Exceptional Amyloid β Peptide Hydrolyzing Activity of Nonphysiological Immunoglobulin Variable Domain Scaffolds^,

Abstract

Nucleophilic sites in the paired variable domains of the light and heavy chains (V_L and V_H domains) of Ig can catalyze peptide bond hydrolysis. Amyloid β (Aβ)-binding Igs are under consideration for immunotherapy of Alzheimer disease. We searched for Aβ-hydrolyzing human IgV domains (IgVs) in a library containing a majority of single chain Fv clones mimicking physiological V_L-V_H-combining sites and minority IgV populations with nonphysiological structures generated by cloning errors. Random screening and covalent selection of phage-displayed IgVs with an electrophilic Aβ analog identified rare IgVs that hydrolyzed Aβ mainly at His¹⁴-Gln¹⁵. Inhibition of IgV catalysis and irreversible binding by an electrophilic hapten suggested a nucleophilic catalytic mechanism. Structural analysis indicated that the catalytic IgVs are nonphysiological structures, a two domain heterodimeric V_L (IgV_L2-t) and single domain V_L clones with aberrant polypeptide tags (IgV_L-t′). The IgVs hydrolyzed Aβ at rates superior to naturally occurring Igs by 3-4 orders of magnitude. Forced pairing of the single domain V_L with V_H or V_L domains resulted in reduced Aβ hydrolysis, suggesting catalysis by the unpaired V_L domain.Ångstrom level amino acid displacements evident in molecular models of the two domain and unpaired V_L domain clones explain alterations of catalytic activity. In view of their superior catalytic activity, the V_L domain IgVs may help attain clearance of medically important antigens more efficiently than natural Igs.

The antigen-combining sites of immunoglobulins found in higher organisms are composed of the variable domains of light and heavy chain subunits (V_L and V_H domains). The individual V_L and V_H domains can bind antigens independently of each other, but the paired V_L-V_H structure consistently expresses superior antigen binding affinity because of cooperative antigen-binding forces contributed by the two domains (1). High affinity Igs are generated by adaptive V domain sequence diversification over the course of B lymphocyte differentiation, a process in which antigen binding to mutated B cell receptors (surface Igs associated with signal-transducing proteins) drives the selective expansion of the cells. Adaptive Ig maturation entails the use of one each of ∼50 inherited V_L and V_H genes, diversification at the junctions of the V_L-D_L gene segments and V_H-D_H-J_H gene segments, and somatic mutation over the entire length of the V domains.

Following initial noncovalent binding of antigen, some Igs proceed to catalyze its chemical transformation. Examples of Ig-catalyzed reactions include hydrolysis of polypeptide antigens (2, 3), hydrolysis of nucleic acids (4, 5), and various acyl transfer reactions of other antigen classes (6). Proteolytic Igs that utilize serine protease-like covalent hydrolytic pathways have been described (7, 8). Serine protease-like catalytic triads have been identified in the V domains of Igs by site-directed mutagenesis and crystallography (9, 10). The catalytic mechanism involves nucleophilic attack on the electrophilic carbonyl of peptide bonds. Electrophilic phosphonate diesters originally developed as covalent probes for the nucleophilic site of serine proteases bind catalytic Igs irreversibly and inhibit their catalytic activity (7, 11, 12). The strength of Ig-antigen noncovalent binding often exceeds that of enzyme-substrate binding. An important limitation holding back the application of catalytic Igs for clearance of undesirable antigens is that their catalytic rate constants (turnover number; k_cat) are small compared with enzymes. Evidently, Ig adaptive selection is geared toward noncovalent immune complexation (the ground state stabilization step), and the ability of Igs to recognize the high energy transition state complex that must be stabilized to accelerate chemical reactions is limited. This is supported by observations that IgMs, the first and least diversified Ig class produced during B cell differentiation, express superior catalytic rate constants than IgGs produced by the cells at later stages of their adaptive differentiation (12).

Accumulation of amyloid β peptide (Aβ)2 aggregates in the brain is thought to be a central contributor to neurodegenerative changes underlying Alzheimer disease (AD). Administration of monoclonal IgGs that bind Aβ reversibly to transgenic mice overexpressing human Aβ clears brain Aβ deposits and improves cognitive function (13, 14). Suggested mechanisms explaining the favorable effect of peripherally administered IgG are as follows: (a)Aβ containing immune complexes formed by small amounts of IgGs that cross the blood-brain barrier are removed by Fc receptor-mediated uptake by resident macrophages in the brains, the microglia (14); (b)Aβ binding to the IgG constrains the peptide into a nonaggregable conformation (15); (c) IgG bound to FcRn receptors on the blood-brain barrier accelerates Aβ exit from the brain to periphery blood (16); and (d) binding of peripherally circulating Aβ by IgG disrupts equilibrium between the central and peripheral compartments, causing compensatory Aβ release from the brain (17). In principle, Igs that catalyze the hydrolysis of Aβ can be applied to clear Aβ. We reported naturally occurring IgMs and isolated Ig light chain subunits (IgLs) that hydrolyze Aβ impede Aβ aggregation and inhibit Aβ-induced neurotoxicity (18, 19). However, these Igs hydrolyze Aβ slowly, and development of more efficient catalysts will help advance the use of catalytic Igs for Aβ clearance.

We report here the search for efficient Aβ-hydrolyzing Ig fragments in a human IgV domain (IgV) library in which the majority of clones are single chain Fv constructs (scFv-t; a V_L domain attached via a linker peptide to a V_H domain). The scFv scaffold mimics the physiological structure of antigen-combining sites. A minority of clones in the library are nonphysiological V domain structures generated by repertoire cloning errors. Unexpectedly, the nonphysiological two domain and single domain IgV_L fragments expressed exceptional Aβ hydrolyzing efficiency. scFv-t derivatives obtained by repairing a high activity single domain IgV_L displayed reduced catalytic activity. The observations suggest that novel Ig structures freed of constraints imposed by the physiological organization of V domains can be the source of efficient catalysts to medically important antigens.

