Functional structure of the somatomedin B domain of vitronectin

Abstract

The N-terminal somatomedin B domain (SMB) of vitronectin binds PAI-1 and the urokinase receptor with high affinity and regulates tumor cell adhesion and migration. We have shown previously in the crystal structure of the PAI-1/SMB complex that SMB, a peptide of 51 residues, is folded as a compact cysteine knot of four pairs of crossed disulfide bonds. However, the physiological significance of this structure was questioned by other groups, who disputed the disulfide bonding shown in the crystal structure (Cys5–Cys21, Cys9–Cys39, Cys19–Cys32, Cys25–Cys31), notably claiming that the first disulfide is Cys5–Cys9 rather than the Cys5–Cys21 bonding shown in the structure. To test if the claimed Cys5–Cys9 bond does exist in the SMB domain of plasma vitronectin, we purified mouse and rat plasma vitronectin that have a Met (hence cleavable by cyanogen bromide) at residue 14, and also prepared recombinant human SMB variants from insect cells with residues Asn14 or Leu24 mutated to Met. HPLC and mass spectrometry analysis showed that, after cyanogen bromide digestion, all the fragments of the SMB derived from mouse or rat vitronectin or the recombinant SMB mutants are still linked together by disulfides, and the N-terminal peptide (residue 1–14 or 1–24) can only be released when the disulfide bonds are broken. This clearly demonstrates that Cys5 and Cys9 of SMB do not form a disulfide bond in vivo, and together with other structural evidence confirms that the only functional structure of the SMB domain of plasma vitronectin is that seen in its crystallographic complex with PAI-1.

Vitronectin, a 70-kDa adhesive glycoprotein, is abundant in plasma and the extracellular matrix (Tomasini and Mosher 1991; Preissner and Seiffert 1998). The mature human vitronectin molecule of 459 amino acid residues is composed of several functionally unique domains. From the N terminus, these are the SMB domain, the Arg-Gly-Asp motif, a connecting segment, and two hemopexin-like domains. Vitronectin interacts with numerous cellular and extracellular proteins and regulates several biological processes, such as cell adhesion, pericellular proteolysis, tissue invasion, angiogenesis, and metastasis (Dano et al. 1985; Meredith et al. 1996; Stefansson and Lawrence 1996; Loskutoff et al. 1999; Andreasen et al. 2000; Bajou et al. 2001; Chapman and Wei 2001; Look et al. 2002). In particular, vitronectin, through its N-terminal SMB domain, binds to the urokinase receptor on the cell surface, promoting cell adhesion (Blasi 1997; Chapman and Wei 2001). PAI-1 (Gils and Declerck 2004), the primary inhibitor of both tissue- and urinary-type plasminogen activators, can detach cells from the extracellular matrix by competing for the SMB binding site with the urokinase receptor (Seiffert and Loskutoff 1991; Sigurdardottir and Wiman 1994; Deng et al. 1996). In circulation all the active PAI-1 is bound to vitronectin (Wiman et al. 1988), and the binding stabilizes PAI-1's activity (Declerck et al. 1988; Mimuro and Loskutoff 1989; Salonen et al. 1989; Sigurdardottir and Wiman 1990; Seiffert and Loskutoff 1991).

