Ontogeny of Proteolytic Immunity IgM SERINE PROTEASES*

We report the chemical activity of immunoglobulin (cid:1) and (cid:2) / (cid:3) subunits expressed on the surface of B cells and in secreted IgM antibodies (Abs) found in the preimmune repertoire. Most of the nucleophilic reactivity of B cells measured by formation of covalent adducts of a hapten amidino phosphonate diester was attributed to (cid:1) and (cid:2) / (cid:3) subunits of the B cell receptor. Secreted IgM Abs displayed superior nucleophilic reactivity than IgG Abs. IgM Abs catalyzed the cleavage of model peptide substrates at rates up to 344-fold greater than IgG Abs. Catalytic activities were observed in polyclonal IgM Abs from immunologically naı¨ve mice and humans without immunological disease, as well as monoclonal IgM Abs to unrelated antigens. Comparison of several IgM Abs indicated divergent activity levels and substrate prefer-ences, with the common requirement of a basic residue flanking the cleavage site. Fab fragments of a monoclonal IgM Ab expressed catalytic activity, confirming the V domain location of the catalytic site. The catalytic reaction was inhibited by the covalently reactive hapten probe and diisopropylfluorophosphate, suggesting a serine protease-like mechanism. These observations indicate the existence

adaptive sequence diversification processes or genuine enzymes with important functional roles. The major reason is that the turnover (k cat ) of antigen-specific IgG Abs is low. Some catalytic Abs express catalytic efficiencies (k cat /K m ) comparable to conventional enzymes, but this is due to high affinity recognition of the antigen ground state (reviewed in Ref. 8).
Certain enzymes cleave peptide bonds by a mechanism involving the formation of a transient covalent intermediate of the substrate and a nucleophilic residue present in the active site. The nucleophiles are generated by intramolecular activation mechanisms, e.g. the activation of Ser/Thr side chain hydroxyl groups by hydrogen bonding to His residues, and can be detected by covalent binding to electrophilic phosphonate diesters (9,10). Using these compounds as covalently reactive analogs of antigens (CRAs), we observed that IgG Abs express nucleophilic reactivities comparable to trypsin (11). Despite their nucleophilic competence, IgG Abs display low efficiency proteolysis, presumably due to deficiencies in steps occurring after formation of the acyl-Ab intermediate, viz., water attack on the intermediate and product release. Occupancy of the B cell receptor (BCR, surface Ig complexed to ␣ and ␤ subunits along with other signal transducing proteins) by the antigen drives B cell clonal selection. Proteolysis by the BCR is compatible with clonal selection, therefore, only to the extent that the release of antigen fragments is slower than the rate of antigeninduced transmembrane signaling necessary for induction of cell division. Immunization with haptens mimicking the charge characteristics of the transition state (12) has been suggested as a way to surmount the barrier to adaptive improvement of catalytic rate constants. Catalysis by designer IgG Abs obtained by these means, however, also proceeds only slowly.
In mice and humans, the initial Ab repertoire consists of ϳ100 heritable VL and VH genes. Adaptive maturational processes expand the repertoire by several orders of magnitude. The initial BCR complex on the pre-B cell surface contains V-(D)-J rearranged Ig chains as a complex with surrogate L chains (reviewed in Ref. 13). Precise assignment of the B cell differentiation stage at which cell division becomes antigen-dependent is somewhat ambiguous, but it is generally believed that non-covalent antigen binding to the pre-BCR is not required for initial cell growth. / chains replace the surrogate L chain at the later stages of antigen-driven B cell differentiation, which is accompanied by diversification via somatic hypermutation processes and continued gene rearrangements (14,15). V-(D)-J gene rearrangements allow development of specificity for individual antigens by IgM (16) but antigen binding affinities tend to be low compared with IgG Abs. Somatic mutations accumulating in the V domains following isotype switching to IgG promote high affinity antigen recognition. In some anatomic locations, IgM Abs can be extensively mutated and can display high affinity antigen binding (17). Loss of a membrane anchoring peptide at the C terminus of the H chain results in production of secreted IgM and IgG Abs.
Very little information is available about the developmental aspects of Ab catalysis. Here, we report the nucleophilic reactivity of secreted IgM and the Ig subunits expressed on the surface of B cells. Cell surface and / chains were the major sites of covalent reaction of a hapten CRA with B cells, and the magnitude of nucleophilic and proteolytic activities of secreted IgM Abs was consistently superior to IgG Abs.

