Plasmin-mediated release of the guidance molecule F-spondin from the extracellular matrix.

Serine proteases are implicated in a variety of processes during neurogenesis, including cell migration, axon outgrowth, and synapse elimination. Tissue-type plasminogen activator and urokinase-type activator are expressed in the floor plate during embryonic development. F-spondin, a gene also expressed in the floor plate, encodes a secreted, extracellular matrix-attached protein that promotes outgrowth of commissural axons and inhibits outgrowth of motor axons. F-spondin is processed in vivo to yield an amino half protein that contains regions of homology to reelin and mindin, and a carboxyl half protein that contains either six or four thrombospondin type I repeats (TSRs). We have tested F-spondin to see whether it is subjected to processing by plasmin and to determine whether the processing modulates its biological activity. Plasmin cleaves F-spondin at its carboxyl terminus. By using nested deletion proteins and mutating potential plasmin cleavage sites, we have identified two cleavage sites, the first between the fifth and sixth TSRs, and the second at the fifth TSR. Analysis of the extracellular matrix (ECM) attachment properties of the TSRs revealed that the fifth and sixth TSRs bind to the ECM, but repeats 1-4 do not. Structural functional experiments revealed that two basic motives are required to elicit binding of TSR module to the ECM. We demonstrate further that plasmin releases the ECM-bound F-spondin protein.

environment, enabling growth through an impeding substrate (2). It was anticipated by Krystosek and Seeds (3) that release of extracellular proteases by the axonal growth cone may facilitate its movement by digesting cell-cell and cell-matrix contacts that block the path of the advancing growth cone. Over the past several years, it became evident that extracellular serine proteases, such as plasminogen, tissue-type plasminogen activator (tPA), 1 urokinase-type plasminogen activator (uPA) (for review, see Ref. 4), thrombin (5), and neurotrypsin (6), are expressed in the nervous system. They have been implicated in a variety of processes during neurogenesis, including cell migration, axon outgrowth, and synapse elimination (7,8). They also play a critical role in the adult nervous system by mediating neuronal plasticity (9,10), apoptosis (11), and peripheral nerve regeneration (12).
Localization of plasmin activity to neuronal growth cones was initially demonstrated by digestion of a fibrin clot overlay (3). It was further demonstrated that plasmin cleaves the ECM molecules: collagen, fibronectin (13) and laminin (for review, see Ref. 4). In addition to cleaving ECM molecules directly, the extracellular serine proteases may act indirectly by releasing latent proteases and growth factors from the matrix. It was demonstrated that metalloproteases (14), transforming growth factor-␤ (15), vascular endothelial growth factor (16), fibroblast growth factor (17), platelet-derived growth factor (18), and hepatic growth factor/scatter factor (15) are produced as matrix-attached latent proteins, subjected to cleavage and subsequently to activation by plasmin.
F-spondin, a gene expressed in the floor plate, encodes a secreted, ECM-attached protein (19). It plays a dual role in patterning axonal trajectory in the spinal cord by promoting outgrowth of commissural axons (20) and inhibiting outgrowth of motor axons (21). F-spondin protein is processed in vivo to yield an amino half protein, which contains regions of homology to reelin and mindin, and a carboxyl half protein, which contains either six or four thrombospondin type I repeats (TSRs) (20,22). F-spondin expression in the nervous system overlaps with expression of several serine proteases. In the floor plate, F-spondin is expressed together with tPA and uPA (19,(22)(23)(24), whereas in the hippocampus, F-spondin is coexpressed with tPA and neurotrypsin (6,25). In addition, several tPA-expressing neurons extend axons toward or through an F-spondin-rich milieu. Embryonic motor neurons are exposed to the floor plate-derived F-spondin (21). As motor axons emerge from the spinal cord and up-regulate the expression of tPA (26), they encounter the somite-derived F-spondin (27); subsequently, at the peripheral nerve, they are ensheathed by Schwann cellexpressing F-spondin (22). Similarly, embryonic sensory neurons and sympathetic ganglia neurons expressing tPA are also surrounded by F-spondin-expressing cells in the ganglia and along their axonal path (22,26,28).
In the current study we provide evidence demonstrating that F-spondin is a substrate for plasmin. Plasmin cleaves F-spondin at two sites, the first located between the fifth and sixth TSRs and the second at the fifth TSR. The cleavage sites are located between the extracellular matrix binding TSRs (repeats 5 and 6) and the nonbinding repeats (repeats [1][2][3][4]. In accordance, treatment of F-spondin with plasmin yields a diffusible, ECM-free, TSR domain protein containing TSRs 1-4.

