Aβ Immunotherapy: Intracerebral Sequestration of Aβ by an Anti-Aβ Monoclonal Antibody 266 with High Affinity to Soluble Aβ

Amyloid β (Aβ) immunotherapy is emerging as a promising disease-modifying therapy for Alzheimer's disease, although the precise mechanisms whereby anti-Aβ antibodies act against amyloid deposition and cognitive deficits remain elusive. To test the “peripheral sink” theory, which postulates that the effects of anti-Aβ antibodies in the systemic circulation are to promote the Aβ efflux from brain to blood, we studied the clearance of 125I-Aβ1-40 microinjected into mouse brains after intraperitoneal administration of an anti-Aβ monoclonal antibody 266. 125I-Aβ1-40 was rapidly eliminated from brains with a half-life of ∼30 min in control mice, whereas 266 significantly retarded the elimination of Aβ, presumably due to formation of Aβ-antibody complex in brains. Administration of 266 to APP transgenic mice increased the levels of monomer Aβ species in an antibody-bound form, without affecting that of total Aβ. We propose a novel mechanism of Aβ immunotherapy by the class of anti-Aβ antibodies that preferentially bind soluble Aβ, i.e., intracerebral, rather than peripheral, sequestration of soluble, monomer form of Aβ, thereby preventing the accumulation of multimeric toxic Aβ species in brains.


Introduction
Immunization with amyloid ␤ peptide (A␤), the pathogenic protein in Alzheimer's disease (AD), or passive transfer of anti-A␤ antibodies, have been shown to be effective at reducing the amyloid burden and reversing the memory deficiency phenotype in APP transgenic (Tg) mice Janus et al., 2000;Morgan et al., 2000). These observations bolstered the clinical development of A␤ immunotherapy for treatment of AD, although a long-term follow-up study of A␤ immunization in patients with AD showed that clearance of A␤ deposits did not necessarily prevent progressive neurodegeneration (Holmes et al., 2008). Anti-A␤ antibodies are presumed to be the effector molecules in A␤ immunotherapy, although the precise mechanisms, as well as the site of action of anti-A␤ antibodies in immunotherapy remain elusive. A class of anti-A␤ antibodies directed to the A␤ N terminus have been shown to act within the CNS by binding to A␤ aggregates in plaques, triggering microglial phagocytotic clearance of amyloid plaques through an Fc receptor-mediated mechanism (Bard et al., 2000), or inhibiting aggregation or neurotoxicity of A␤ (McLaurin et al., 2002;Mamikonyan et al., 2007). Non-Fc-mediated mechanisms may be involved in clearance of A␤ plaques because F(abЈ) 2 fragments that lack the Fc region of the antibody also are effective . Another proposed mechanism of A␤ clearance is that the site of antibody action is in the periphery, where the anti-A␤ antibodies would sequestrate soluble forms of A␤ in the peripheral circulation and drive an efflux of A␤ from the brain to the blood plasma, providing a "peripheral sink" for A␤ clearance (DeMattos et al., 2001(DeMattos et al., , 2002aLemere et al., 2003). This hypothesis was proposed based on data on passive transfer of an anti-A␤ monoclonal antibody (mAb) 266, which recognizes the midportion of A␤ and has a high affinity to soluble A␤ but not to aggregated ␤-amyloid (Seubert et al., 1992;DeMattos et al., 2001), in PDAPP Tg mice. Intraperitoneal administration of 266 resulted in a rapid and robust increase in the level of plasma A␤, and chronic treatment with 266 reduced A␤ deposition in the brains of PDAPP mice, leading to the notion that 266 might have reduced brain A␤ by changing the equilibrium of A␤ levels between brain interstitial fluids and blood, thereby accelerating the A␤ efflux and causing an acute improvement of learning and memory (DeMattos et al., 2001;Dodart et al., 2002). Other studies have not observed cerebral amyloid reduction with 266 therapy, and attributed the increased plasma A␤ levels to a reduced clearance rate of A␤ complexed to antibody (Seubert et al., 2008). Moreover, direct evidence for whether anti-A␤ antibodies in the blood acutely accelerate the efflux of A␤ from the brain, has not been shown.
In this study, we conducted further examination of the mechanism of action of mAb 266 in vivo, using brain injection of radiolabeled A␤ and biochemical analysis of A␤ in the brains of APP transgenic mice, and showed that mAb 266 acts within the brain parenchyma by binding to and stabilizing the soluble, monomeric form of A␤. Our observation argues against the peripheral sink theory and proposes a novel mechanism whereby A␤ antibodies act in A␤ immunotherapy, i.e., sequestration of soluble A␤ within the CNS, not in the peripheral blood stream.

