Molecular Cloning of a Proteolytic Antibody Light Chain*

The cDNA for an antibody light chain raised by immu- nization against vasoactive intestinal peptide (VIP) was cloned in a bacterial expression vector, and the recom- binant light chain was purified to electrophoretic homo-geneity. The light chain catalyzed the hydrolysis of VIP efficiently owing to its comparatively high affinity for the substrate. In control experiments, the catalytic ac- tivity was preserved at a constant level after further chromatography of the light chain on anion-exchange and gel-filtration fast protein liquid chromatography columns, and it was removed by immunoadsorption with immobilized anti-mouse light chain antibody. The amide bond linking methylcoumarinamide (MCA) and arginine in a tripeptide unrelated in sequence to VIP was cleaved by the light chain with lower affinity and kinetic efficiency (kJK,). Hydrolysis of the peptidyl- MCA conjugate was inhibited competitively by the alternate substrate, VIP. The Ki and K , values for VIP were in the same range, indicating that peptide-MCA and VIP hydrolysis occurs at a common catalytic site in the light chain. Molecular modeling suggested the presence of a serine protease-like site in the light chain. This was supported by inhibition of the hydrolytic activity by serine protease

The cDNA for an antibody light chain raised by immunization against vasoactive intestinal peptide (VIP) was cloned in a bacterial expression vector, and the recombinant light chain was purified to electrophoretic homogeneity. The light chain catalyzed the hydrolysis of VIP efficiently owing to its comparatively high affinity for the substrate. In control experiments, the catalytic activity was preserved at a constant level after further chromatography of the light chain on anion-exchange and gel-filtration fast protein liquid chromatography columns, and it was removed by immunoadsorption with immobilized anti-mouse light chain antibody. The amide bond linking methylcoumarinamide (MCA) and arginine in a tripeptide unrelated in sequence to VIP was cleaved by the light chain with lower affinity and kinetic efficiency (kJK,). Hydrolysis of the peptidyl-MCA conjugate was inhibited competitively by the alternate substrate, VIP. The Ki and K , values for VIP were in the same range, indicating that peptide-MCA and VIP hydrolysis occurs at a common catalytic site in the light chain. Molecular modeling suggested the presence of a serine protease-like site in the light chain. This was supported by inhibition of the hydrolytic activity by serine protease inhibitors, but not by inhibitors of other classes of proteases. These observations suggest a poorly discriminatory catalytic site, with specificity for VIP arising chiefly by means of the antigen recognition function of the light chain combining site.
Efficient catalysis by autoantibodies to vasoactive intestinal polypeptide (VIP)' (1) and DNA (2) has been reported, but the genesis of these activities and the identity of antigens inciting formation of the antibody catalysts are not clear. Since transition state lifetimes are very short, an antitransition state specificity is an unlikely explanation for the catalytic activity of autoantibodies. In analogy with the evolution of enzymes, sequence diversification occurring in antibody variable region genes (3) over the course of the immune response could result i n the formation of catalytic active sites. The key test of this HL44126 and AI31268 and by a contract with IGEN Inc. The costs of * This work was supported by National Institutes of Health Grants publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 'U.S.C. Section 1734 solely to indicate this fact.
$ Performed this work in partial fulfillment of the requirements for a Ph.D. degree.
( hypothesis is that immunization with natural antigens, as opposed to commonly used transition state mimics (41, should lead to development of catalytic antibodies. A monoclonal antibody raised to VIP displays a peptidolytic activity, but this activity is expressed only in very dilute solutions (see Ref. 5 for discussion). Consequently, this antibody is unsuitable for detailed study of catalysis and demonstration of turnover, a defining feature of a true catalyst. Here, we report efficient catalysis by the recombinant light chain subunit of the anti-VIP antibody. The activity shows preference for VIP at the substrate binding step, but turnover rates for VIP and a nonhomologous protease substrate are nearly equivalent.
The presence of a serine protease-like catalytic triad in the light chain was suggested by molecular modeling and supported by inhibition of the activity by serine protease inhibitors. These observations indicate a poorly discriminatory catalytic site in the light chain, with specificity for VIP arising chiefly by means of the high affinity antigen binding function.

