Inhibition of the Proteolytic Activity of Pregnancy-associated Plasma Protein-A by Targeting Substrate Exosite Binding*

The metalloproteinase pregnancy-associated plasma protein-A (PAPP-A) cleaves both insulin-like growth factor (IGF)-binding protein 4 (IGFBP-4) and -5 at a single site in their central domain causing the release of bioactive IGF. Inhibition of IGF signaling is relevant in human disease, and several drugs in development target the IGF receptor. However, inhibition of PAPP-A activity may be a valuable alternative. We have generated monoclonal phage-derived single chain fragment variable (scFv) antibodies which selectively inhibit the cleavage of IGFBP-4 by PAPP-A, relevant under conditions where cleavage of IGFBP-4 represents the final step in the delivery of IGF to the IGF receptor. None of the antibodies inhibited the homologous proteinase PAPP-A2, which allowed mapping of antibody binding by means of chimeras between PAPP-A and PAPP-A2 to the C-terminal Lin12-Notch repeat module, separated from the proteolytic domain by almost 1000 amino acids. Hence, the antibodies define a substrate binding exosite that can be targeted for the selective inhibition of PAPP-A proteolytic activity against IGFBP-4. In addition, we show that the Lin12-Notch repeat module reversibly binds a calcium ion and that bound calcium is required for antibody binding, providing a strategy for the further development of selective inhibitory compounds. To our knowledge these data represent the first example of differential inhibition of cleavage of natural proteinase substrates by exosite targeting. Generally, exosite inhibitors are less likely to affect the activity of related proteolytic enzymes withsimilar active site environments. In the case of PAPP-A, selective inhibition of IGFBP-4 cleavage by interference with exosite binding is a further advantage, as the activity against other known or unknown PAPP-A substrates, whose cleavage may not depend on binding to the same exosite, is not targeted.

The insulin-like growth factors (IGF-I and -II) 2 are polypeptides of ϳ70 residues with auto-and paracrine effects on cell proliferation, migration, and differentiation (1). The IGFs bind to the IGF-1 receptor (IGFR) (2), but six homologous binding proteins, IGFBP-1-6, are able to sequester IGF from IGFR due to higher affinities for IGF-I and -II (3,4). However, bioactive IGF can be released from such complexes by means of proteolytic cleavage of the binding protein, causing the generation of IGFBP fragments with diminished affinity for IGF (5,6). Specific, limited proteolysis represents the principal mechanism of IGF activation.
Several lines of evidence have demonstrated that the metalloproteinase pregnancy-associated plasma protein-A (PAPP-A, pappalysin-1, EC 3.4.24.79) (7,8) functions in the IGF system. PAPP-A specifically cleaves IGFBP-4 (9) and IGFBP-5 (10), thereby releasing sequestered IGF or causing binding protein inactivation (6). PAPP-A knock-out mice are proportional dwarfs, reduced to a body mass of 60% compared with wild type littermates (11). This phenotype is similar to the phenotype of IGF-II knock-out mice (12), supporting the hypothesis that IGF-II activity requires PAPP-A activity in early fetal development (11) with cleavage of IGFBP-4 responsible for the final delivery of IGF to the receptor (6,13). Postnatally, PAPP-A and IGFBP-4 appear to be implicated in many different processes of cell proliferation, such as implantation (14), myoblast proliferation and differentiation (15), and bone formation in vivo (16). PAPP-A is produced by vascular smooth muscle cells after angioplasty (17), suggesting that it promotes neointimal cell proliferation, and it is abundantly synthesized in unstable atherosclerotic plaques (18).