Materials and Methods
Animals. Seven-to nine-week-old male C57BL/6J mice were purchased from Charles River Laboratories. A7 are transgenic mice that overexpress human APP695 harboring K670N, M671L, and T714I FAD mutations in neurons under the control of Thy1.2 promoter. The expression level of the transgenic human mutant APP is ϳ1.4-fold that of endogenous murine APP. These mice develop progressive amyloid deposition in cerebral cortices at the age of ϳ9to 12-month-old in an age-dependent manner (supplemental Fig. S1, available at www.jneurosci.org as supplemental material). All animals were maintained on food and water with a 12 h light/dark cycle. All experiments were approved by the Institutional Animal Care and Use Committee of the Graduate School of Pharmaceutical Sciences, The University of Tokyo.
Brain microinjection of 125 I-A␤ 1-40 and assay of efflux from brain. 10D5, 266, and LB509 were administrated by intraperitoneal injection at a dose of 300 g. One, twenty-four, or one hundred twenty hours after antibody administration, 125 I-A␤ 1-40 was injected into the cerebral cortices, and its clearance was monitored as previously described (Kakee et al., 1996). Briefly, mice were anesthetized with an intraperitoneal injection of pentobarbital (100 mg/kg) and placed on the heating pad to maintain the body temperature. A burr hole was made right above the parietal cortex area 2 (Par2) region, and a unilateral injection of the applied solution (0.30 l) containing [ 125 I]-A␤ 1-40 (0.007-0.02 Ci) and [ 14 C]-inulin (0.20 Ci) in an extracellular fluid (ECF) buffer (122 mM NaCl, 25 mM NaHCO 3 , 3 mM KCl, 1.4 mM CaCl 2 , 1.2 mM MgSO 4 , 0.4 mM K 2 HPO 4 , 10 mM D-glucose, and 10 mM HEPES, pH 7.4) was made using a Hamilton syringe over a period of 30 s. The target coordinate (Par2) was 0.0 mm anterior/posterior and 3.9 mm lateral to bregma, and 2.5 mm ventral to the surface of the skull, determined by using a stereotaxic frame (SR-6; Narishige). The precise location of the target region was fixed according to Franklin and Paxinos (2007). Par2 region was selected because the transfer of the molecule to CSF or contralateral hemisphere is reported to be minimal upon injection into this region (Kakee et al., 1996). The concentration of 125 I-A␤ in the solution was ϳ10 -30 nM. After microinjection, the microsyringe was kept in place for an additional 4 min before gentle withdrawal to minimize backflow. At designated times, the injected side of the brain was excised and dissolved in 2 ml of 2 M NaOH at 60°C for 1 h. The [ 125 I] radioactivity of the samples was measured in a gamma counter (ART300, Aloka). The samples were then mixed with 14 ml of Hionic-fluor (Perkin-Elmer), and the [ 14 C] radioactivity was measured in a Figure 1. Brain injection and clearance of 125 I-A␤ 1-40 after peripheral administration of anti-A␤ antibodies in mice. A, Time course of intraperitoneal administration of antibodies and 125 I-A␤ injection. 125 I-A␤ was injected into the cortex of mice brains 1, 24, or 120 h after antibody administration. The radioactivities of A␤ remaining in the brain were calculated as 100 Ϫ BEI (%) at 15, 30, 60, or 120 min after injection of 125 I-A␤. B, Time course of elimination of 125 I-A␤ from brains 120 h after intraperitoneal administration of 10D5, 266, or LB509 as a control IgG. The means Ϯ SEMs in three to eight independent assays are shown. **p Ͻ 0.01, ANOVA. C, The remaining 125 I-A␤ at 60 min after microinjection of 125 I-A␤ at 1, 24, or 120 h after intraperitoneal administration of 266 or LB509. The means Ϯ SEMs in three to eight independent assays are shown. *p Ͻ 0.05, **p Ͻ 0.01, by Student's t test.