EXPERIMENTAL PROCEDURES
Splenocyte-CRA Binding-Synthesis of compounds I-IV and confirmation of their chemical identity have been published (11,18). Compounds I, III, and IV are diphenyl phosphonate esters reactive with nucleophilic sites (9,10,18). Biotin incorporated in these compounds allowed the visualization of Ab-CRA adducts. Diisopropyl fluorophosphate (DFP) was from Sigma. BALB/c mice (5-6 weeks, female, Jackson Laboratories) were euthanized by cervical dislocation and splenocytes were prepared in RPMI 1640 (Invitrogen, Life Technologies, Inc.) by teasing apart the spleen and removing undissociated tissue (unit gravity sedimentation). Erythrocytes were lysed in hypotonic ammonium chloride (5 min; ACK Lysis Buffer, Cambrex, Walkersville, MD) and the cells washed twice with 10 mM sodium phosphate, pH 7.5, 137 mM NaCl, 2.7 mM KCl (PBS). B cells were isolated from splenocytes using a B cell negative selection isolation kit (Miltenyi, Auburn, CA) according to manufacturer's instructions and verified to be Ͼ95% CD19ϩ by flow cytometry as described below. Viability was determined using 0.05% Trypan Blue (90 -95%). The cells (2-5 ϫ 10 6 cells) were incubated with hapten CRA I or compound II (37°C; final Me 2 SO concentration 1%) in 0.5 ml of PBS, washed thrice and treated with 100 l of anti-CD16/32 Ab (10 g/ml l; BD Pharmingen, San Diego, CA; 5 min, 4°C) to block Ab binding to Fc receptors. Staining was with FITC-conjugated streptavidin (1 g/ml or as stated; Molecular Probes, Eugene, OR) and PEconjugated rat monoclonal Ab to CD19 (10 g/ml; Caltag, Burlingame, CA) in 100 l for 20 min at 4°C. Following further washing with PBS (2ϫ), the cells were fixed with 2% paraformaldehyde (1 h, 4°C), washed once and resuspended in PBS. In control incubations, an equivalent concentration of PE-conjugated isotype-matched rat Ab to an irrelevant antigen (BD Pharmingen) replaced the anti-CD19 Ab. Deconvolution microscopy was performed employing an Olympus IX-70 inverted microscope and Applied Precision Delta work station (SoftWoRx TM software; Ref. 19). Stained cells were subjected to multiple acquisitions at a thickness of 0.25 m, and the images were stacked. The images were subjected to deconvolution (5 iterations) for each probe (FITC; ex 488 nm, em 525 nm; DAPI; ex 350 nm, em 470 nm; phycoerythrin; ex 565 nm, em 578 nm). Flow cytometry was performed in the Baylor Medical College Core Facility (EPICS XL-MCLs Beckman-Coulter flow cytometer, EXPO32 software). Instrument calibration to minimize cross-detection of PE and FITC was done using cells stained individually with these fluorochromes. Forward and side scatter measurements allowed exclusion of dead cells from the gated cell population. CRA-stainable cells were identified as the population showing staining above the level observed for compound II staining. CRA-stainable CD19ϩ cells were estimated by subtraction of background observed using the isotypematched Ab. Cell extraction was by treatment with the detergent CHAPS (12 mM, 2 h at 4°C). The extract was centrifuged (10,000 ϫ g, 30 min), the supernatant diluted with PBS to 1 mM CHAPS and then subjected to affinity chromatography using goat polyclonal Abs (IgG) to mouse , ␥, ␦, , and chains (Caltag) immobilized on protein G-Sepharose columns (100 l of settled gel; 0.6 ϫ 5 cm columns; Amersham Biosciences). For this purpose, the Abs (50 g) were mixed with the protein G gel in a column (15 min, 4°C) in PBS containing 1 mM CHAPS (PBS-CHAPS), the gel allowed to settle, the unbound fraction collected and the columns washed with PBS-CHAPS. The cell extract (1.4 ml; diluted to 1 mM CHAPS; from 3 ϫ 10 6 cells) was passed through the column, the column washed with PBS-CHAPS (9 volumes) and bound proteins were eluted with 100 mM glycine-HCl, pH 2.7 (8 column volumes) and subjected to reducing SDS-polyacrylamide gel electrophoresis (4 -20%, Bio-Rad). Protein-CRA adducts were visualized by staining nitrocellulose electroblots of the gels with streptavidin-peroxidase as in (11). For immunoblotting, the blots were stained with goat polyclonal Abs (IgG) to mouse , ␥, ␦, , and chains followed by peroxidase-conjugated rabbit anti-goat IgG (Fc specific, 1:1000; Pierce) as in Ref. 11. Nominal mass values were computed by comparison with standard proteins (14 -220 kDa; Amersham Biosciences).
