Specific Amyloid β Clearance by a Catalytic Antibody Construct*

Background: Naturally occurring catalytic antibodies (catabodies) can hydrolyze peptide bonds. Results: A catabody engineered from innate immunity principles hydrolyzed amyloid β (Aβ) specifically, dissolved Aβ aggregates, and cleared brain Aβ deposits without evident toxicity. Conclusion: The catabody could potentially be developed as a therapy for Alzheimer disease. Significance: The innate catabody repertoire may be a source of useful catabodies to toxic amyloids. Classical immunization methods do not generate catalytic antibodies (catabodies), but recent findings suggest that the innate antibody repertoire is a rich catabody source. We describe the specificity and amyloid β (Aβ)-clearing effect of a catabody construct engineered from innate immunity principles. The catabody recognized the Aβ C terminus noncovalently and hydrolyzed Aβ rapidly, with no reactivity to the Aβ precursor protein, transthyretin amyloid aggregates, or irrelevant proteins containing the catabody-sensitive Aβ dipeptide unit. The catabody dissolved preformed Aβ aggregates and inhibited Aβ aggregation more potently than an Aβ-binding IgG. Intravenous catabody treatment reduced brain Aβ deposits in a mouse Alzheimer disease model without inducing microgliosis or microhemorrhages. Specific Aβ hydrolysis appears to be an innate immune function that could be applied for therapeutic Aβ removal.

According to the "amyloid hypothesis," soluble and fibrillar amyloid ␤ peptide (A␤) 3 aggregates contribute causally in the pathogenesis of Alzheimer disease (AD). The aggregates activate microglial inflammatory processes, exert direct neurotoxic effects, and disrupt the brain anatomic architecture (1). In addition to deposits of A␤(1-42) (A␤42) that damage the brain parenchyma, accumulation of A␤(1-40) (A␤40) in blood vessel walls causes microvasculature-related neuroinflammation and compromised blood-brain barrier (BBB) integrity (2), resulting in cerebral amyloid angiopathy (CAA) in nearly all AD patients (3). Intravenous administration of brain-penetrating A␤-binding monoclonal IgGs was proposed for AD therapy (4 -6). Such IgGs exert competing favorable and unfavorable effects (7,8). Whereas the A␤-IgG immune complexes are cleared via the Fc receptor-dependent uptake pathway by phagocytic cells (the microglia) (4), the activated cells release inflammatory mediators and neurotoxic factors (5,9). Moreover, A␤-binding monoclonal IgGs clear parenchymal A␤42, but they induce increased A␤40 deposition in blood vessel walls, enhancing the incidence of microhemorrhages and CAA (6,7) thought to be correlated with cognitive impairments (10). Reminiscent of the exogenous IgG effect, the appearance of A␤-binding autoantibodies in the cerebrospinal fluid of AD patients correlates with exacerbated CAA (11).
Catalytic antibodies (catabodies) hold potential for digesting the antigen into harmless soluble fragments with no dependence on accessory inflammatory cells. Conventional immunization procedures based on acquired immunity principles do not produce catabodies with hydrolytic rates sufficient for medical use. Recent studies suggest that catalysis is an innate property of the germ line immunoglobulin variable (V) domains that have developed over the course of Darwinian evolution (12). Degradation of several self-antigens by catabodies in autoimmune disease was reported (12,13). The innate V domain rep-ertoire expressed prior to contact with an antigen is very large, containing diverse light and heavy chain V domains (V L and V H domains) that hold potential for specific recognition of individual antigenic epitopes. We reported the catalytic immunoglobulin V domain (IgV) construct 2E6 isolated from a human IgV library (14). Here we present evidence showing that the IgV degrades and clears A␤ specifically with no evidence of microglial activation or microhemorrhages.
The IgVs were purified from bacterial periplasmic extracts by binding of the C-terminal His 6 tag to metal affinity columns followed by acid elution (pH 5.0, designated aIgV), yielding the 30-kDa intact IgV and its 18-kDa fragment in electrophoresis gels (14). Unfractionated culture supernatants were prepared by centrifugation (5,000 ϫ g, 30 min) of IgV-secreting and control bacteria grown to equivalent density as before (ϳ0.8 A 600 units). Diagnostic anion exchange chromatography of culture supernatants (33 ml) was in a neutral pH buffer (MonoQ HR 5/5 FPLC column, 1 ml/min (GE Healthcare); 20-min 0 -1 M NaCl gradient in 50 mM Tris-HCl, pH 7.4, 0.1 mM CHAPS). The partially fractionated IgV 2E6 preparation was obtained on a larger scale by two cycles of anion exchange FPLC at neutral pH (designated nIgV; from 1 liter of culture supernatant concentrated 25-fold on a 10-kDa Pellicon 2 membrane followed by dialysis against chromatography buffer). The concentrated culture supernatant contained 160 mg of total protein and 1.1 mg of IgV 2E6, determined, respectively, by the micro-BCA method and dot blotting with antibody to the c-Myc peptide tag (14). The first FPLC cycle was in the pH 7.4 chromatography buffer as before (Hitrap Q FF column, GE Healthcare; 3 ml/min). The unbound fraction containing the anti-c-Myc-reactive IgV 2E6 (retention volume, 4 -16 ml) was subjected to the second FPLC cycle on the same column at more alkaline pH using a 0 -1 M NaCl gradient over 20 min (50 mM Tris-HCl, 0.1 mM CHAPS, adjusted to pH 8.0 with Tris base). Anti-amyloid tests were done using the resultant bound nIgV 2E6 fraction (protein content 1.8 mg, IgV content 0.11 mg). The culture supernatant from control bacteria harboring empty pHEN2 vector was fractionated identically by two chromatography cycles. Control IgV MMF6 eluted in the bound fraction from the first chromatography cycle conducted as described for nIgV 2E6 (retention volume, 44.1-56.2 ml; protein content, 3.8 mg; nIgV MMF6 content, 0.11 mg). nIgV 2E6 requires bound divalent metal for maintenance of the catalytic activity (18). To avoid diluting the protein, catalytically inactive nIgV 2E6 was prepared by dialysis for 24 h against PBS containing 0.1 mM CHAPS and 10 mM EDTA, followed by dialysis against the same buffer without EDTA (to remove the chelator).
