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.

Abstract

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.

Introduction

According to the “amyloid hypothesis,” soluble and fibrillar amyloid β peptide (Aβ)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 Aβ(1–42) (Aβ42) that damage the brain parenchyma, accumulation of Aβ(1–40) (Aβ40) 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 Aβ-binding monoclonal IgGs was proposed for AD therapy (4–6). Such IgGs exert competing favorable and unfavorable effects (7, 8). Whereas the Aβ-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, Aβ-binding monoclonal IgGs clear parenchymal Aβ42, but they induce increased Aβ40 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 Aβ-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 repertoire 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 Aβ specifically with no evidence of microglial activation or microhemorrhages.

MATERIALS AND METHODS

Antibodies

IgV clones 2E6 and MMF6 were isolated from an IgV library cloned in pHEN2 vector from peripheral blood lymphocytes of three lupus patients without amyloid disease (14). The two IgVs contain the same C-terminal V_L2 domain and different N-terminal V_L1 domains (GenBank^TM accession numbers FJ231715 and KF018653, respectively). The mutant IgV 2E6 gene was synthesized by Mutagenex Inc. (Hillsborough, NJ) by replacing the V_L1 domain framework regions (FRs; Kabat residues 1–23, 35–49, 57–88, and 98–107) with the corresponding FRs of IgV MMF6 V_L1 domain and cloned into pHEN2 vector (NcoI/XhoI sites). IgV expression in culture supernatants after 24-h induction with isopropyl-β-d-thiogalactopyranoside was 1.5–5.4 mg/liter. Aβ-binding IgG1 59 contains the V_L-V_H domain pair of a single chain Fv fragment that binds Aβ42 aggregates and clears brain Aβ deposits in a transgenic mouse model (15). The polymerase chain reaction-amplified V_L and V_H domain cDNAs of the single chain Fv were cloned in pTT5 expression vector adjacent to the murine κ and γ1 constant domain segments, respectively, and coexpressed transiently in HEK 293 cells (16, 17). Assembled IgG1 59 in tissue culture supernatant (∼0.6 mg of IgG/liter) was concentrated 10 times, purified using immobilized Protein A, and dialyzed against 10 mm sodium phosphate, pH 7.4, 137 mm NaCl, 2.7 mm KCl (PBS), and the protein concentration was determined (16).

The IgVs were purified from bacterial periplasmic extracts by binding of the C-terminal His₆ 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 × g, 30 min) of IgV-secreting and control bacteria grown to equivalent density as before (∼0.8 A₆₀₀ 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).

Proteolysis Assays

Synthetic Aβ40 from EZBiolab (Carmel, IN) treated with 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP; Sigma-Aldrich) was radiolabeled with ¹²⁵I at Tyr¹⁰. Aβ40 forms particulate aggregates slowly compared with Aβ42. Routine hydrolysis tests were conducted using ¹²⁵I-Aβ40 (∼30,000 cpm) treated with nIgV or aIgV preparations (3–24 h) in PBS containing 0.1 mm CHAPS and 0.1% bovine serum albumin, followed by trichloroacetic acid precipitation to separate the intact peptide from fragments (14). Reported values are means ± S.D. (2 or 3 replicates). Formic acid was added to 60% (v/v) prior to FPLC gel filtration of ¹²⁵I-Aβ40 (50,000 cpm; ∼0.2 nm) digested with IgVs (2.5 μg/ml, 24 h; Superdex peptide column (GE Healthcare), run in 60% formic acid; 0.5-ml fractions). The nominal peptide mass was computed from the retention volumes of these markers: aprotinin (6,512 Da), vasoactive intestinal peptide (3,326 Da), N-tert-butoxycarbonyl-O-benzyl-Glu-Ala-Arg-7-amino-4-methylcoumarin (721 Da), and Ala-7-amino-4-methylcoumarin (246 Da). Apparent K_m and turnover number (k_cat) were estimated from initial hydrolysis rates measured at a constant ¹²⁵I-Aβ40 amount (∼300,000 cpm) mixed with increasing Aβ40 concentrations (0.1 nm to 25 μm), assuming an IgV mass of 30 kDa (14).