Secreted Ab-CRA Binding-Human serum Abs were from subjects without evidence of infection or immunological disease (2 females, 3 males; ages 23-45 years). Murine serum Abs were from BALB/c mice (purchased from Harlan, Indianapolis, IN; pooled from 150 mice; 8 -12 weeks). Murine monoclonal IgM Abs used here are directed against major histocompatibility antigens (clones corresponding to catalog 8702, 8704, 9008, 9010, 9020; cell-free ascites; Cedarlane, Ontario, Canada). Monoclonal IgM Yvo is from a patient with Waldenstrom's macroglobulinemia (20). All monoclonal IgM Abs contain chains. The 4 murine monoclonal IgG Abs used here were: clone c23.4 (anti-VIP; Ref. 6), clone c39.1 (anti-glucagon), 2 ATCC clones HP6045 (anti-Fab 2 , ␥), and ATCC clone HP6054 (anti-Ig chain). All monoclonal IgG Abs contain ␥2a heavy chains and light chains. Serum or ascites fluid (1 ml) was mixed for 1 h with 1 ml of Sepharose 4B conjugated rat anti-mouse IgM Abs (settled gel; Zymed Laboratories Inc., San Francisco, CA) or agarose-conjugated goat anti-human IgM Abs (Sigma) with IgM binding capacities 0.8 and 3 mg, respectively, in 50 mM Tris-HCl, pH 7.5, 0.1 mM CHAPS (buffer A). The unbound fraction was recovered and the gel washed with 20-buffer A volumes taking care that protein in the effluent had returned to undetectable levels prior to elution (A 280 Ͻ 0.001). Elution was with 100 mM glycine pH 2.7 (0.5 ml/fraction into 25 l of 1 M Tris-HCl, pH 9.0). Further purification was on a Superose-6 FPLC gel filtration column (1 ϫ 30 cm; 0.25 ml/min; Amersham Biosciences) in two different solvents: 50 mM Tris-HCl, pH 7.7, 0.1 M glycine, 0.15 M NaCl, 0.025% Tween-20 (buffer B), or 6 M guanidine hydrochloride in buffer B adjusted to pH 6.5 with HCl (buffer C). Prior to column fractionation, the affinity purified IgM was dialyzed against buffer C. Column calibration was with thyroglobulin (660 kDa), IgG (150 kDa), and albumin (67 kDa). IgM with M r 900 kDa eluted close to the void volume of the column. IgM was renatured following buffer C chromatography by dialysis against buffer B (21). IgM Yvo, a cryoglobulin, was purified from serum by repetitive warming (37°C) and cooling (4°C; 3 cycles; Ref. 20) followed by affinity chromatography on the antihuman IgM column. IgG was purified on protein G-Sepharose columns (21) using as starting material the unbound fraction from the anti-IgM columns or cell-free ascites. FPLC gel filtration of IgG was as described for IgM except that a Superose 12 column was employed. Fab fragments were prepared by digesting IgM (300 l, 1 mg/ml) with agarose-conjugated pepsin (0.6 ml of gel, 30 min, 37°C) in 100 mM sodium acetate, pH 4.5, 150 mM NaCl, 0.05% NaN 3 , 0.1 mM CHAPS) as recommended by the manufacturer (Pierce). The unbound fraction was dialyzed against buffer B, purified by FPLC gel filtration on a Superose 12 column and dialyzed against 50 mM Tris-HCl, pH 7.7, 0.1 M glycine, 0.1 mM CHAPS. Total protein was determined by the bicinchoninic acid method (Pierce). Immunoblotting of SDS-gels contanining murine Abs was as in the preceding section. Human Ab gels were immunoblotted using peroxidase-conjugated goat anti-human , anti-human , and anti-human Abs (Sigma).
Purified Abs were treated with the biotinylated CRAs in 50 mM Tris, HCl, 100 mM glycine, 0.1 mM CHAPS, pH 7.7 at 37°C. Formation of Ab-CRA adducts was determined by SDS-electrophoresis as in the preceding section. Band intensities are expressed in arbitrary area units (AAU) determined by densitometry (11). Initial velocities were computed as the slopes of progress curves plotted as a function of time (initial 60 min).