EXPERIMENTAL PROCEDURES
DNA Constructs-DNA plasmids were constructed by PCR as indicated in Table I. Forward and backward primers were used for PCR using a "template plasmid." The PCR products were subcloned into a suitable plasmid (cloning vector) into the restriction sites indicated in the table. The mutant plasmids PL1m, PL2m, and DM were generated as follows. Two PCRs were set up with two sets of primers (the upper and lower row of primers in the table). The PCR products of the two reactions were combined, and an additional PCR was performed with the forward primer of the upper row and the backward primer of the lower row. The PCR products were digested and subcloned as indicated in the table.
Plasmin Cleavage Assay-HEK293 T cells were transfected with the various plasmids using the liposome-mediated transfection reagent DOTAP (Roche, Manheim, Germany), and LipofectAMINE (Life Technologies, Inc.). Conditioned medium was collected after 2-4 days and treated with the appropriate reagents at 37°C for 1 h. For plasmin cleavage assays, conditioned medium was treated with plasmin (Chromogenix, Sweden) at the indicated concentrations and the chromogenic substrate specific for plasmin, S-2251 (Val-Leu-Lys-p-nitroanilide, Chromogenix) in 100 mM Tris-HCl, pH 7.4, in a final volume of 100 l, in microtiter plates for 1 h at 37°C. Plasmin activity was measured by monitoring the increase of absorbance at 405 nm, using a Thermomax thermostat plate reader (Molecular Devices Corp.). In other cases, plasmin was generated from its zymogen Glu-plasminogen, 10 g/ml (Chromogenix), and activated by 100 pM recombinant single chain tPA (Actylase, Roche, Manheim) using 20 g/ml fibrin (Chromogenix) as a cofactor. In some cases, an inhibitor to serine proteases, such as aprotinin (Sigma), was used at a final concentration of 10 g/ml, or an inhibitor to metalloproteinases such as EGTA (Merck) at a final concentration of 20 mM was added to the reaction mixture.
Preparation of ECM-coated Dishes-Bovine corneal endothelial cells (second to fifth passages) were plated in 35-mm tissue culture dishes at an initial density of 2 ϫ 10 5 cells/ml and cultured as described above, except that 4% Dextran T-40 was included in the growth medium. Na 2 35 SO 4 (25 Ci/ml) (Amersham Pharmacia Biotech) was added on days 2 and 5 after seeding, and the cultures were incubated with the label without medium change. On day 12, the subendothelial ECM was exposed by dissolving the cell layer with phosphate-buffered saline containing 0.5% Triton X-100 and 20 mM NH 4 OH followed by four washes with phosphate-buffered saline. The ECM remained intact, free of cellular debris and firmly attached to the entire area of the tissue culture dish. Nearly 80% of the ECM radioactivity was incorporated into heparin sulfate proteoglycans.
Binding of the TSRs to ECM-ECM plates were prepared as described above. Conditioned medium of transfected HEK293 ⌻ cells was incubated on the ECM for 90 min at room temperature. The plates were washed for 5 min three times with HABA (Hanks' balanced salt solution, 0.5 mg/ml BSA, 0.1% NaN 3 , 20 mM Hepes, pH 7.0), 5 min twice with phosphate-buffered saline, and 2 min with AP buffer (29). Bound alkaline phosphatase was detected by incubation with p-nitrophenyl phosphate for 2 h at room temperature.
Binding of Plasmin-treated TSRs to ECM-TSR2-6 conditioned medium was treated with 50 g/ml plasmin for 1 h at 37°C. The reactions were stopped by adding 10 g/ml aprotinin, then incubated in 96-wells plates covered with ECM, for 90 min. Wells were washed with phosphate-buffered saline, blocked with 1% bovine serum albumin, and then incubated with 9E10 mAb overnight at 4 0 c, followed by incubating with horseradish peroxidase-conjugated anti-mouse secondary antibody. Proteins bound to the ECM were detected using the colorimetric reagent 3,3Ј,5,5Ј-tetramethylbenzidine (Chemicon).