Figure 2.
Formation of 125 I-A␤/antibody complex in mouse brains. A, Peripheral administration of antibodies 10D5, 266, or LB509 followed by intracerebral injection of 125 I-A␤ and protein G precipitation of antibodies from brain extracts. 125 I-A␤ was injected 5 d after intraperitoneal administration of antibodies. After extensive transcardiac perfusion with PBS, brains were removed and extracted with RIPA buffer and the resulting supernatants were precipitated by protein G agarose. The radioactivity derived from 125 I-A␤ bound to the protein G-precipitated antibody as a percentage of total injected 125 I-A␤ is shown. The means Ϯ SEMs in three independent assays are shown. ***p Ͻ 0.001, ANOVA. B, Coinjection of 125 I-A␤ with antibodies 10D5, 266, or LB509 (1 mg/ml) into brains. Apparent elimination rate constant (k app,el %) of 125 I-A␤ efflux is indicated as a percentage of that with LB509 (control antibody) as 100%. The means Ϯ SEMs in three independent assays are shown. **p Ͻ 0.01, ANOVA.
liquid scintillation counter (Perkin-Elmer). The brain efflux index (BEI), which represents the percentage of 125 I-A␤ effluxed from the brain, was defined by Equation 1, and the percentage of 125 I-A␤ remaining in the ipsilateral cerebrum (100 Ϫ BEI) was determined using Equation 2: BEI (%) ϭ 125 I-A␤ undergoing efflux at the BBB 125 I-A␤ injected in the brain ϫ 100 (1) 100 Ϫ BEI (%) ϭ amount of 125 I-A␤ in the brain amount of 14 C-inulin in the brain amount of 125 I-A␤ injected amount of 14 C-inulin injected ϫ 100 (2) In the brain coinjection study, 125 I-A␤ was mixed with 600 ng/ml antibodies (10D5, 266, or LB509) and was injected into the Par2 as above. The apparent elimination rate constant (k app,el ) was determined from the slope given by fitting a semilogarithmic plot of (100 Ϫ BEI) versus time (Shiiki et al., 2004). The 125 I-A␤ level in blood plasma at 60 min after brain microinjection in control IgG-and 266-treated mice were 8.65 Ϯ 0.53% and 7.34 Ϯ 0.69% (n ϭ 4, each), respectively, which were considerably lower than the predicted levels of 125 I-A␤ to be detected outside the brain (ϳ85% in control IgG-and ϳ55% in 266treated, respectively, based on the BEI results), presumably due to the rapid uptake and degradation by peripheral organs.
Determination of binding of antibodies to fibrillized A␤. The binding of anti-A␤ antibodies to fibrillized A␤ was determined by ELISA. Microtiter wells were coated with fibrillized A␤ 1-42 at 4°C as previously described (Osada et al., 2005). Unoccupied sites were blocked with Block Ace (Snow Brand), and washed with PBS containing 0.05% Tween 20. Wells were then incubated with anti-A␤ antibodies (10D5 or 266) or LB509 for 3 h. After incubation with a horseradish peroxidase-tagged secondary antibody (GE Healthcare Life Sciences), the levels of antibodies bound to fibrillized A␤ were quantitated by development using TMB microwell system (KPL).
Brain extraction and guanidine HCl denaturation. A7 transgenic mice (4 months old) were killed 120 h after intraperitoneal injection of 600 g of antibodies (LB509 or 266). Mice were extensively perfused with PBS before brain excision. The brains were homogenized in RIPA buffer containing Complete protease inhibitor cocktail (Roche). Homogenates were centrifuged at 75,000 ϫ g for 20 min at 4°C, and the resultant supernatant was collected as brain RIPA extract. The RIPA extract was mixed with 8 M guanidine HCl (Gdn) to a final concentration of 4 M Gdn and incubated at room temperature for 30 min. These Gdn-denatured RIPA extracts of brains were diluted to 500 mM Gdn and A␤ level was analyzed by ELISA.

Figure 3. Differential extraction of brains of APP transgenic mice following peripheral administration of anti-A␤ antibodies. A,
Procedures of ELISA quantitation of A␤ in two sequentially prepared fractions from brains of APP transgenic mice. The level of "monomer A␤" is defined as the concentration of A␤ in RIPA fraction detected by monomer-specific two-site ELISA, and "Gdndissociated soluble A␤" as that of A␤ in the Gdn fraction. The levels of monomer A␤ 1-40 (B) and A␤ 1-42 (C) in the RIPA fractions of 4-month-old A7 mice 120 h after intraperitoneal administration of 266 or LB509 were quantitated by ELISA with (shaded) or without (blank) immunodepletion by protein G. The means Ϯ SEMs in four independent assays are shown. ***p Ͻ 0.001, ANOVA. D, The levels of Gdn-dissociated monomer A␤ in the Gdn-fraction of the brains of 4-month-old A7 mice 120 h after intraperitoneal administration of 266 (shaded) or LB509 (blank) were quantitated by ELISA. The means Ϯ SEMs in four independent assays are shown. E, The total A␤ and its precursor C99 in the RIPA fractions were analyzed by immunoblotting with 82E1 after immunoprecipitation with the same antibody.
Statistics. Statistical analyses were performed using Student's t test or ANOVA.