MATERIALS AND METHODS

Electrophilic Compounds—Syntheses and Ig binding characteristics of these compounds are reported: E-hapten 1 and 2 (20) and E-hapten 3 (12). Bt-E-Aβ40 was prepared by reacting biotinylated Aβ-(1-40) (Aβ40; 10 mg, 2.1 μmol) with diphenyl-N-[O-(3-sulfosuccinimidyl)suberoyl]-amino(4-amidinophenyl)methane phosphonate (10.6 mg, 12.5 μmol) in DMSO. The reaction mixture was purified by RP-HPLC (Waters) and lyophilized. Its identity and purity were confirmed by RP-HPLC (retention time 36.36 min, purity >99.9%, Vydac C4 column; 0.05% trifluoroacetic acid in water, 0.05% trifluoroacetic acid in MeCN 90:10-40:60 in 50 min, 1.0 ml/min; 220 nm absorbance) and electrospray ionization-mass spectrometry (ESI-MS; observed m/z, 1427.6, 1142.6 and 952.4; calculated (M + H)⁴⁺,(M + 2H)⁵⁺, and (M + 3H)⁶⁺ for C₂₆₆H₃₈₀N₆₂O₇₁P₂S₂, 1427.2, 1141.9, and 951.8).

Igs—Methods for IgV preparation and characterization have been described (7, 21). Briefly, the library consists of 1.4 × 10⁷ IgVs cloned in the phagemid vector pHEN2 prepared from the peripheral blood lymphocyte of patients with lupus. A His₆ sequence and a c-myc epitope are located at the IgV C terminus. Expression levels were 1-3 mg of IgV/liter of bacterial culture, determined by anti-c-myc immunoblotting. Soluble IgVs were purified from periplasmic extracts of HB2151 cells by metal affinity chromatography. Further purification was by anion exchange FPLC (MonoQ HR 5/5 column; 0-1 m NaCl in 50 mm Tris buffer, pH 7.4, containing 0.1 mm CHAPS). Purity was determined by SDS-gel electrophoresis and immunoblotting. IgV phages (10¹² colony-forming units) were packaged using the hyperphage method (22) and incubated (2 h, 37 °C) with Bt-E-Aβ40 in 0.07 ml of 10 mm sodium phosphate, 137 mm NaCl, 2.7 mm KCl, pH 7.4 (PBS). Phages with bound Bt-E-Aβ40 were captured using anti-biotin antibody coupled to agarose gel (0.22 ml settled gel; Sigma) and washed with 100 ml of PBS containing 0.1% bovine serum albumin. Reversibly bound phages were eluted by incubation of the gel in 0.2 ml of 100 μm Aβ40 for 1 h with slow mixing, and the residual phages covalently complexed to Bt-E-Aβ were eluted with 0.4 ml of 0.1 m glycine, pH 2.7. scFv-t derivatives of single domain IgV_L-t′ 5D3 were prepared by inserting the deleted V_H residues 8-115 (Kabat numbering). For this, full-length V_H cDNA was amplified by PCR using as template the IgV-pHEN2 DNA library and back/forward primers containing ApaLI/NotI restriction sites (respectively, GGTAGTGCACTTCAGGTGCAGCTGTTGCAGTCT/ATGTGCGGCCGCGGGGAAAAGGGTTGGGGGCATGC), and the cDNA digested with ApaLI/NotI was ligated into similarly digested plasmid IgV_L-t′ 5D3 DNA with T4 DNA ligase (Invitrogen). Full-length light chain L-t′ 5D3 was prepared by Mutagenex by a chimeragenesis method (23) using as starting materials the V_L domain of IgV_L-t′ 5D3 and human κ chain constant domain (obtained from pLC-huCκ (24)) and cloned into pHEN2 vector as the NcoI/NotI-digested fragment. cDNAs for the homodimeric IgV_L2-t form of clone 5D3 were prepared by PCR by Mutagenex. Briefly, V_L cDNA was amplified by PCR from the IgV_L-t′5D3 template using back/forward primers containing ApaLI/NotI restriction sites (respectively, AAAGTGCACTTGAAATTGTGTTGACGCAGTCTC/AAAGCGGCCGCGCGTTTGATCTCCAGCTTGGT), and the cDNA digested with ApaLI/NotI was ligated into pHEN2 vector. The nucleotide sequence of all constructs determined by dideoxy nucleotide sequencing in the 5′ to 3′ and 3′ to 5′ directions was identical (Applied Biosystems, ABI PRISM® 3100 Genetic Analyzer). Following electroporation of IgV phagemid DNA into HB2151 cells, soluble IgVs were purified from periplasmic extracts as before. Total protein was determined by the microBCA kit (Pierce). For mass spectroscopy (25), the IgV band was excised from the SDS-electrophoresis gel stained with GelCode Blue (Pierce), subjected to dehydration in 50% acetonitrile and SpeedVac drying, reduced (dithiothreitol) and alkylated (iodoacetamide), and digested with sequencing grade trypsin (Promega) and Lys-C (Wako) for 20 h at 37 °C. Following extraction of gel fragments with acetonitrile/formic acid, digested peptides obtained by ZipTip C18 (Millipore) fractionation using 5 μl of aqueous 50% acetonitrile containing 2% formic acid were analyzed by mass spectrometry using a matrix of α-cyano-4-hydroxycinnamic acid (ABI 4700 MALDI-TOF/TOF mass spectrometer). Predicted monoisotopic peptide mass values were obtained using MS-Fit for protein data base searches (Protein Prospector, University of California, San Francisco). IgM was purified from human sera as described (18).

Hydrolysis and Binding Assays—Hydrolysis of ¹²⁵I-Aβ40 was determined as described (18). Briefly, ¹²⁵I-Aβ40 prepared by the chloramine-T method was purified by RP-HPLC (2.2 Ci/μmol). The ¹²⁵I-Aβ40 (∼0.1 nm, ∼30,000 cpm/tube) was treated with IgVs in PBS containing 0.1 mm CHAPS and 0.1% (w/v) bovine serum albumin; intact peptide was separated from fragments by precipitation with trichloroacetic acid, and acid-soluble radioactivity was counted and corrected for background values in control assay tubes incubated in diluent without Ig (mean ± S.D., 18 ± 6%; n = 7 assays). This procedure affords estimates of hydrolysis concordant with RP-HPLC separation of the reaction mixtures. Apparent kinetic parameters were estimated by fitting hydrolysis rates observed at varying Aβ40 concentrations mixed with a constant amount of ¹²⁵I-Aβ40 to the following equation: v = (V_max[Aβ40])/K_m + [Aβ40], where V_max is the maximum velocity at saturating Aβ40 concentrations, and K_m is the concentration at which half-maximal velocity was observed. To identify the reaction products, reaction mixtures of nonradiolabeled synthetic Aβ40 or Aβ42 (100 μm; American Peptide Co.) incubated with IgVs in PBS/CHAPS were desalted by gel filtration (Bio-Rad micro Bio-spin 6 columns), lyophilized, and subjected to MALDI-TOF MS with α-cyano-4-hydroxycinnamic acid as matrix (positive ion mode, 20,000 V). RP-HPLC of Aβ40-Ig reaction mixtures and ESI-MS identification of the product have been described previously (18). Hydrolysis of the amide bond linking 7-amino-4-methylcoumarin (AMC) to the C-terminal amino acid of peptide-AMC substrates (Peptides International) was measured in PBS/CHAPS buffer by fluorimetry with authentic AMC as reference (λem 470 nm; λex 360 nm (12)). Hydrolysis of biotinylated proteins was determined by SDS-electrophoresis using peroxidase-conjugated streptavidin to stain blots of the gels (18). IgV covalent binding to biotinylated E-hapten 2 or E-hapten 3 was assayed in PBS/CHAPS (12). The reaction mixtures were boiled in SDS in reducing buffer (5 min) and subjected to SDS-electrophoresis, and blots of the gels were stained with the streptavidin-peroxidase conjugate. IgV binding to immobilized Bt-Aβ40 was determined by enzyme-linked immunosorbent as described (18) except that anti-c-myc antibody (1:100) was employed to detect IgVs bound to immobilized antigens (26).