RESULTS
The Carboxyl Terminus of F-spondin Is Cleaved by Plasmin-To test whether F-spondin is cleaved by tPA, we incubated conditioned medium of carboxyl-terminal Myc-tagged Fspondin (Fig. 1A) together with elements of the tPA proteolysis complex. Incubation of F-spondin with plasmin (the tPA-activated plasminogen) (Fig. 1B, lane 2) or with all of the proteolysis components: plasminogen, fibrin, and tPA (Fig. 1B, lane 8), caused elimination of the 110-kDa protein (Fig. 1B, lane 1). Incubation with the isolated components: Fb (Fig. 1B, lane 5), tPA (Fig. 1B, lane 6), tPA and plasminogen (Fig. 1B, lane 7), and tPA and fibrin (Fig. 1B, lane 9), did not result in cleavage of F-spondin. On the other hand, a combination of plasminogen and fibrin was active in the processing (Fig. 1B, lane 4). This activity is probably mediated by the tPA that is produced by HEK293 ⌻ cells (30,31).
To test the specificity of the cleavage, the serine protease inhibitor aprotinin and the metalloproteinase inhibitor EGTA were added. Addition of aprotinin abolished the plasmin-mediated cleavage of F-spondin (Fig. 1C, lanes 3 and 9), whereas addition of EGTA (Fig. 1C, lanes 4 and 10) did not inhibit the degradation of F-spondin (compare with the non-inhibited plasmin-treated samples (Fig. 1C, lanes 2 and 8)). Thus, F-spondin is cleaved specifically by plasmin at its carboxyl terminus.
Plasmin Cleaves F-spondin Carboxyl at the Fourth TSR Domain-To map the plasmin cleavage sites we generated nested deletion constructs. F-spondin expression constructs containing the reelin/spondin domain and the reelin/spondin plus nested TSRs ( Fig. 2A) were transfected into HEK293 ⌻ cells. The conditioned medium was subjected to plasmin proteolysis, and the protein products were analyzed by the anti-reelin domain antibody R8. Except for TSR5a, the size of all the plasmin-treated proteins was unchanged. TS5a protein, however, was reduced, and thus the treated protein migrated faster than its untreated counterpart. This suggests that the cleavage site is carboxyl to the fourth TSR.
To visualize the two protein cleavage products, an anti-TSR domain antibody (R2) was used. Proteins containing TSRs 1-6 and TSRs 2-6 ( Fig. 2B) were subjected to plasmin treatment. The 55-kDa TSR1-6 protein yielded a 40-kDa protein and a 16-kDa protein. The 46-kDa TSR2-6 protein yielded a 28-kDa and a 16-kDa protein (Fig. 2B) The 16-kDa proteolytic protein migrated to the same extent in both proteins, suggesting that the cleavage site in these proteins is identical. It is difficult to ascertain the precise cleavage site by the molecular masses of the plasmin proteolytic fragments because an N-linked glycosylation site is present at amino acids 681-683, within the fifth TSR.
A Highly Plasmin-sensitive Site Is Located between the Fifth and Sixth TSRs-To pinpoint the plasmin cleavage site, nested deletions of 5 amino acids each, covering the 30-amino acid region that interspaced repeats 5 and 6, were generated. All of the proteins were equipped with the Myc epitope at the carboxyl terminus. The proteins were subjected to plasmin digestion and analyzed with anti-Myc mAb, to visualize the cleaved product, and anti-TSR antibody to detect the unprocessed and processed protein. TSR1-5ϩ5 was resistant to 10 g/ml plasmin, whereas TSR1-5ϩ10, TSR1-5ϩ15, and TSR1-5ϩ20 were sensitive to plasmin (Fig. 3, A and B). Thus, a sensitive site (designated PL1) to plasmin is located carboxyl to the fifth TSR. It is plausible that another site is located amino to the PL1 site, within the fifth TSR. To test this hypothesis we incubated the TSR1-5ϩ5 and TSR1-5ϩ20 with increasing con- Xba-Xba centrations of plasmin. TSR1-5ϩ20 was cleaved at 10 g/ml, and TSR1-5ϩ5 was fully cleaved at 50 g/ml plasmin (Fig. 3C).
The size products of the fully cleaved proteins were identical.
Hence, it appears that a second site (PL2) is located amino to PL1. In the absence of PL1, the PL2 site is less sensitive to plasmin. We assume that an intermediate protein product appears with the TSR1-5ϩ20 protein. Presumably, at 10 g/ml plasmin, only the PL1 and not the PL2 site is digested. The fact that no intermediate size protein is detected suggests that the PL2 site renders high sensitivity after the PL1 site is cleaved. Plasmin cleaves after arginine or lysine. Two arginines are located at amino acids 730 and 732, between TSR1-5ϩ5 and TSR1-5ϩ10. Thus, these two arginines are potential cleavage sites for plasmin. The two arginines were mutated to proline and serine (Fig. 4A). The mutated protein, designated PL1m, was analyzed with 9E10 and R2 antibodies (Fig. 4, A and B).