Proteolysis Assays-Cleavage of the amide bond linking aminomethylcoumarin to the C-terminal amino acid in peptide-AMC substrates (Peptide International, Louisville, KY or Bachem, King of Prussia, PA) was measured in 50 mM Tris HCl, pH 7.7, 0.1 M glycine, 0.025% Tween-20 at 37°C in 96-well plates by fluorimetry ( ex 360 nm, em 470 nm; Varian Cary Eclipse) (21). Authentic aminomethylcoumarin was used to construct a standard curve. Kinetic parameters were obtained by fitting rate data obtained at increasing concentrations of peptide-AMC substrates to the Michaelis-Menten-Henri Equation 1.

Irreversible CRA-B Cell Binding-Hapten
CRAs such as compound I (Fig. 1) react irreversibly with nucleophilic sites in conventional serine proteases and Abs (9 -11,18). To evaluate the nucleophilic reactivity expressed on the surface of B cells in the preimmune repertoire (viz., the repertoire developed spontaneously without purposeful immunological challenge), viable splenocytes from BALB/c mice were treated with hapten CRA I. The control compound II is identical in structure to hapten CRA I, except that the phosphonate group is not esterified, which results in loss of covalent reactivity with nucleophilic residues (11,18). Treatment with hapten CRA I resulted in staining of most of cells at levels greater than compound II, with a minority of the cells displaying intense staining (11 Ϯ 2, n ϭ 3 experiments; determined by counting 400 lymphocytes using a UV microscope). All of the CRA I-stained cells displayed lymphocytic morphology, with no evident staining of monocytes or the occasional basophil. No loss of viability of the cells was evident following incubation with CRA I or compound II, as determined by trypan blue exclusion. Flow cytometry confirmed the microscopy results. Seventy-nine percent of the CRA I-treated cells displayed fluorescence intensities exceeding the compound II-treated cells, including a minority subpopulation with very high fluorescence intensity (14%; subpopulation 2 in Fig 2A). In three repeat experiments, the proportion of CRA I-stained cells that were positive for the B cell marker CD19 was 82 Ϯ 4% (Fig. 2B). Deconvolution microscopy indicated that the fluorescence pattern due to hapten CRA I binding was nearly coincident with the anti-CD19 Ab fluorescence pattern (Fig. 2, C-E). Most of the CRA fluorescence was restricted to the surfaces of the B cells (Fig. 2F).
To identify the nucleophilic molecules on the cell surface, purified B cells were labeled with CRA I, detergent extracts of the cells were boiled and then analyzed by SDS-electrophoresis. Only limited CRA-containing proteins were evident (Fig.  3A). As expected, silver staining revealed the presence of heterogeneous species, reflecting the complex protein constitution of the cells. The mass of the predominant CRA adduct band was 70 kDa, and this band was stainable by anti-chain Ab (Fig.  3B). Smaller amounts of CRA-containing bands were evident at 25, 40, 50, 55-60, 90 -135, and 140 kDa. The bands at 55-60 kDa and 140 kDa were stainable by the anti-Ab, and the bands at 25 and 50 kDa were stainable with anti-/ Ab. The anomalous and / bands at mass range different from the full-length monomer proteins presumably represent unreduced oligomers, breakdown products and truncated B cell Ig products, as also observed in previous studies of secreted Abs and B cell extracts (22)(23)(24). The minor bands at 40 and 90 -135 kDa that were not stainable with Abs to , ␥, /␥ (Fig. 3B) and ␦ chains (not shown) presumably represent non-Ig proteins. No CRA-containing adduct corresponding to Ig␥ chains were detected. Immunoblotting of the cell extracts identified a band at 50 kDa stainable with anti-␥ Ab, but the band was visible only in highly overexposed gels, suggesting that only small amounts of ␥ chains were present in the extract.
Confirmation that the CRA I adducts contain Ig subunits was by affinity chromatography on columns of immobilized Abs to , ␦, ␥, and / chains followed by SDS-electrophoresis (Fig.  3C). CRA-containing and / bands were evident in eluates from the anti-and anti-/ columns. Recovery of CRA-containing chains in the eluate from the anti-/ column can be explained by the presence of disulfide bonded light and heavy chain complexes on the cell surface. No CRA-containing bands were evident in eluates from the anti-␥ and anti-␦ columns (not shown), but this can not be interpreted to reflect deficient ␥/␦ chain nucleophilic reactivity, as these proteins are expressed only at low levels in B cells from immunologically naïve mice.
