Pregnancy-associated Plasma Protein-A2 (PAPP-A2), a Novel Insulin-like Growth Factor-binding Protein-5 Proteinase*

A novel metalloproteinase with similarity to pregnan-cy-associated plasma protein-A (PAPP-A), which we de-noted PAPP-A2, has been identified. Through expression in mammalian cells we showed that recombinant PAPP-A2 polypeptide of 1558 residues resulted from processing of a 1791-residue prepro-protein. Unlike PAPP-A, PAPP-A2 migrated as a monomer (of 220 kDa) in non-reducing SDS-polyacrylamide gel electrophore-sis. The prepro-parts of PAPP-A2 and PAPP-A are not homologous, but mature PAPP-A2 shares 45% of its residues with PAPP-A. Because PAPP-A specifically cleaves insulin-like growth factor-binding protein (IGFBP)-4, one of six known modulators of IGF-I and -II, we looked for a possible PAPP-A2 substrate among the members of this family. We showed that PAPP-A2 specifically cleaved IGFBP-5 at one site, between Ser-143 and Lys-144. In contrast to the cleavage of IGFBP-4 by PAPP-A that strictly requires the presence of IGF, the cleavage of IGFBP-5 by PAPP-A2 was IGF-independent. Recent solved in solvent A, and further diluted three times in 50 m M Tris, pH 7.5. This material was prepared for experiments to demonstrate that cleavage of IGFBP-5 by PAPP-A2 (using ; 0.1 m g purified PAPP-A2/2 m g IGFBP-5) occurs in the absence of IGF and is not promoted by IGF.

Pregnancy-associated plasma protein-A (PAPP-A) 1 has recently been shown to specifically cleave insulin-like growth factor-binding protein-4 (IGFBP-4) (1), one of six modulators of IGF-I and -II activity (2). Cleavage of IGFBP-4 causes release of bound IGF that in turn interacts with its cellular receptor (3). Interestingly, the cleavage of IGFBP-4 by PAPP-A strictly depends on the presence of IGF (1). It has been established that PAPP-A is the IGFBP-4 proteinase secreted from fibroblasts (1), osteoblasts (1,4), marrow stromal cells (1), and vascular smooth muscle cells (5) and is present in pregnancy serum (6) and ovarian follicular fluid (7).
PreproPAPP-A is synthesized as a 1627-residue protein, which is processed into mature PAPP-A of 1547 residues (8,9). In most systems, PAPP-A appears to exist as a homodimer of 400 kDa (1). However, in pregnancy serum, and possibly elsewhere, Ͼ99% of PAPP-A is found as a disulfide bound 2:2 complex with the highly and unusually glycosylated proform of eosinophil major basic protein (proMBP) (10 -12). In this covalent complex, proMBP functions as an inhibitor of the proteolytic activity of PAPP-A (6).
No proteins with global homology to PAPP-A have been described. Here we report the identification of a novel protein, PAPP-A2, which is a homologue of PAPP-A. Expression of recombinant PAPP-A2 allowed us to establish that PAPP-A2 is an active metalloproteinase, that it specifically cleaves IGFBP-5, and that it therefore likely functions in the same growth regulatory system as PAPP-A.

Identification and Cloning of a Nucleotide Sequence Encoding PAPP-A2
Searching (13) public databases for DNA sequences that when translated showed homology to the preproPAPP-A protein sequence 2 revealed two genomic clones, AL031734 and AL031290, containing coding sequence stretches corresponding to ϳ25% of the N-terminal amino acid sequence of PAPP-A (residues 96 -493) 2 , and 15% of the C-terminal (residues 1393-1593). Based on their co-localization on chromosome 1 (1q24), we hypothesized the existence of a novel protein similar to PAPP-A, which we named PAPP-A2. We first established a coding cDNA sequence aligning with the ϳ60% remaining, central region of PAPP-A; cDNA was synthesized using human placental mRNA as a template and a primer derived from AL031290 (5Ј-GCTCACACACCA-CAGGAATG-3Ј). With primers derived from AL031734 (5Ј-GGCTGAT-GTGCGCAAGACCTG-3Ј) and AL031290 (5Ј-GCATTGTATCTTCAG-GAGCTTG-3Ј), a PCR product was generated that corresponded to the central region, a total of 908 amino acids (PAPP-A2 residues 665-1572). Manual inspection of the genomic sequence of AL031734 revealed that the open reading frame of the sequence stretch aligning with PAPP-A continued further in the 5Ј direction; nt 102646 -103566 encode a polypeptide sequence of 307 residues that starts with a methionine residue. Thus, the same cDNA preparation was used to obtain a contiguous cDNA sequence encoding the N-terminal portion of PAPP-A2 using primers from AL031734 (5Ј-GAAGTTGACTTCTGGTTCTGTAG-3Ј) and from the central region (5Ј-CCCTGGGAAGCGAGTGAAGCC-3Ј).