Aβ42 (90 μg) was labeled with ¹²⁵I using 1,3,4,6-tetrachloro-3α,6α-diphenylglycouril-coated tubes (Pierce), and free ¹²⁵I was removed (Sep-Pak Vac C18 cartridge, Waters (Milford, MA)). The acetonitrile-eluted ¹²⁵I-Aβ42 was dried, mixed with non-radiolabeled Aβ42 (1 mg), retreated with HFIP for 1 h, lyophilized, and stored (−80 °C). After reconstitution in dimethyl sulfoxide, the ¹²⁵I-Aβ42 (5 mm) was diluted to a concentration of 50 μm in PBS and incubated for 92 h (37 °C), and fibrillar Aβ42 was collected by centrifugation (16,000 × g, 20 min). The radioactivity content of fibrillar ¹²⁵I-Aβ42 was 2.3 × 10⁷ cpm/μmol of Aβ. Fibrillar ¹²⁵I-Aβ42 (2.2 nmol/0.1 ml) was incubated with nIgV samples in PBS containing 0.1 mm CHAPS. Formic acid was added to 60% (v/v) in the supernatants collected by centrifugation (16,000 × g, 20 min), and FPLC gel filtration in formic acid was conducted as before. The pelleted ¹²⁵I-Aβ42 was dissolved in PBS containing 60% formic acid and analyzed by FPLC gel filtration similarly.

For epitope identification, non-radioactive Aβ40, Aβ(1–16), Aβ(17–42), or Aβ(19–40) (Anaspec, Fremont, CA) was included as a competitive inhibitor in the ¹²⁵I-Aβ40 hydrolysis reaction mixture (0.2 μg/ml nIgV 2E6; 3 h, 37 °C). Hydrolysis of soluble FLAG-tagged APP751 (83 kDa, 0.1 μm; Origene Technologies, Rockville, MD), glutathione S-transferase (GST)-tagged amphiphysin (153 kDa, 0.1 μm), and GST-tagged zinc finger protein 154 (43 kDa, 0.1 μm; Abnova, Taipei, Taiwan) treated with 2.5 μg/ml nIgV (24 h) was estimated by SDS electrophoresis and staining with horseradish peroxidase-conjugated anti-FLAG or anti-GST IgG antibodies (Sigma-Aldrich) (16). Minor low mass protein bands visible in the amphiphysin reaction mixtures in addition to the major intact amphiphysin band (153 kDa) are impurities from the starting protein preparation. Biotin-conjugated transthyretin (TTR) (1.5 mol of biotin/mol of TTR monomer; conjugation done as described (14)) was preaggregated for 5 days in 100 mm sodium acetate, pH 4.2, 5 mm sodium phosphate, 68 mm NaCl, 51 mm KCl containing 1 mm EDTA, followed by buffer exchange to PBS containing 0.1 mm CHAPS by ultrafiltration (Amicon Ultra-4 centrifugal filter, 10 kDa cut-off). TTR aggregation was confirmed by development of turbidity (400 nm) and increased thioflavin T (ThT) fluorescence (λ_ex = 440 nm; λ_em = 485 nm; photomultiplier tube, 600 V) compared with the starting non-aggregated TTR (turbidity, 0.61 ± 0.01 and 0.02 ± 0.01, respectively; ThT fluorescence, 144.9 ± 3.2 and 17.4 ± 6.9, respectively). The preaggregated TTR (0.1 μm) was incubated with 40 μg/ml IgV (60 h), and the reaction mixtures were boiled in SDS and analyzed by electrophoresis (16.5% gels). Under these conditions, the physiological tetramer TTR species and amyloid TTR aggregates dissociate into the TTR monomer band (14 kDa), and a faint dimer band is detected (19).

Anti-amyloid Assays

Preaggregated fibrillar Aβ42 (starting Aβ42 peptide concentration, 20 μm; prepared as described for ¹²⁵I-Aβ42) was treated with nIgVs (24 h, 37 °C) in PBS, 0.1 mm CHAPS, and 1% dimethyl sulfoxide. ThT (5 μm) was added, and fluorescence emission was measured after 30 min. Inhibition of fibrillization was determined similarly by treating non-aggregated Aβ42 (20 μm) with the nIgVs. The data were corrected for background ThT fluorescence of the same antibody without Aβ42 (<5–19 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 Aβ42 (5 μm) 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 μl 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 Aβ42 (50 μm) in phenol red-free Ham's F-12 medium (4 °C, 24 h) (21). The oligomer preparation (total Aβ peptide concentration 40 μm) was treated (24 h, 37 °C) with nIgV 2E6 or MMF6 (3 μg/ml) in Ham's medium/PBS containing 0.1 mm CHAPS (4:1, v/v). SDS and β-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 Aβ species mass (monomers, oligomers, 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-Aβ monoclonal IgG 6E10, IgG 4G8 (both from Covance, Princeton, NJ), and IgG 6D4 (MyBioSource, San Diego, CA) (directed to Aβ(1–17), Aβ(17–24), and the Aβ C terminus, respectively). Freshly dissolved Aβ42 without prior oligomerization was mixed immediately with the nIgVs (3 μg/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). Aβ42 binding by IgV 2E6 and IgG1 59 was tested by immunoblotting of oligomerized Aβ42 electrophoresed in SDS gels (1 μg of Aβ42/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-Aβ immune complexes (14, 15). Bound IgG1 59 was detected by staining with peroxidase-conjugated anti-mouse IgG (22).