IgV Modeling—The two V_L domains of the heterodimeric IgV_L2-t 2E6 located, respectively, on the N- and C-terminal side of the linker (designated VL1 and VL2) were initially modeled as monomers by sequence alignment to the most-homologous V_L domains in the Protein Data Bank (PDB codes, respectively, 1MCB and 2BX5; 85-95% sequence identity) and homology modeling with DS 1.7 (Accelrys; modeler module followed by minimization in CHARMm force field; 1000 cycles). The VL1 and VL2 structures were then refined in dimeric form using as template the light chain dimer PDB 1MCW. The flexible interdomain linker peptide and C-terminal tag region were incorporated into the model, and minor steric clashes were removed by energy minimization using CNS 1.1 (200 cycles; see Ref. 27). The V_L domain of the single domain IgV_L-t′ 5D3 was initially modeled in its paired V_L-V_H scFv-t 5D3-E6 form by the WAM server. V_L models in the IgV_L-t′, IgV_L2-t, and full-length light chain L-t forms of the molecule were prepared by superimposing the V_L domain from the WAM model to the coordinates of the homodimeric light chain crystal structure (PDB 1B6D; 85% sequence identity). The structures were submitted to steepest-descent energy minimization using the Adopted Basis-set Newton-Raphson method under the CHARMm force field (2000 cycles) until an r.m.s. deviation of 0.1 kcal/mol/Å was obtained. The quality of all structures was checked using PROCHECK. Percent V_L residues in the final models located in the most favored or generous regions of the Ramachandran plot were as follows: IgV_L2-t 2E6, 90.8%; IgV_L-t′ 5D3, 96.5%; scFv-t 5D3-E6, 97.7%; IgV_L2-t 5D3 homodimer, 95.5%; L-t 5D3, 96.9%. No unacceptable atomic collisions were detected. The van der Waals energy was negative, suggesting the absence of bad non-bonded contacts. Superimposition and determination of global r.m.s. deviation and translational Cα-Cα movements were performed using PyMOL (DeLano Scientific LLC). The feasibility of Aβ40 interactions with IgV_L2-t 2E6 was assessed by molecular docking as described (18) using ZDOCK with Gln¹⁵ of Aβ40 constrained within 12 Å of each putative nucleophilic residue (VL1 domain: Ser^27a, Tyr⁸⁷, Ser⁹³, and Thr¹⁰⁵; VL2 domain: Thr⁸⁵, Ser⁷⁶, and Ser⁹¹). The docked model containing VL1 domain Thr¹⁰⁵ apposed to Aβ40 Gln¹⁵ was the energetically most favored structure.

RESULTS

Aβ40 Hydrolyzing IgVs—We reported previously the hydrolysis of Aβ40 by polyclonal IgM purified from humans without dementia (18). Here we searched for Aβ40-hydrolyzing human IgVs in a library composed of ∼10⁷ clones. A majority of the clones in the library are scFv-t constructs with the domain organization V_L-Li-V_H-t, where Li denotes the 16-residue peptide SS(GGGGS)₂GGSA joining the V_L domain C terminus to the V_H domain N terminus, and t denotes the 26-residue C-terminal peptide containing the c-myc peptide and His₆ tags (7). A minority of clones possess unusual IgV structures generated by cloning errors (see below). Sixty three IgVs purified from the periplasmic extracts of randomly picked clones by His₆ binding to nickel affinity columns were tested for ¹²⁵I-Aβ40 hydrolyzing activity. Two IgVs with activity markedly superior to the remaining clones were identified (Fig. 1A). The activity of the empty vector control extract (pHEN2 devoid of an IgV insert) was within the assay error range (50 cpm/h, corresponding to mean background acid soluble radioactivity + 3 S.D.). The phagemid DNA of the high activity IgV clone 2E6 was re-expressed in 15 individual bacterial colonies. All recloned colonies secreted IgV with robust ¹²⁵I-Aβ40 hydrolyzing activity (Fig. 1B), ruling out trivial sample preparation variations as the cause of proteolytic activity. As before, the purified extract of the control empty vector clone did not hydrolyze ¹²⁵I-Aβ40.

FIGURE 1.

¹²⁵I-Aβ40 hydrolysis by randomly picked IgV clones. A, identification of rare Aβ-hydrolyzing IgVs. Sixty three IgVs were purified from periplasmic extracts (15 ml) of bacterial colonies picked randomly from the IgV library grown on an agar plate. Aliquots of eluates (75 μl) obtained by metal affinity chromatography of the periplasmic extracts (total eluate volume 0.9 ml) were tested for Aβ hydrolyzing activity. Values are means of closely agreeing duplicates (S.D. < 7.3% of mean). Each point represents one clone. ¹²⁵I-Aβ40, 0.1 nm; 18 h of incubation. B, validation of IgV 2E6 hydrolytic activity by recloning. Phagemid DNA isolated from the high activity clone 2E6 was electroporated into bacteria, and purified IgV from 15 individual colonies obtained from the recloning procedure were analyzed for Aβ hydrolyzing activity as in A along with the identically purified empty vector control (pHEN2).

