 with NHS to permit complement activation, Western blots for Ab exhibited bands for Ab monomer and multiple Ab oligomer species (Fig. 3C, left lane), consistent with preaggregation of the peptide. Blots of the same solution that were immunoreacted for C3/iC3b (Fig. 3C, right lanes) showed bands parallel to those for Ab at w48 to 60 kD and .110 kD. These bands putatively represent iC3b fragments covalently bound to Ab because (1) they are absent in samples treated with EDTA, which blocks iC3b formation and opsonization; (2) they remain present even under SDS/reducing conditions, consistent with the covalent binding of complement opsonins; and (3) they are shifted up from the normal molecular weights for iC3b fragments to match corresponding bands for Ab. As expected, there was also heavy labeling of bands corresponding to fragments of C3, one of the most abundant proteins in blood. For example, under SDS/ reducing conditions, the two disulfide-linked chains that comprise C3 (C3a and C3b) are observed (Fig. 3C, right  lanes). Because C3 is endogenous in NHS and does not require complement activation for its generation, immunoreactive bands at the normal molecular weights for C3a and C3b remain present in the blot regardless of whether EDTA is used.
A faint band at approximately 34 kD was also detected and may correspond to C3d, another C3 fragment generated by further cleavage of iC3b by the protease, factor I. The band is at the correct molecular weight for C3d and was abolished with EDTA treatment. Alternatively, the slow kinetics of C3d formation may not be consistent with the time scale of the present experiment [30]. Finally, the absence of any complement immunoreactivity associated with monomeric Ab, despite the large amounts that were present, confirms our previous finding that Ab monomer does not activate complement and that increasing fibrilization of Ab enhances complement activation [16].

Erythrocytes capture Ab through complementdependent processes
Once opsonized by complement, circulating pathogens in primates are bound by erythrocytes via CR1 expressed at the erythrocyte surface. When incubated with erythrocytes, Ab that had been exposed to NHS to permit complement activation and opsonization was captured by the erythrocytes in a significant dose-dependent fashion (R 5 0.99, P , .001) (Fig. 4A). To confirm this result using a second technique and to assess its specificity with respect to CR1 and the different potential ligands for CR1 (i.e., C1q, C4b, C3b), we next performed erythrocyte tip plate adhesion assays that have been previously used to assess CR1/ligand interactions [28]. Here, Ab42 was coated to the bottom of wells, exposed to NHS (to permit complement activation and opsonization) or, as controls, exposed to C1-depleted NHS, C4depleted NHS, EDTA-treated NHS, or heat-inactivated NHS, all of which inhibit complement activation and opsonization at different stages. Treated wells were then washed and incubated with erythrocytes, followed by washing to remove nonadherent cells. Incubation of Ab42 with NHS produced significant dose-dependent adherence of erythrocytes to Ab-coated wells (R 5 0.91, P 5.03; Fig. 4B). By contrast, incubation of Ab42 with heat-inactivated NHS or EDTAtreated NHS, which eliminate both alternative and classical pathway complement reactions and provide an estimate of non-complement-mediated mechanisms, reduced erythrocyte binding to Ab-coated wells to background. C1depleted serum, which abolishes classical complement pathway reactions but still permits C3b opsonization of Ab via the alternative or lectin pathways, also reduced erythrocyte adherence to background. Similarly, C4-depleted serum, which eliminates classical pathway generation of Fig. 2. Antibody-independent activation of the complement cascade. (A) Aggregated Ab42 was incubated with normal human serum and then assayed by enzyme-linked immunosorbent assay for production of C3a, a cleavage product generated from C3 following C3 activation. A significant dose-dependent response was obtained. Incubation of Ab and serum with 10-mM ethylenediaminetetraacetic acid (EDTA), which blocks complement activation, abolished the response to Ab and gave only background readings. Ab40 gave similar results. (B) These findings were also extended to the terminal step in classical and alternative pathway activation, formation of C5b-9, the membrane attack complex, and its soluble form, sC5b-9. Significant dose-dependent activation was observed in all experiments. Abbreviation: Ab, amyloid b peptide.
C4b and C3b opsonins, reduced erythrocyte adherence to background (Fig. 4B). These findings show that erythrocyte/Ab binding is predominantly mediated by classical complement pathway-dependent mechanisms (i.e., immune adherence) and depends on the generation of appropriate CR1 ligands.