Anti-amyloid Assays-Preaggregated fibrillar A␤42 (starting A␤42 peptide concentration, 20 M; prepared as described for 125 I-A␤42) was treated with nIgVs (24 h, 37°C) in PBS, 0.1 mM CHAPS, and 1% dimethyl sulfoxide. ThT (5 M) was added, and fluorescence emission was measured after 30 min. Inhibition of fibrillization was determined similarly by treating non-aggregated A␤42 (20 M) with the nIgVs. The data were corrected for background ThT fluorescence of the same antibody without A␤42 (Ͻ5-19 fluorescence units). For potency comparisons, ThT fluorescence for reaction mixtures containing nIgV 2E6 or IgG1 59 was expressed as a percentage of the value of the control antibody with the same scaffold structure (64 -94 fluorescence units; nIgV MMF6 and IgG1 SKT03 directed to gp120) (20). Dissolution of preaggregated fibrillar A␤42 (5 M) treated with an equal volume of IgV-containing tissue culture supernatant was monitored by transmission electron microscopy using a JEOL 1400 microscope at 120 kV (5 l of reaction mixture adsorbed for 1 min on glow-discharged 300-mesh Formvar carbon grids followed by three 30-s washes with water, negative staining for 1 min with 1% uranyl acetate in water, and another three 30-s washes with water). Oligomers were prepared by incubating A␤42 (50 M) in phenol red-free Ham's F-12 medium (4°C, 24 h) (21). The oligomer preparation (total A␤ peptide concentration 40 M) was treated (24 h, 37°C) with nIgV 2E6 or MMF6 (3 g/ml) in Ham's medium/PBS containing 0.1 mM CHAPS (4:1, v/v). SDS and ␤-mercaptoethanol were added (final concentrations, 2% and 0.46 M, respectively), and the reaction mixtures were analyzed by SDS-gel electrophoresis without prior boiling. The A␤ species mass (monomers, oligo-mers, and proteolytic fragments) was computed by comparison with reference proteins (1.4 -27 kDa and 14 -97 kDa ladders; Bio-Rad). Oligomer disappearance was monitored by densitometry of the SDS-stable trimer, tetramer, and high mass oligomer bands (45 kDa band, 64 -84 kDa smear) following staining of gel blots with a mixture of mouse anti-A␤ monoclonal IgG 6E10, IgG 4G8 (both from Covance, Princeton, NJ), and IgG 6D4 (MyBioSource, San Diego, CA) (directed to A␤(1-17), A␤ (17)(18)(19)(20)(21)(22)(23)(24), and the A␤ C terminus, respectively). Freshly dissolved A␤42 without prior oligomerization was mixed immediately with the nIgVs (3 g/ml) to test the nIgV effect on oligomer accumulation over 24 h of incubation at 4°C. Experimental and control antibody effects were studied using equivalently analyzed reaction mixtures (including equivalent gel staining and imaging procedures). A␤42 binding by IgV 2E6 and IgG1 59 was tested by immunoblotting of oligomerized A␤42 electrophoresed in SDS gels (1 g of A␤42/lane). Bound IgV was visualized by staining with mouse anti-c-Myc IgG followed by peroxidase-conjugated anti-IgG. The procedure was validated previously to detect IgV-A␤ immune complexes (14,15). Bound IgG1 59 was detected by staining with peroxidaseconjugated anti-mouse IgG (22).
Brain A␤ Removal-In the first study, nIgV 2E6 or MMF6 (1 g/2 l of PBS) was injected into the right brain hemisphere neocortex of 5XFAD mice (5 months old). These mice express mutant human presenilin 1 (M146L and L286V mutations) and wild type A␤ produced from mutant human A␤ precursor protein (APP) (K670N, M671L, I716V, and V717I mutations associated with familial AD (23)). Another age-matched mouse group received similar PBS injections. Coronal brain sections of the right and left hemispheres obtained 7 days thereafter were stained with a mixture of anti-A␤ monoclonal IgGs 6E10 and 4G8, and the A␤ deposits were quantified (BIOQUANT Image Analysis Corp., Nashville, TN) (24). The A␤ plaque burden is defined as the percentage area occupied by the stained reaction product in a 640 ϫ 480-m rectangle surrounding the injection site (estimated in five 40-m sections spaced 200 m apart). The relative A␤ burden in the injected right hemisphere was computed as a percentage of the A␤ plaque burden in the corresponding non-injected left hemisphere neocortical area within the same sections from the same mouse.