In previous studies, electrophilic phosphonate groups incorporated within polypeptides were bound covalently by catalytic Ig nucleophilic sites, with noncovalent binding at the peptide epitopes conferring specificity to the reaction (21). We employed the biotinylated Aβ40 analog containing phosphonates at Lys¹⁶ and Lys²⁸ side chains to isolate Aβ40 catalysts displayed on phage surface (Bt-E-Aβ40; Fig. 2A). Phage IgVs treated with Bt-E-Aβ40 were captured using immobilized antibiotin antibody, and noncovalently bound phage IgVs were eluted by treatment with excess Aβ40 (designated noncovalently selected IgVs), and the irreversible phage IgV immune complexes were eluted by acid disruption of the biotin-antibiotin antibody complexes (designated covalently selected IgVs). The frequency of IgVs with robust Aβ hydrolyzing activity was increased by covalent selection (Fig. 2B). Four of 7 IgVs obtained by covalent selection at 2 μm Aβ40 hydrolyzed ¹²⁵I-Aβ40 at rates >400 cpm/h, compared with 2 of 63 IgVs with this level of activity identified by random screening. Phage selections conducted at increased Aβ40 concentration (10 μm) yielded less active IgVs (Fig. 2B), consistent with the prediction of more efficient selection of catalysts at the lower ligand concentration. Eighteen IgVs recovered by noncovalent selection displayed no or little hydrolytic activity.

FIGURE 2.

Covalent selection of Aβ-hydrolyzing IgVs. A, structure of Bt-E-Aβ40. The electrophilic phosphonates placed within Aβ40 permit specific covalent binding to nucleophilic Igs facilitated by noncovalent Aβ40 recognition. B, frequent Aβ-hydrolyzing IgVs obtained by covalent selection. The IgV-phage library was treated with Bt-E-Aβ40 (2 or 10 μm), and covalently bound phages were recovered as described under “Materials and Methods,” and hydrolysis of ¹²⁵I-Aβ40 (0.1 nm) by purified IgVs obtained from selected clones was measured as in Fig. 1A. Each point represents an individual IgV clone.

IgV Primary Structure and Activity Validation—The cDNAs for IgV clone 2E6 obtained without phage selection and three covalently selected IgVs with the greatest Aβ40 hydrolyzing activity (clones 5D3, 1E4, and 5H3) were sequenced. Identical nucleotide sequences were obtained for each clone sequenced from the 5′ to 3′ direction and the 3′ to 5′ direction. The cDNA sequences indicated that IgV 2E6 is a dimer of two different V_L domains with the intervening linker peptide and the expected C-terminal tag (designated heterodimeric IgV_L2-t, Fig. 3A; sequences in supplemental Fig. S1). IgVs 5D3, 1E4, and 5H3 are single domain V_L clones with unexpected C-terminal polypeptide segments, designated IgV_L-t′ clones, where t′ denotes the following: (a) the expected linker peptide, (b) a 15-28-residue aberrant peptide sequence in place of the V_H domain composed of ∼115 residues, and (c) the expected 26 residue peptide sequence containing the c-myc and His₆ sequences (Fig. 3A and supplemental Fig. S1).

FIGURE 3.

Structure and hydrolytic activity of IgV_L2-t 2E6 and IgV_L-t′ 5D3. A, schematic representation of IgV structures. scFv-t, V_L, and V_H domains connected by a peptide linker with t, the His⁶/c-myc tag located at the C terminus; IgV_L2-t 2E6, a heterodimer of two V_L domains with the linker and t; IgV_L-t′ 5D3, a single domain V_L with t′ at the C terminus, corresponding to the peptide linker, a short peptide region in place of the V_H domain and the t tag. B, SDS-electrophoresis of IgVs. Lanes 1 and 2, respectively, IgV_L2-t 2E6 purified by metal affinity chromatography followed by anion exchange FPLC and stained with silver and anti-c-myc antibody. Lanes 3 and 4, respectively, IgV_L2-t 2E6 purified by two cycles of metal affinity chromatography and stained with silver and anti-c-myc antibody. Lanes 5 and 6, respectively, IgV_L-t′ 5D3 purified by metal affinity chromatography followed by anion exchange FPLC and stained with silver and anti-c-myc antibody. Lanes 7 and 8, respectively, IgV_L-t′ 5D3 purified by two cycles of metal affinity chromatography and stained with silver and anti-c-myc antibody. C, ¹²⁵I-Aβ40 hydrolysis (cpm/h/μg Ig; means ± S.D.) by monoclonal IgM Yvo, a pooled human polyclonal IgM preparation, IgV_L2-t 2E6, and IgV_L-t′ 5D3. ¹²⁵I-Aβ40, ∼30,000 cpm (∼0.1 nm). The monoclonal and polyclonal IgM preparations are described previously (18). Data are from assays at 90-405 μg/ml IgMs and 0.075-0.55 μg/ml IgVs.

Two clones were studied further, IgV_L2-t 2E6 and IgV_L-t′ 5D3. Their deduced protein masses predicted from the cDNA sequences are, respectively, 27 and 17 kDa. Denaturing electrophoresis of the IgVs purified by 2 cycles of nickel-affinity chromatography and anion exchange FPLC (IgV_L2-t 2E6 and IgV_L-t′ 5D3 fractions corresponding, respectively, to retention times 10-11 and 23-23.5 min; supplemental Fig. S2) revealed silver and anti-c-myc stainable protein bands close to the predicted mass of the monomer proteins (IgV_L2-t 2E6, 29 kDa; IgV_L-t′ 5D3, 18 kDa; Fig. 3B). The presence of the c-myc peptide epitope confirms that these are IgV bands. In view of its unusual structure, the identity of IgV_L-t′ 5D3 monomer band was confirmed further by tryptic digestion and mass spectroscopy (supplemental Table S1). All observed spectroscopic signals originated from peptides within the predicted IgV_L-t′ structure deduced from the cDNA sequence, and the peptide signals were consistent with deletion of V_H residues 8-115 predicted from the cDNA sequence. Additional IgV bands were detected prior to anion exchange chromatography (low mass IgV_L2-t 2E6 band at 18 kDa; high mass IgV_L-t′ 5D3 bands at 36, 50, 58, 67, and 74 kDa). All of these bands were stained by anti-c-myc antibody (Fig. 3B). As irrelevant proteins are not stained by the anti-c-myc antibody, there is no evidence of non-IgV contaminants. We concluded that the anomalous low mass band is an IgV_L2-t self-degradation product, and the high mass bands are IgV_L-t′ aggregates. ¹²⁵I-Aβ40 hydrolyzing activities of both IgVs remained constant after one and two cycles of nickel-affinity chromatography but were increased following further FPLC purification that removed the degradation product and aggregates (Table 1; by 3.1- and 31.2-fold, respectively, for the IgV_L2-t and IgV_L-t′). These observations are consistent with Aβ hydrolysis by the unaggregated IgVs.