The mutated protein was resistant to low concentrations of plasmin (Fig. 4B). At higher concentrations, 50 g/ml, a cleaved product was apparent (Fig. 4, C and D). The size of the cleaved product was identical to the size of the cleaved control protein TSR1-5a. Thus, mutating the arginines at positions 730 -732 created a PL1-resistant protein.
A Second Plasmin Cleavage Site Is Located within the Fifth TSR-To locate the PL2 site, a similar approach was taken. Nested deletion proteins, of 5-amino acid intervals, in the fifth TSR, between TSR4ϩ5 and TSR4ϩ30 were generated. TSR4ϩ5 (Fig. 5A) TSR4ϩ10 and TSR4ϩ15 (data not shown) were resistant to plasmin at concentrations ranging from 10 to 100 g/ml, as assessd by the anti-TSR antibody R2, and the anticarboxyl end 9E10. TSR1-4ϩ30 (Fig. 5B) and TSR1-4ϩ20 (Fig. 5C) were partially cleaved by 10 g/ml and fully cleaved by 50 -100 g/ml. Thus, the PL2 site is located between TS4ϩ15 and TS4ϩ20. There are two lysines between the TSR1-4ϩ15 and TSR1-4ϩ20 at positions 682 and 686. The lysines were mutated to glycine and glutamic acid to generate Partial cleavage is also obtained by plasminogen ϩ fibrin (lane 4). Panel C, the plasmin-mediated cleavage of F-spondin is inhibited by plasmin-specific inhibitor. Inhibitors to either plasmin (aprotinin) or to metalloproteinase (EGTA) were added to the plasmin system. The protein products were analyzed by Western blotting with the 9E10 mAb. Aprotinin fully blocks F-spondin cleavage (lanes 3 and 9), whereas EGTA does not inhibit its cleavage (lanes 4 and 10). PL2m (Fig. 6A). A double mutant, DM, containing the PL1m and PL2m was also constructed. Analysis of the sensitivity of the proteins to plasmin with the 9E10 antibody, revealed that the carboxyl end of the wild type, TSR1-5a protein and the mutant PL2m protein were cleaved at low concentrations of plasmin. The PL1m and the DM were both resistant to low plasmin, with the DM protein being more resistant to higher concentrations of plasmin than PL1m (Fig. 6B). The high sensitivity of PL2m, as judged by the elimination of the carboxyl end, is caused by the presence of a PL1-sensitive site in this protein. Analysis of the cleaved product with the R2 antibody revealed that there is a hierarchy in the appearance of the cleaved protein. The wild type protein is the most sensitive, followed by the PL2m, PL1m with the DM being the most resistant protein (Fig. 6, C and D). The super-resistance of the DM demonstrates that mutating the lysines at positions 682 and 686 reduced the plasmin sensitivity of F-spondin. Nevertheless, the DM protein is not completely resistant. There are two arginines at positions 693 and 695 (Fig. 6A). Mutating those two sites did not change the plasmin sensitivity of Fspondin (data not shown). It is conceivable that mutating the 682 and 686 sites changed the conformation of the protein and exposed arginines 693 and 695 to plasmin.
ECM Binding Properties of the TSRs-We have shown previously that the TSR domain of F-spondin binds to the ECM (19,25). To examine whether the processing of F-spondin by plasmin modulates its interaction with the ECM, we tested the binding properties of the TSRs. Fusion proteins containing an alkaline phosphatase (AP) fused to various combinations of TSRs were generated (Fig. 7A). The conditioned media of HEK293 ⌻ transfected cells were incubated with tissue culture plates coated with bovine corneal endothelial cell ECM. The amount of bound protein was measured by an AP colorimetric reaction. All of the fusion proteins that contained TSR 5 or 6 or FIG. 3. A highly plasmin-sensitive site is located between the fifth and the sixth TSRs. Plasmids containing TSRs 1-5 plus nested deletions (5-amino acid intervals) of the 20 amino acids from the "a" region were generated. All of the proteins were equipped with the Myc epitope at their carboxyl terminus. The conditioned media of HEK293 ⌻ transfected cells were subjected to proteolysis with 10 g/ml plasmin (Plm) and analyzed with the 9E10 (panel A) and R2 (panel B) antibodies. Panel A, analysis with 9E10 antibody. Only the uncleaved protein is detected with the anti-Myc epitope antibody. TSR1-5ϩ15 and ϩ20 disappear after plasmin digestion. Most of TSR1-5ϩ10 protein disappears as well. Only the carboxyl terminus of TSR1-5ϩ5 is retained after plasmin treatment. Panel B, analysis of the panel A blot with R2 antibodies. The conditioned media of TSR1-5ϩ10, ϩ15, and ϩ20 contain two bands. The upper band represents the unprocessed protein, and the