The 400-kDa PAPP-A contains two subunits of 1547 residues (7,19 and belongs to the metzincin superfamily of metalloproteinases (8, 20). A laminin G-like module of unknown function is present N-terminal to the proteolytic domain (21), and five complement control protein (CCP) modules, enabling PAPP-A to bind to the cell surface (22), are located in the C-terminal end of the subunit (see Fig. 1). Additionally, PAPP-A contains three Lin12-Notch repeat (LNR) modules, which are unique to PAPP-A, its homologue PAPP-A2 (23), and the family of Notch receptors (24). In PAPP-A and PAPP-A2, two of the LNR modules (LNR1 and -2) are inserted into the proteolytic domain (8, 25), whereas the third (LNR3) is located C-terminal to the CCP modules (8) (see Fig. 1).
The IGFs are involved in both normal physiology and human disease, e.g. cancer (26,27) and cardiovascular disease (28,29), and therefore, strategies for the direct inhibition of IGF signaling have been developed (30 -33). However, specific inhibition of growth promoting proteolytic activity may represent a valuable alternative, in particular because unintended interference with other signaling pathways, e.g. insulin signaling, is avoided by such approach.
In the present study we have generated monoclonal scFv antibodies which bind to LNR3 of PAPP-A and efficiently inhibit proteolysis of IGFBP-4 but not -5. Thus, LNR3 represents a substrate binding exosite (34,35) that can be targeted for selective proteolytic inhibition. Our data represents to our knowledge the first example of differential inhibition of cleavage of natural proteinase substrates by exosite targeting.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-An expression construct encoding a His-tagged form of the C-terminal portion of murine PAPP-A, starting at module CCP1 ( Fig. 1), was made by overlap extension PCR (36). Briefly, a nucleotide fragment encoding a signal peptide (MVLLTAVLLLLAAYAGPAQS) was derived by PCR from a construct encoding human IGFBP-5 (GenBank TM accession number NP_000590) (10,23). The primers were 5Ј-GACGAAGCTTATGGTGTTGCTCACCGCGGT-3Ј (an introduced HindIII site is shown in bold), and 5Ј-TTCTGGG-CAGTCGGCGCTCTGGGCCGGCCCCGCAT-3Ј (nucleotides encoding residues 1129 -1133 of murine PAPP-A are underlined 3 ). A second nucleotide fragment encoding residues 1129 -1545 of murine PAPP-A followed by a His tag was made using pcDNA3.1-mPA (37), encoding murine PAPP-A, as a template. The primers were 5Ј-GGGCCGGCCCAGAGCGC-CGACTGCCCAGAACTGGC-3Ј (murine PAPP-A-encoding nucleotides are underlined) and 5Ј-CATTTCTAGATCAAT-GATGATGATGATGATGTCCTGAGCCATGGCTATATC-CCCGAAGATCTTTCC-3Ј (an XbaI site is shown in bold, nucleotides encoding residues Gly-Ser followed by six histidine residues are shown in italics, and murine PAPP-A-encoding nucleotides are underlined). The PCR product resulting from the combination of the two fragments was cloned into the HindIII/XbaI sites of pcDNA3.1(ϩ) (Invitrogen) to obtain pmPA(CCP1-C-His), encoding murine PAPP-A residues 1129 -1545 followed by the amino acid sequence GSHHH-HHH. A similar construct, encoding residues 1133-1547 of human PAPP-A, had previously been made (38). This construct did not include a His tag.
Cell Culture and Expression of Protein in Mammalian Cells-Human embryonic kidney 293T cells (293tsA1609neo) (41) were maintained in high glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, nonessential amino acids, and gentamicin (Invitrogen). Cells were plated onto 6-cm tissue culture dishes and transiently transfected 18 h later by calcium phosphate co-precipitation (42) using 10 g of plasmid DNA prepared by QIAprep Spin kit (Qiagen). After 48 h, the culture media were harvested and cleared by centrifugation. Cells transfected with cDNA encoding murine PAPP-A(1129 -1545), human PAPP-A-(1133-1547), and human PAPP-A(E483Q) were further cultured in serum-free medium to facilitate purification.