To determine the proportion of overall cellular CRA staining attributable to complexation with Ig subunits, the B cell extract was fractionated on a single affinity column composed of immobilized Abs to and / chains. Eighty percent of the total CRA content of the cells was adsorbed by the column (not shown), determined by densitometry of the biotin-containing bands in the unbound fraction and the extract loaded on the column. Taken together, these observations indicate that most of the CRA staining of intact B cells is attributable to irreversible binding to surface Ig, with the chain accounting for most of the covalent reactivity.
Nucleophilic Reactivity of Secreted IgM-The initial velocity for formation of hapten CRA I adducts by IgM purified from the pooled serum of immunologically naïve BALB/c mice was 40fold greater than by IgG (Fig 4A; values are sums of velocities for the reactions occurring at the two Ab subunits expressed per unit concentration of intact Abs). The velocity difference is 8-fold when expressed per unit combining site concentration 3 (10 and 2 combining sites, respectively, in IgM and IgG). Three CRA-containing bands were observed in reducing SDS-gels of the IgM reaction mixtures at 70, 50, and 25 kDa (Fig. 4B). The 70 and 25 kDa bands were stainable with anti-and anti-/ Abs, respectively. The 50-kDa band was stainable with anti-Ab and presumably represents a breakdown product. Two CRA-containing bands corresponding to ␥ and / chains were observed in reducing gels of the IgG reaction mixtures. Similar results were obtained with a panel of 6 randomly selected monoclonal IgM Abs (5 murine and 1 human) and 4 monoclonal IgG Abs (all murine). The monoclonal IgM Abs uniformly displayed superior rates of irreversible CRA I binding compared with the IgG Abs (Fig 4C; mean Ϯ S.E.: 62.6 Ϯ 24.4 ϫ 10 4 and 1.9 Ϯ 0.4 ϫ 10 4 AAU/M Ab/hour, respectively; p Ͻ 0.01, Mann-Whitney U test, 2-tailed). Consistent with the polyclonal Ab experiments, the chain accounted for most of the covalent binding in the polyclonal and monoclonal IgMs, but smaller levels of binding at the / chain subunit were also observed for every Ab preparation (for clarity, chain and the corresponding / chain data points from individual IgM preparations are connected in Fig. 4C; data are expressed per micromolar subunit concentration to allow ready comparison). The four monoclonal IgG Abs contain ␥2a heavy chains, and all monoclonal IgM/IgG Abs contain light chains. No attempt was made to determine the nucleophilic reactivity of various ␥ chain isotypes. However, the polyclonal Ab data indicate that the average nucleophilic reactivity of the IgG isotype mixture in blood is lower than the IgM reactivity. A similar argument can be presented in regard to antigenic specificity. The 5 murine IgM Abs and 4 IgG Abs were raised by experimental immunization and bind different antigens (MHC antigens, VIP, glucagon, Ig subunits; Refs. 6 and 25 and specifications provided by the manufacturers). The sixth monoclonal IgM was from a patient 3 However, all 10 IgM valencies are usually not filled (e.g., Ref. 44).  with Waldenström's macroglobulinemia with unknown antigenic specificity (20). The monoclonal IgM Abs uniformly displayed superior reactivity to IgG Abs, suggesting that divergent antigenic specificities do not account for the reactivity difference.
One of the monoclonal IgM Abs, Yvo, was employed to help define the structural requirements favoring hapten CRA covalent binding. Compound II, which contains the unesterified phosphonate, did not form adducts with the IgM at incubation times up to 3 h (reaction conditions as in Fig. 3B). Similarly, the neutral hapten CRA III devoid of the amidino group and the hapten CRA IV with weak leaving groups (methyl instead of phenyl groups) failed to form detectable adducts with this IgM Ab. These reactivity characteristics are similar to those of IgG Abs reported previously (11).