A stop codon was present in the sequence of AL031290 corresponding to PAPP-A residue 1618. Therefore, cDNA was synthesized using placental mRNA and a primer originating from AL031290 further 3Ј to this stop codon (5Ј-GCATTTCTTATAAGATCCTTCATGC-3Ј). A contiguous * This work was supported by grants from the Danish Medical Research Council, the Novo Nordic Foundation, and the Alfred Benzon Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF311940.

Identification of EST Sequences
A cluster of EST sequences matching the genomic sequence of AL031290 was identified around nt 64000 -66000 of AL031290 starting ϳ1.2 kb from the 3Ј end of the PAPP-A2 encoding sequence. The existence of mRNA connecting the coding region of PAPP-A2 and this cluster was verified by PCR using primers from AL031290 (5Ј-GGAAA-GAGCAGAGTTCACCCAT-3Ј, nt 64900 -64879 of AL031290), the PAPP-A2 encoding sequence (5Ј-CCGTCTTAGTCCACTGCATCC-3Ј, nt 20499 -20519 of AL031290 and nt 5171-5191 of AF311940), and oligo(dT)-primed placental cDNA as a template (12). As expected, the size of the resulting product was 2.2 kb, further demonstrating the existence of a PAPP-A2 mRNA with a 3Ј untranslated region of ϳ3 kb.

Plasmid Construction
pPA2, Encoding Wild-type PreproPAPP-A2-The NotI-BamHI fragment from p2C was cloned into pBluescriptIISKϩ (Stratagene) to obtain p2CBlue. The NotI-SpeI fragment from p2N and the SpeI-BclI fragment from p2Mid were ligated into the NotI/BclI sites of p2CBlue to obtain p2NMidCBlue containing the entire PAPP-A2 cDNA. The NotI-ApaI fragment of pBluescriptIISKϩ was ligated into the NotI/ApaI sites of the mammalian expression vector pcDNA3.1ϩ (Invitrogen) to obtain a modified version of this vector, pcDNA-NA. The full-length cDNA was then excised from p2NMidCBlue with NotI and XhoI and cloned into pcDNA-NA to obtain pPA2.
pPA2-mH-Two primers (5Ј-GAGGGCCTGTGGACCCAGGAG-3Ј, nt 4906 -4926 of AF311940, and 5Ј-GACGTAAAGCTTCTGATTTTCT-TCTGCCTTGG-3Ј, nt 5373-5354 of AF311940, preceded by a HindIII site, AAGCTT) were used in a PCR with pPA2 as the template to generate a nucleotide fragment encoding the C-terminal 156 residues of PAPP-A2 with the stop codon replaced by a HindIII site for in-frame ligation into the vector pcDNA3.1/Myc-His(-)A. The PCR product was digested with EcoRI and HindIII and cloned into the EcoRI/HindIII sites of the vector to generate pPA2C-mH. The NotI-XbaI fragment and the XbaI-EcoRI fragment were excised from pPA2 and cloned into the NotI/EcoRI sites of pPA2C-mH. The resulting construct, pPA2-mH, encoded PAPP-A2 followed by residues KLGP, the c-myc epitope (EQKLISEEDL), residues NSAVD, and six histidine residues.
pPA2-KO-mH-A variant of pPA2-mH was constructed with a Glu-734 3 Gln substitution; the NotI-KpnI fragment of pPA2-KO was swapped into the NotI-KpnI sites of pPA2-mH to generate pPA2-KO-mH.