Erythrocyte capture of Ab in vivo
To extend the previous findings to in vivo conditions, two cynomolgus monkeys were infused with either 183 mg/kg or 366 mg/kg of Ab40. Saphenous vein blood samples were taken at baseline and at intervals from 2 to 60 minutes thereafter. Plasma and erythrocyte Ab40 levels were tightly correlated (R 5 0.98, P , .001 and R 5 0.85, P 5 .004 for the 186 mg/kg and 366 mg/kg Ab doses, respectively; Fig. 5), with an immediate spike at 2.5 minutes and a return to near baseline within 20 minutes. These kinetics are comparable to previous studies of immune adherence with bacterial pathogens wherein .90% of plasma and erythrocyte clearance is observed within the first 10 to 20 minutes after intravenous injection [25]. Previous studies in monkeys by Mackic et al. [8,9] reported that some 97% of infused, radiolabeled Ab40 was sequestered in other organs, including brain, with only w3% to 4% retrievable in plasma. Our studies, using a direct ELISA of Ab40, gave almost identical results, including the spike and rapid fall in plasma Ab in the first 20 minutes after infusion. Clearance of Ab through the erythrocyte pathway appeared to operate on demand, such that the higher, 366 mg/kg dose of Ab was reduced to near baseline as quickly as the lower, 188 mg/kg dose. Although erythrocyte Ab40 levels were typically only w1% to 3% of plasma levels at any given and a Western blot of the same solution using an antibody that reacts with C3 and iC3b (right two lanes). C3 is abundantly present whether complement activation has occurred or not. In sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS/PAGE) gels under reducing conditions, its two major, disulfide-linked chains, C3a and C3b, therefore dominate the gel and, as endogenous constituents, are not affected by EDTA. By contrast, generation and covalent binding of iC3b to activating substrates such as Ab requires complement activation and is sensitive to EDTA. Thus, putative immunoreactivity for iC3b and its fragments (brackets) is present when complement activation is permitted (2EDTA) and absent when activation is inhibited (1EDTA). Abbreviation: Ab, amyloid b peptide. time, the erythrocyte immune adherence pathway is clearly capable of clearing normal circulating levels of Ab. For example, from 2.5 to 20 minutes after infusion of 366 mg/kg Ab, 3 ng/mL Ab was removed from the erythrocyte compartment, which is some six-fold greater than typical blood Ab levels in the monkeys (and humans).

Erythrocyte-mediated clearance of circulating Ab to the liver
In primates, specialized macrophages called Kupffer cells line the hepatic sinusoids, where they recognize and strip off complement-tagged pathogens from erythrocytes [11,12]. Low power, high power, and electron micrographs of human AD, Parkinson's disease, and NDE liver reveal Ab immunoreactivity in Kupffer cells colocalized with Ab-immunoreactive erythrocytes within the sinusoid (Fig. 6). Notably, Ab clearance to Kupffer cells was observed in all patient diagnostic conditions, suggesting that this pathway is normally used for peripheral clearance of Ab, as it is for other toxins.

Pathophysiologic significance of immune adherence in AD
We previously reported AD and MCI deficits in erythrocyte Ab levels in a small (N 5 36), retrospective set of AD, MCI, and NDE participants [26]. Here, we confirm this finding in a prospective set of samples from 140 wellannotated, well-matched AD, MCI, and NDE participants. As in our previous study [26], AD (F 5 18.1, P , .001) and MCI (F 5 7.5, P 5 .006) groups each exhibited significantly lower levels of erythrocyte Ab compared with the NDE group (Fig. 7A), and there was a significant correlation of erythrocyte-captured Ab with Mini-Mental State Examination scores (R 5 0.216, P 5 .012) (Fig. 7B). A major point left unclear by the present replication, however, is whether MCI patients are developing or have fully evolved deficits in erythrocyte capture of Ab. In our original study, MCI patients exhibited levels of erythrocyte Ab that were significantly intermediate between AD and NDE patients [26], whereas in the present study, MCI deficits were similar to those of AD patients. In either case, the findings are consistent with the growing consensus Fig. 4. Erythrocyte binding and capture of Ab. (A) Following complement activation, Ab is opsonized, tagging it for immune adherence reactions with erythrocytes. Here, NHS was incubated with various concentrations of Ab42 to permit complement activation and opsonization and then incubated with erythrocytes from the same subject. Erythrocytes captured the Ab42 in a significant dose-dependent manner. (B) When Ab42 was incubated with NHS, permitting complement activation and opsonization of the Ab, significant dose-dependent binding to erythrocytes was observed (R 5 0.91, P 5 .03). By contrast, binding was reduced to background with heat-inactivated NHS, which abolishes complement reactions. To control for the possibility that heat inactivation might also somehow inhibit non-complement-mediated erythrocyte capture of Ab, we also included use of C1-depleted and C4-depleted NHS, which are specific to complement reactions, and EDTA treatment, a standard inhibitor of complement reactions. These conditions also abolished erythrocyte binding to Ab, showing that complement mediation and formation of CR1 ligands is a primary mechanism for Ab binding to erythrocytes. (C) An excess of C3b, one of the ligands for CR1, was used to block erythrocyte C3b/CR1 binding sites, resulting in 62% inhibition of erythrocyte adhesion to Ab42-coated plates. Abbreviations: Ab, amyloid b peptide; NHS, normal human serum. that AD treatment strategies need to be inaugurated as early as possible.