lower band represents the cleaved product that results from the HEK293 ⌻ cell endogenous tPA. Adding plasmin fully converts the unprocessed proteins to processed. The TS1-5ϩ5 is fully resistant. Panel C, TSR1-5ϩ5 and ϩ20 were incubated with increasing concentrations of plasmin. TSR1-5ϩ20 is cleaved at 10 g/ml, whereas TSR1-5ϩ5 is cleaved at 50 g/ml. Analysis was performed with the R2 antibody. The mutations were generated in the background of a plasmid containing the 1-5 TSRs and the 30 amino acids of the "a" region, followed by the Myc epitope -TSR1-5a. Panel B, analysis with the 9E10 antibody. The wild type protein TSR1-5a is cleaved at 10 g/ml plasmin (Plm). The mutated protein PL1m is detected even after incubation with 50 g/ml plasmin, even though the intensity of the band does decrease. Panel C, analysis of A with the R2 antibody. A cleaved product is detected in the conditioned media of TS1-5a. A similar size of processed fragment is also detected at elevated concentrations of plasmin, in the mutated PL1m protein. The amount of the cleaved PL1m protein is lower than the uncleaved PL1m, even at 50 g/ml plasmin. Panel D, a densitometry of the bands in panel C was performed using NIH Image software. The ratio between the intensities of uncleaved versus cleaved proteins was plotted in a logarithmic scale (y) as a function of plasmin concentrations (x). both (AP-TSR1-6, AP-TSR1-5, AP-TSR5, AP-TSR6, and AP-TSR5-6) bound to the ECM (Fig. 7B). Proteins restricted to repeats 1-4 (AP-TSR1, AP-TSR1-2, AP-TSR1-4) did not bind to the ECM (Fig. 7B). The fifth TSR is more adhesive than the sixth. The binding levels of AP-TSR1-5 and AP-TSR1-6 are lower than the isolated fifth and sixth TSRs. This suggests that repeats 1-4 might reduce the affinity of repeats 5 and 6 in the context of the nonprocessed protein. Nevertheless, even binding the entire TSR domain is significantly greater than the processed 1-4 TSR protein. Thus, repeats 5 and 6 are required and sufficient for binding to the ECM.
What are the distinctive properties of repeats 5 and 6 which enable ECM binding? We have shown that F-spondin ECM binding is blocked by heparin sulfate and chondroitin sulfate (19,25). This suggests that the binding of F-spondin is mediated by proteoglycans. Potential proteoglycan binding sites, BBXB (32), are present in all six TSRs (Fig. 7C). It was shown that each of the two TSRs of the heparin-binding protein HB-GAM has a ␤-sheet structure composed of three antiparallel ␤-strands (33). In F-spondin, repeats 1-4 contain one stretch of basic amino acids, a potential binding site in the second antiparallel ␤-strand, whereas the fifth and sixth repeats have two potential proteoglycan binding sites in the second and third antiparallel ␤-strands (Fig. 7C). A hypothetical model that accounts for the different binding properties of the TSRs is that there is a requirement for two basic domains in the second and third antiparallel ␤-strands to facilitate binding to the ECM. To test this hypothesis we replaced the third antiparallel ␤-strand of repeats 4 and 5 generating the (4-4-5) chimeric repeat 4 (AP-TSR4&5) and the reciprocal (5-5-4) chimeric repeat 5 (AP-TSR5&4) (Fig. 7A). The chimeric repeat 4 bound to the ECM, but the chimeric repeat 5 did not (Fig. 7D). This demonstrates that the third basic antiparallel ␤-strand of repeat 5 is required to elicit ECM binding of the fifth TSR and sufficient to confer binding properties to the fourth TSR. This supports the hypothesis that two basic motifs are required to enhance the binding of the F-spondin TSRs to the proteoglycans.
Plasmin Releases F-spondin from the ECM-The cleavage of F-spondin at the PL1 site and subsequently at the PL2 site should generate an ECM-free TSR protein (composed of repeats 1-4 and the first antiparallel ␤-strand of repeat 5). To assess this theory, we tested whether plasmin generates an ECM-free protein. Because alkaline phosphatase was found to be degraded by plasmin (data not shown), we generated an amino Myc-tagged TSR protein. Four copies of the Myc epitope were cloned upstream of TSRs 2-6, to generate TSR2-6 protein (Fig.   8A). The protein was preincubated with plasmin and subsequently plated on ECM. The protein was detected by an anti-Myc antibody followed by secondary horseradish peroxidaseconjugated antibody and colorimetric horseradish peroxidase reagent. Pretreatment of the TSR2-6 protein with plasmin significantly reduced the binding to ECM compared with the uncleaved untreated protein (Fig. 8B).