Secreted IgM Catalytic Activity-The catalytic activity of polyclonal IgM and IgG prepared from pooled mouse serum was initially measured using Glu-Ala-Arg-AMC as substrate (Fig. 5A). Cleavage of the amide bond linking the AMC to the C-terminal Arg residue of this peptide has been validated as a surrogate for peptide bond hydrolysis by IgG Abs (21). Cleavage of Glu-Ala-Arg-AMC by polyclonal murine and human IgM fractions proceeded at rates 344-fold and 237-fold greater, respectively than the IgG fractions from the same sera (computed from initial velocity data; expressed per unit intact Ab concentration). If all 10 IgM valencies 2 and both IgG valencies are filled, the velocities for individual combining sites of murine and human IgM are 69-and 47-fold greater than the corresponding IgG velocities. Consistent with the irreversible binding data in the preceding section, Glu-Ala-Arg-AMC cleavage by murine polyclonal IgM was inhibited by hapten CRA I (Fig.  5B) and the serine protease inhibitor diisopropylfluorophosphate (not shown; 63 and 93% inhibition at 30 and 100 M DFP, respectively after 12 h). The deviation of the progress curve from linearity in the presence of CRA I suggests an irreversible inhibition mode (26). Progressively increasing inhibition of the murine IgM activity (9 -100%) at increasing hapten CRA I concentrations (10 -300 M) was evident (IC 50  Contamination of IgM with conventional proteases was studied by methods employed previously to validate IgG and Ab light chain enzymatic activities (21,27). The IgM obtained by affinity chromatography on the anti-column displayed essentially identical levels of catalytic activity as the 900-kDa IgM fraction obtained by further purification by FPLC gel filtration (Fig. 6A). This fulfills the criterion of purification to constant specific activity required for assignment of enzymatic activity to IgM. Next, we examined IgM treated with 6 M guanidine hydrochloride to dissociate any noncovalently associated contaminants. For this purpose, the affinity purified IgM was subjected to three cycles of gel filtration in 6 M guanidine hydrochloride (Fig. 6B) and the 900 kDa fraction from the final gel filtration cycle was renatured by dialysis. Time-dependent Glu-Ala-Arg-AMC cleavage by IgM subjected to these procedures was observed (Fig. 6C).
Substrate selectivity of the polyclonal IgM preparations and 6 monoclonal IgM Abs was studied using a panel of 10 peptide- AMC conjugates. The rates shown in Table I were computed as slopes of the progress curves. Only substrates containing a basic residue at the cleavage site were hydrolyzed by the IgM Abs. No hydrolysis was detected with substrates containing acidic and neutral residues at the cleavage site. All 6 monoclonal IgM Abs displayed catalytic activity, but the activity levels for different Abs were not identical (varying, for example, over a 24-fold range with Glu-Ala-Arg-AMC as substrate). The Abs displayed different substrate selectivity profiles. For example, the ratio of Glu-Ala-Arg-AMC and Ile-Glu-Gly-Arg-AMC cleavage rates varied from 0.9 to 30.0 for the 5 murine monoclonal IgMs, and the human monoclonal IgM cleaved the former substrate at a robust rate without cleaving the latter substrate detectably (Fig. 7). Hydrolysis of Gly-Gly-Arg-AMC and Gly-Gly-Leu-AMC by IgM 9020 was compared to confirm the requirement for a basic residue at the cleavage site. These substrates are identical except for the Arg-AMC/Leu-AMC linkage, eliminating the possibility of confounding remote residue effects. Cleavage of Gly-Gly-Arg-AMC was detectable, but cleavage of Gly-Gly-Leu-AMC was not (12.6 Ϯ 0.6 and Ͻ 0.13 M AMC/M Ab/hour, respectively).