TABLE 1.

¹²⁵I-Aβ-(1-40) hydrolyzing activity of IgVs following metal affinity and anion exchange FPLC purification

Recombinant IgV preparations purified by one round of metal affinity chromatography on nickel-agarose (preparation MA1) were subjected either to a second round of metal affinity chromatography (MA2) or anion exchange chromatography (AEQ) on a Mono Q FPLC column. Aβ hydrolysis was assayed as in Fig. 2. Shown are specific activities expressed in cpm/h/μg of protein.

Catalyst	Specific activity, cpm/h/μg of protein IgV (mean ± S.D.)
	MA1	MA2	AEQ
IgVL2-t 2E6	1318 ± 375	1220 ± 134	4495 ± 1035
IgVL-t′ 5D3	2660 ± 114	2273 ± 308	61401 ± 5274

Sequencing of 24 randomly picked clones indicated that most IgVs in the library are scFv-t constructs (83.3% clones), with only rare representation of IgV_L2-t and IgV_L-t′ constructs (respectively, 12.5 and 4.2% clones; supplemental Table S2). The cumulative probability that all four Aβ40-hydrolyzing IgVs identified in the present study are IgV_L2-t or IgV_L-t′ clones by random chance is very small (p = 0.9 × 10^-5; computed as 0.125 × 0.042³). It may be concluded that the rare IgV structures favor expression of Aβ hydrolysis compared with the physiological V_L-V_H paired structure of scFv-t clones. This is supported by comparisons of IgV catalytic activities with the previously reported polyclonal human IgM preparations and a monoclonal IgM from a patient with Waldenstrom's macro-globulinemia (18). IgV_L2-t 2E6 and IgV_L-t′ 5D3 hydrolyzed ¹²⁵I-Aβ40 with potencies superior to the IgMs by 3-4 orders of magnitude (Fig. 3C).

Repaired IgV 5D3 Versions—The aberrant t′ region of IgV_L-t′ 5D3 contains a deletion of V_H domain residues 8-115 (Kabat numbering). Four scFv-t constructs were generated from the IgV_L-t′ by inserting the deleted residues derived from full-length V_H domains represented in the library (supplemental Fig. S3). The repaired scFv-t 5D3 derivatives migrated at the expected mass in electrophoresis gels (30 kDa, example in Fig. 4A, inset). Their ¹²⁵I-Aβ40 hydrolyzing activity was consistently lower than the parent IgV_L-t′ (by ∼82-167-fold, computed by rate comparisons in the linear region of the hydrolysis curves; Fig. 4A). This suggests that Aβ40 hydrolysis occurs at an autonomous catalytic site in the IgV_L-t′ V_L domain that is suppressed by pairing with V_H domains.

FIGURE 4.

¹²⁵I-Aβ40 hydrolysis by single domain IgV_L-t′ 5D3 and its paired V domain derivatives. A, hydrolytic activity of the IgV_L-t′ (○) and its scFv-t derivatives (•), ¹²⁵I-Aβ40, ∼30,000 cpm, (∼0.1 nm). Values are means ± S.D. Incubation, 18 h. Inset, lanes 1 and 2, respectively, SDS-electrophoresis gels of the IgV_L-t′ stained with silver and anti-c-myc antibody. Lanes 3 and 4, respectively, an example scFv-t derivative of 5D3 (clone 5D3-E6) stained with silver and anti c-myc antibody. The bands at 18 and 30 kDa are, respectively, the IgV_L-t′ and scFv-t. B, hydrolytic activity of full-length l-t 5D3 and homodimeric IgV_L2-t 5D3. ¹²⁵I-Aβ40 hydrolysis assayed as in A. Inset, lanes 1 and 2, respectively, SDS-electrophoresis gels of the L-t stained with silver and anti-c-myc antibody. Lanes 3 and 4, respectively, the IgV_L2-t stained with silver and anti-c-myc antibody. The bands at 30 and 27 kDa are, respectively, the L-t and IgV_L2-t.

Intermolecular noncovalent bonding between isolated light chains can generate dimeric light chain structures (29). To assess whether noncovalently associated IgV_L-t′ dimers containing paired V_L-V_L structures might account for the hydrolytic activity, we prepared the homodimeric IgV_L2-t molecule containing two 5D3 V_L domains connected by the peptide linker. Homodimeric IgV_L2-t 5D3 hydrolyzed ¹²⁵I-Aβ40 poorly (Fig. 4B). In comparison, the full-length light chain containing the 5D3 V_L domain linked to the κ constant domain (designated L-t), displayed detectable hydrolytic activity. In denaturing electrophoresis gels, the homodimeric IgV_L2-t 5D3 migrated as a single protein band at the predicted 27-kDa position (Fig. 4B, inset). As expected, L-t 5D3 migrated as a mixture of monomers and S-S-bonded dimers with nominal mass, respectively, 30 and 60 kDa. These observations suggest that the monomeric, unpaired state of the V_L domain is required to maintain catalytic site integrity.

Catalytic Properties—The hydrolytic activity of both IgV clones was saturable with increasing Aβ40 concentrations (1-100 μm nonradioactive Aβ40 mixed with 0.1 nm ¹²⁵I-Aβ40). Kinetic parameters are reported in Table 2. MALDI-MS of nonradioactive Aβ40 treated with IgV_L2-t 2E6 or IgV_L-t′ 5D3 indicated similar product profiles (Fig. 5, A and B). The prominent products were Aβ-(1-14) and Aβ-(15-40), suggesting hydrolysis of the His¹⁴-Gln¹⁵ peptide bond (Fig. 5E; observed and calculated m/z values in Fig. 5 legend). Smaller signals for the following peptide products were also detected as follows: Aβ-(1-15), Aβ-(1-20), Aβ-(16-40), and Aβ-(21-40). The corresponding scissile bonds in Aβ40 are Gln¹⁵-Lys¹⁶ and Phe²⁰-Ala²¹. Similar product profiles were evident in reaction mixtures of the longer Aβ42 peptide treated with the IgV_L2-t 2E6 (Fig. 5, C and D), except that additional minor products suggesting cleavage at the Lys²⁸-Gly²⁹ and Gly²⁹-Ala³⁰ bonds were evident. As accurate product quantification by MALDI-MS is difficult, we also conducted RP-HPLC of Aβ40 treated with IgV_L2-t 2E6. This indicated depletion of the intact Aβ40 peak, accompanied by appearance of a major peptide product absent in control chromatograms of the IgV alone or Aβ40 alone (supplemental Fig. S4A). The product peak was identified as Aβ-(1-14) by ESI-MS, confirming the His¹⁴-Gln¹⁵ bond as the major cleavage site.