Generation of Chicken Polyclonal Antibodies-Murine PAPP-A(1129 -1545) was purified from serum-free medium by nickel affinity chromatography using Chelating-Sepharose Fast Flow beads (1 ml) (Amersham Biosciences). The column was washed with 1 M NaCl, 50 mM sodium phosphate, pH 5.5, and bound protein was eluted with phosphate-buffered saline containing 20 mM EDTA. Pooled fractions were further purified by heparin affinity chromatography using a 1-ml HiTrap Heparin HP column (Amersham Biosciences) equilibrated in phosphate-buffered saline. The column was washed with phosphate-buffered saline (PBS) containing 250 mM NaCl, and bound protein was eluted with PBS containing 1 M NaCl.
Three chickens (Isa Brown) were immunized by intramuscular injections in the breast muscle using 15 g of purified protein for each round. Complete Freund's adjuvant (Sigma) was used in the first round of injections, and incomplete Freund's adjuvant (Sigma) was used for successive rounds (43). Eggs were collected daily and kept at 4°C. Titers were evaluated at regular intervals by purification of IgY from yolk (43) followed by a direct ELISA, in which murine PAPP-A(1129 -1545) was coated onto plastic. Detection was done using horseradish peroxidase-conjugated anti-chicken-IgY (Sigma), and preparations of IgY with the highest titers were used for further experiments.
Measurement of Proteolytic Activity-Proteolytic activity of PAPP-A against IGFBP-4 and -5 was analyzed as previously described in detail (44). In brief, purified substrates quantified by amino acid analysis were labeled with 125 I (Amersham Biosciences). Cleavage reactions were carried out in 50 mM Tris-HCl, 100 mM NaCl, 1 mM CaCl 2 , pH 7.5, in the absence (IGFBP-5) or presence (IGFBP-4) of a 10-fold molar excess of IGF-II (Diagnostic Systems Laboratories). In some reactions purified antibodies were added as specified. After incubation at 37°C, reactions were quenched by the addition of EDTA (10 mM) and stored at Ϫ20°C. Cleavage products were separated by 10 -20% SDS-PAGE and visualized by autoradiography. The degree of cleavage was determined by quantification of band intensities using a Typhoon imaging system (GE Healthcare), and background levels (mock signals) were subtracted. Proteolytic activity of PAPP-A2 against IGFBP-5 in the absence of IGF was analyzed similarly, as previously detailed (23). Analysis of peptidolytic activity against a 26-residue synthetic peptide derived from IGFBP-4 was carried out as previously described (45). Residues on the N-terminal and C-terminal side of the cleavage site were modified with o-aminobenzoic acid and substituted with 3-nitrotyrosine, respectively. The reaction buffer was 50 mM Tris, pH 8.0, 0.01% Tween 20. Light at 310 nm was used for excitation, and emission was detected at 420 nm. Quantitative analyses were carried out with Prism 5.0 (Graphpad software) using the equation for competitive inhibition. In the case of partial inhibition, sigmoidal dose-response curves were fitted to the data.
Screening of Semi-synthetic Phage Libraries-Human PAPP-A-(1133-1547) was immobilized (1 h at 37°C) to 3.5-ml Immunotubes (Nunc Maxisorp), which were coated overnight at 4°C with polyclonal PAPP-A antibodies (46) (5 g/ml) contained in 100 mM sodium bicarbonate, pH 9.8, and blocked with 3% skimmed milk powder in 20 mM Tris, 150 mM NaCl, pH 7.5 (TBS). We used a combination of two semi-synthetic phage libraries (Tomlinson I ϩ J) (47), and capture of phages (10 12 from each library) was carried out for 2 h at room temperature while rotating gently. After capture the tubes were rinsed 10 times in TBS containing 1 M NaCl and 0.1% Tween 20 (TBST), further washed at 4°C with 200 ml of TBST for 1 h using a peristaltic pump, and finally rinsed 5 times with TBS. Elution of phages was carried out for 10 min at room temperature using 0.5 ml of DPPC-treated trypsin (1 mg/ml) (Sigma) diluted in TBS.