In the second study, the aIgV proteins purified by metal affinity chromatography were injected intravenously on day 0 and day 3 (50 g of aIgV in PBS/injection) in TgSwDI mice (n ϭ 8/group, 7-9 months old). The mice express human A␤ with mutations outside the IgV-sensitive epitope and scissile bond region (E22Q and D23N, corresponding to the E693Q and D694N mutations in APP770) (25,26). In addition to diffuse parenchymal A␤ deposits, TgSwDI mice develop vascular A␤ deposits at an early age compared with other human A␤-expressing mouse models. Cortical A␤ burden was quantified on day 10 in five randomly selected sections as before using an MBF StereoInvestigator. The analyzed neocortical area was dorsomedial from the cingulate cortex and extended ventrolaterally to the rhinal fissure within the right hemisphere (measured field 700 ϫ 700 m). Left brain hemispheres after removing the olfactory bulb were homogenized as described (24). The hemispheres were weighed and homogenized (10%, w/v) in 20 mM Tris base, 250 mM sucrose, 1 mM EDTA, 1 mM EGTA, 100 mM phenylmethylsulfonyl fluoride, 5 g/ml pepstatin A, and a 25-fold dilution of cOmplete protease inhibitor mixture as recommended by the supplier (Roche Applied Science). Total A␤ levels were determined after solubilizing particulate A␤ in the homogenates mixed with cold formic acid (1:2.2 (v/v)) in duplicate using the A␤ 3-plex Ultrasensitive immunoassay kit (Meso Scale Discovery, Gaithersburg, MD). Soluble A␤ levels were measured similarly by ELISA in the soluble fractions of brain homogenates after treatment with cold 0.4% diethylamine containing 100 mM NaCl (24). The difference between total and soluble A␤ brain contents represents particulate A␤ dissolved by formic acid. Cerebral microhemorrhages were measured by staining with 5% potassium ferrocyanide in 10% hydrochloric acid for 30 min (Perl's iron stain; 20 sections/animal, 40-m sections spaced 400 m apart throughout the brain) (24). Cortical microgliosis was assessed according to a semiquantitative scale in 15-20 similarly obtained coronal sections per animal, stained with antibody to Iba-1 (0, a few resting microglia; 1ϩ, a few activated microglia; 2ϩ, a moderate number of activated/ phagocytic microglia; 3ϩ, numerous activated/phagocytic microglia) (24).
Brain entry of intravenously injected IgV in B6SJLF1/J mice (Jackson Laboratory, Bar Harbor, ME) was studied using 125 Ilabeled aIgV 2E6 prepared by the chloramine-T method followed by gel filtration in PBS containing 0.1 mM CHAPS and 1% BSA (Econo-Pac 10DG column, Bio-Rad; specific activity, 3.7 ϫ 10 6 cpm/g aIgV) (27). Essentially all recovered radioactivity was present in the aIgV bands identified by electrophoresis and autoradiography. Following injection of 125 I-labeled aIgV into the tail vein (1.1 g/110 l/mouse, 6.3 ϫ 10 6 cpm), the radioactivity/g of whole blood from the retroorbital plexus or whole brain obtained at euthanasia was measured using a ␥ counter (n ϭ 3 mice/time point). The nominal half-life was computed from single phase decay kinetics (cpm ϭ (cpm max ) ϫ exp(Ϫk ϫ t), where k is the decay constant and t1 ⁄ 2 ϭ ln2/k), and cpm and cpm max correspond to the observed radioactivity values at varying time points and the extrapolated maximum radioactivity value obtained by curve fitting, respectively.
Statistical Analysis-p values were from the unpaired twotailed Student's t test.

IgV 2E6 Hydrolytic Properties-Recombinant
IgV 2E6 is a single-chain heterodimer of V L domains (Fig. 1A, inset) that expressed A␤ hydrolyzing activity following purification of the IgV by acid elution from a metal affinity column (aIgV 2E6) (14). Per unit IgV 2E6 mass, the A␤ hydrolytic activity of the native IgV secreted into the bacterial supernatant was substantially superior to the acid-purified aIgV and at least comparable with neprilysin (Table 1), an enzyme with promiscuous hydrolytic activity not restricted to A␤ that is a proposed therapeutic agent for AD (28). Nanogram IgV 2E6 amounts in the unfractionated culture supernatant hydrolyzed radiolabeled 125 I-A␤40 with negligible contribution from bacterial proteases, shown by lack of hydrolytic activity of supernatants containing non-catalytic IgV MMF6 with the same scaffold structure as IgV 2E6 and mutated IgV 2E6 containing the IgV MMF6 frame-work regions (Fig. 1, A-C). Sub-ångstrom conformational transitions can reduce the catalytic activity of IgVs (14) and enzymes (29). The superior catalytic efficiency of native IgV 2E6 compared with the aIgV suggests compromised catalytic site integrity due to acid-induced conformational perturbations. For most substrate specificity and anti-amyloid tests, we used the highly catalytic IgV 2E6 fractionated partially by ion exchange chromatography at neutral pH (nIgV 2E6). The procedure removed small amounts of bacterial proteases in the supernatants without appreciable loss of catalytic activity due to conformational perturbations (Fig. 1, D-F; hydrolytic activity of the nIgV and native IgV in the bacterial supernatant, respectively, 211,555 and 181,827 cpm/h/g). The previously characterized aIgV 2E6 with lesser activity was employed for certain confirmatory anti-amyloid tests. Like the aIgV (14), the native IgV in the culture supernatant was composed of the catalytically active 30-kDa intact V L1 -V L2 IgV construct and an inactive co-purifying fragment containing the C-terminal V L2 domain (Fig. 1G). A bound divalent metal is required for maintenance of IgV 2E6 catalytic activity (18). Some anti-amyloid tests were conducted using nIgV 2E6 that had been irreversibly inactivated by EDTA chelation of the bound metal (Fig. 1H).