Brain Aβ Removal

In the first study, nIgV 2E6 or MMF6 (1 μg/2 μl 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 Aβ produced from mutant human Aβ 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-Aβ monoclonal IgGs 6E10 and 4G8, and the Aβ deposits were quantified (BIOQUANT Image Analysis Corp., Nashville, TN) (24). The Aβ plaque burden is defined as the percentage area occupied by the stained reaction product in a 640 × 480-μm rectangle surrounding the injection site (estimated in five 40-μm sections spaced 200 μm apart). The relative Aβ burden in the injected right hemisphere was computed as a percentage of the Aβ 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 μg of aIgV in PBS/injection) in TgSwDI mice (n = 8/group, 7–9 months old). The mice express human Aβ 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 Aβ deposits, TgSwDI mice develop vascular Aβ deposits at an early age compared with other human Aβ-expressing mouse models. Cortical Aβ 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 × 700 μm). 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 μg/ml pepstatin A, and a 25-fold dilution of cOmplete protease inhibitor mixture as recommended by the supplier (Roche Applied Science). Total Aβ levels were determined after solubilizing particulate Aβ in the homogenates mixed with cold formic acid (1:2.2 (v/v)) in duplicate using the Aβ 3-plex Ultrasensitive immunoassay kit (Meso Scale Discovery, Gaithersburg, MD). Soluble Aβ 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 Aβ brain contents represents particulate Aβ 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-μm sections spaced 400 μm 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; 1+, a few activated microglia; 2+, a moderate number of activated/phagocytic microglia; 3+, numerous activated/phagocytic microglia) (24).

Brain entry of intravenously injected IgV in B6SJLF1/J mice (Jackson Laboratory, Bar Harbor, ME) was studied using ¹²⁵I-labeled 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 × 10⁶ cpm/μg aIgV) (27). Essentially all recovered radioactivity was present in the aIgV bands identified by electrophoresis and autoradiography. Following injection of ¹²⁵I-labeled aIgV into the tail vein (1.1 μg/110 μl/mouse, 6.3 × 10⁶ cpm), the radioactivity/g of whole blood from the retroorbital plexus or whole brain obtained at euthanasia was measured using a γ counter (n = 3 mice/time point). The nominal half-life was computed from single phase decay kinetics (cpm = (cpm_max) × exp(−k × t), where k is the decay constant and t_½ = 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 two-tailed Student's t test.

RESULTS

IgV 2E6 Hydrolytic Properties

Recombinant IgV 2E6 is a single-chain heterodimer of V_L domains (Fig. 1A, inset) that expressed Aβ hydrolyzing activity following purification of the IgV by acid elution from a metal affinity column (aIgV 2E6) (14). Per unit IgV 2E6 mass, the Aβ 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 Aβ that is a proposed therapeutic agent for AD (28). Nanogram IgV 2E6 amounts in the unfractionated culture supernatant hydrolyzed radiolabeled ¹²⁵I-Aβ40 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 framework 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/μg). 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).

FIGURE 1.