The SMB domain of vitronectin (∼44–50 residues) has also been isolated from human serum as a separate soluble protein (Standker et al. 1996). It contains eight Cys residues arranged in four disulfide bonds. Treatment of the isolated SMB domain with reducing agents abolished its PAI-1 binding, and conversion of any single Cys residue in the SMB domain into alanine destroyed its PAI-1-binding activity (Deng et al. 1996). All eight Cys residues are strictly conserved in the vitronectin molecules from all known species (Fig. 1), and also in other SMB-like proteins, such as plasma cell membrane glycoprotein PC-1 (Buckley et al. 1990) and autotaxin (Murata et al. 1994). Our previous crystallographic studies (Zhou et al. 2003) using recombinant SMB showed that SMB stabilizes PAI-1's activity by restricting the expansion of its β-sheet A and that it binds to PAI-1 through both hydrophobic and ionic interactions (Fig. 1). The identified interface is consistent with previous mutagenesis studies with PAI-1 (Lawrence et al. 1994; Padmanabhan and Sane 1995; Jensen et al. 2002; Schroeck et al. 2002) and SMB (Deng et al. 1996). The structure also reveals that the key interaction of the binding interface is through Arg101 of PAI-1, with the side chain of Arg101 being completely buried within the SMB domain (Fig. 1C). Subsequent mutagenesis studies of PAI-1 showed that mutation of Arg101 to Ala abolished the binding of PAI-1 to vitronectin (Jensen et al. 2004; Xu et al. 2004). Moreover, our structure demonstrated how the SMB domain is folded into a compact cysteine-knot structure by four pairs of crossed disulfide bonds (Cys5–Cys21, Cys9–Cys39, Cys19–Cys32, and Cys25–Cys31, termed here pattern Xtal) (Fig. 1).

Figure 1.

(A) Sequence alignment of SMB homologs from various species and the proposed disulfide patterns of SMB. The eight cysteines and all the residues of the PAI-1/SMB interface are conserved. Human, rabbit, mouse, and rat vitronectin have Met51, which can be cleaved by CNBr. Rat and mouse have two extra Mets of residues 14 and 36. The disulfide bonding pattern Chem1, where the disulfide bonds are linearly arranged, was proposed by Kamikubo et al. (2002) using recombinant SMB. Pattern Chem2 is proposed based on the studies of plasma SMB (Horn et al. 2004). Pattern Xtal is from the crystal structure of the PAI-1/SMB complex (PDB 1OC0) using recombinant SMB (Zhou et al. 2003). According to patterns Chem1 and Chem2, CNBr cleavage of rat or mouse vitronectin would release the fragment (residues 1–14) of 1539.7 Da. (B) Construct for expressing the vitronectin fragment (VN-sc) in insect cells. The human vitronectin fragment (residues 1–132) is downstream from the Bip signal peptide (SP) with the expected SP cleavage site indicated (filled reverse triangle). The recombinant protein is expressed following copper sulfate induction and is secreted into the culture medium. There is a glycosylation site (Asn67, marked in diamonds) located within the connecting region. SMB (residues 1–51) will be released from VN-sc after CNBr cleavage. Two variants of VN-sc were also prepared from insect cells, where Asn14 and Leu24 of SMB were replaced by Met, respectively. Cleavage of these mutants with CNBr would also distinguish the disulfide patterns shown in A. (C) Ribbon diagram showing the compact SMB structure stabilized by four pairs of crossed disulfide bonds (pattern Xtal, PDB 1OC0). The SMB is colored in rainbow from the N terminus (blue) to the C terminus (red). The key interaction of the PAI-1/SMB interface is from Arg101 of PAI-1 with its side chain (in sticks) forming a strong ionic interaction with Asp22 of SMB, which are sandwiched by the aromatic rings of F13 and Y28 of SMB. (D) The disulfide pattern Xtal can also be seen in a mouse SMB-like protein. (E) The NMR structure (PDB 1SSU) with a linear disulfide arrangement (pattern Chem1) shows no stabilizing interactions between the N- and C-terminal halves of the molecule. The NMR structure of plasma-derived SMB (PDB 1S4G) is not compared here as it was refined with a very small number of restraints (Mayasundari et al. 2004). The pictures were prepared in PyMOL (Delano 2002) based on PDB files as indicated.