Specificity-aIgV 2E6 cleaved the A␤ His 14 -Gln 15 bond (14). nIgV treatment of 125 I-A␤40 generated a hydrolytic product with mass close to the predicted A␤(1-14) radiolabeled product (Fig. 2, A and B; 1,654 Da; the C-terminal A␤(15-40) fragment does not contain the radiolabel), indicating retention of the scissile bond specificity regardless of the IgV preparation method. APP fulfills an essential role in physiological neurotransmission (30). A single chain antibody fragment engineered from a catalytic V L domain degraded APP (31), highlighting the risk of interference in APP function. Intact, fulllength APP is found physiologically in two forms: the membrane-bound form and the soluble, secreted form (32,33). nIgV 2E6 did not digest the purified soluble APP substrate detectably (Fig. 2C), suggesting that the A␤ region of soluble APP assumes an IgV-insensitive conformation. Sequence-independent recognition of a generic ␤-sheet amyloid epitope could explain A␤ degradation. nIgV 2E6, however, did not digest TTR amyloid aggregates (Fig. 2C). If the specificity for A␤ derives only from His 14 -Gln 15 peptide unit recognition, the IgV is predicted to degrade other His-Gln-containing proteins. The nIgV did not degrade amphiphysin and zinc finger protein 154 containing the His-Gln unit at positions 110 and 111 and positions 57 and 58, respectively (Fig. 2C). In previous specificity studies (14), aIgV 2E6 did not degrade self-antigens and B cell superantigens containing numerous antibody-binding epitopes and protease-sensitive bonds. We analyzed synthetic A␤ peptides to assess the contribution of noncovalent epitope recognition in IgV specificity. A␤ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) containing the His 14 -Gln 15 dipeptide unit did not inhibit 125 I-A␤40 hydrolysis by the nIgV, but the C-terminal A␤ (29 -40), A␤ , and control full-length A␤40 were equipotent inhibitors (Fig. 2D). The results indicate specific noncovalent recognition of the A␤(29 -40) epitope followed by hydrolysis at the remote His 14 -Gln 15 , with no requirement for the two A␤42 C-terminal residues (Fig. 2D, inset). In membrane-bound APP and its C99 fragment generated upon ␥-secretase cleavage, the His 14 -Gln 15 scissile bond is located in the juxtamembrane extracellular protein region, and the A␤(29 -40) epitope is buried within the lipid bilayer (34). In addition to epitope conformational factors, steric inaccessibility of the epitope due to membrane burial will restrict the IgVcatalyzed hydrolysis of membrane-anchored APP.
Amyloid Dissolution-A␤ self-assembles into ␤-sheet fibrils and soluble oligomers. Treatment of prefibrillized A␤42 with tissue culture supernatants containing IgV 2E6 but not IgV MMF6 dissolved the peptide fibrils nearly completely in two repeat experiments, leaving only sparse individual fibrils visible by electron microscopy (Fig. 3A, images 1-4). Likewise, a small concentration of the catalytically efficient nIgV 2E6 preparation dissolved prefibrillized A␤42, judged by the ThT-binding test for ␤-sheet-containing aggregates (Fig. 3B). The aIgV 2E6 with lesser A␤40 hydrolytic activity also dissolved prefibrillized A␤42 but with lower potency compared to nIgV 2E6 ( Fig. 3C; 17-fold difference between dissolution potency, computed as the ratio of nIgV and aIgV concentrations needed to reduce ThT binding by 20 fluorescence units compared with the value prior to incubation with the IgV; difference in A␤ hydrolytic  (14) was also present in the acid-purified aIgV 2E6 (lanes 3 and 4, stained with anti-c-Myc antibody and silver, respectively) and the aIgV subjected to an additional ion exchange chromatography step (lanes 5 and 6, stained with anti-c-Myc and silver, respectively (14)). Lane 7, assembled IgG1 59 150 kDa band stained with silver. H, EDTA-inactivated nIgV 2E6. The nIgV was treated with the metal chelator EDTA (10 mM, 24 h) or diluent, and EDTA was removed prior to the hydrolysis assay. The EDTA-inactivated nIgV did not hydrolyze 125 I-A␤40 detectably (Ͻ161 cpm/h). Hydrolytic activity was measured at 0.15 g nIgV/ml over 18 h. Error bars, S.D.