TABLE 2.

Apparent kinetic parameters for IgV-catalyzed Aβ40 hydrolysis

Reaction rates were measured following incubation of increasing Aβ40 (1-100 μM) containing a constant amount of ¹²⁵I-Aβ40 (~300,000 cpm) with IgV_L2-t2E6 (3.75 μg/ml) or IgV_L-t′ 5D3 (0.15 μg/ml). Kinetic constants were obtained from nonlinear regression fits to the Michaelis-Menten equation (r² for IgV_L2-t 2E6 and IgV_L-t′ 5D3, respectively, 0.99 and 0.98).

Catalyst	Km	kcat	kcat/Km
	M × 10−5	min−1	M−1 min−1 × 103
IgVL2-t 2E6	8.0 ± 1.2	0.3 ± 0.1	3.7 ± 0.6
IgVL-t′ 5D3	1.7 ± 0.1	1.0 ± 0.1	59.0 ± 4.5

FIGURE 5.

Peptide bonds in Aβ40 and Aβ42 hydrolyzed by IgV_L2-t 2E6. A and B, MALDI-TOF spectra of Aβ40 incubated, respectively, with diluent or IgV_L2-t 2E6. C and D, MALDI-TOF spectra of Aβ42 incubated, respectively, with diluent or IgV_L2-t 2E6. E, scissile bonds deduced from the spectra (arrows). Aβ, 100 μm; IgV_L2-t 2E6, 1 μm; 89-h incubation. Peak identities are as follows: 1, Aβ-(1-14) (calculated (M + H)⁺ 1698.7, observed (M + H)⁺ 1698.8); 2, Aβ-(1-15) (calculated (M + H)⁺ 1826.8, observed (M + H)⁺ 1826.9); 3, Aβ-(21-40) (calculated (M + H)⁺ 1886.0, observed (M + H)⁺ 1886.1); 4, Aβ-(1-20) (calculated (M + H)⁺ 2461.2, observed (M + H)⁺ 2461.2); 5, Aβ-(16-40) (calculated (M + H)⁺ 2520.4, observed (M + H)⁺ 2520.4); 6, pGluAβ-(15-40) (calculated (M + H)⁺ 2631.4, observed (M + H)⁺ 2631.5); 7, Aβ-(15-40) (calculated (M + H)⁺ 2648.5, observed (M + H)⁺ 2648.5); 8, full-length Aβ40 (calculated (M + H)⁺ 4328.2, observed (M + H)⁺ 4328.1); 9, full-length Aβ40 (calculated (2 M + H)⁺ 2164.6, observed (2 M + H)⁺ 2164.6); 10, Aβ-(15-29) (calculated (M + H)⁺ 1638.9, observed (M + H)⁺ 1638.9); 11, Aβ-(15-28) (calculated (M + H)⁺ 1581.8, observed (M + H)⁺ 1581.8); 12, Aβ-(21-42) (calculated (M + H)⁺ 2070.1, observed (M + H)⁺ 2070.2); 13, pGluAβ-(15-42) (calculated (M + H)⁺ 2814.6, observed (M + H)⁺ 2814.8); 14, Aβ-(15-42) (calculated (M + H)⁺ 2832.6, observed (M + H)⁺ 2832.6); 15, full-length Aβ42 (calculated (M + H)⁺ 4512.3, observed (M + H)⁺ 4512.1); 16, full-length Aβ42 (calculated (2 M + H)⁺ 2256.7, observed (2 M + H)⁺ 2256.6).

The ¹²⁵I-Aβ40 hydrolysis measurements were conducted using a small amount of the Aβ40 substrate (0.1 nm) mixed with excess albumin (1 mg/ml; 15 μm), which can serve as an alternate substrate for promiscuous catalysts. As hydrolysis of ¹²⁵I-Aβ40 was detected readily, the IgVs do not appear to be nonspecific catalysts. In addition, there was no evidence for hydrolysis of several irrelevant biotinylated polypeptides (ovalbumin, soluble extracellular domain of the epidermal growth factor receptor, human immunodeficiency virus gp120, protein A; supplemental Fig S4B). Previous reports have identified promiscuous Igs present in human blood using model fluorigenic peptide substrates (12). IgV_L2-t 2E6 and IgV_L-t′ 5D3 failed to hydrolyze the model peptide substrate appreciably (Table 3), whereas a representative human polyclonal IgM preparation 9010 hydrolyzed large amounts of Arg/Lys-containing peptide substrates. The data indicate specific Aβ hydrolysis by the IgVs.

TABLE 3.

Negligible hydrolysis of model peptide substrates by Aβ hydrolyzing IgVs

ND indicates not detected; one-letter amino acid code is used. Peptide-AMC substrates, 200 μM; purified IgVs and polyclonal human IgM 9010, 30 μg/ml; 37 °C. Values (means of three replicates ± S.D.) are slopes of progress curves monitored by fluorimetry as in Ref. 12.

Substrate	Hydrolysis, μM AMC/h/μM Ig
	IgV1,2-t 2E6	IgVL-t′ 5D3	IgM 9010
AE-AMC	ND	ND	ND
AAA-AMC	ND	ND	ND
IIW-AMC	ND	ND	ND
AAPF-AMC	ND	ND	ND
EKK-AMC	0.030 ± 0.001	0.104 ± 0.001	13.1 ± 1.4
VLK-AMC	ND	ND	6.9 ± 0.4
EAR-AMC	0.005 ± 0.001	0.023 ± 0.001	61.9 ± 6.9
IEGR-AMC	0.010 ± 0.001	0.052 ± 0.001	2.0 ± 0.7
PFR-AMC	ND	ND	37.4 ± 0.7

Previous reports have indicated that proteolytic Igs utilize a serine protease-like catalytic mechanism entailing nucleophilic attack on the electrophilic carbonyl of peptide bonds (7, 9). This was the basis for the covalent phage IgV selection in this study. To confirm the mechanism, we studied the reactivity of IgV_L2-t 2E6 and IgV_L-t′ 5D3 with the electrophilic phosphonate diester E-hapten-1 (Fig. 6A), a compound originally developed as a covalent inhibitor of serine proteases (30). E-hapten-1 inhibited ¹²⁵I-Aβ40 hydrolysis by both clones (Fig. 6B). The biotin-containing version of the phosphonate diester, E-hapten-2, formed 30-kDa covalent adducts with IgV_L2-t 2E6 that were stable to boiling and SDS treatment (Fig. 6B, inset). Control hapten-3, a poorly electrophilic analog of E-hapten-2, did not form detectable adducts with the IgV_L2-t. The results support a nucleophilic catalytic mechanism.