However, there were controversies about the disulfide arrangements of the SMB domain. Initially Kamikubo et al. (2002) prepared recombinant SMB using a similar Escherichia coli expression system to ours and determined the disulfide pattern by partial reduction and amino acid sequencing. It was suggested that these disulfides are organized in a linear uncrossed pattern (Cys5–Cys9, Cys19–Cys21, Cys25–Cys31, Cys32–Cys39, termed here pattern Chem1) (Fig. 1). Their subsequent NMR study of SMB found that the backbone of the SMB structure is consistent with that of the crystal structure; however, they claimed that the active SMB of vitronectin could have multiple disulfide arrangements including patterns Xtal and Chem1 (Kamikubo et al. 2004). The issue was further complicated by the studies from another group where the SMB domain was derived from plasma vitronectin (Horn et al. 2004; Mayasundari et al. 2004). In combining the results of their chemical and NMR studies, they proposed another disulfide pattern for SMB (Cys5–Cys9, Cys19–Cys31, Cys21–Cys32, Cys25–Cys39, termed pattern Chem2) and a completely different SMB structure. It was suggested by these investigators that the crystal and NMR structures of E. coli-expressed recombinant SMB might represent disulfide scrambled intermediate isomers of the native SMB domain of plasma vitronectin (Horn et al. 2004; Mayasundari et al. 2004). More recently, another group chemically synthesized SMB with three different disulfide patterns and compared their PAI-1-binding affinities. It was found that the SMB of pattern Xtal is, indeed, functional, while other isoforms of SMB are largely inactive (Li et al. 2007).

To distinguish these disulfide patterns and test if multiple disulfide arrangements exist in the SMB domain of vitronectin, vitronectin isolated from rat and mouse was treated with cyanogen bromide (CNBr) and analyzed by HPLC and mass spectrometry. In both mouse and rat vitronectin, there is a Met (cleavable by CNBr) at residue 14, located between the second and third cysteines of the SMB (Fig. 1). If the first two cysteines (Cys5 and Cys9) of the SMB did, as claimed by others, form a pair of disulfide bonds, the fragment of SMB containing residue1–14 should be released upon CNBr cleavage. As a further test, a new eukaryotic expression system (insect cells) was adopted to prepare recombinant vitronectin fragments that contain mutations in the SMB (Asn14Met or Leu24Met). These mutants were expressed and secreted into culture medium by the insect cells and were similarly isolated and treated with CNBr.

Results

CNBr digestion and LC-MS analysis of rat and mouse vitronectins

Mouse vitronectin has a similar molecular weight of ∼70 kDa as human plasma vitronectin, and it has Met14, Met36, and Met51 (Fig. 1A). After CNBr digestion of mouse vitronectin, the disulfide pattern Chem1 or Chem2 should yield two fragments of 1539.7 Da (residues 1–14 with an internal disulfide bond) and 4193.9 Da (residues 15–36 linked with 37–51 of the SMB), respectively. However, LC-MS analysis showed no such fragments. The search for ions corresponding to these two peptides over the whole mass spectrum revealed no peaks. An SMB fragment of 5733.6 Da (m/z at 1434.6 and 1912.7) (Supplemental Fig. S2A), corresponding to the pattern Xtal, was readily identified (Fig. 2A). After reduction with Tris(2-carboxyethyl) phosphine hydrochloride (TCEP), the peptides (a, b, c) of mouse SMB derived from CNBr cleavage were separated by HPLC and detected by mass spectrometry with corresponding masses of 1540.9, 2509.6, and 1690.8 Da, respectively (Table 1; Fig. 2A; Supplemental Fig S2A). Specific modification of Cys by N-ethylmaleimide (NEM) confirmed there were two cysteines in peptide a and one cysteine in peptide c with each modification resulting in a mass increase of ∼125 Da (Table 1). Similarly after CNBr cleavage of rat vitronectin, only the SMB fragment (5761.6 Da) corresponding to the pattern Xtal was observed, and subsequent TCEP reduction released the CNBr-cleaved peptides a (1541.8 Da), c (1690.9 Da), and d (2537.8 Da) of rat SMB (Table 1). These results indicate that in mouse and rat SMB, Cys5 and Cys9 are not paired, which is incompatible with both pattern Chem1 and pattern Chem2 (Fig. 1A).

Figure 2.

CNBr digestion and LC-MS analysis on the analytic HPLC C18 column (Sigma) using a 10%–45% acetonitrile gradient in 0.1% TFA. (A) Mouse SMB/CNBr (the major peak of the top panel) from the HPLC column was collected and reduced with TCEP, and then analyzed by HPLC (lower panel). The samples of human SMB variants (B) N14M and (C) L24M were analyzed under the same conditions. (Top panel) Purified CNBr-treated SMB variants; (middle panel) SMB variants reduced with TCEP; (bottom panel) the reduced variants were alkylated with NEM. The mass of each peptide is shown in Table 1 and Supplemental Figure S2.