activity of the two IgV preparations, 36-fold). The control noncatalytic nIgV MMF6 and aIgV MMF6 preparations did not dissolve fibrillar A␤. IgV 2E6 is a metal-dependent serine protease. The EDTA-treated nIgV and aIgV 2E6 preparations without A␤ hydrolytic activity also failed to dissolve fibrillar A␤, suggesting catalysis as the dissolution mechanism. Enzymatic hydrolysis at the IgV-sensitive His 14 -Gln 15 bond was described to generate the soluble, non-amyloidogenic, and non-neurotoxic A␤ (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14) and A␤  fragments (35). A soluble peptide fragment with mass approximating that of the anticipated A␤(1-14) product was released upon nIgV 2E6 treatment of preformed fibrillar 125 I-A␤42 (Fig. 3D). The residual particulate fraction following nIgV 2E6 treatment contained only the non-degraded 125 I-A␤42 peak, shown by recovery of essentially all of its radioactivity content at a retention volume of 9.7 ml corresponding to the full-length peptide (not shown). Detection of SDS-stable aggregates in electrophoresis gels run is a diagnostic test for A␤42 oligomerization (21) (Fig.  3E, inset, lane 1). Treatment with nIgV 2E6 reduced the content of the 3-mer A␤42 band (13 kDa), 4-mer A␤42 band (17 kDa), and higher mass A␤42 bands (45 kDa, 64 -84 kDa) in preoligomerized A␤42, accompanied by the appearance of a 2.5-kDa product fragment stainable by anti-A␤ antibodies (Fig. 3E). The monomer A␤42 band intensity does not serve as a useful index of monomer A␤42 hydrolysis because the monomer can be generated by dissociation of certain A␤42 oligomers at the SDS treatment step (36) after completion of the nIgV hydrolysis reaction. The appearance of minor 10 kDa and 19 -27 kDa bands in the nIgV reaction mixture suggests the presence of disaggregation intermediates containing full-length A␤ along with variable product fragment amounts. The small mass A␤(1-14) fragment generated by His 14 -Gln 15 hydrolysis was not detected, consistent with the anomalous electrophoresis behavior of this peptide fragment (37). The 2.5-kDa product band probably corresponds to the A␤(15-42) fragment.
We also tested the anti-amyloid effect of nIgV 2E6 using A␤42 that had not been subjected to prior aggregation. A␤42 aggregates rapidly. In diluent, ThT binding was measurable at the earliest time point studied (bar labeled 30 min) and was increased at 48 h, indicating continued peptide fibrillization (Fig. 4A). The ThT binding values for A␤42 treated with diluent or control IgV MMF6 for 48 h were similar, but ThT binding was reduced significantly by treatment with nIgV 2E6 for the same duration (Fig. 4A). Likewise, the content of each SDSstable oligomer species was reduced significantly by nIgV 2E6 treatment of A␤42 that had not been subjected to prior oligomerization, accompanied by the appearance of the 2.5-kDa hydrolytic product (Fig. 4B). Small nIgV 2E6 concentrations were sufficient to inhibit the accumulation of A␤42 fibrils and oligomers in the foregoing tests (0.13-4 g/ml). Consistent with its lesser hydrolytic activity, the concentrations of an aIgV 2E6 preparation required to inhibit A␤ fibril and oligomer accumulation were larger (e.g. at 27 g of aIgV/ml, inhibition by 63. 3 Ϯ 3.1 and 70.4 Ϯ 18%, respectively). We compared the anti-amyloid effect of nIgV 2E6 with the reference A␤-binding IgG (IgG1 59). The IgG contains V domains that clear brain A␤ in a mouse AD model (15). The amyloid inhibition effect was evident at 1 g/ml nIgV 2E6, whereas a 30-fold larger concentration of the A␤-binding IgG1 59 was without effect (Fig. 4C). No aIgV binding to the A␤42 monomer or oligomers was evident by immunoblotting, whereas the reference IgG1 59 displayed A␤ binding activity (Fig. 4C, inset). Previous ELISAs also failed to reveal measurable IgV-A␤ complexes, suggesting that hydrolysis of A␤ and product release is too rapid for detection of the IgV-A␤ complexes. Together, the data suggest catalytic A␤ degradation as the mechanism of IgV anti-amyloid effects without the participation of accessory phagocytic cells that clear immune complexes by the Fc receptor uptake pathway.
Amyloid Removal in Transgenic Mice-For in vivo clearance tests, we first analyzed brain sections 7 days after administering the nIgV 2E6 (1 g) directly into the right brain neocortex of 5XFAD transgenic mice, an AD model characterized by robust accumulation of human A␤ in the brain parenchyma. Consistent with previous findings, the needle track was visible from increased A␤ deposition, an effect caused by physical trauma to the tissue (38). The needle track terminus was assumed to represent the delivery site of injected nIgV 2E6, nIgV MMF6, and control PBS. Decreased A␤ plaque burden surrounding the nIgV 2E6 delivery site was evident compared with the noninjected left brain neocortex of the same mice (Fig. 5, A and B). The A␤ clearance effect was limited to the immediate vicinity of the injection site, consistent with administration of a small IgV volume (2 l). There was no decrease of local A␤ burden following nIgV MMF6 or PBS injection into the right brain neocortex compared with the non-injected left brain neocortex.