Hydrolytic properties of IgV 2E6. A, ¹²⁵I-Aβ40 hydrolysis by unfractionated bacterial culture supernatants (1:10 diluted, 32–38 ng of IgV/assay). Each symbol represents an independent culture well. Hydrolysis by supernatants from control bacteria without the vector (bacteria alone) or with empty vector (pHEN2) was negligible. Inset, schematic IgV structure. Filled and unfilled boxes, FRs and CDRs, respectively. B, concentration-dependent ¹²⁵I-Aβ40 hydrolysis by native IgV 2E6-containing bacterial culture supernatant. Hydrolysis by culture supernatants from control bacteria expressing IgV MMF6, bacteria without vector, or bacteria with empty vector was negligible. C, hydrolytically inactive mutant of IgV 2E6 with its FR1–FR4 segments replaced by the corresponding FRs of IgV MMF6 (¹²⁵I-Aβ40 treated for 18 h with 1:30 diluted unfractionated culture supernatants). D, diagnostic anion exchange FPLC. Top, protein absorbance (A₂₈₀) profiles for supernatants of bacteria transformed with the IgV 2E6 vector or empty vector. Bottom, the ¹²⁵I-Aβ40 hydrolytic activity of native IgV 2E6 in the bacterial supernatant was recovered in the unbound fractions. The corresponding unbound fractions containing basal proteases secreted by bacteria harboring the empty vector displayed little or no hydrolytic activity (<9% hydrolysis). Aβ hydrolysis per unit IgV 2E6 mass by the pooled unbound fractions was comparable with the unfractionated bacterial supernatant (170,337 and 158,857 cpm/h/μg of IgV, respectively). E, nIgV 2E6 preparation by a second anion exchange FPLC cycle. For details of the first FPLC cycle, see “Materials and Methods.” The bound fraction (gray) contained 160 mg of total protein and 1.1 mg of nIgV 2E6, determined by immunoblotting for the c-Myc tag (retention volume, 26–34 ml), corresponding to ∼10-fold catabody enrichment compared with the unfractionated culture supernatant. F, superior ¹²⁵I-Aβ40 hydrolytic activity of nIgV 2E6 from E compared with acid-purified aIgV. Control nIgV MMF6 fractionated by anion exchange chromatography at neutral pH did not hydrolyze Aβ. G, reducing SDS-electrophoresis of IgV 2E6. The bacterial culture supernatant contained the anti-c-Myc antibody-stained intact 30-kDa IgV 2E6 V_L1-V_L2 heterodimer and an 18-kDa IgV fragment containing the V_L2 region (lane 1), along with other silver-stained non-IgV proteins (lane 2). The catalytically inactive 18-kDa fragment (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 ¹²⁵I-Aβ40 detectably (<161 cpm/h). Hydrolytic activity was measured at 0.15 μg nIgV/ml over 18 h. Error bars, S.D.

TABLE 1.

Apparent kinetic parameters for Aβ40 hydrolysis by IgV 2E6

Increasing Aβ40 (0.1 nm to 25 μm) containing a constant ¹²⁵I-Aβ40 amount (∼300,000 cpm) was treated with the acid-purified aIgV (3.8 μg/ml) or native IgV in the bacterial culture supernatant (0.1 μg/ml, determined by c-Myc dot blotting). aIgV and neprilysin data are from Refs. 61 and 29, respectively. The Michaelis constant (K_m) and turnover number (k_cat) were estimated by fitting hydrolysis rates versus Aβ40 concentration to the Michaelis-Menten equation (r² = 0.99).

Catalyst	Km	kcat	kcat/Km
	m × 10−5	min−1	m−1 min−1 × 103
aIgV	8.0 ± 1.2	0.3 ± 0.1	3.7 ± 0.6
Native IgV	2.5 ± 0.5	8.2 ± 1.0	330 ± 66
Neprilysin		2.2

Specificity

aIgV 2E6 cleaved the Aβ His¹⁴-Gln¹⁵ bond (14). nIgV treatment of ¹²⁵I-Aβ40 generated a hydrolytic product with mass close to the predicted Aβ(1–14) radiolabeled product (Fig. 2, A and B; 1,654 Da; the C-terminal Aβ(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, full-length 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 Aβ region of soluble APP assumes an IgV-insensitive conformation. Sequence-independent recognition of a generic β-sheet amyloid epitope could explain Aβ degradation. nIgV 2E6, however, did not digest TTR amyloid aggregates (Fig. 2C). If the specificity for Aβ derives only from His¹⁴-Gln¹⁵ 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 Aβ peptides to assess the contribution of noncovalent epitope recognition in IgV specificity. Aβ(1–16) containing the His¹⁴-Gln¹⁵ dipeptide unit did not inhibit ¹²⁵I-Aβ40 hydrolysis by the nIgV, but the C-terminal Aβ(29–40), Aβ(17–42), and control full-length Aβ40 were equipotent inhibitors (Fig. 2D). The results indicate specific noncovalent recognition of the Aβ(29–40) epitope followed by hydrolysis at the remote His¹⁴-Gln¹⁵, with no requirement for the two Aβ42 C-terminal residues (Fig. 2D, inset). In membrane-bound APP and its C99 fragment generated upon γ-secretase cleavage, the His¹⁴-Gln¹⁵ scissile bond is located in the juxtamembrane extracellular protein region, and the Aβ(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 IgV-catalyzed hydrolysis of membrane-anchored APP.

FIGURE 2.