MATERIALS AND METHODS
cDNAAmpZification and Cloning-Ahybridoma cell line (clone c23.5) secreting a monoclonal antibody was developed from a mouse hyperimmunized with a VIP-keyhole limpet hemocyanin conjugate (5). cDNA preparation was by the reverse transcriptase-polymerase chain reaction method from RNA isolated from 3 x lo7 c23.5 hybridoma cells (6) using forward (GAGTCATTCTGCGGCCGCCTCATTCCT-GTTGAAGCTCTTGAC) and back (GTCCTCGCAACTGCGGCCAGC-CGGCCATGGCCGAYGTNGTNATGACNCAGAC) primers corresponding to constant region residues 206213 and framework 1 residues 1-8 (Asp-Val-Val-Met-Thr-Gln-Thr-Pro, determined by N-terminal sequencing of the reduced and alkylated light chain purified from antibody c23.5 (7)). The cDNA (801 base pairs) was inserted into pCANTAB5his, (8,9) via SfiI andNotI restriction sites (underlined). Sequencing was by the dideoxy chain termination method (30) using vector-specific primers located in the vector on the 5'-side (CAGGAAACAGCTATGAC) and 3'-side (GAATTTTCTGTATATGGG) of the insert. Transformation of Escherichia coli (HB2151) was by electroporation; recombinants were grown in ampicillin; expression was induced with 1 m M isopropyl-l-thio-6-D-galactopyranoside (2.5 h, 30 "C); and a periplasmic extract was prepared in 10 m M sodium phosphate, 1 m M EDTA, pH 7. stainable light chain were dialyzed and rechromatographed on chelating Sepharose (bed volume of 10 ml). Light chain expression levels were 1-2 mglliter of bacterial culture, and the yield of the pure protein was 350 pglliter. Isoelectric focusing was on Phast IEF gels, pH 3-9, and SDS-polyacrylamide gel electrophoresis was on 8-25'; Phast acrylamide gels. SDS gels were diffusion-blotted onto nitrocellulose membranes for 30 min and stained with anti-c-mvr antibody. The Nterminal amino acid sequence of the recombinant light chain (2.5 pg) adsorbed on a polyvinylidene difluoride memhrane (ProSpin cartridge, Applied Biosystrms Inc.) was determined by automated Edman degradation as described (7). Anion-exchange chromatography was on a Mono-Q column (Pharmacia Biotech Inc.; 0-1 31 NaCl(30 min) in 50 mhl "is-HCI, pH 7.4.0.025% Tween 201, and gel-filtration chromatomaphy was on a Superose 12 column (Pharmacia Biotech Inc.; 0.5 ml/min) (10). Light chain recovered from the gel filtration column (retention time of 28 min) and from the Mono-Q column (elution position at 0.43 SI NaCl) was dialyzed against assay diluent prior to determination of hydrolytic activity. Immunoadsorption ofthe light chain (7.5 pg) was for 17 h ( 4 "CI in 50 mM Tris-HCI. 100 mhl glycine. pH 7.7, 0.025'7 Tween 20, 0.02?4 sodium azide with the following antibodies immobilized on Sepharose ( 3 ml ofsettled gel): mouse anti-c-my (clone 9E10). rat anti-mouse K-chain (Zymed Laboratories, Inc.), and rat anti-mouse IgGi (Zymed 1,aboratories, Inc.). Preparation of immobilized anti-c-mvc and the immunoadsorption method were as descrihed (11 ). The light chains from reduced and alkylated antibody c23.5 and a control nonimmune antihody (from myeloma IgGCr?,,~; UPC10, Sigma) wrrr purified in 6 Y guanidine hydrochloride and renatured by dialysis as descrihed (7). to their catalytic counterparts and allowed to relax with no constraints on the system. N-terminal Asp' was not well aligned with Asp3' of subtilisin, but when all 3 light chain residues were template-forced to the catalytic triad of the enzyme followed by relaxation of the structure, good superposition was evident. The manipulations did not unduly perturb the light chain backbone conformation, and the conformations of the canonical CDRs (L1 and L3) were retained (16).