Small proportions of intravenously administered full-length IgG (39) and IgM (40) molecules are documented to permeate the BBB in mouse AD models. We observed restricted brain entry of intravenously injected 125 I-labeled aIgV 2E6 in wild type mice (e.g. at 2 h, the radioactivity/g of brain tissue was 5.2 Ϯ 0.8% of the radioactivity/g of peripheral blood; nominal aIgV half-life in brain and blood, 2.6 and 1.8 h, respectively; Fig.  6A). A␤ clearance following intravenous aIgV 2E6 administration was analyzed in the TgSwDI mouse model characterized by the predominant vasculotropic deposition of the transgene-encoded mutant A␤40 peptide, which causes the CAA-like state (25,26). These mice are suited for testing induction of microvascular dysfunction observed upon infusion of A␤-binding monoclonal IgGs that penetrate the BBB (8,41). Moreover, impaired A␤ peptide egress from the brain in the TgSwDI model is thought to minimize compensatory brain A␤ release incidental to peripheral A␤ clearance by A␤-binding antibod- ies, with the result that any observed brain A␤ clearance effect is probably due to antibodies that cross the BBB (42). Ten days after intravenous treatment with aIgV 2E6 (100 g total IgV/ mouse), the right hemisphere neocortical A␤ deposits were reduced significantly compared with the control aIgV MMF6 treatment, as judged by immunohistochemical staining of brain sections (28% reduction, p Ͻ 0.05; Fig. 6, B and C). The A␤-clearing effect was confirmed from the reduced hippocampal A␤ deposits (Fig. 6D). ELISA measurements in whole left brain hemisphere extracts showed modest but significant reductions of the soluble and insoluble A␤40 and A␤42 levels in the aIgV 2E6-treated mice ( Table 2), suggesting a widespread A␤ clearing effect. A␤38 was not detected reliably at the brain extract concentrations tested (Ͻ16 and Ͻ0.5 pg/mg insoluble and soluble A␤38, respectively). Treatment with aIgV 2E6 did not induce microglial activation (Fig. 7, A and B) or microhemorrhages (Fig. 7C). Together, the studies suggest significant brain A␤ clearance by the brief intravenous aIgV 2E6 treatment. In comparison, prolonged intravenous treatment with milligram A␤-binding IgG amounts is required for brain A␤ clearance in various mouse models of Alzheimer disease (4, 6).
trating IgGs (e.g. bapineuzumab) (6 -8) and increased A␤ egress from the brain induced by A␤ binding to IgGs in peripheral blood (e.g. solanezumab) (43). The catalytic IgV rapidly digested A␤ into non-amyloidogenic soluble fragments without forming stable immune complexes. As for enzymes, nonspecific proteolysis is a significant risk. IgV 2E6 did not degrade substrates without structural similarity to A␤, transthyretin amyloid containing the ␤-sheet amyloid motif, or the A␤ precursor protein, suggesting sufficiently specific A␤ removal. The same Fc receptor-dependent microglial interactions with A␤-IgG immune complexes that result in beneficial A␤ clearance also hold potential for damaging neurons and other brain cells (5)(6)(7)(8). Side-by-side toxicity studies of catalytic IgV and A␤-binding IgGs remain to be conducted, but it is noteworthy  3 and 4). Scale bar, 200 nm. The dense A␤ plaques visible after diluent or IgV MMF6 treatment were absent after IgV 2E6 treatment, and only rare individual A␤ fibrils were evident (image 4 shows a magnified solitary fibril). IgV concentration was 0.13 g/ml. B, ThT binding to prefibrillized A␤42 following treatment with nIgV 2E6 (24 h). ThT binding was reduced to levels below the starting fibrillar A␤42 prior to incubation with the nIgVs (dashed line, 103 Ϯ 10 fluorescence units (FU)). Increased ThT binding is observed over 24 h of diluent and IgV MMF6 treatment due to continued A␤ fibrillization. ThT binding by A␤42 immediately after dissolving the peptide in buffer was 68 Ϯ 11. *, p Ͻ 0.05; **, p Ͻ 0.01. C, ThT binding to preformed fibrillar A␤42 following treatment with aIgV 2E6 (24 h). ThT binding was reduced to levels below the starting fibrillar A␤42 prior to incubation with the aIgVs (dashed line, 103 Ϯ 10 fluorescence units). ThT binding by A␤42 immediately after dissolving the peptide in buffer was 68 Ϯ 11. *, p Ͻ 0.05. D, FPLC gel filtration profile of the supernatant following treatment of fibrillar 125 I-A␤42 with nIgV 2E6. Resuspension of fibrillar 125 I-A␤42 in nIgV 2E6 (10 g/ml) resulted in release of the soluble 1654-Da fragment into the supernatant. The intact 125 I-A␤42 (4,629 Da) peak represents spontaneous peptide release that occurred at equivalent levels in reaction mixtures of fibrillar A␤42 treated with PBS (not shown), nIgV 2E6, and control nIgV MMF6. Inset, time-dependent release of 1,654-Da product fragment into the supernatant by nIgV 2E6 treatment. E, dissolution of preformed A␤42 oligomers. Treatment of preoligomerized A␤42 with nIgV 2E6 but not nIgV MMF6 (3 g/ml, 24 h) depleted the SDS-stable A␤ trimers, tetramers, and high mass oligomers (45 kDa band, 64 -84 kDa region). *, p Ͻ 0.005. Inset, A␤42 oligomers stained with a mixture of A␤-binding monoclonal antibodies after treatment with nIgV MMF6 (lane 1) or nIgV 2E6 (lane 2). Migration of protein standards is shown by mass values on the right. Error bars, S.D.
that the catalytic IgV dissolved fibrillar A␤ with no requirement for microglia, and there was no evidence for microglial activation or cerebral microhemorrhages in the IgV-treated mice.