Table 1.

Mass spectrometry analysis of SMB after CNBr digestion

Preparation and characterization of recombinant SMB expressed by insect cells

The N-terminal fragment (amino acids 1–132) of human vitronectin, coding the SMB domain and the connecting region (VN-sc), was expressed in insect cells and secreted into the culture medium. It migrates in an SDS gel as a single band of ∼26 kDa under reducing conditions and ∼30 kDa under nonreducing conditions. The identities of these bands were confirmed by Western blot using anti-vitronectin antibody (data not shown). The fragment with an expected mass of 15,613 Da from amino acid sequence, has a measured molecular weight of 16,673 Da by electrospray ionization mass spectrometry. The difference in mass (∼1060 Da) is possibly due to glycosylation at Asn67 (Fig. 1B). A similar 1060-Da mass increase was also observed with both variants (VN-sc-N14M and L24M).

After digestion with CNBr, the recombinant vitronectin fragment was analyzed by LC-MS analysis. As expected, wild-type SMB from VN-sc had a mass of 5762 Da. It had a mass of 5770 Da under reducing conditions and a mass of 6770 Da after subsequent NEM treatment (Table 1). These results confirmed that eight cysteines of SMB formed four pairs of disulfide bonds and could be modified by eight NEM molecules after reduction. The homoserine form of SMB in which the last residue Met was converted to a homoserine instead of homoserine lactone was also observed with a mass of 5780 as expected. The analytical RP-HPLC could readily differentiate peptides with different disulfide connectivities (Wu et al. 2003; Kamikubo et al. 2006). We showed previously that our E. coli-expressed SMB had the same retention time as plasma SMB on HPLC columns (Zhou et al. 2003). Here we further compared the recombinant SMB expressed from either E. coli or insect cells and their samples spiked with plasma SMB on a HPLC column. The elution traces (Supplemental Fig. S1) demonstrated that the recombinant SMBs from different expression systems were the same as plasma-derived SMB. Thus the discrepancy of the disulfide pattern of SMB is unlikely to be due to the sources of human SMB.

CNBr digestion and LC-MS analysis of SMB variants (N14M and L24M)

Two recombinant variants (VN-sc-N14M and VN-sc-L24M) in which Asn14 and Leu24 of SMB were replaced by Met, respectively, were also prepared from insect cells. VN-sc-N14M was digested with CNBr and then analyzed by LC-MS as described above. There was only one SMB fragment observed with a mass of 5750 Da (Fig. 2B). Two peptides corresponding to residues 1–14 (peptide e, 1542.5 Da) and 15–51 (peptide f, 4215.9 Da) of SMB could only be identified after reduction with TCEP. After subsequent NEM modification, the mass increases confirmed there were two and six cysteines in the corresponding peptides (Table 1; Supplemental Fig. S2B). This, again, indicates that Cys5 and Cys9 are not paired in human SMB.

Similarly, both peptides of SMB-L24M from CNBr cleavage were found still linked together with a mass of 5751.2 Da (Fig. 2C). Subsequent TCEP treatment released these two peptides (g, 2706.0 Da and h, 3052.9 Da). They can be modified by four NEM molecules with a mass of 3026.3 Da and 3551.9 Da, respectively (Table 1; Fig. 2; Supplemental Fig. S2C). This result further excludes the linear disulfide bonding arrangement of pattern Chem1.

Structures of the SMB domain of vitronectin and its homolog

Here we compared the structures of SMB from crystallographic and NMR studies. As shown in Figure 1C, the four pairs of disulfide bonds of the crystal structure (Zhou et al. 2003) are required to hold the SMB in a compact scaffold (PDB 1OC0). The critical interaction of Arg101 from the PAI-1/SMB-binding interface is shown in sticks. This compact structure and the disulfide pattern were confirmed by the NMR study of a mouse SMB-like protein (PDB 2CQW). This protein has a relatively low sequence identity with human SMB (Fig. 1A), but it has the eight conserved cysteines, and its disulfide linkages are identical to pattern Xtal (Fig. 1D). The NMR structure of SMB (PDB 1SSU) proposed by Kamikubo et al. (2006) has a similar backbone trace as that of crystal structure; however, this fold is unlikely to be stable with their linear disulfide arrangement as there is no stabilizing interaction (either hydrophobic or hydrogen bonding) between the N- and the C-terminal half of the peptide (Fig. 1E).