Specificity of IgV 2E6. A, FPLC gel filtration of ¹²⁵I-Aβ40 treated with nIgV 2E6. Intact ¹²⁵I-Aβ40 (9.8 ml) was depleted, and the product peak mass is consistent with the predicted ¹²⁵I-Aβ1–14 fragment (retention time, 12.2 ml; observed mass, 1,654 Da; computed mass, 1,699 Da). Control nIgV MMF6 did not hydrolyze ¹²⁵I-Aβ40. B, time-dependent 1,654-Da product accumulation following ¹²⁵I-Aβ40 treatment with nIgV 2E6 as in A. C, no hydrolysis of APP, TTR amyloid, His-Gln-containing proteins, and irrelevant proteins. Shown are electrophoresis gels of nIgV 2E6-treated APP (lane 2, 83 kDa), TTR amyloid (which dissociates into 14-kDa TTR monomers under the electrophoresis conditions; lane 4), amphiphysin (AMPH; lane 6, 176 kDa), zinc finger protein 154 (ZNF154; lane 8, 47 kDa), or gp120 (lane 10, 95 kDa). Lanes 1, 3, 5, 7, and 9 are the diluent-treated proteins, respectively. Arrows, intact proteins. Electrophoresis was done on 4–20% gels except for the TTR reaction mixtures, which were analyzed on 16.5% gels. Migration of protein standards is shown by the mass values on the left (4–20% gels) and right (16.5% gels). D, noncovalent IgV epitope. Synthetic Aβ40, Aβ29–40, and Aβ17–42 inhibited the hydrolysis of ¹²⁵I-Aβ40 by nIgV 2E6 equipotently, indicating Aβ(29–40) as the IgV recognition epitope. ¹²⁵I-Aβ40 hydrolysis without competitor peptide was 2,592 ± 534 cpm). Inset, the His¹⁴-Gln¹⁵ scissile bond (arrow) is distant from the noncovalent epitope (red font). Error bars, S.D.

Amyloid Dissolution

Aβ self-assembles into β-sheet fibrils and soluble oligomers. Treatment of prefibrillized Aβ42 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 Aβ42, judged by the ThT-binding test for β-sheet-containing aggregates (Fig. 3B). The aIgV 2E6 with lesser Aβ40 hydrolytic activity also dissolved prefibrillized Aβ42 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 Aβ hydrolytic activity of the two IgV preparations, 36-fold). The control non-catalytic nIgV MMF6 and aIgV MMF6 preparations did not dissolve fibrillar Aβ. IgV 2E6 is a metal-dependent serine protease. The EDTA-treated nIgV and aIgV 2E6 preparations without Aβ hydrolytic activity also failed to dissolve fibrillar Aβ, suggesting catalysis as the dissolution mechanism. Enzymatic hydrolysis at the IgV-sensitive His¹⁴-Gln¹⁵ bond was described to generate the soluble, non-amyloidogenic, and non-neurotoxic Aβ(1–14) and Aβ(15–42) fragments (35). A soluble peptide fragment with mass approximating that of the anticipated Aβ(1–14) product was released upon nIgV 2E6 treatment of preformed fibrillar ¹²⁵I-Aβ42 (Fig. 3D). The residual particulate fraction following nIgV 2E6 treatment contained only the non-degraded ¹²⁵I-Aβ42 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 Aβ42 oligomerization (21) (Fig. 3E, inset, lane 1). Treatment with nIgV 2E6 reduced the content of the 3-mer Aβ42 band (13 kDa), 4-mer Aβ42 band (17 kDa), and higher mass Aβ42 bands (45 kDa, 64–84 kDa) in preoligomerized Aβ42, accompanied by the appearance of a 2.5-kDa product fragment stainable by anti-Aβ antibodies (Fig. 3E). The monomer Aβ42 band intensity does not serve as a useful index of monomer Aβ42 hydrolysis because the monomer can be generated by dissociation of certain Aβ42 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 Aβ along with variable product fragment amounts. The small mass Aβ(1–14) fragment generated by His¹⁴-Gln¹⁵ 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 Aβ(15–42) fragment.

FIGURE 3.