Culture and transient transfection of human embryonic kidney 293T cells (293tsA1609neo) was performed as described earlier (6) except that the cells were maintained for another 48 h in serum-free medium (293 SFM II, Life Technologies, Inc.).

Miscellaneous Procedures
SDS-PAGE was performed in Tris/glycine gels (10 -20% or 16%). A metal chelate affinity column (1 ml, Amersham Pharmacia Biotech) was charged with nickel ions and used for affinity purification of His-tagged proteins in serum-free medium. Bound protein was eluted with 10 mM EDTA in phosphate-buffered solution containing 500 mM NaCl and further dialyzed against 20 mM HEPES, 100 mM NaCl, and 1 mM CaCl 2 , pH 7.4. For cleavage site determination, affinity-purified rIGFBP-5 (20 g) was digested (37°C for 12 h) with purified rPAPP-A2 (ϳ1 g) immobilized via anti-c-myc to recombinant protein G-agarose (Life Technologies, Inc.). Edman degradation was performed as described earlier (16). The amount of purified protein was determined by amino acid analysis (17). Estimation of the amount of rPAPP-A2 and rIGFBP-5 when non-purified was done by comparing the responses in Western blotting (using anti-c-myc) with the responses of known molar amounts of purified rIGFBP-5.
To ensure the complete absence of IGF-I or -II in the preparation of rIGFBP-5, affinity-purified material (100 g) was further loaded onto a Superdex 75 (Amersham Pharmacia Biotech), equilibrated, and eluted with 50% formic acid at 0.5 ml/min. At the acidity of the solvent (pH Ͻ 1), any possible bound IGF would dissociate from rIGFBP-5 and elute separately (18,19). Next, IGFBP-5-containing fractions were loaded directly onto a reversed-phase high pressure liquid chromatography (RP-HPLC) column (4 ϫ 250 mm Nucleosil C4 500-7, Macherey-Nagel). A linear gradient was formed from 0.1% (v/v) trifluoroacetic acid (solvent A) and 0.075% (v/v) trifluoroacetic acid in 90% (v/v) acetonitrile (solvent B), increasing the amount of solvent B by 2.3%/min. The column was equilibrated with 5% solvent B, operated at 50°C and with a flow rate of 0.5 ml/min, and the separation was monitored at 226 nm. Bound IGFBP-5 eluted at ϳ42% solvent B. The HPLC step alone was not sufficient as IGF-II (Bachem) eluted close to IGFBP-5 on this column as seen in separate runs. The pure protein was lyophilized, redissolved in solvent A, and further diluted three times in 50 mM Tris, pH 7.5. This material was prepared for experiments to demonstrate that cleavage of IGFBP-5 by PAPP-A2 (using ϳ0.1 g purified PAPP-A2/2 g IGFBP-5) occurs in the absence of IGF and is not promoted by IGF.

RESULTS AND DISCUSSION
We have identified and cloned a novel cDNA sequence encoding a protein of 1791 residues with homology to PAPP-A as detailed under "Experimental Procedures". We name this protein preproPAPP-A2 3 . Alignment of preproPAPP-A2 with pre-proPAPP-A (Fig. 1) demonstrates its similarity to mature PAPP-A; there is no homology between the prepro-peptides. In this alignment, 46% of the residues of mature PAPP-A are also present in PAPP-A2. Like PAPP-A, the amino acid sequence of PAPP-A2 contains three lin-notch motifs (LNR1-3) and five short consensus repeats (SCR1-5) (Fig. 1). All 82 cysteine residues of mature PAPP-A are also found in PAPP-A2, but PAPP-A2 has four additional cysteines not present in PAPP-A. In addition, the zinc binding site and a putative Met-turn are conserved between the two proteins ( Fig. 1), classifying PAPP-A and PAPP-A2 as metzincins (20). Interestingly, the residue immediately following the third histidine of the zinc binding consensus (Val-744 in PAPP-A2) and one residue close to the methionine (Asn-805 in PAPP-A2) are conserved within each of the four recognized metzincin superfamily members: the astacins, the reprolysins, the serralysins, and the matrix metalloproteinases (20). However, PAPP-A and PAPP-A2 do not fit in this pattern, they do not show homology to any members of the four families, and further, the linear distance between the third histidine and the methionine is much longer in PAPP-A and PAPP-A2. PAPP-A and PAPP-A2 are therefore reasonably classified as members of a new, fifth metzincin family, which we tentatively designate the pappalysins.