The A␤ recognition properties of IgV 2E6 are similar to the previously documented single domain IgV catabody fragment (clone 5D3) (14). While the present manuscript was under review, we reported that prolonged brain expression of IgV 5D3 by means of gene transfer induced significant A␤ clearance with no noticeable inflammatory or vascular effects (44). The gene transfer approach may eliminate the need for repeated antibody infusions, but its clinical use awaits long term safety analyses. In comparison, there are ample precedents for intravenous antibody therapy. A brain A␤-clearing effect was evident after a brief intravenous catalytic IgV treatment in mice. Certain shortcomings of the IgV can be foreseen. First, like other antibody fragments devoid of the Fc region (45,46), the IgV displayed a comparatively short blood half-life. Second, like full-length IgGs, the intravenously injected IgV gained only limited entry into the brain. Routes toward improving the IgV longevity and BBB penetration properties are available (e.g. attachment of an appropriate polypeptide tag) (47)(48)(49).
Other potential factors governing the catabody clearance capacity are (a) the A␤ aggregation status, (b) steric access to FIGURE 5. Reduced brain A␤ following intrabrain nIgV 2E6 injection. A, local plaque burden was reduced 7 days after nIgV 2E6 but not nIgV MMF6 or PBS injection into the right brain neocortex, determined by staining with anti-A␤ antibody (1 g of nIgV; n ϭ 7 5XFAD mice in the nIgV 2E6 group, n ϭ 4 mice each in the nIgV MMF6 and PBS groups). The change in plaque burden following test treatments was computed as the percentage of plaque burden in the autologous untreated left hemisphere region. p values are indicated. B, images 1 and 2 show representative sections of the nIgV 2E6-injected right hemisphere neocortex and the corresponding tissue area in the control untreated left neocortex, respectively. Rectangles, area in which plaque burden was determined. Error bars, S.E.  Ͼ 0.1). B, inhibition of A␤42 oligomerization. The intensities of the SDS-stable trimer and tetramer were reduced by nIgV 2E6 treatment (24 h) of A␤42 that had not been subjected to prior oligomerization compared with an equivalent nIgV MMF6 concentration (3 g/ml). Higher order SDS-stable oligomers were less abundant than in preoligomerized A␤42 preparations. Only the 74-kDa high mass oligomer band was evident, which was reduced significantly by treatment with nIgV 2E6. **, p Ͻ 0.005; *, p Ͻ 0.03. Inset, A␤42 oligomers stained with a mixture of A␤-binding monoclonal antibodies following treatment with nIgV MMF6 (lane 1) or nIgV 2E6 (lane 2). The 2.5-kDa product fragment is visible. C, superior inhibition of A␤42 fibrillization by nIgV 2E6 compared with IgG1 59. A␤42 fibrillization tested as in A was inhibited by nIgV 2E6 (1 g/ml) but not the A␤-binding IgG1 59 (1 or 30 g/ml). Shown are ThT fluorescence data as a percentage of the values in the presence of control antibodies with the same scaffold structure as nIgV 2E6 and IgG1 59 (nIgV MMF6 and IgG1 SKT03, respectively). *, p Ͻ 0.01. Inset, IgV 2E6 at a concentration displaying readily detectable A␤ hydrolysis and anti-amyloid effects (4 g/ml) did not bind the A␤42 monomer or oligomers determined by immunoblotting (lane 4). Lane 5, control immunoblot after treatment of the same A␤42 preparation with an equivalent control IgV MMF6 concentration. Lane 1, silver-stained trimer and tetramer bands in the A␤42 preparation. Lanes 2 and 3, the A␤42 oligomers stained with the A␤-binding IgG1 59 and control nonimmune IgG1 SKT03, respectively. Error bars, S.D.
the A␤ noncovalent binding epitope and scissile bonds, and (c) conformational transitions in A␤. Small IgV concentrations were sufficient to hydrolyze particulate A␤, but diffusional restrictions decelerate enzymatic digestion of particulate substrates compared with soluble substrates (50,51). We acknowledge the potential of varying clearance rates for A␤ coaggregated with other proteins and A␤ aggregates containing post-translational chemical modifications. Catalysis requires topographically precise IgV interactions at the A␤(29 -40) epitope and the His 14 -Gln 15 scissile bond (present study) (14,29). Conformational factors and limited access to A␤(29 -40) epitope located in the transmembrane APP region probably restrict the proteolytic rate for the membrane-anchored APP. The A␤(29 -40) epitope of APP is helical (34) and acquires increasing ␤-sheet character with increasing aggregation of the mature A␤40/42 peptides (52,53). Such a conformational transition may explain poor utilization of the soluble APP substrate by the IgV.