Dissolution of preformed Aβ aggregates by IgV 2E6. A, transmission electron micrographs of preformed Aβ42 fibrils (2.5 μm) treated for 24 h with diluent (image 1) or tissue culture supernatant containing IgV MMF6 (image 2) or IgV 2E6 (images 3 and 4). Scale bar, 200 nm. The dense Aβ plaques visible after diluent or IgV MMF6 treatment were absent after IgV 2E6 treatment, and only rare individual Aβ fibrils were evident (image 4 shows a magnified solitary fibril). IgV concentration was 0.13 μg/ml. B, ThT binding to prefibrillized Aβ42 following treatment with nIgV 2E6 (24 h). ThT binding was reduced to levels below the starting fibrillar Aβ42 prior to incubation with the nIgVs (dashed line, 103 ± 10 fluorescence units (FU)). Increased ThT binding is observed over 24 h of diluent and IgV MMF6 treatment due to continued Aβ fibrillization. ThT binding by Aβ42 immediately after dissolving the peptide in buffer was 68 ± 11. *, p < 0.05; **, p < 0.01. C, ThT binding to preformed fibrillar Aβ42 following treatment with aIgV 2E6 (24 h). ThT binding was reduced to levels below the starting fibrillar Aβ42 prior to incubation with the aIgVs (dashed line, 103 ± 10 fluorescence units). ThT binding by Aβ42 immediately after dissolving the peptide in buffer was 68 ± 11. *, p < 0.05. D, FPLC gel filtration profile of the supernatant following treatment of fibrillar ¹²⁵I-Aβ42 with nIgV 2E6. Resuspension of fibrillar ¹²⁵I-Aβ42 in nIgV 2E6 (10 μg/ml) resulted in release of the soluble 1654-Da fragment into the supernatant. The intact ¹²⁵I-Aβ42 (4,629 Da) peak represents spontaneous peptide release that occurred at equivalent levels in reaction mixtures of fibrillar Aβ42 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 Aβ42 oligomers. Treatment of preoligomerized Aβ42 with nIgV 2E6 but not nIgV MMF6 (3 μg/ml, 24 h) depleted the SDS-stable Aβ trimers, tetramers, and high mass oligomers (45 kDa band, 64–84 kDa region). *, p < 0.005. Inset, Aβ42 oligomers stained with a mixture of Aβ-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.

We also tested the anti-amyloid effect of nIgV 2E6 using Aβ42 that had not been subjected to prior aggregation. Aβ42 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 Aβ42 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 SDS-stable oligomer species was reduced significantly by nIgV 2E6 treatment of Aβ42 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 Aβ42 fibrils and oligomers in the foregoing tests (0.13–4 μg/ml). Consistent with its lesser hydrolytic activity, the concentrations of an aIgV 2E6 preparation required to inhibit Aβ fibril and oligomer accumulation were larger (e.g. at 27 μg of aIgV/ml, inhibition by 63.3 ± 3.1 and 70.4 ± 18%, respectively). We compared the anti-amyloid effect of nIgV 2E6 with the reference Aβ-binding IgG (IgG1 59). The IgG contains V domains that clear brain Aβ in a mouse AD model (15). The amyloid inhibition effect was evident at 1 μg/ml nIgV 2E6, whereas a 30-fold larger concentration of the Aβ-binding IgG1 59 was without effect (Fig. 4C). No aIgV binding to the Aβ42 monomer or oligomers was evident by immunoblotting, whereas the reference IgG1 59 displayed Aβ binding activity (Fig. 4C, inset). Previous ELISAs also failed to reveal measurable IgV-Aβ complexes, suggesting that hydrolysis of Aβ and product release is too rapid for detection of the IgV-Aβ complexes. Together, the data suggest catalytic Aβ 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.

FIGURE 4.

Reduced Aβ aggregation in the presence of nIgV 2E6. A, inhibition of Aβ42 fibrillization. Freshly dissolved Aβ42 substrate treated with nIgV 2E6 (4 μg/ml) but not nIgV MMF6 showed reduced ThT binding after 48 h. *, p < 0.01 versus IgV MMF6. ThT fluorescence values of Aβ treated with nIgV MMF6 and diluent after 48 h were indistinguishable (p > 0.1). B, inhibition of Aβ42 oligomerization. The intensities of the SDS-stable trimer and tetramer were reduced by nIgV 2E6 treatment (24 h) of Aβ42 that had not been subjected to prior oligomerization compared with an equivalent nIgV MMF6 concentration (3 μg/ml). Higher order SDS-stable oligomers were less abundant than in preoligomerized Aβ42 preparations. Only the 74-kDa high mass oligomer band was evident, which was reduced significantly by treatment with nIgV 2E6. **, p < 0.005; *, p < 0.03. Inset, Aβ42 oligomers stained with a mixture of Aβ-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 Aβ42 fibrillization by nIgV 2E6 compared with IgG1 59. Aβ42 fibrillization tested as in A was inhibited by nIgV 2E6 (1 μg/ml) but not the Aβ-binding IgG1 59 (1 or 30 μg/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 < 0.01. Inset, IgV 2E6 at a concentration displaying readily detectable Aβ hydrolysis and anti-amyloid effects (4 μg/ml) did not bind the Aβ42 monomer or oligomers determined by immunoblotting (lane 4). Lane 5, control immunoblot after treatment of the same Aβ42 preparation with an equivalent control IgV MMF6 concentration. Lane 1, silver-stained trimer and tetramer bands in the Aβ42 preparation. Lanes 2 and 3, the Aβ42 oligomers stained with the Aβ-binding IgG1 59 and control nonimmune IgG1 SKT03, respectively. Error bars, S.D.