Discussion

The crystallographic structure of the SMB domain (Zhou et al. 2003) shows a clear and unequivocal presence of four disulfide linkages (Cys5–Cys21, Cys9–Cys39, Cys19–Cys32, Cys25–Cys31). The validity of this disulfide bonding was challenged by others, who claimed in particular that the first linkage was formed by Cys5–Cys9 rather than the Cys5–Cys21 linkage observed in the crystal structure (Horn et al. 2004; Kamikubo et al. 2004, 2006; Mayasundari et al. 2004). Their results were obtained by a partial reduction process and subsequent analysis of the alkylated SMB isoforms by direct amino acid sequencing (Kamikubo et al. 2002) and mass spectrometry analysis (Horn et al. 2004). However, such approaches are susceptible to disulfide scrambling and the inadequately purified isoforms, especially so with Cys-rich proteins, where the cysteines are clustered together, as seen here with SMB with eight cysteines within its first 39 residues (Fig. 1A). The CNBr cleavage of the mouse and rat SMB as well as that of Met-containing recombinants shown here contraindicates the claimed Cys5–Cys9 linkage, thus excluding both disulfide patterns proposed by chemical methods. The finding together with many previous biochemical studies of the PAI-1/vitronectin complexes (Deng et al. 1996; Schroeck et al. 2002; Jensen et al. 2004; Xu et al. 2004) makes it highly probable that the structure of SMB and its disulfide linkages, as seen in its complex with PAI-1, does, indeed, represent the only functionally relevant structure of the domain in vivo. Independent support for this comes from the synthesis studies of the SMB (Li et al. 2007), and further confirmation is from the NMR structure of a mouse SMB-like protein (PDB 2CQW), which has <30% amino acid sequence identity (Fig. 1A) with human SMB but identical disulfide linkages (Fig. 1D).

Over the last 10 years, an enormous amount of research has shown that the interplay of vitronectin with uPAR, integrin, and PAI-1 plays a crucial role in orchestrating tumor cell migration (Loskutoff et al. 1999; Chapman and Wei 2001; Blasi and Carmeliet 2002). Our published crystal structure of SMB complexed with PAI-1, further supported here, not only provides the binding interface for targeting vitronectin's interactions, but also gives a perfect lead (mimetics of Arg101) in searching agents to block vitronectin's binding as free arginine does, indeed, dissociate the PAI-1/vitronectin complex (Sigurdardottir and Wiman 1992; Seiffert and Wagner 1997; Ragno 2006).

Materials and Methods

Preparation of plasma vitronectin-derived SMB peptide (plasma SMB)

Plasma human, mouse, and rat vitronectin was purified in the presence of NEM using a modified urea method as previously described (Yatohgo et al. 1988; Bittorf et al. 1993). Mouse plasma vitronectin was also purchased from Molecular Innovations. To prepare human plasma SMB (residues 1–51 of vitronectin), vitronectin (30 mg) was dissolved in 0.1 M hydrochloride with 1000-molar excess CNBr (∼5 M in acetonitrile; Aldrich). The mixture was kept at dark for ∼16 h and then loaded directly onto a C18 HPLC column (Prodigy ODS3, 25 × 200 mm; Phenomenex). The peptides were eluted with an acetonitrile gradient (20%–40%) in 0.1% trifluoroacetic acid (TFA). Fractions containing SMB were pooled and lyophilized. There were two peaks of SMB from the HPLC, one corresponding to SMB (lactone form) with Met51 converted to homoserine lactone and a minor peak eluted earlier corresponding to SMB (homoserine form) with Met51 converted to a homoserine. Both peaks were collected, and the final yield of plasma SMB is ∼1.5 mg from 30 mg of plasma vitronectin. The recombinant SMB was prepared from an E. coli expression system as previously described (Zhou et al. 2003).