Escherichia coli (TG1) infected (30 min at 37°C) with the eluted phages were plated on TYE (10 g peptone, 5 g yeast extract, 5 g NaCl, and 16 g agar per liter) plates supplemented with 1% glucose and ampicillin (100 g/ml). Colonies were transferred to 96-well culture plates with 2ϫTY (16 g peptone, 10 g yeast extract, and 5 g NaCl per liter) medium containing 1% glucose and ampicillin (100 g/ml) and incubated (overnight at 37°C). A replicate of each plate was incubated for 3 h, and KM13 helper phages (48) were added (10 9 to each well). After incubation for 1 h, the medium was changed to 2ϫTY containing ampicillin (100 g/ml) and kanamycin (50 g/ml), the plates were incubated for 20 h at 30°C, and phage-containing supernatants were then analyzed by ELISA for binding to human PAPP-A(E483Q), immobilized in 96-well plates with polyclonal PAPP-A antibodies. The plates were blocked with 2% bovine serum albumin (Sigma), washing was done with TBS containing 0.1% Tween 20, and detection was performed using horseradish peroxidase-conjugated anti-M13 (GE Healthcare). Phagemid DNA (pIT2) from selected clones was prepared and sequenced.
To also obtain phage antibodies which bind the region of PAPP-A N-terminal to CCP1, a similar screening was carried out using full-length human PAPP-A. Clone PAC5 was obtained and used as a control antibody in the scFv format. Clone PAC33, with specificity irrelevant to PAPP-A, was also used as a control antibody.
Expression and Purification of scFv Antibodies and Analysis of Inhibitory Potential-E. coli HB2151 (nonsuppressor of amber stop codon) was infected with selected phages for the production of monoclonal scFv antibodies. Cultures of 1 liter were induced for 4 -16 h with 1 mM isopropyl 1-thio-␤-D-galactopyranoside, and expressed proteins were purified by nickel affinity chromatography on Chelating-Sepharose Fast Flow beads (5 ml) (GE Healthcare) after sonication. Washing was carried out with 20 mM imidazole, 100 mM NaCl, 50 mM sodium phosphate, pH 8.0. For elution, the concentration of imidazole was increased to 300 mM. Eluted protein was dialyzed into 20 mM Tris, 100 mM NaCl, pH 8. A Mono Q column (GE Healthcare) or a protein L column (Pierce) was used for further purification. The purified protein was dialyzed into the relevant buffers before functional analysis. Analysis of binding to PAPP-A, immobilized as described above, was carried out in 96-well plates using a horseradish peroxidase-conjugated anti-His-tag antibody (Sigma) for detection.
Preparations of purified antibodies were quantified by amino acid analysis, and the effects of antibodies on PAPP-A activity were analyzed by the addition of controlled amounts to cleavage reactions. PAC33 was used as a negative control.
Surface Plasmon Resonance Analysis-Surface plasmon resonance experiments were carried out on a BIAcore T100 instrument (BIAcore AB, Uppsala, Sweden) using series S CM5 sensor chips and coupling reagents supplied by the manufacturer. Affinity-purified PAPP-A(E483Q), 10 g/ml in 10 mM sodium acetate, pH 5.0, was immobilized (at a level of 500 resonance units) to the activated chip at 25°C. Remaining activated groups were blocked by 1 M ethanolamine. Purified antibodies (0.35-11 nM) diluted in 10 mM HEPES, 150 mM NaCl, 1 mM CaCl 2 , 0.05% Tween 20, pH 7.4, were injected over the sensor chip for 2 min at a flow rate of 30 l/min at 37°C. Recorded signals were defined as the difference between the signal of a ligand-coupled and an unliganded channel. Sensorgrams were plotted after subtraction of the recorded signal determined by the injection of an irrelevant antibody. For the analysis of calcium ion dependence, some experiments were carried out using 175 nM PAC1 in the absence or presence of 10 mM EDTA. Data were analyzed using the BIAcore T100 evaluation software Version 1.1.