FIGURE 6.

Inhibition of IgV catalysis by E-hapten-1. A, E-hapten-1 is an active site-directed inhibitor of serine proteases. E-hapten-2, a biotinylated analog of E-hapten-1, was used to detect covalent IgV adducts. E-hapten-3, the unesterified phosphonic acid analog of E-hapten-1, is the poorly electrophilic control compound. B, inhibition of IgV_L2-t 2E6 and IgV_L-t′ 5D3 catalyzed ¹²⁵I-Aβ40 hydrolysis by E-hapten-1. IgV_L2-t 2E6 (0.55 μg/ml) or IgV_L-t′ 5D3 (0.075 μg/ml) was treated with E-hapten-1 for 2 h followed by determination of ¹²⁵I-Aβ hydrolytic activity. Inset, streptavidin-peroxidase-stained blots of SDS gels showing IgV_L2-t 2E6 (20 μg/ml) treated for 18 h with 0.1 mm E-hapten-2 (lane 2) or E-hapten-3 (lane 3). Lane 1, IgV_L2-t 2E6 stained with silver.

IgV_L2-t 2E6 and IgV_L-t′ 5D3 (20 μg/ml) did not bind detectably to immobilized Bt-Aβ40 in enzyme-linked immunosorbent assay tests (A₄₉₀ < 0.06). Under similar conditions, the reference anti-Aβ monoclonal IgG (6E10, 0.1 μg/ml) afforded readily detected binding activity (A₄₉₀, 0.7 ± 0.01).

Molecular Models—Intramolecular H-bonding between triads and dyads of amino acids enhances the nucleophilicity of certain side chains responsible for enzymatic catalysis. Examples are the hydroxyl side chains of Ser, Thr, and Tyr residues activated by spatially neighboring general bases contributed by His, Lys, Arg, Tyr, Glu, and Asp residues (31-34). We screened molecular models of IgV_L2-t 2E6 and IgV_L-t′ 5D3 for side chain hydroxyls located within 4 Å of atoms that can serve as general bases as described in Ref. 31. One triad and several dyads fulfilling this requirement were found in each of the catalysts (supplemental Fig. S5 and supplemental Table S3). The presence of the potential nucleophiles is consistent with the mechanism of IgV catalysis suggested by electrophilic inhibitor studies. Several complexes containing the candidate nucleophilic residues of IgV_L2-t 2E6 apposed to Gln¹⁵ of the major Aβ40 scissile bond were evaluated by molecular modeling. Among these, the complex containing VL1 domain Thr¹⁰⁵ apposed to the scissile bond was the energetically most favored structure. This complex also contained various noncovalent stabilizing interactions between IgV_L2-t 2E6 and Aβ40, including VL1 domain Gly⁴¹ and Ala⁴³ backbone atoms hydrogen bonded with, respectively, Aβ40 Asp²³ backbone and side chain atoms.

The V_L domain of clone 5D3 is highly catalytic in the unpaired IgV_L-t′ state and poorly catalytic when paired with a second V domain in scFv-t or homodimeric IgV_L2-t states. Sub-angstrom movements of electronegative atoms can weaken or strengthen H-bonds and thereby modulate the nucleophilic and proteolytic activities (35, 36). Frequent backbone displacements on the order of 0.5-1.7 Å were evident by energy minimization of the V_L domain modeled in the unpaired state versus the paired scFv-t or IgV_L2-t states (Fig. 7, A and B). Spatial displacement of amino acid side chains that influence H-bonding strength and increase or decrease the nucleophilic reactivity could also occur by virtue of rotation around single bonds (see supplemental Table S3 for changes of inter-residue distances within the potential nucleophilic sites in the unpaired and paired V_L domain states). The modeling results therefore suggest the feasibility of altered nucleophilic reactivity and provide a rational explanation for unequal catalysis by various V_L domain-containing molecules.

FIGURE 7.

Models of the V_L domain of IgV_L-t′ 5D3 (black) superimposed on the VL domains of its heterodimeric scFv-t derivative clone 5D3-E6 (red, A) and its homodimeric IgV_L2-t 5D3 derivative (red, B). C-α atoms belonging to amino acids with r.m.s. deviation >1.5 Å are identified, and overall r.m.s. deviations the superimposed models are indicated.

DISCUSSION

Our observations indicate the superior Aβ hydrolyzing activity of V_L domains expressed in the IgV_L2-t and IgV_L-t′ scaffolds compared with scFv-t constructs mimicking physiological antigen-combining sites. The catalytic activity is also strikingly superior to previously reported catalytic IgMs that contain fully natural Aβ-combining sites (18). Several IgV_L2-t and IgV_L-t′ clones with exceptional Aβ hydrolyzing activity were identified from the library. Inclusion of full-length V_H domains in these clones uniformly suppressed the catalytic activity. Certain previous studies have presented evidence for hydrolysis of polypeptides (37) and nucleic acid (38) by isolated Ig light chain subunits. In contrast to catalysis, high affinity antigen binding by noncovalent means depends on cooperative V_L-V_H domain interactions, and heterodimeric V_L-V_H domain complexes invariably bind antigens noncovalently at levels superior to the binding activity of either V domain alone. It may be concluded that the catalytic sites are frequently present in IgV_L domains, but the physiological Ig scaffold does not favor expression of catalytic activity.