Expression and purification of recombinant vitronectin fragments in insect cells

The DNA coding sequence for human vitronectin fragment (amino acids 1–132, containing the SMB domain and the Connecting region, termed VN-sc), together with a Met at the N terminus and a six-histidine tag at the C terminus, was amplified by PCR and cloned into a Drosophila expression vector, pMT/Bip/V5-his-A (Invitrogen), by restriction enzyme digestion sites of BglII and XhoI. This fragment was placed downstream from the insect Bip gene secretion signal sequence and could be secreted into the culture medium after expression (Fig. 1). Two variants of VN-sc, in which the Asn14 and Leu24 of SMB were mutated to Met, respectively, were generated by a Quikchange mutagenesis kit (Stratagene). The expression plasmid (pMT-VN-sc) was cotransfected into Drosophila Schneider S2 cells with the selection plasmid pCoBlast (Invitrogen), which coded a Blasticidin-resistant gene, using Insect Genejuice transfection reagents (Merck Biosciences). After selection with 25 μg/mL Blasticidin S in Drosophila serum-free medium, the stable cell lines were cultured in suspension in 1-L Erlenmeyer flasks, and the expression of VN-sc was induced by adding 0.5 mM copper sulfate into the medium. The medium was collected after ∼6 d of induction and mixed with an equal volume of 20 mM Tris-HCl (pH 7.4), 5 mM EDTA, and loaded onto a Hitrap Q column. Bound proteins were eluted with a NaCl gradient of 0–0.5 M in 10 mM Tris-HCl (pH 7.4). Fractions containing VN-sc were identified by Western blot using polyclonal anti-vitronectin antibody, and loaded directly onto a 5-mL HisTrap column (GE Biosciences). The VN-sc was eluted by an imidazole gradient (20–200 mM imidazole in 20 mM phosphate buffer at pH 7.8, 0.5 M NaCl) and was further purified by a gel filtration column (Suphedex 200, 16 × 600 mm; GE Biosciences) in 10 mM Tris-HCl, 0.15 M NaCl (pH 7.4). VN-sc-N14M and VN-sc-L24M were purified using the same procedure.

CNBr digestion and liquid chromatography-mass spectrometry (LC-MS) analysis

VN-sc or its variants (∼0.75 mg) was diluted in 0.5 mL of 0.1 M hydrochloride containing a 1000-fold excess of CNBr and incubated for ∼16 h at room temperature. The sample was then lyophilized and dissolved in 500 μL of 20 mM phosphate buffer, 0.5 M NaCl, and 20 mM imidazole (pH 7.8). Subsequently 100 μL of Ni Sepharose slurry (GE Biosciences) was added to adsorb the peptides containing the His-tag. The supernatant was collected for subsequent analysis without further purification. Rat or mouse vitronectin (200 μg) was digested with CNBr as above, lyophilized, and then resuspended in the 100 μL of water. For online LC-MS analyses, the HPLC system was coupled to a ThermoFinnigan LCQ ion-trap mass spectrometer. Samples were loaded onto an analytic Discovery C18 HPLC column (4.6 mm × 250 mm; Supelco). Bound proteins were eluted by an acetonitrile gradient of 10%–45% in 0.1% TFA over 30 min at 1 mL/min. Mass spectra were acquired over the mass-over-charge (m/z) range of 600–2000. To reduce all the disulfide bonds and modify the cysteines of SMB, SMB was incubated with 4 M guanidine chloride, 20 mM TCEP, and 50 mM NEM in 0.1 M Tris-HCl buffer for 30 min at 37°C before LC-MS analysis. To compare the behavior of the recombinant SMB (E. coli-expressed, or insect cell-expressed) with that of plasma on the same HPLC column, recombinant SMB was spiked with plasma SMB and analyzed by LC-MS. The detailed HPLC analysis of SMB from various sources (Supplemental Fig. S1) and the electrospray mass spectrum analysis of peptides derived from SMB (Supplemental Fig. S2) are included in the Supplemental material.