Amyloid Removal in Transgenic Mice

For in vivo clearance tests, we first analyzed brain sections 7 days after administering the nIgV 2E6 (1 μg) directly into the right brain neocortex of 5XFAD transgenic mice, an AD model characterized by robust accumulation of human Aβ in the brain parenchyma. Consistent with previous findings, the needle track was visible from increased Aβ 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 Aβ plaque burden surrounding the nIgV 2E6 delivery site was evident compared with the non-injected left brain neocortex of the same mice (Fig. 5, A and B). The Aβ clearance effect was limited to the immediate vicinity of the injection site, consistent with administration of a small IgV volume (2 μl). There was no decrease of local Aβ burden following nIgV MMF6 or PBS injection into the right brain neocortex compared with the non-injected left brain neocortex.

FIGURE 5.

Reduced brain Aβ 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-Aβ antibody (1 μg of nIgV; n = 7 5XFAD mice in the nIgV 2E6 group, n = 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.

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 ¹²⁵I-labeled aIgV 2E6 in wild type mice (e.g. at 2 h, the radioactivity/g of brain tissue was 5.2 ± 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). Aβ clearance following intravenous aIgV 2E6 administration was analyzed in the TgSwDI mouse model characterized by the predominant vasculotropic deposition of the transgene-encoded mutant Aβ40 peptide, which causes the CAA-like state (25, 26). These mice are suited for testing induction of microvascular dysfunction observed upon infusion of Aβ-binding monoclonal IgGs that penetrate the BBB (8, 41). Moreover, impaired Aβ peptide egress from the brain in the TgSwDI model is thought to minimize compensatory brain Aβ release incidental to peripheral Aβ clearance by Aβ-binding antibodies, with the result that any observed brain Aβ clearance effect is probably due to antibodies that cross the BBB (42). Ten days after intravenous treatment with aIgV 2E6 (100 μg total IgV/mouse), the right hemisphere neocortical Aβ deposits were reduced significantly compared with the control aIgV MMF6 treatment, as judged by immunohistochemical staining of brain sections (28% reduction, p < 0.05; Fig. 6, B and C). The Aβ-clearing effect was confirmed from the reduced hippocampal Aβ deposits (Fig. 6D). ELISA measurements in whole left brain hemisphere extracts showed modest but significant reductions of the soluble and insoluble Aβ40 and Aβ42 levels in the aIgV 2E6-treated mice (Table 2), suggesting a widespread Aβ clearing effect. Aβ38 was not detected reliably at the brain extract concentrations tested (<16 and <0.5 pg/mg insoluble and soluble Aβ38, 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 Aβ clearance by the brief intravenous aIgV 2E6 treatment. In comparison, prolonged intravenous treatment with milligram Aβ-binding IgG amounts is required for brain Aβ clearance in various mouse models of Alzheimer disease (4, 6).

FIGURE 6.

Reduced brain Aβ following intravenous aIgV 2E6 injection. A, aIgV blood and brain levels plotted as a function of time following intravenous injection of the ¹²⁵I-aIgV. n = 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 μg/injection) compared with mice treated equivalently with aIgV MMF6 (n = 8 TgSwDI mice/group). *, p < 0.05. C, top and bottom images, anti-Aβ antibody stained neocortex sections from aIgV MMF6-treated and aIgV 2E6-treated mice, respectively. Horizontal bar, 0.25 mm. Arrow, a plaque. D, hippocampal Aβ plaque burden. Aβ removal in aIgV 2E6-treated mice from B was confirmed in hippocampus (Hippo) sections (n = 3 mice/group; neocortex data from the same mice are included for reference). *, p < 0.05 versus aIgV MMF6. Error bars: in A, S.D.; in B and D, S.E.

TABLE 2.