Determination of Antibody Specificity and Mapping of Antibody Binding-Mapping of antibody binding was carried out by ELISA using PAPP-A/PAPP-A2 chimeric proteins and mutated variants of PAPP-A. The proteins were immobilized with polyclonal PAPP-A antibodies (46), and binding of phage antibodies was analyzed as detailed above.

Inhibition of Proteolysis by Polyclonal Antibodies Raised against a C-terminal Fragment of PAPP-A-
The only known physiological inhibitor of PAPP-A proteolytic activity is the proform of eosinophil major basic protein (proMBP), which inactivates PAPP-A by the formation of a covalent disulfidebased 2:2 complex, denoted PAPP-A/proMBP (49 -51). Inhibition by proMBP is irreversible, but the process of complex formation is relatively slow and sensitive to redox potential (49). Of synthetic low molecular weight compounds, such as many inhibitors of the related matrix metalloproteinases, none has been identified for PAPP-A.
However, PAPP-A activity toward both IGFBP-4 and -5 was effectively inhibited by polyclonal antibodies (Fig. 2A). Because we have previously hypothesized that the LNR3 region of PAPP-A is involved in substrate binding (38), we wanted to test the inhibitory properties of antibodies raised against the C-terminal region of PAPP-A. We made a plasmid construct encoding residues 1129 -1545 of murine PAPP-A, including the five CCP modules and LNR3 (Fig. 1), and recombinant protein, expressed in 293T cells, was purified (Fig. 2B) for immunization. Because PAPP-A LNR3 is highly conserved (100% identity between human and murine PAPP-A LNR3 (37)), chickens were chosen for immunization. The resulting polyclonal IgY antibodies were found to effectively inhibit PAPP-A proteolysis of IGFBP-4, whereas cleavage of IGFBP-5 was less affected (Fig.  2C). No effect on proteolysis of the two substrates was observed with preimmune IgY (not shown).
Although many epitopes outside the LNR3 region exist on the ϳ60-kDa C-terminal fragment used for immunization, this experiment supports a hypothesis that LNR3 functions as a substrate binding exosite. Furthermore, the finding that such polyclonal antibodies cannot efficiently inhibit cleavage of IGFBP-5 suggests a differential mode of proteinase-substrate interaction and that an inhibitor which selectively targets PAPP-A cleavage of IGFBP-4 can be obtained.
Selection by Phage Display of PAPP-A Monoclonal scFv Antibodies Inhibitory of IGFBP-4 Proteolysis-To obtain monoclonal antibodies with selective inhibitory activity against PAPP-A cleavage of IGFBP-4, a phage antibody library was screened for binding to the 60-kDa C-terminal fragment of human PAPP-A (residues 1133-1547), as detailed under "Experimental Procedures." To increase chances of obtaining a phage antibody which binds to the conserved LNR3 region, we used a combination of two semi-synthetic phage libraries (47), built on a framework of commonly used single VH and VL human gene segments. Bound phages were cloned, their binding to full-length PAPP-A was evaluated by ELISA, and scFv antibodies from selected phages were then produced in E. coli and assessed for inhibitory activity. We obtained two inhibitory scFv antibodies, PAC1 and PAC2, which by sequence analysis were found to be unique clones (not shown). For further characterization, PAC1 and PAC2 were expressed at a larger scale, purified, and quantified by amino acid analysis. Both scFv antibodies effectively inhibited PAPP-A cleavage of IGFBP-4, as illustrated with PAC1 (Fig. 3A). After 1 h of incubation, only a faint cleavage product was observed in the presence of PAC1. Based on such preliminary analysis, PAC1 appeared to be a slightly better inhibitor than PAC2 (not shown), which was, therefore, not further quantitatively analyzed.
Inhibition of human PAPP-A was further analyzed by plotting relative initial velocities (44) as a function of antibody concentration (Fig. 3B), and the inhibitory constant of PAC1 toward human PAPP-A was determined (K i ϭ 1.2 nM). Antibody binding to immobilized PAPP-A was further analyzed by surface plasmon resonance, from which we obtained an equilibrium dissociation constant for PAC1 (K D ϭ 0.25 nM, Fig. 3C), in fair agreement with the solution phase experiments.