To study whether plasmin releases F-spondin from the ECM we performed the converse experiment. To circumvent the nonspecific digest of ECM component (which might anchor F-spondin to the ECM) by plasmin (34), we have sensitized the assay by using limited amounts of plasmin and a higher plasmin-sensitive form of F-spondin. Because the PL2 site at the fifth TSR domain is less sensitive to plasmin than the PL1 site, we generated a TSR domain protein, TSR1-6⌬5, with a deletion of the fifth TSR (and subsequently deletion of PL2 site).

FIG. 5. A second plasmin site is located within the fifth TSRs.
Plasmids (Plm) containing TSRs 1-4 plus 5 (TSR1-4ϩ5), 20 (TSR1-4ϩ20), or 30 (TSR1-4ϩ30) amino acids from the fifth TSR were generated. The plasmids also include the Myc epitope at the carboxyl end. The conditioned media of transfected HEK293 ⌻ cells were digested with plasmin and analyzed by Western blotting with the 9E10 mAb (panels A and C) and the R2 antibody (panel B). Panel A, analysis with the 9E10 antibody of TSR1-4ϩ5 and TSR1-4ϩ30. TSR1-4ϩ30 is cleaved at 50 g/ml, whereas TS1-4ϩ5 is not. Panel B, analysis of panel A with R2 antibody. TSR1-4ϩ30 is cleaved at 50 g/ml, whereas TSR1-4ϩ5 is not. Panel C, analysis of TS1-4ϩ20 with 9E10 antibody. TSR1-4ϩ20 is partially cleaved at 10 g/ml and fully at 100 g/ml plasmin. PL2m and DM were constructed in the backbone of TSR1-5a; hence, they have the Myc epitope at the carboxyl end. Panel B, analysis of plasmin (Plm) digestion using the 9E10 antibody. TSR1-5a and PL2m are sensitive to 10 g/ml, whereas PL1m and DM are resistant to 10 and 20 g/ml plasmin. A slight decrease of the band intensity is apparent with the 50 g/ml plasmin. Panel C, analysis of panel A with the R2 antibody. The wild type protein is the most sensitive followed by PL2m, PL1m, and DM being the most resistant protein. Panel D, a densitometry of the bands in panel C was performed using NIH Image software. The ratio between the intensities of uncleaved to cleaved proteins was plotted in a logarithmic scale (y) as a function of plasmin concentrations (x).
Treatments with increasing concentrations of plasmin yield a gradual decrease of the ECM-bound TSR1-6⌬5 protein (Fig.  8C). At 40 g/ml about 70% of the bound protein was released. Western analysis of the released fraction demonstrates that only the cleaved protein, but not the uncleaved, is released from the ECM after plasmin cleavage (Fig. 8D). Thus, plasmin is generating a diffusible, ECM-free TSR-domain protein.

DISCUSSION
In this work we have identified plasmin as the protease that cleaves F-spondin at the carboxyl terminus. Two cleavage sites have been identified, each with a different sensitivity to plasmin. The cleavage at the more sensitive PL1 site renders high sensitivity to the less sensitive site, PL2. The PL2 site delaminates between the ECM binding modules of F-spondin, repeats 5 and 6, and the nonbinding modules, repeats 1-4. Thus, plasmin activity mediates release of a protein domain, containing TSRs 1-4, from the ECM.
Which Protease Cleaves F-spondin in Vivo?-Serine protease expression overlaps F-spondin expression in the nervous system. The protease that is mostly coexpressed with F-spondin is tPA. In addition, tPA is also expressed in neurons that encounter F-spondin at their place of origin as well as along their axonal path (23,26). The partially cleaved F-spondin in the conditioned medium of transfected HEK293 ⌻ cells (Figs. 3B and 6C) suggests that the HEK293 ⌻ cell-derived tPA contains a moderate proteolytic activity for F-spondin. Yet, the fully cleaved F-spondin is evident, in vitro, only after activation of plasminogen by tPA, or by plasmin. In vivo, growth cone and floor plate cells may utilize tPA, uPA, or activated plasminogen to cleave F-spondin. The activity of serine proteases is tightly regulated during development by serine protease inhibitors. Neuroserpin, a serine protease inhibitor, is expressed in a dynamic fashion in the floor plate and on motor neurons (35). Neuroserpin may account for the partial resistance of the floor plate-derived F-spondin because Western analysis of protein extracts from the spinal cord revealed that both the short -4 TSRs, and the long -6 TSRs, proteins are detected (20,21).