Finally, as in our previous research [26], plasma Ab42 levels tended to be higher in the AD group (mean 6 standard error of the mean 5 1284 6 190 pg/mL) compared with the NDE group (mean 6 standard error of the mean 5 995 6 88 pg/mL) but did not differ significantly, consistent with the mixed results for plasma Ab reported by other investigators [31][32][33][34]. By contrast, the fraction of Ab42 captured in the erythrocyte compartment relative to the amount available in the plasma compartment was significantly lower for the AD group than the NDE group (F 5 13.2, P , .001) (Fig. 8), again confirming an AD deficit in erythrocyte clearance. As shown in the figure, it may also be notable that the amount of Ab42 in the erythrocyte compartment in all the patient groups was at least equal to or higher than (P 5 .07) the amount of Ab42 in the plasma compartment, consistent with the hypothesis that erythrocyte clearance of Ab, immune adherence, is a major player in handling circulating loads of Ab.

Discussion
The present research demonstrates for the first time that all steps in the classical immune adherence pathway are fulfilled with respect to clearance of circulating Ab and confirms our previous finding that this mechanism is deficient in AD and MCI patients compared with NDE patients [26]. Ab inoculated into human serum dose-dependently activated complement, forming complement-opsonized complexes. When incubated with human erythrocytes, complementopsonized Ab was captured in a dose-dependent manner. Specificity to complement-and CR1-dependent mechanisms rather than to nonspecific binding was demonstrated by abolition of Ab/erythrocyte adherence after heat inactivation of complement, EDTA treatment, or depletion of classical pathway components. Because these manipulations eliminate the formation of the ligands for CR1, they indirectly demonstrate that erythrocyte binding of Ab is most likely to be dependent on CR1. This conclusion was directly demonstrated by inhibiting erythrocyte CR1 binding to Ab by blocking CR1 binding sites with anti-CR1 antibody and recombinant C3b, a CR1 ligand. At the terminal end of the immune adherence pathway, Ab-immunoreactive Kupffer cells and apposed erythrocytes could be localized to the hepatic sinusoids by electron microscopy. In primates, Kupffer cells are specialized to capture complement-opsonized complexes carried by erythrocytes [11,12].
Several lines of evidence suggest that erythrocyte CR1dependent mechanisms play a pathophysiologic role in Ab clearance. First, our experiments show that peripheral erythrocyte capture of Ab is highly dependent on complement reactions with Ab and, moreover, is dependent on interactions of complement-opsonized Ab with the receptor for complement-opsonized ligands expressed on erythrocytes, CR1. Although these findings do not necessarily exclude other mechanisms for Ab capture by erythrocytes, they strongly suggest that complement/CR1 mediation is predominant.
Recent, nonoverlapping genome-wide association studies provide a second important connection of the erythrocyte immune adherence pathway to AD pathophysiology. Namely, multiple studies have found single-nucleotide polymorphisms in CR1 to be among the top genetic risk factors for AD [20][21][22][23][24]. Perhaps, because AD is a brain disorder, an underlying CR1 mechanism in brain has understandably been sought in previous studies [35][36][37][38][39][40][41]. Although the findings from these endeavors clearly confirm a role for CR1 in AD, the cell types expressing CR1 in brain-and even the presence of CR1 in brainremain controversial. Although it is possible that the CR1 in brain is modified in some way that conceals, removes, or alters epitopes to conventional CR1 antibodies [42], the fact remains that the vast majority of human CR1 indisputably resides in the erythrocyte compartment, where CR1 expression is unequivocally and universally detected [11,12]. The most parsimonious underlying basis for CR1 as an AD risk factor is, therefore, likely to be its role in peripheral clearance of Ab, where CR1 expression by erythrocytes has evolved to amplify pathogen clearance in primates and is most abundantly expressed toward that end. Consistent with this view, AD and MCI patients  [8,9], plasma levels of Ab40 spiked almost immediately after inoculation and rapidly declined over the next 15 to 20 minutes (top panel). Erythrocyte levels followed a nearly identical pattern and were significantly correlated with plasma levels at both doses of Ab (bottom panel). Clearance from the erythrocyte pathway was rapid, also consistent with previous studies [25], and was capable of reducing the high dose of Ab from nearly 10-fold normal levels to normal levels in 15 to 20 minutes (bottom panel). Abbreviation: Ab, amyloid b peptide. exhibited significant deficits in CR1-mediated erythrocyte capture of Ab, a finding we previously reported [26] and confirmed in this report with a much larger sample. Deficits in erythrocyte clearance of pathogens have also been reported in several other human disorders, including leprosy, lupus, and malaria [11,12]. No consensus mechanism for this association with disease-whether by erythrocyte immunosenescence, decreased erythrocyte CR1 expression, impaired binding of CR1 to its ligands, or other mechanisms-has been accorded, and we are exploring these possibilities in the context of AD and the reported polymorphisms of CR1.  [26], erythrocyte capture of Ab42 is significantly deficient in both AD and MCI patients. (B) Also in concert with our previous findings [26], there was a significant correlation of erythrocyte Ab42 levels with cognitive status (MMSE) score. Although the data clearly exhibit too much scatter to make erythrocyte Ab a definitive prognostic for AD, they do strongly suggest that it has pathophysiologic relevance to clinical AD progression. Abbreviations: Ab, amyloid b peptide; AD, Alzheimer's disease, MCI, mild cognitive impairment; MMSE, Mini-Mental State Examination; NDE, nondemented elderly.  ). We note that, like most anti-Ab antibodies, the 4G8 antibody used here also reacts with APP. However, the punctate, granular, intracytoplasmic labeling in these micrographs appears to be more characteristic of Ab than its precursor, amyloid precursor protein. Abbreviations: AD, Alzheimer's disease; Ab, amyloid b peptide; KC, Kupffer cell; PD, Parkinson's disease.
In addition to peripheral clearance of Ab, complement opsonization of Ab in the brain itself is also likely to have pathophysiologic significance-although not necessarily through CR1 mechanisms. Our laboratory, in fact, first reported colocalization of complement opsonins with Ab in brain [15], a finding that has been recently extended by Hong et al. [43]. The latter demonstrated that opsonized Ab associated with synapses helps target the synapses for microglial engulfment via microglial expression of another complement receptor, CR3. We have also shown direct attack of neurites in the vicinity of Ab plaques by the terminal component of complement, C5b-9, the membrane attack complex [27].
Immune adherence has historically been studied in the context of immune complexes (i.e., antigen-antibody complexes) that activate and are bound by complement opsonins, as opposed to antigens, such as Ab, that directly activate and are bound by complement in the absence of antibody. Although Ab is a relatively potent antibody-independent complement activator [14][15][16][17], antibody-dependent classical pathway activation is typically much more powerful [18]. As such, circulating endogenous anti-Ab antibodies, which are common in human subjects [44][45][46][47][48], and anti-Ab antibodies introduced by immunotherapy, as reviewed in [49], should theoretically enhance complement activation and peripheral clearance of Ab. We are also now exploring these possibilities as a means to better understand the putatively beneficial mechanisms and adverse consequences of Ab immunization.
In summary, immune adherence, the clearance of pathogens through an erythrocyte CR1-mediated process, appears to be an important mechanism for removing Ab from the circulation in humans, and deficiencies in this mechanism are likely to be pathologically relevant to AD.

RESEARCH IN CONTEXT
1. Systematic review: Although there are data on peripheral amyloid b peptide (Ab) levels after intravenous Ab inoculation in monkeys, the mechanisms for removing Ab from the blood in primates remain unclear.
2. Interpretation: Our findings demonstrate that circulating Ab is subject to a highly efficient, wellstudied pathway for clearing complement-opsonized antigens-immune adherence. This mechanism is unique to primates, deficient in Alzheimer's disease, and dependent on erythrocyte complement receptor 1, single-nucleotide polymorphisms in which are a consistent risk factor for Alzheimer's disease. Failure to remove peripheral Ab is likely to provide an unfavorable concentration gradient for its clearance from brain and to have deleterious effects on the vasculature and other organ systems in which it becomes sequestered.
3. Future directions: Peripheral Ab clearance by immune adherence should be enhanced by Ab antibodies introduced in the course of Ab immunotherapy, providing an additional explanation for why this treatment strategy helps remove Ab in brain. Fig. 8. Erythrocyte capture of Ab relative to the amount of Ab available in the plasma compartment. Note that in each group, mean erythrocyte Ab42 levels tend to be higher than those in plasma. Abbreviations: Ab, amyloid b peptide; AD, Alzheimer's disease, MCI, mild cognitive impairment; NDE, nondemented elderly.