PAPP-A Activity against IGFBP-5 Is Only Partially Inhibited by Antibody PAC1-Antibody PAC1 showed much less inhibitory activity against PAPP-A cleavage of IGFBP-5 compared with IGFBP-4, as illustrated by an end-point assay (Fig. 4A) and further analyzed quantitatively (Fig. 4B). At saturating concentrations of PAC1, PAPP-A still showed about 50% activity against IGFBP-5, possibly because PAPP-A activity against IGFBP-5 is reduced by steric hindrance rather than direct interference with enzyme-substrate interaction.
Interestingly, PAPP-A cleavage of a synthetic peptide derived from IGFBP-4 (45) was not inhibited by PAC1 (Fig. 5, A and B), suggesting that binding of PAC1 to PAPP-A does not cause an altered structure of the active site environment. This is in con-  JUNE 13, 2008 • VOLUME 283 • NUMBER 24

JOURNAL OF BIOLOGICAL CHEMISTRY 16775
trast to the inhibitory properties of a PAPP-A monoclonal antibody (mAb), PA-1A (44), which efficiently inhibited cleavage of this peptide (Fig. 5, C and D). However, mAb PA-1A was a poor inhibitor of both intact IGFBP-4 ( Fig. 6A) and IGFBP-5 (Fig. 6B). In both cases about 40% of PAPP-A activity remained at saturating concentrations of mAb PA-1A, in striking contrast to the efficient inhibition against IGFBP-4 cleavage obtained with PAC1 (Fig. 7A). Most likely, mAb PA-1A recognizes an epitope of PAPP-A, which is located at or close to the active site. In agreement with this, mAb PA-1A did not bind to the C-terminal fragment of human PAPP-A (residues 1133-1547) (not shown).
Antibody PAC1 Binds to LNR3 in a Calcium Ion-dependent Manner-Efficient inhibition of murine PAPP-A cleavage of IGFBP-4 was found with PAC1 ( Fig. 7A) and PAC2 (not shown), in accordance with the high degree of sequence conservation between murine and human PAPP-A. However, we found that neither PAC1 nor PAC2 had any effect on the activity of the homologue PAPP-A2 toward its two known substrates, IGFBP-3 (not shown) and IGFBP-5 (Fig. 7B). Hence, to delineate the binding of PAC1 and PAC2 to PAPP-A, a set of chimeras in which a variable portion of the C terminus was exchanged with the sequence of PAPP-A2 was analyzed. Only constructs in which the sequence of LNR3 was derived from PAPP-A showed binding of PAC1 (Fig. 8A) and PAC2 (not shown). Also, neither of the antibodies showed any binding to a mutant of PAPP-A truncated on the N-terminal side of LNR3 (Fig. 8A).
Because the LNR modules of PAPP-A have been suggested to coordinate calcium ions (40), the possible dependence of calcium ions for antibody binding was assessed by surface plasmon resonance. Binding of PAC1 to immobilized PAPP-A was observed in the presence of calcium ions (Figs. 8B and 3C), but no binding was seen when both the sample and the flow cell were equilibrated with the chelator EDTA (not shown). After EDTA treatment, re-equilibration of the flow cell using calcium ion-containing buffer restored its ability to bind PAC1 (Fig. 8B,  dashed line).
When PAC1 was injected with EDTA onto a chip equilibrated with the calcium ion-containing buffer (Fig. 8B, lower  sensorgram), initial association was observed, but early dissociation occurred because the injected EDTA caused the LNR3 module to lose its bound calcium ion. Hence, calcium ions are required for PAC1 binding to PAPP-A, and the removal of calcium from LNR3 by EDTA is a reversible process.