Proteases belonging to other families are also expressed in the floor plate. Metalloproteinase MMP-11 is expressed in the floor plate (36), and MT5-MMP is expressed ubiquitously in the spinal cord (37). The furin protease SPC4 is also expressed in the floor plate (38). These proteases may cleave F-spondin between the reelin/spondin domain and the TSR domain. Furin and plasmin may be indirectly involved in this activity by activating latent metalloproteinase.
The colocalization of F-spondin, serine proteases, and growth factors in the ECM suggests that they all may interact to facilitate their activities. It was shown that thrombospondin-1 (Tsp-1) binds plasminogen and its activators (39 -41). Tsp-1 also binds and activates latent transforming growth factor-␤ (42). Of special interest is the fact that the recognition signal in Tsp-1 required for activation of transforming growth factor-␤, the KRFK motif (42), is also present in the sixth TSR of Fspondin. Hence, specific binding among F-spondin, plasminogen, uPA, tPA, and growth factors might enhance their biological activities.
Binding Properties of F-Spondin TSRs to the ECM-The thrombospondins are a family of proteins widely found in the embryonic and adult extracellular matrix. Both cell and matrix binding motifs have been identified in the TSRs of Tsp-1, so it has been hypothesized that the properties of these diverse proteins may also depend on the presence of these repeats (43). Proteoglycans have been implicated as mediating the adhesion properties of the TSR modules. Other biological activities of the TSR of Tsp-1 are mediated by the receptors CD36 (44,45), a 50-kDa protein from A549 lung carcinoma cells (46), and integrins (47).
Studies aimed at the identification of the amino acids required for cell adhesion have yielded discrepant results. Various motifs, all conserved among the thrombospondin family and present in F-spondin, have been proposed as the core adhesion motif that binds to the ECM via interaction with proteoglycans. These include: the WSXW motif present in the amino part of the repeats (the first antiparallel ␤-strand) (48,49); The CSVTCG motif, assumed to be in the junction between the first and second antiparallel ␤-strands (50); and the RXR motif present in the second antiparallel ␤-strand (51). In all of these studies, synthetic peptides were used either as a substrate for cell adhesion or as inhibitor that blocked the binding of cells to TSR-bearing proteins.
Other studies, using native recombinant TSR motifs, yield contradictory results. Some of the TSRs, such as the TSRs of Tsp1, did not bind to heparin (52), whereas others, like the TSRs of ADAMTS-1 (53), HB-GAM (33), and the TSR of the malaria circumsporozoite protein (50, 54) do bind heparin. Structural studies of the recombinant HB-GAM, using heteronuclear NMR (33), revealed that the native HB-GAM structure is essential for heparin binding. Reduction of the disulfide bonds dramatically reduced heparin binding. The HB-GAMheparin complex revealed that heparin binds to the ␤-sheet structure of the TSR, but not to the lysine-rich amino-and carboxyl-terminal tails. This implies that heparin binding requires a specific tertiary structure (33). Basic amino acids are probably essential for binding, but only in the context of the tertiary structure.
Our results obtained with native proteins support the requirement of a specific tertiary structure to elicit binding to the ECM. Repeats 1-4 contain all of the adhesive motifs (WSXW, CSVTCG, and RXR) that were identified in other thrombospondin proteins. Nevertheless, these modules do not bind to the ECM or to cells (data not shown). The second antiparallel ␤-strand of repeats 1-4 are rich in basic amino acids. The inability of repeats 1-4 to bind to the ECM might be the result of a different tertiary structure imposed by the acidic amino acids in the third antiparallel ␤-strand. The results of the reciprocal replacement of the third antiparallel ␤-strand of repeats 4 and 5 supports the hypothesis that basic residues in the third antiparallel ␤-strand are required for generating an ECM binding module. In support of this, the TSRs HB-GAM and ADAMTS-1, which mediate cell binding, are basic in their third antiparallel ␤-strand.
F-spondin TSRs represent a unique combination of ECM binding and ECM nonbinding modules. The nonbinding 1-4 TSRs are likely to interact with other receptors because they retain their biological activity (21). The ECM-bound TSRs (twothirds of the fifth and the sixth) might be involved in activation of latent transforming growth factor-␤ via the KRFK motive in the sixth TSR, as it was demonstrated for Tsp-1 (42).