RESULTS AND DISCUSSION
Catalytic Activity of the Light Chain-Polymerase chain reaction-amplified light chain cDNA was inserted into a bacterial expression vector, and the recombinant protein was purified by means of its affinity for metals (Fig. 1). The purified light chain preparations contained a single protein migrating with an apparent PI between 5.2 and 5.8 on isoelectric focusing gels. The molecular mass of the light chain was 27 kDa as assessed by SDS-polyacrylamide gel electrophoresis, and the protein was immunoblotted by anti-light chain and anti-c-myc antibodies (the recombinant protein contains a 10-residue cmyc tag). N-terminal sequencing of the purified protein yielded a single peptide sequence (Asp-Val-Val-Met-Thr) corresponding to the previously reported sequence of the light chain purified from reduced and alkylated antibody c23.5 (7). A linear increase in ['251-Tyr'o]VIP hydrolysis was observed (5.3-49.4% available peptide) at increasing recombinant light chain concentrations (1.5-15 m). The reaction was saturable at increasing substrate concentration, and the initial rates were consistent with Michaelis-Menten kinetics ( Table I). The K, value of the light chain (0.2 w) was 1900-fold lower than that of trypsin, a nonspecific protease, determined under identical conditions. The kinetic efficiency (kc,JK,) of light chain-catalyzed VIP hydrolysis is only 53-fold lower than that of trypsin. Treatment of 30 w nonradioactive VIP with 2 pl light chain for 6 h resulted in 32% peptide hydrolysis estimated using a n HPLC assay: corresponding to 4.8 catalyst turnovers. This value is close to the turnover determined by radiometric analyses.
The specificity of catalysis was examined by screening for hydrolysis of synthetic peptide-MCA conjugates, a reaction accompanied by increased fluorescence due to accumulation of 7-amino-4-methylcoumarin (17). Very slow hydrolysis of some of the single amino acid-MCA conjugates examined (Phe-MCA, Met-MCA, Lys-MCA, and Arg-MCA) was observed ( Table 11). Hydrolysis of the dipeptide-MCA and tripeptide-MCA substrates was more rapid. The hydrolysis of Pro-Phe-Arg-MCA, a substrate bearing no sequence similarity to VIP, was examined in detail. The light chain recognized this peptide with low af-&versed-phase HPLC was as described (10). Peptide hydrolysis was estimated based on the decrease in area of the intact VIP peak eluted at 46% acetonitrile. Product identities are to be reported elsewhere.
Hydrolysis of peptide-MCA substrates by the anti-VIP light chain  The N terminus was derivatized with a benzoyl group. The N terminus was derivatized with a t-butyloxycarbonyl group.
finity (Table I; K, 57-fold greater than for VIP). The turnover rates (kcat) for Pro-Phe-Arg-MCA and VIP were similar, suggesting that the hydrolytic site does not discriminate between the specific antigen and the nonhomologous substrate at the transition state stabilization step of the reactions. This conclusion assumes a common site t o be responsible for hydrolysis of VIP and Pro-Phe-Arg-MCA. The validity of this assumption was proved by demonstration of competitive inhibition of peptidyl-MCA hydrolysis by VIP, the tight-binding alternate substrate (Fig. 2). The apparent K, (0.33 p l ) and K, (0.2 PM) values for VIP were in the same range, suggesting that inhibition of Pro-Phe-Arg-MCA hydrolysis is due to the occupancy of the light chain-binding site by VIP. These data indicate a poorly discriminatory substrate-hydrolyzing site located close to or within the VIP-binding site, In control experiments, the light chain immunoadsorbed with immobilized antibodies to c-myc and antibodies to mouse K-chain displayed near-complete loss of VIP hydrolyzing activity (11 and 15% of the activity observed by immunoadsorption with irrelevant anti-IgG, antibody, respectively). The specific activities of the recombinant protein after sequential chromatography on a metal affinity column, a Mono-$ anion-exchange column (elution position at 0.43 M NaCl), and a Superose 12 gel-filtration column (retention time of 28 min) were 22.8,24.8, and 25.87 cpm x 103/h/pg of light chain with ['251-Tyr'o]VIP as substrate and 4.8, 4.5, and 4.3 AFl'hfpg of light chain with Pro-Phe-Arg-MCA as substrate, respectively, where AF denotes the increase in fluorescence in arbitrary units relative to substrate incubated without catalyst. The constancy of these specific activity values suggests catalyst homogeneity. Metal aninity-purified fractions of periplasmic extracts of bacteria infected with the control cloning vector without the light chain insert did not display catalytic activity. These data, taken together with the functional specificity of the catalyst preparations for VIP, show that the hydrolytic activities are due to the recombinant light chain, The light chain was also prepared from reduced and alkylated antibody c23.5 in denaturing solvent and renatured by dialysis against assay diluent (7). The levels of VIP hydrolyzing activity of several preparations were observed to vary as a function of the concentration at which the light chain was held during the renaturation step (181, suggesting a negative influence of protein aggregation on assumption of a catalytically active conformation. This type of activity variability was not observed in three recombinant light chain preparations, each of which was purified under nondenaturing conditions. Kinetic parameters for the reduced and alkylated light chain renatured at 0.16 PM were as follows: K, = 1. x min-l. A nonimmune light chain (from myeloma I~G,,K; UPC10, Sigma) purified and renatured under identical conditions was devoid of catalytic activity.