Based on these data, we analyzed antibody binding to mutants of LNR3 by ELISA, in which the three amino acid residues predicted to coordinate a calcium ion were individually substituted with alanine. In agreement with the mapping data using PAPP-A/PAPP-A2 chimeras and with the surface plas-  mon resonance experiment data, PAC1 and PAC2 showed no binding to these mutants (Fig. 8C).

DISCUSSION
Using a recombinant C-terminal fragment of human PAPP-A (Fig. 1), we have generated inhibitory scFv antibodies, PAC1 and PAC2, by selection from two semi-synthetic phage libraries. PAC1 efficiently inhibits (K i ϭ 1.2 nM) the proteolytic activity of PAPP-A against one of its substrates, IGFBP-4 ( Fig.  3B), whereas the activity against the other substrate, IGFBP-5, is only moderately reduced at saturating concentrations (Fig.  4B). This strategy was based on the initial observation that polyclonal antibodies against full-length PAPP-A efficiently inhibit proteolysis of both IGFBP-4 and -5 ( Fig. 2A), whereas polyclonal antibodies, raised against a C-terminal PAPP-A fragment, only efficiently inhibited cleavage of IGFBP-4 (Fig. 2C).
We found that neither PAC1 nor PAC2 was inhibitory against the related proteinase PAPP-A2 (Fig. 7B). This allowed a mapping strategy based on chimeric proteins between PAPP-A and PAPP-A2 within the 400-residue fragment used for phage selection. Both PAC1 and PAC2 show binding only to chimeras in which the C-terminal LNR3 module (Fig. 1) was derived from PAPP-A (Fig. 8A). Furthermore, the antibodies bound PAPP-A only in the presence of calcium ions, as shown by surface plasmon resonance (Fig. 8B), and the substitution into alanine of single LNR3 residues, predicted to coordinate a calcium ion (40), abrogated antibody binding (Fig. 8C). Thus, the epitopes recognized by both PAC1 and PAC2 include amino acids within the 26-residue LNR3 module. LNR3 calcium ion binding may be required for maintaining a native conformation of this module, essential for antibody binding. Alternatively, the putative calcium ion coordinating residues of LNR3 (Asp-1484, Asp-1499, and Asp-1502) may be directly involved in antibody binding, or the antibodies may provide a ligand that completes the coordination sphere of the calcium ion. Although the PAPP-A LNRs probably form units of three modules with LNR1 and two from one PAPP-A subunit and LNR3 from the other (38), our data suggest that LNR3 is correctly folded in the isolated C-terminal PAPP-A fragment and, therefore, does not depend on LNR1 and -2 (Fig. 8B). The epitopes of PAC1 and PAC2 are most likely very similar. However, based on sequence analysis of the variable regions, the two antibodies are unique.
Interestingly, at saturating concentrations of PAC1, PAPP-A shows only residual activity against IGFBP-4, but the activity against IGFBP-5 is still ϳ50% (Figs. 3 and 4). This suggests that cleavage of IGFBP-4 by PAPP-A is competitively inhibited. The partial inhibition of IGFBP-5 cleavage is probably a result of steric hindrance rather than an allosteric effect, in particular because peptide cleavage in the presence of PAC1 is not compromised (Fig. 5B). Such interpretation agrees with a model in which PAPP-A LNR3 forms a substrate binding exosite required for cleavage of IGFBP-4 but not IGFBP-5. Hence, the efficient inhibition of IGFBP-4 cleavage is caused by a direct interference with the binding of IGFBP-4 to this module of PAPP-A, whereas the reduction in activity against IGFBP-5 is likely caused by steric hindrance of substrate binding. Hence, a future compound of lower molecular weight targeting LNR3 may not show any inhibition of IGFBP-5 cleavage. Such hypothesis can be tested by the selection of e.g. phage-displayed peptides against the C-terminal PAPP-A fragment, which require calcium ions for binding.