To obtain recombinant PAPP-A2 (rPAPP-A2), 293T cells were transiently transfected with an expression vector encoding c-myc tagged PAPP-A2 (pPA2-mH). Western blotting of the culture medium showed a single band of ϳ220 kDa (Fig. 2, lane  2), absent in medium from mock transfected cells (Fig. 2, lane  1). Reduction of disulfide bonds did not cause a visible change in band migration (Fig. 2, lane 4). Thus, in contrast to PAPP-A, PAPP-A2 is not a disulfide bound dimer.
Edman degradation, performed on purified rPAPP-A2 (Fig.  2, lane 5) blotted onto a PVDF membrane, revealed the N-FIG. 1. Amino acid sequence of preproPAPP-A2 aligned with preproPAPP-A. The amino acid sequence of preproPAPP-A2 3 (PA2) was aligned with the sequence of preproPAPP-A 2 (PA) using CLUSTALW (29). Because the prepro-part of PAPP-A did not show significant identity with the corresponding region of PAPP-A2, the alignment was manually adjusted to emphasize difference in length of pro-peptides. Arrows indicate the N-terminal residues of the mature proteins as found earlier for PAPP-A (8) (Glu-81), and here for PAPP-A2 (Ser-234). Putative signal peptides, strongly predicted using SignalP V2.0 (30), are shown with lowercase letters. The pro-part of PAPP-A2 contains one other candidate initiation codon corresponding to Met-168, but no signal peptide was predicted following this residue using SignalP. The sequence motifs of PAPP-A (8) are also found in PAPP-A2; the catalytic zinc binding motif (residues 733-743) and residues of the putative Met-turn (around Met-807) are underlined and bolded in both sequences. Lin-notch motifs (LNR1-3) and short consensus repeats (SCR-1-5) are boxed. All cysteines (shaded) of mature PAPP-A are also found in PAPP-A2. In addition, the secreted form of PAPP-A2 has four cysteine residues (Cys-343, Cys-533, Cys-618, and Cys-1268) with no counterpart in PAPP-A.
terminal sequence of the secreted rPAPP-A2 to be 234 SP-PEESNQ, corresponding to cleavage after the four residues RVKK 233 (Fig. 1). This supports the prediction that PAPP-A2 is synthesized as a prepro-protein. The absence of an arginine residue in the P1 position indicates that the proprotein processing enzyme responsible for this cleavage is not furin but likely a different proprotein convertase (21).
Because the two motifs responsible for the metalloproteolytic activity of PAPP-A are conserved in PAPP-A2, we hypothesized that PAPP-A2 is a proteolytic enzyme, and because IGFBP-4 is the only known PAPP-A substrate, we looked for a PAPP-A2 substrate among the six known IGFBPs (Fig. 3). IGFBP-1, -2, -4, and -6 were not cleaved; IGFBP-3 showed limited degradation. However, complete proteolysis of IGFBP-5 is evident from this experiment.
To further analyze the cleavage of IGFBP-5 by PAPP-A2, IGFBP-5 was expressed in mammalian cells. Wild-type rPAPP-A2 (Fig. 4, lane 3), but not a variant mutated in the zinc binding site (E734Q) (Fig. 4, lane 2), efficiently degraded rIGFBP-5. As expected, 1,10-phenanthroline and EDTA but not 3,4-dichloroisocoumarin abolished wild-type rPAPP-A2 activity (Fig. 4, lanes 4 -6). Of interest, in contrast to the proteolysis of IGFBP-4 by PAPP-A (1), cleavage of IGFBP-5 by PAPP-A2 did not require the addition of IGF-I or -II. To verify that cleavage was not induced by traces of IGF bound to IGFBP-5, we further purified this by acidic gel filtration and (RP-HPLC). At low pH, any bound IGF would dissociate from IGFBP-5 (18,19) and elute separately according to their different molecular weights. Using this material, cleavage of IGFBP-5 by purified PAPP-A2 was still observed without added IGF (Fig. 5, lane 2). Interestingly, added IGF-II did not have any apparent effect on the rate of cleavage as seen with shorter incubation resulting in incomplete digestion (Fig. 5, lanes 3 and 4).