Ig light chain homodimers overproduced by cancerous B cells in multiple myeloma patients can bind certain antigens (39, 40). There is no naturally occurring Ig homolog of IgV_L2-t 2E6, a heterodimer of two V_L domains. Homodimeric IgV_L2-t constructs containing the individual V_L domains of IgV_L2-t 2E6 were without appreciable Aβ hydrolyzing activity, suggesting that both V_L domains in the heterodimer are important in maintaining the integrity of the catalytic site. Three Aβ-hydrolyzing clones with the IgV_L-t′ scaffold were also identified. In energy-minimized molecular models of one such clone, the small V_H domain peptide in the C-terminal segment was revealed as a disordered region without the β sheet structure typical of the Ig fold. Moreover, the proximity of the V_H peptide region to the V_L domain was insufficient to anticipate that it contributes to Aβ recognition by the V_L domain catalytic site. Noncovalent intermolecular association of the single domain IgV_L-t′ can be hypothesized to generate homodimeric V_L-V_L-combining sites. However, the stable homodimeric IgV_L2-t derivative containing its two V_L domains was devoid of catalytic activity, supporting attribution of catalysis to the unpaired V_L domain. No natural Igs with an unpaired, functionally active V_L domain are known. In extant organisms, the closest functional homolog of the unpaired catalytic V_L domain are certain jawed fish and camelid Igs containing a single V_H domain, which is thought to bind antigen in its unpaired state (41, 42). The V_L and V_H domains express appreciable sequence identity with each other, and modern Igs have likely evolved by duplication and sequence diversification of a common primordial gene encoding the Ig fold (43). The phylogenic origin of Ig catalysis and deterioration or improvement of the catalytic function over the course of evolution of the immune system remains to be examined.

Electrophilic compounds that react irreversibly with the active site of serine proteases inhibited the Aβ hydrolyzing activity and formed irreversible complexes with the catalytic IgVs. This suggests a nucleophilic catalytic mechanism as deduced for other proteolytic Igs from inhibitor, mutagenesis, and crystallography studies (7, 9, 10). Protein nucleophilic reactivity is generated by intramolecular activation reactions within dyads and triads formed by precisely positioned amino acids, e.g. by hydrogen bonding between the Ser hydroxyl side chain and an imidazole nitrogen. Even small, sub-Å side chain movements can weaken the bonding and induce loss of active site nucleophilic reactivity. Noncovalent antigen binding, on the other hand, is mediated by weak and more numerous interactions at several contact residues in Ig-combining sites (1). Loss of any single contact because of a minor conformational change may weaken noncovalent antigen binding, but an abrupt transition from the binding state to a nonbinding state is less likely. Molecular modeling of the single domain IgV_L-t′ 5D3 suggested the likelihood of minor structural perturbations upon pairing the catalytic V_L domain with another V domain, helping explain the poor catalytic activity of the scFv-t and homodimeric IgV_L2-t versions of the molecule. Compared with the suppressive effect of V_L-V_H and V_L-V_L pairing, the catalytic activity of the single domain V_L was tolerant to inclusion of the C-terminal κ constant domain. This is significant, because it opens the route to inclusion of C-terminal moieties that reduce IgV clearance, e.g. polyethylene glycol or the Ig Fc fragment (44-46). Engineering stable catalyst versions with sufficient longevity in vivo is an important goal for clinical applications. scFv-t constructs have short half-lives in peripheral blood (47), and in view of their small size, the IgVs may also be subject to rapid clearance in vivo. Another route to prolonging the lifetime of IgVs is their recloning within the physiological IgG scaffold (48). The two domain IgV_L2-t 2E6 is a heterodimeric structure that should allow development of a IgG-like structure with a combining site formed by the two V_L domains.

Drugs currently employed to treat AD do not arrest the underlying pathology and progressive cognitive decline. Aβ oligomer accumulation is thought to be a major cause of neuronal death and dysfunction in the AD brain (49, 50). Small amounts of peripherally infused Aβ binding monoclonal IgGs traverse the blood-brain barrier, and IgG-facilitated Aβ clearance has emerged as a novel therapeutic strategy with the potential to halt cognitive decline in AD patients (51). Reversibly binding IgGs can at best bind two antigen molecules. Large quantities of stoichiometrically binding monoclonal IgGs are usually required for immunotherapy. Catalytic Igs hold the potential of clearing Aβ efficiently by virtue of the specific Aβ degrading activity. For example, from its k_cat value in Table 1, a single IgV_L-t′ 5D3 molecule is predicted to digest 4320 Aβ molecules in 3 days at excess Aβ concentration. The enzyme neprilysin has received attention as a potential Aβ-clearing AD drug (52). The k_cat of neprilysin for Aβ is comparable with the IgVs reported here (53). Neprilysin, however, also hydrolyzes irrelevant polypeptides (54), whereas the IgVs did not degrade non-Aβ polypeptides detectably. Aβ degradation by an IgM at a substantially lower rate than the IgVs was previously reported to inhibit Aβ aggregation and Aβ-induced neurotoxicity (18). The IgVs hydrolyze the His¹⁴-Gln¹⁵ bond and, at lower levels, other peptide bonds located in the central Aβ region. The aggregability of various synthetic Aβ fragments is generally weaker than full-length Aβ (55-57), but the precise functional effects of IgV-catalyzed Aβ hydrolysis remain to be examined. Concerns have been raised that Aβ binding IgGs can induce inflammation (58) and vascular microhemorrhages (59, 60) caused, respectively, by immune complex-stimulated release of microglial inflammatory mediators and IgG-stimulated Aβ deposition in cerebral blood vessels. The IgVs reported here do not form immune complexes detectably. They degrade Aβ permanently, minimizing the risks of inflammatory mediator release and Aβ re-deposition in the vascular wall. The recently reported phase II clinical trial of a reversibly binding anti-Aβ monoclonal IgG in patients with mild-to-moderate AD patients highlights the importance of searching for safer and more effective immunotherapeutic reagents (61). A dose-limiting incidence of vasogenic edema in magnetic resonance images was evident in this trial. At lower doses of the IgG, cognitive performance tended to improve, but the effect did not reach statistical significance in the intent-to-treat population. However, upon exclusion of patients homozygous for the apolipoprotein E4 allele, post-hoc analysis suggested significantly improved cognitive functions in the remaining patient subgroup. The apolipoprotein E4 allele is known to predispose AD patients to increased amyloid accumulation (62).

In summary, the specific and efficient Aβ degrading activity of the IgVs supports evaluation of their efficacy and safety in attaining Aβ clearance. The potential medical utility of catalytic Igs to microbial antigens and cancer-associated antigens has been discussed previously (28), but the low catalytic rate constants of physiological Igs have been a barrier to their clinical application. Our observations suggest that enhanced catalysis can be achieved by placing the V_L domains within nonphysiological two domain and single domain scaffolds. If this finding proves generally applicable, development of efficient catalysts specific for other clinically important antigens should be feasible.