The principle of exosite inhibition is highly relevant for many proteolytic enzymes involved in human disease, in particular multidomain enzymes such as the large group of matrix metalloproteinases and related enzymes (34,35,52). Inhibition by targeting a substrate binding exosite has at least two advantages compared with direct active site inhibition: specificity and selectivity. First, an inhibitor directed toward a unique exosite is unlikely to influence the activity of other related proteinases with similar active site architecture. Second, different substrates of a given proteinase may not use the same exosite, and their cleavage can, therefore, be differentially inhibited.
There are examples of proteolytic inhibition by exosite targeting in the literature. For example, matrix metalloproteinase-2 cleavage of type I gelatin and type IV collagen has been inhibited by targeting the collagen binding domain using a synthetic peptide (53), and a substrate binding pocket distinct from the catalytic site of ␤-amyloid precursor protein cleaving enzyme has been targeted by synthetic peptides (54). Also, many exosite inhibitors have been developed toward the coagulation enzyme factor VIIa (55), but although specific for this proteinase, all of its several biological substrates are targeted by such inhibitors (56). Therefore, to the best of our knowledge, the inhibition of IGFBP-4 proteolysis by PAPP-A represents the first example of selective inhibition of one natural proteinase substrate by exosite targeting.
An obvious advantage of selective inhibition of PAPP-A  The dashed line shows binding of PAC1 after EDTA treatment and reequilibration of the flow cell with the calcium ion-containing running buffer, demonstrating that PAPP-A LNR3 binds calcium reversibly. RU, resonance units. C, binding of PAC1, PAC2, and PAC5 to mutants of PAPP-A LNR3, in which residues Asp-1484, Asp-1499, and Asp-1502, predicted to coordinate a calcium ion (40), are substituted individually with alanine, analyzed by ELISA. The absence of binding (Ϫ) or binding (ϩ) is indicated. PAC5, which was used as a positive control, binds to an epitope of PAPP-A located N-terminal to CCP1 (not shown).
activity against IGFBP-4 is that this is thought to represent the final step of IGF delivery to the IGF receptor (6,13). Other PAPP-A substrates, including the known substrate IGFBP-5 and other potential, yet unknown PAPP-A substrates whose cleavage may not depend on exosite binding to LNR3, will be unaffected by this strategy of inhibition. Interestingly, the mAb PA-1A efficiently inhibits PAPP-A cleavage of a synthetic peptide (Fig. 5, C and D), but this antibody was a poor inhibitor of both IGFBP-4 and -5 (Fig. 6, A and B). In contrast, whereas PAC1 efficiently inhibited cleavage of intact IGFBP-4, PAPP-A activity against the synthetic peptide was not reduced by PAC1 (Fig. 5, A and B). This suggests that exosite inhibition by targeting LNR3 may not only provide increased specificity but may also be the preferred strategy for efficient inhibition of IGFBP-4 cleavage.
Targeting IGF signaling is highly relevant in human disease, notably in cancer (26,27,33) and cardiovascular disease (28,29,32). Specific inhibition of growth promoting proteolytic activity may be a valuable alternative to the direct inhibition of IGF signaling, in particular because such inhibition of IGF receptor stimulation is unlikely to interfere with e.g. insulin signaling. Our finding that PAC1 also efficiently inhibits murine PAPP-A will allow testing of the hypothesis that IGF signaling can be inhibited therapeutically in vivo by means of targeting PAPP-A activity. Suitable murine models for the role of PAPP-A in the development of atherosclerotic plaques (57) and of IGFdependent tumor development (58) have been reported. Additionally, selective inhibition of PAPP-A activity will also potentially allow us to delineate the in vivo roles of PAPP-A activity toward different substrates, such as IGFBP-4 and -5, in normal physiology. Although both are PAPP-A substrates, the different roles of these binding proteins within the IGF system are not completely understood (6).
In conclusion, our data represent the first example of a proteinase inhibitor that causes differential inhibition of two physiological substrates of a proteinase by targeting a substrate binding exosite. Our work represents a potentially valuable alternative to the direct inhibition of IGF receptor signaling, which is relevant in human disease.