Aβ content of TgSWDI mouse brain extracts following treatment with aIgV 2E6 and MMF6

Aβ40 and Aβ42 contents were measured by ELISA and are expressed per mg of brain tissue (mean ± S.D.). Particulate Aβ contents represent the difference between the observed total and soluble Aβ content. aIgV effect size corresponds to the reduction of Aβ content in mice treated with aIgV 2E6 expressed as a percentage of the Aβ content in mice treated with control aIgV MMF6. p values are from Student's t test, two-tailed.

Aβ species	aIgV MMF6	aIgV 2E6	aIgV 2E6 effect size	p
	pg/mg	pg/mg	%
Soluble Aβ40	15.5 ± 2.9	11.2 ± 2.6	21.3	<0.04
Soluble Aβ42	1.2 ± 0.2	0.8 ± 0.2	33.3	<0.01
Particulate Aβ40	3,453.5 ± 645.0	2,544.9 ± 472.0	26.3	<0.03
Particulate Aβ42	369.8 ± 68.2	240.1 ± 75.3	32.6	<0.01

FIGURE 7.

Failure of aIgV to induce microglial activation and microbleeds. Data are from the Fig. 6B mice. A, microgliosis scores in the neocortex of mice that received intravenous aIgV 2E6 or MMF6 injections determined by staining for Iba-1. p > 0.05 versus aIgV MMF6. B, neocortex sections stained with antibody to Iba-1 following intravenous aIgV MMF6 (top image) or aIgV 2E6 (bottom image) treatment. There was no evidence of increased Iba-1 expression or microglial clustering. Scale bar, 250 μm. C, no induction of cerebral microhemorrhages by intravenous aIgV 2E6. The number of microhemorrhages in neocortex sections was determined by Perl's iron staining. p > 0.05 versus aIgV MMF6. Error bars, S.E.

DISCUSSION

Two major brain Aβ clearance mechanisms were proposed for intravenously administered Aβ-binding monoclonal IgGs: the microglia-dependent removal of brain Aβ by BBB-penetrating IgGs (e.g. bapineuzumab) (6–8) and increased Aβ egress from the brain induced by Aβ binding to IgGs in peripheral blood (e.g. solanezumab) (43). The catalytic IgV rapidly digested Aβ 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 Aβ, transthyretin amyloid containing the β-sheet amyloid motif, or the Aβ precursor protein, suggesting sufficiently specific Aβ removal. The same Fc receptor-dependent microglial interactions with Aβ-IgG immune complexes that result in beneficial Aβ clearance also hold potential for damaging neurons and other brain cells (5–8). Side-by-side toxicity studies of catalytic IgV and Aβ-binding IgGs remain to be conducted, but it is noteworthy that the catalytic IgV dissolved fibrillar Aβ with no requirement for microglia, and there was no evidence for microglial activation or cerebral microhemorrhages in the IgV-treated mice.

The Aβ 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 Aβ 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 Aβ-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) (47–49).

Other potential factors governing the catabody clearance capacity are (a) the Aβ aggregation status, (b) steric access to the Aβ noncovalent binding epitope and scissile bonds, and (c) conformational transitions in Aβ. Small IgV concentrations were sufficient to hydrolyze particulate Aβ, but diffusional restrictions decelerate enzymatic digestion of particulate substrates compared with soluble substrates (50, 51). We acknowledge the potential of varying clearance rates for Aβ coaggregated with other proteins and Aβ aggregates containing post-translational chemical modifications. Catalysis requires topographically precise IgV interactions at the Aβ(29–40) epitope and the His¹⁴-Gln¹⁵ scissile bond (present study) (14, 29). Conformational factors and limited access to Aβ(29–40) epitope located in the transmembrane APP region probably restrict the proteolytic rate for the membrane-anchored APP. The Aβ(29–40) epitope of APP is helical (34) and acquires increasing β-sheet character with increasing aggregation of the mature Aβ40/42 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 Aβ(29–40) recognition by IgV 2E6. First, the IgV originated from non-aged humans without pathological 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 Aβ hydrolytic V_L domain with a V_H domain suppressed the catalytic activity (14). Fourth, Aβ recognition without acquired immunity processes is not limited to catabodies; selective noncovalent binding of the C-terminal Aβ 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 Aβ 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 Aβ hydrolysis as an innate immune property.

Microbial B-cell superantigens provide a precedent for specific Aβ 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 Aβ 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 (∼4 × 10⁹ 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, 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 Aβ 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 Aβ and antibodies are ancient molecules. The human Aβ(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.