The constant domain scaffold in the 5 murine monoclonal IgM Abs is identical. Observations of divergent catalytic activity levels and substrate selectivities suggested that the catalytic site is located in the V domains. To confirm this, IgM Yvo was digested with immobilized pepsin and Fab fragments were purified by gel filtration as the 55 kDa protein peak (Fig. 8A). Concentration-dependent cleavage of Glu-Ala-Arg-AMC by the Fab fragment was observed (Fig. 8B). Next, we considered the possibility that pepsin released from the column could be re-sponsible for the observed Fab activity. The pH optimum of pepsin is 1.5-2.7 depending on the substrate (28). The catalysis assays were repeated in 0.1 M glycine, pH 2.7, 1 mM CHAPS. At Fab concentrations affording readily detectable catalytic activity at neutral pH (Fig. 8B), no detectable cleavage of Glu-Ala-Arg-AMC by the Fab was evident at pH 2.7. The cleavage site preference of pepsin (hydrolysis on the C-terminal side of aromatic and hydrophobic residues) is dissimilar to the basic residue preference of IgM Abs. Purified pepsin did not cleave   Glu-Ala-Arg-AMC under conditions yielding readily detectable catalysis by the Fab (675 nM pepsin; other reaction conditions as in Fig. 8B). These data indicate that pepsin contamination is not a factor in the observed Fab activity. Determination of reaction rates for 4 IgM preparations at increasing Glu-Ala-Arg-AMC concentrations indicated typical enzymatic kinetics (polyclonal murine and human IgM, monoclonal IgM 9020 and IgM 9008). The rates were saturable at excess substrate concentration and consistent with the Michaelis-Menten-Henri kinetics (Table II). Observed K m values were in the high micromolar range, reminiscent of the recognition characteristics of conventional proteases. Catalytic antibodies that are adaptively specialized to recognize individual antigens, on the other hand, display K m values in nanomolar to low micromolar range, e.g. IgG c23.5 shown in Table II cleaves the autoantigen VIP with K m 0.4 nM (from Ref. 6). Apparent k cat values for the IgMs exceed those reported previously for IgG (21). DISCUSSION IgM Abs, the first class of Abs produced by B cells, displayed superior nucleophilic and catalytic reactivities compared with IgG Abs. This contrasts with the noncovalent antigen binding function of Abs, which improves adaptively over the course of the immune response. The nucleophilic and catalytic IgM activities were identified in unimmunized mice and healthy humans. Preferential IgM recognition of the positive charged group adjacent to the phosphonate group of the hapten CRAs and the scissile bond in peptide substrates presumably reflect an intrinsic property of the preimmune Ab repertoire. This noncovalent recognition motif enables low affinity interactions with the peptide substrates, whereas traditional noncovalent epitope-paratope binding is characterized by high affinity antigen recognition (nanomolar range K d ).
Hapten CRA I was validated previously as a probe for nucleophilic reactivities expressed by serine proteases, including IgG Abs (11,28). The extent of irreversible CRA binding activity correlates approximately with the catalytic activity (11,29). In the present study, hapten CRA I adducts were located in close proximity to CD19 on the surface of B cells. The latter protein fulfills a signal transducing role as a component of the BCR throughout B cell development (30). Immunochemical and affinity chromatography studies suggested that the majority of the B cell surface staining is attributable to covalent binding by Ig subunits, with the chain providing the dominant contribution, and / chains, a smaller contribution. This is consistent with the superior nucleophilic reactivity of the chain subunit of secreted IgMs. The control phosphonic acid hapten, which stained the cells poorly, does not react with nucleophiles due to the poor electrophilicity of the phosphorus atom (11,18). Monoclonal BCRs were not included here, but all six monoclonal IgM Abs examined expressed nucleophilic reactivity, suggesting that the reactivity may also be expressed by a significant proportion of BCRs. A minority of the B cells was stained intensely by the CRA. These cells are of interest as a potential source of catalysts in future studies. Observations that both Ig subunits express nucleophiles are consistent with the ability of light and heavy chains to independently catalyze the cleavage of peptide bonds in the absence of their partner subunit (31). Site-directed mutagenesis studies have indicated a serine protease-like catalytic triad in the light chain of an IgG Ab (32) and the heavy chain of other IgG Abs is reported to contain nucleophilic Ser residues (e.g. Ref. 33).
Functional roles for serine protease activities have been deduced in B cell developmental processes, but the molecules responsible for the activities have not been identified to our knowledge. The serine protease inhibitors DFP and ␣-1 antitrypsin inhibit mitogen induced B cell division (34,35) and up-regulate the synthesis of certain Ab isotypes by cultured B cells (35). The DFP-sensitive enzyme is B cell-associated and prefers Arg-containing substrates (36). Serine protease inhibitors are reported to inhibit anti-IgM induced BCR signal transduction (37), and anti-IgM mediated B cell activation is correlated with the appearance of a serine protease activity on the cell surface (38). Undoubtedly, conventional serine proteases may contribute to B cell regulation, but it remains that the major CRA binding components on the B cell surface evident in the present study are the BCRs themselves. It is logical to hypothesize, therefore, that stimulation of BCR nucleophilic sites may influence B cell development. Such compounds include naturally occurring serine protease inhibitors and reac-  tive carbonyl compounds capable of irreversible binding to nucleophilic amino acids (39).