For cleavage site determination, purified rIGFBP-5 (Fig. 6, lane 1) was digested with purified PAPP-A2 and analyzed by SDS-PAGE (Fig. 6, lane 2). Edman degradation of blotted material showed that both distinct, visible degradation products (Fig. 6, lane 2) contained the N-terminal sequence 144 KFVGGA 4 . Both of the two distinct bands represent intact C-terminal cleavage fragments because they also contain the C-terminal c-myc tag (Fig. 6, lane 3); they are likely to be differently glycosylated in accordance with the heterogeneity of purified rIGFBP-5 (Fig. 6, lane 1). Both bands contained a second sequence at lower level (45%), 1 LGXFVH, corresponding to the N-terminal sequence of IGFBP-5. The absence of Ser, expected in the third cycle of Edman degradation, was taken as evidence for carbohydrate substitution of Ser-3. O-linked gly- can on the N-terminal cleavage fragment is likely to cause it to smear around the two distinct, C-terminal fragments. Sequence analysis directly on the reaction mixture (Ͼ100 pmol) without SDS-PAGE separation showed only the same two IGFBP-5 sequences in equimolar amounts. Thus, PAPP-A2 cleaves IGFBP-5 at one site between Ser-143 and Lys-144.
Probing a blot with mRNA from several human tissues, previously used for PAPP-A (12), resulted in a signal from the human placenta only (not shown). However, using the blast algorithm (13), a total of 98 human EST sequences were identified that matched the 3Ј untranslated region of the PAPP-A2 mRNA sequence. Of these, 39% originated from placenta, 21% from pregnant uterus, 11% from fetal liver/spleen, and 5% from kidney. Several other tissues were represented but with fewer EST sequences. Hence, like PAPP-A (12), PAPP-A2 expression is neither limited to the placenta nor to pregnancy.
The recent identification of PAPP-A as the IGFBP-4 proteinase in fibroblasts (1), osteoblasts (1,4), ovarian follicular fluid (7), pregnancy serum (6), and vascular smooth muscle cells (5) firmly establishes PAPP-A and IGFBP-4 as an important functional pair in several systems. No other substrate has been found for PAPP-A, and no other proteinase has been shown to cleave IGFBP-4 physiologically. It is therefore very tempting to speculate that the pair of PAPP-A2 and IGFBP-5 plays an analogous role in a number of the tissues mentioned above. Interestingly, incubating IGFBP-5 with smooth muscle cellsconditioned medium resulted in cleavage between Ser-143 and Lys-144 (24), the same cleavage site as found here with purified PAPP-A2. This immediately identifies PAPP-A2 as an obvious candidate IGFBP-5 proteinase for this tissue.
After completion of the experimental work presented here, data base searching revealed that additional genomic sequences (AC027620 and AL139282) and a partial cDNA sequence (AJ278348) had appeared. From none of these, however, can the complete cDNA sequence of PAPP-A2 be deduced.
In conclusion, we have identified, cloned, and expressed a novel protein with homology to PAPP-A, and we have demonstrated the putative proteolytic activity of this protein. We named the protein PAPP-A2 to signify its close relationship with PAPP-A both structurally and functionally. With PAPP-A, PAPP-A2 defines a new, fifth family of the metzincin super-family of metalloproteinases, the pappalysins. Further, we have identified a natural substrate for PAPP-A2, IGFBP-5, analyzed its cleavage, and indicated several tissues where PAPP-A2 may be of physiological relevance.
Note Added in Proof-Following the acceptance of this manuscript we have become aware that the partial sequence of database entry AJ278348 has been presented in a publication (Farr et al., (2000) Biochim. Biophys. Acta 1493, 356 -362). In this paper, residues 147-1791 of PAPP-A2 is named PAPP-E.