Peripheral administration of an anti-A␤ monoclonal antibody 266 retards the A␤ clearance from mouse brains
To examine whether passive immunization by peripheral administration of anti-A␤ antibodies results in a "sink effect," i.e., acute promotion of the A␤ efflux from the brain, we microinjected 125 I-A␤ 1-40 into the cerebral neocortex of mice after intraperitoneal injection of anti-A␤ or control (LB509; anti-␣-synuclein) murine mAbs, and analyzed the time course of labeled-A␤ clearance from the brain (Fig. 1 A). The radioactivity derived from 125 I-A␤ microinjected into the cerebrum of nontreated or control mAbtreated mice was rapidly eliminated from the brain with a half-life of ϳ30 min (Fig. 1 B), in agreement with the previous reports (Shibata et al., 2000;Shiiki et al., 2004;Ito et al., 2007). Mice treated with 10D5, which recognizes the N terminus and preferentially binds to fibrillized form of A␤ (supplemental Fig. S3, available at www.jneurosci.org as supplemental material), showed no change in the elimination of 125 I radioactivity compared with the rate in control mice. In sharp contrast, administration of 266, an antibody that recognizes the midportion of A␤ and prefers the soluble form of A␤, caused a marked retardation in the elimination of 125 I radioactivity from the brain (Fig. 1 B). The A␤ retention effects by 266 were similarly observed at shorter intervals of 1 or 24 h after intraperitoneal administration of antibodies (Fig. 1C). Altogether, these results indicate that passive immunization with the anti-A␤ antibodies does not cause an acute promotion of A␤ efflux from the brain, and that administration of 266 actually caused retention of A␤ in the brain.

Anti-A␤ monoclonal antibody 266 enters the brain parenchyma and binds to A␤
To examine whether peripherally administrated 266 entered the brain parenchyma and formed a complex with injected 125 I-A␤, we captured the Ig moiety of the A␤-antibody complex in the RIPA buffer lysates of brains by protein G after an extensive perfusion. Approximately 45% of the 125 I-A␤ brain-injected radioactivity was retrieved from the protein G precipitates of the RIPA-extracts of brains of 266-treated mice, suggesting that 266 formed a complex with 125 I-A␤ in the brain parenchyma (Fig.  2 A). In contrast, only ϳ10% of 125 I radioactivity was precipitated in 10D5-treated mice, a level comparable to controls. To further examine whether the formation of A␤/266 complex delayed A␤ efflux, we coinjected anti-A␤ antibodies with 125 I-A␤ into the cortices of C57BL/6J mice. Coinjection of 266 with 125 I-A␤ significantly reduced the elimination rate constant (k app,el ) of 125 I-A␤ efflux from the brain, by ϳ70% compared with those by administration of 10D5 or a control mAb (Fig. 2 B). Together, the data strongly suggest that a fraction of peripherally administered 266 entered the brain and captured 125 I-A␤, thereby inhibiting the brain clearance of A␤.