Biological Significance of Modulation of F-spondin Binding to the ECM by Plasmin-F-spondin protein was shown to accumulate in the ECM that underlies the floor plate, the endoneurial ECM of the embryonic peripheral nerve, and of the regenerating sciatic nerve (20,22). An antibody raised against the spondin domain was used in these studies. The anti-TSR domain antibodies that we raised failed to detect the protein in immunohistochemistry studies. We assume that the TSR domain colocalizes with the reelin/spondin domain to the ECM. This assumption is supported by the binding properties of the unprocessed TSR domain protein to the ECM. F-spondin expression in the central nervous system overlaps the expression of other ECM proteins that are targets for plasmin cleavage. Both F-spondin and laminin are expressed in the hippocampus. Thus, F-spondin cleavage by plasmin may also account for the cell death in the hippocampus after seizure and the subsequent elevation of tPA levels, as was demonstrated for laminin (11). In addition, F-spondin may be a target for tPA during activitydependent forms of synaptic plasticity in the hippocampus and thus mediates the late phase, long term potentiation in both Schaffer collateral and mossy fiber pathways, which is interfered with by tPA inhibitors and in the tPA null mouse (9, 10). F-spondin processing by the neuronally derived tPA may account for the retarded migration of granule cells in the cerebellum of the tPA null mouse (55). F-spondin is expressed in the internal granule layer of the cerebellum during early postnatal days. 2 F-spondin may serve as a repulsive protein for granule neurons, as was demonstrated for the migrating neural crest cells (27). Granule neurons expressing tPA would then be able to clear a path through the F-spondin-rich milieu as 2 Y. Feinstein, A. Klar, and E. Soriano, unpublished data.

FIG. 8. Plasmin mediated release of F-spondin from the ECM.
Panel A, schematic drawing of the constructs used in these experiments. The gray boxes represent TSRs 1-4, the black boxes are TSRs 5 and 6. Black triangles represent the Myc epitope. The arrows point to the plasmin cleavage sites. Panel B, plasmin cleavage prevents Fspondin binding to the ECM. The TS42-6 protein was preincubated for 1 h with 50 g/ml plasmin followed by binding to ECM coated wells. The bound protein was detected by the 9E10 mAb followed by the second horseradish-peroxidase (HRP)-conjugated antibody and colorimetric horeseradish peroxidase reagent. The binding of TS42-6 to the ECM was abolished after plasmin cleavage (n ϭ 8 for each protein). Panel C, plasmin releases F-spondin from the ECM. Conditioned medium of TSR1-6⌬5 was incubated in ECM-coated wells for 1.5 h followed by a 1-h incubation with plasmin. The ECM-bound protein was detected as in panel B (n ϭ 6 for each plasmin concentration). Panel D, the cleaved F-spondin is released by plasmin. The conditioned medium of the plasmin-treated ECM-bound protein (of panel C) was analyzed by Western blotting with the 9E10 antibody. No protein was detected in the control, untreated sample (lane 1). A cleaved product is detected in all of the plasmin-treated samples (lane 2, 10 g/ml; lane 3, 30 g/ml; lane 4, 40 g/ml). The TSR1-6⌬5 protein used in this assay contains a major full-length band (black arrow) and a minor cleaved band (gray arrow). they migrate from the external granule layer to the internal granule layer.
F-spondin plays a dual role in patterning neuronal connectivity in the embryonic nervous system. It promotes outgrowth of commissural neurons (20) and inhibits outgrowth of motor neurons (21). Motor neurons, in rodents, are exposed to Fspondin from different sources during their exogenesis. As they start to extend axons away from the ventral midline of the spinal cord, they are exposed to the floor plate-derived F-spondin (21). Subsequently, as they elongate in the periphery, motor axons are ensheathed with Schwann cell-expressing F-spondin (22). The presence of motor axons in the peripheral nerve poses a potential dilemma. How can motor axons elongate during development and regeneration in an F-spondinrich milieu? One possibility is that they might change their responsiveness to F-spondin during development. As axons start to extend from the spinal cord, they might be repelled by F-spondin, but as they encounter Schwann cells they might alter their responsiveness. The temporal expression of tPA in motor neurons may suggest that tPA is modifying F-spondin attachment to the ECM in the peripheral nerve (26). tPA expression in motor neurons is elevated as they extend their axons in the peripheral nerve. Hence, it is conceivable that the motor neuron-derived tPA cleaves the Schwann cell-derived F-spondin, thereby releasing it from the ECM. Thus, it may act to clear a path for the motor axons.