Observations of divergent levels of catalytic activity of monoclonal IgM Abs, their differing substrate preference and retention of the activity in the Fab fragments suggest that the catalytic site is located in the V domains. The catalysis assays were conducted in solution phase and at excess concentrations of the small peptide substrate. These conditions will not support binding of a single peptide molecule by more than one Ab valency. Dissociation of antigen bound reversibly at the individual combining sites may increase antigen availability for neighboring sites. However, such an effect will influence the rate of catalysis only when initial antigen concentration is limiting, and there will be no change in the observed V max . Therefore, multivalent binding by non-interacting sites (avidity effects) is an unlikely explanation for the superior activity of decavalent IgM compared with the divalent IgG. The following explanations can be presented for the superior IgM activity. First, loss of catalytic activity may be attendant to V domain somatic diversification after isotype switching from IgM to IgG. Second, distinctive IgM constant domain characteristics may be important in maintaining the integrity of the catalytic site, in which case isotype switching itself may result in reduced catalytic activity. These explanations are not mutually exclusive. Both explanations are consistent with the argument that catalysis is a disfavored phenomenon in the advanced stages of B cell development (as efficient BCR catalysis will result in reduced BCR occupancy). We did not attempt to address these points experimentally in the present study. However, the monovalent Fab studies suggested that disruption of the constant domain architecture of IgM is deleterious for catalysis. The Fab preparations displayed ϳ10-fold lower activity than computed for the individual combining sites of pentameric IgM. Pepsin employed to prepare Fab cleaves chains on the Cterminal side of the CH2 domain (40), which is distinguished by its conformational flexibility (41). Alterations of antigen binding activity when the same V domains are expressed as fulllength IgG Abs belonging to different isotypes are described (e.g. 42), but we are not aware of IgM-IgG V domain swapping experiments in the literature. Allosteric combining site activation due to filling of individual Ab valencies has previously been considered in the case of divalent IgG preparations (43). The temporal sequence of events as the individual IgM combining sites bind antigen has not been elucidated. At excess antigen, only 5 of the 10 IgM combining sites are thought to be filled (e.g. Ref. 44), suggesting that favorable allosteric effects on antigen binding, if present, must be restricted to conditions of limiting antigen concentrations.
Our screening experiments were restricted to a few IgMs and a few commercially available substrates. Additional studies are necessary to define the physiological substrates for IgM Abs and the upper limit for catalytic rates. However, certain conclusions can be reached from the available data. Apparent turnover numbers (k cat ) for the IgM preparations were as high as 2.8/min. Serum IgM concentrations (1.5-2.0 mg/ml; ϳ2 M) are ϳ3-4 orders of magnitude greater than conventional enzymes (for example, thrombin found at ng-g/ml in serum as a complex with antithrombin III; Ref. 45), and IgM k cat values are ϳ2 orders of magnitude smaller than conventional serine proteases. If catalysis proceeds at the rate observed in vitro, 2 M human IgM with turnover 2.8/min will cleave ϳ24,000 M antigen present at excess concentration (Ͼ ϾK m ) over 3 days (corresponding to the approximate half-life of IgM in blood). Maximal velocity conditions can be approached in the case of antigens present at high concentrations, e.g. albumin and IgG in blood; polypeptides accumulating at locations close to their synthetic site, such as thyroglobulin in the lumen of thyroid follicles; and bacterial and viral antigens in heavily infected locations. Inhibitors regulate the activity of conventional proteases in vivo such as the enzymes responsible for blood coagulation. Unregulated catalysis may lead to disruption of homeostasis. Inhibitory mechanisms regulating conventional enzymes are conceivable in regard to IgM proteolysis.
Identification of promiscuous IgM proteolytic activities in the preimmune repertoire raises important question concerning the existence antigen-specific catalytic IgM Abs. Under conditions of limiting antigen concentration, catalyst competence is measured as the k cat /K m parameter (K m Ϸ K d ). 4 As illustrated for the anti-VIP IgG in Table II, large gains in catalytic competence occur due to enhanced antigen binding affinity (reduced K m ). Certain polypeptides are recognized by IgM Abs present in the preimmune repertoire with high affinity, for example, the superantigens staphylococcal protein A and HIV gp120 5 are recognized by IgM Abs containing VH3 family domains with K d in the nanomolar range (46,47). Moreover, specific IgM Abs with improved affinity for individual antigens emerge by adaptive V domain maturation processes (16,48). Similarly, future study of catalytic IgMs specialized to recognize individual autoantigens is of interest. IgM Abs from patients with autoimmune disease express glycosidase activity (49). Autoimmune humans and mice tend to synthesize catalytic Abs at increased levels (50 -53), and a proteolytic IgG preparation to VIP is shown to interfere with the physiological smooth muscle relaxant effect of VIP (54).