In the acquired immunity paradigm, specific antigen binding activity develops over a few weeks by immunogen-driven selection of sequence-diversified V L and V H domains. The paradigm does not explain specific A␤(29 -40) recognition by IgV 2E6. First, the IgV originated from non-aged humans without path-ological amyloid accumulation. Second, antigen recognition developed by acquired immune processes occurs at the extensively mutated complementarity-determining regions (CDRs), whereas the IgV contains minimal CDR mutations (14). Third, V H domain CDR3 and proper V L -V H domain pairing are specificity-defining factors in acquired immunity (54), but IgV 2E6 does not contain a V H domain, and pairing of an A␤ hydrolytic V L domain with a V H domain suppressed the catalytic activity (14). Fourth, A␤ recognition without acquired immunity processes is not limited to catabodies; selective noncovalent binding of the C-terminal A␤ epitope by antibody fragments from "non-immune" humans was described by another group (55). Fifth, IgMs from non-aged healthy humans, the first antibodies produced by B cells, hydrolyzed A␤ specifically, whereas the IgGs produced by differentiated B cells were poorly catalytic (14). Similarly, IgGs induced by routine immunization with various antigens are poorly proteolytic, and immunization with transition state analogs induced only esterase IgGs that stabilized the transition state noncovalently, not catabodies capable of the complex peptide bond hydrolysis reaction (56). Together, these findings indicate specific A␤ hydrolysis as an innate immune property.
Microbial B-cell superantigens provide a precedent for specific A␤ recognition by subsets of innate catabodies (12) and conventional antibodies (57). In addition, the production of innate amyloid-directed catabodies is not limited to the A␤ target. Healthy humans synthesize catabodies specific for transthyretin amyloid, which is responsible for age-associated systemic amyloidosis (19). Humans produce no more than 100,000 non-antibody proteins, including enzymes and receptors expressing specificity for diverse biological ligands. In comparison, the combinational antibody repertoire derived from the germ line V, D, and J segments is far larger (ϳ4 ϫ 10 9 V L -V H domain pairs derived by recombination of 152 V L with 19 J genes and recombination of 273 V H genes with 34 D and 13 J genes; germ line gene numbers from IgBLAST). It is reasonable, FIGURE 6. Reduced brain A␤ following intravenous aIgV 2E6 injection. A, aIgV blood and brain levels plotted as a function of time following intravenous injection of the 125 I-aIgV. n ϭ 3 mice/time point. B, the plaque burden in the brain neocortex was reduced on day 10 after intravenous aIgV 2E6 injections on days 0 and 3 (50 g/injection) compared with mice treated equivalently with aIgV MMF6 (n ϭ 8 TgSwDI mice/group). *, p Ͻ 0.05. C, top and bottom images, anti-A␤ antibody stained neocortex sections from aIgV MMF6-treated and aIgV 2E6-treated mice, respectively. Horizontal bar, 0.25 mm. Arrow, a plaque. D, hippocampal A␤ plaque burden. A␤ removal in aIgV 2E6-treated mice from B was confirmed in hippocampus (Hippo) sections (n ϭ 3 mice/group; neocortex data from the same mice are included for reference). *, p Ͻ 0.05 versus aIgV MMF6. Error bars: in A, S.D.; in B and D, S.E.

TABLE 2 A␤ content of TgSWDI mouse brain extracts following treatment with aIgV 2E6 and MMF6
A␤40 and A␤42 contents were measured by ELISA and are expressed per mg of brain tissue (mean Ϯ S.D.). Particulate A␤ contents represent the difference between the observed total and soluble A␤ content. aIgV effect size corresponds to the reduction of A␤ content in mice treated with aIgV 2E6 expressed as a percentage of the A␤ content in mice treated with control aIgV MMF6. p values are from Student's t test, two-tailed. therefore, to view the germ line V domains as a source of catabodies with varying selectivity for individual substrates. If amyloid-specific catalysis is an innate immunity function in humans, this implies the existence of selective pressures guiding Darwinian evolution of the antibody germ line genes. Amyloidogenic proteins form harmful aggregates within minutes to days in the test tube, suggesting the existence of homeostatic mechanisms that control amyloid accumulation in vivo. The accumulation of misfolded proteins may begin decades prior to the appearance of disease symptoms (e.g. oligomeric and particulate A␤ in humans (58) and monkeys (59) of reproductive age). Early amyloidosis prior to reproduction will jeopardize survival of the species. This suggests a survival advantage gained from innate amyloid-clearing catabodies. Details of catabody evolutionary history have not been determined, but both A␤ and antibodies are ancient molecules. The human A␤(29 -40) epitope recognized by IgV 2E6 is completely conserved in the cartilaginous fish Narke japonica (corresponding to residues 629 -642 of amyloid precursor protein, GenBank TM accession number BAA24230.1). These fish appeared about 450 million years ago and represent the first extant organisms with antibody V genes bearing discernible sequence identity to the human V genes (60). Together, these arguments support our view of innate amyloid-hydrolyzing catabodies found in humans as functionally important mediators that may be useful for therapy of age-associated amyloidosis.