Molecular Model of the Light Chain-Inspection of the nucleotide sequence and the deduced amino acid sequence of the V, region (Fig. 3) showed it to be a member of K-chain family I1 (19). AV,.V, dimer model of antibody c23.5 was produced, the V, domain of which is shown in Fig. 4. This model is relevant to the properties of the light chain since dilute solutions of monoclonal c23.5 IgG preparations express a weak peptidolytic activity (5); the recombinant single chain V,-V, construct (F,) of this antibody also expresses catalytic activiw; and the structure of V, domains is often largely independent of the mode of association with V, domains, even where the intersubunit contacts are not conserved between the light and heavy domains (20). The model revealed two potential catalytic sites with a composition similar to that found in the catalytic triad of many serine proteases (an Asp, a Ser, and a His located in a hydrophobic pocket). The first site (Fig. 4B) contains the N-terminal aspartic acid residue of the light chain (Asp'), a serine from S. Paul and Q.-S. Gao, unpublished data. CDRl (SerZS), and a histidine from CDR3 (Hisg8). The C a distance geometries and side chain orientations of these residues were similar to those of their counterparts from the catalytic triad of subtilisin (15). The second site involved residues from CDRl ( S e P , His31, and Asp33). Template forcing of this site to the catalytic geometry of subtilisin disrupted the canonical structure of CDR1, rendering it a less likely location for catalysis. The likelihood of a serine protease-like mechanism is supported by observations of reduced light chain-catalyzed VIP hydrolysis in the presence of the serine protease inhibitors diisopropyl fluorophosphate (200 J~M) and aprotinin (0.75 p~) (by 84 and 89%, respectively; assayed as described in the legend to Table I). Inhibitors of other classes of proteases, including iodoacetamide (2 n"), EDTA (2 mM), and pepstatin A (500 p~) , were essentially without effect on the activity. These observations suggest that where efficient catalysis of energetically difficult reactions by antibodies is seen, known enzymatic mechanisms that have evolved over many millions of years will be reproduced.
Conclusion-Polyclonal autoantibody light chains display catalytic activity (21). Some light chains secreted by B-lymphocyte tumors express sequence similarities t o serine proteases (22). Light chains contribute a significant proportion of the antibody-antigen contact surface (231, and they can independently bind antigens (7,24). The light chain described here was raised by immunization with the substrate (VIP), indicating that synthesis of catalytic sites is an intrinsic component of the immune response to polypeptide antigens. The distinctive property of the light chain not shared by proteolytic enzymes is an ability to strongly recognize the ground state of VIP. The lower K,,, of the light chain for VIP compared with the nonhomologous substrate is consistent with the clonal selection theory, i.e. the light chain has been selected during the immune response to VIP because it recognizes the inciting antigen with comparatively high affinity. Polypeptide binding by antibodies can involve contacts at >15 residues in each molecule (23), but most of the binding energy appears to derive from a subset of these contacts (25,26). In the case of a catalyst, such a multiplicity of contacts makes it possible that distinct subsites play important roles in the binding and hydrolyzing functions. This is supported by hydrolysis of Pro-Phe-Arg-MCA by the anti-VIP light chain, a substrate unrelated in sequence to VIP. The turnover rate for VIP is no higher than for the nonhomologous substrate, pointing to the possibility of an independent evolution of the hydrolytic and antigen binding functions. Even so, it is clear that the hydrolytic function is preserved in the affinity-matured light chain. Therefore, the immune system should prove a rich source of catalysts that combine a hydrolytic function with strong substrate binding affinity.