Anti-A␤ antibody 266 increases the level of the monomeric, antibody-bound form of A␤, without affecting that of the total soluble A␤, in the brain
We further analyzed the effects of peripheral administration of 266 on the relatively soluble A␤ species in the brains of young (4-month-old) Tg mice (A7) that overexpress human APP harboring Swedish/Austrian mutations and have not developed amyloid deposits yet. These mice showed an increased level of plasma A␤ upon 266 treatment (supplemental Fig. S4, available at www.jneurosci.org as supplemental material). We quantitated the levels of the monomer form of A␤ in two sequentially prepared fractions, using two-site ELISAs that have been shown to specifically recognize monomer A␤, but not multimers including dimers (Enya et al., 1999) (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). We first extracted brains of A7 Tg mice with RIPA buffer (RIPA fraction), and further denatured the RIPA fraction with 4 M Guanidine HCl (Gdn fraction) (Fig. 3A). Thus, A␤ detected in the RIPA fraction represents A␤ species that were present in monomeric state (monomer A␤) in the brain; it may either be in antibody-bound or unbound forms, because the ELISAs detected 266-bound monomer A␤ at a comparable sensitivity to unbound form (supplemental Fig. S5, available at www.jneurosci.org as supplemental material). In contrast, A␤ detected in the Gdn fraction by the monomer-specific ELISAs should include A␤ that was present in the RIPA-extractable fraction in multimeric state or whose epitope has been masked before Gdn denaturation (Gdn-dissociated monomer A␤).
The levels of monomer A␤ in the RIPA fraction of A7 APP Tg mice were significantly elevated by 266 treatment by approximately twofold for A␤ 1-40 and approximately threefold for A␤ 1-42 (Fig. 3 B, C). These increases in the levels of monomer A␤ 1-40 and A␤ 1-42 were not observed after protein G immunodepletion of the RIPA fraction, suggesting that the increased A␤ was complexed with 266 in the brain parenchyma (Fig. 3 B, C), whereas the levels of Gdn-dissociated monomer A␤ were not altered (Fig.  3D). The total level of A␤ in the RIPA fraction detected by immunoprecipitation by an anti-A␤ mAb 82E1, whose binding is not competed by 266 (supplemental Fig. S4C, available at www. jneurosci.org as supplemental material), also was not altered by 266 treatment (Fig. 3E). Together, we concluded that peripherally administered anti-A␤ mAb 266 entered the brain and selectively bound and stabilized the soluble, monomer form of A␤, without affecting the total level of A␤ in the soluble fraction of brains of APP Tg mice.

Discussion
Previous studies have been mixed as to whether peripheral administration of a high-affinity capture anti-A␤ antibody can reduce central amyloidosis in a transgenic mouse model of AD (DeMattos et al., 2001;Seubert et al., 2008). Furthermore, administration of anti-A␤ antibodies to APP Tg mice did not alter the levels of RIPA-soluble A␤, despite a dramatic increase in plasma A␤ (Levites et al., 2006), questioning the ability of peripheral anti-A␤ antibodies to reduce the levels of soluble A␤ within the brain. Specifically, it has not been directly addressed whether anti-A␤ antibodies in the periphery accelerate the efflux of A␤ from brains. Combining 125 I-A␤ microinjection with passive immunization of anti-A␤ mAbs, we showed that peripheral administration of mAb 266 actually retards the efflux transport of A␤ from brain to blood stream. We further asked what types of A␤ species in multimerization state or solubility are targeted by anti-A␤ mAbs in the brains of APP tg mice using different denaturation techniques and detection by monomer A␤-specific ELISA, and showed that 266 caused an increase in the soluble, monomer form of A␤ in brain (Fig. 3 B, C). Immunodepletion by protein G significantly decreased the levels of A␤ monomer in 266-treated mice, suggesting that the increased monomer was complexed with mAb 266. In contrast, the levels of A␤ in the RIPA fraction detected after Gdn denaturation, that represent a larger pool of soluble A␤ in the brain, were not altered by administration of 266 ( Fig. 3 D, E). These data suggest that 266 binds to and sequesters monomer A␤ in the brain parenchyma, resulting in the reduction in the level of antibody-free A␤, potentially influencing the ability of free A␤ to form into oligomers and amyloid fibrils that may be neurotoxic and causative to functional deterioration (Legleiter et al., 2004) (Fig. 4). The antibody-free A␤ should comprise a subfraction of the protein G-insensitive A␤, although the total levels of the latter fraction were similar regardless of immunodepletion by protein G (Fig. 3B,C). Further studies to detect the A␤ species decreased in the antibody-free fraction, the latter being a mixture of A␤ in different multimerization and interaction states, would be crucial to the understanding of the mechanism of action of 266. It would also be interesting to see whether 266 may have different effects upon administration in older animals after amyloid deposition is established.
In addition to the potential mechanisms of action for immunotherapy including binding to and promoting clearance of insoluble plaques in the brain (Bard et al., 2000) and the perturbation of efflux from the brain by binding to plasma forms of A␤ (DeMattos et al., 2001(DeMattos et al., , 2002a, we must also consider the potential for anti-A␤ antibodies to directly bind soluble A␤ within the CNS. That this binding may have biological importance is supported by acute behavioral improvements noted in APP transgenic models following immunotherapy (Dodart et al., 2002;Kotilinek et al., 2002). Other possible mechanisms may include perturbation of APP processing by binding of 266 (Tampellini et al., 2007). Our findings modify the current under-standing of the mechanism of A␤ immunotherapy, and suggest an important contribution of antibody interactions with soluble A␤ within the CNS.