A Receptor-like Protein-tyrosine Phosphatase PTPζ/RPTPβ Binds a Heparin-binding Growth Factor Midkine

Midkine is a 13-kDa heparin-binding growth factor with 45% sequence identity to pleiotrophin. Pleiotrophin has been demonstrated to bind to protein-tyrosine phosphatase ζ (PTPζ) with high affinity. In this study, we examined the binding of midkine to PTPζ by solid-phase binding assay. Midkine and pleiotrophin binding to PTPζ were equally inhibited by soluble pleiotrophin and also by some specific glycosaminoglycans. For both bindings, Scatchard analysis revealed low (3.0 nm) and high (0.58 nm) affinity binding sites. These results suggested that PTPζ is a common receptor for midkine and pleiotrophin. Midkine is structurally divided into the N- and C-terminal halves, and the latter exhibited full activity for PTPζ binding and neuronal migration induction. The C-terminal half contains two heparin-binding sites consisting of clusters of basic amino acids, Clusters I and II. A mutation at Arg78 in Cluster I resulted in loss of the high affinity binding and reduced neuronal migration-inducing activity, while mutations at Lys83 and Lys84 in Cluster II showed almost no effect on either activity. Chondroitinase ABC-treated PTPζ exhibited similar low affinity binding both to the native midkine and midkine mutants at Arg78. These results suggested that Arg78 in midkine plays an essential role in high affinity binding to PTPζ by interacting with the chondroitin sulfate portion of this receptor.

PTP/RPTP␤ 1 is a receptor-like protein-tyrosine phosphatase, which is abundantly expressed in the central nervous system as a chondroitin sulfate proteoglycan (1)(2)(3)(4). PTP is composed of an N-terminal carbonic anhydrase-like domain, a fibronectin type III domain, a serine, glycine-rich domain that is thought to be chondroitin sulfate attachment region, a transmembrane segment, and two tyrosine phosphatase domains (1,2). There are three splice variants of this molecule: (a) the full-length PTP (PTP-A); (b) the short form of PTP, in which most of the serine, glycine-rich region is deleted (PTP-B); and (c) the secreted form (PTP-S), which corresponds to the extracellular region of PTP-A and is also known as 6B4 proteoglycan/phosphacan (3,5). All these splice variants are expressed as chondroitin sulfate proteoglycans in the brain (6), suggesting that chondroitin sulfate plays an essential role in receptor function.
Several proteins such as contactin, tenascin, L1, NCAM, and TAG1 have been reported to bind PTP (7)(8)(9). Contactin is thought to be a neuronal receptor of PTP expressed on glial cells (7). Recently, we found that PTP binds with pleiotrophin/ heparin-binding growth-associated molecule (10), in that a chondroitin sulfate portion of PTP constitutes a part of the pleiotrophin binding site and regulates the affinity of PTPpleiotrophin binding (10). We further demonstrated that pleiotrophin-induced neurite outgrowth and neuronal migration were suppressed by chondroitin sulfate, polyclonal antibodies against the extracellular domain of PTP, and sodium vanadate, a protein-tyrosine phosphatase inhibitor. These findings suggested that PTP expressed on neurons is a signal transducing receptor for pleiotrophin (10,11).
Pleiotrophin has 45% sequence identity to midkine, forming a new family of heparin-binding growth factors. These molecules share many biological activities (12,13); both proteins promote neurite outgrowth (14 -16), enhance plasminogen activator activity in aortic endothelial cells (17), and oncogenically transform NIH3T3 cells (18,19). These findings suggest that they use a common or highly related receptors.
Midkine and pleiotrophin are structurally composed of two domains (the N-and C-terminal halves), each of which is tightly held through three or two disulfide bridges, respectively (20). The C-terminal half of midkine binds strongly to heparin and exhibits neurite outgrowth-promoting and plasminogen activator-enhancing activities (21,22). On the other hand, the N-terminal half of midkine, which shows relatively weak heparin binding activity, does not promote neurite outgrowth or enhance plasminogen activator activity (21,22). NMR spectroscopy revealed two clusters of basic amino acids in the C-terminal half of midkine, Clusters I and II, both of which interact with heparin oligosaccharides (23). Experiments using various midkine mutants indicated that Cluster II plays an essential role in its plasminogen activator-enhancing effect (22).
In this study, we examined the PTP-midkine interaction using various midkine mutants. Native PTP exhibited high affinity binding to midkine, and the binding properties were essentially the same as those of pleiotrophin. Moreover, PTPmidkine binding was inhibited by the presence of pleiotrophin. These observations suggested that midkine and pleiotrophin share a common binding site on PTP. PTP bound to the C-terminal half of midkine, but not to the N-terminal half. A mutation R78Q in Cluster I reduced the binding affinity, while mutations K83Q, K84Q, and K83Q/K84Q in Cluster II did not affect binding. Furthermore, in these midkine mutants, the strength of binding affinities and the neuronal migration-inducing activities were highly correlated. These findings suggested that basic amino acids in Cluster I of midkine and pleiotrophin are crucial for high affinity binding to PTP to transduce signals in neurons.

EXPERIMENTAL PROCEDURES
Materials-Chondroitin sulfate A from whale cartilage, chondroitin sulfate B from pig skin, chondroitin sulfates C and D from shark cartilage, chondroitin sulfate E from squid cartilage, heparan sulfate from bovine kidney, keratan sulfate from bovine cornea, and chondroitinase ABC were purchased from Seikagaku Corp. Heparin was obtained from Sigma. 125 I-Bolton-Hunter reagent was purchased from DuPont NEN. Chroma Spin columns were obtained from CLONTECH. Maxisorp immunoplates were purchased from Nunc. Dulbecco's modified Eagle's medium, F-12 medium, and B-27 supplement were purchased from Life Technologies, Inc. Transwells TM were obtained from Corning Coster Corp. Micro BCA kit was from Pierce. PTP-S was purified as reported elsewhere (24). The N-and C-terminal half domains of human midkine (1-59 and 60 -121, respectively) were synthesized as described previously (25). Mouse midkine mutants, R78Q, K83Q, K84Q, K83Q/ K84Q and R78Q/K83Q/K84Q were prepared by site-directed mutagenesis (21,22). Mutations are indicated by the amino acid residues (in one-letter code) in the wild-type and the mutant, preceding and following the numbers of the altered residues, respectively. 125 I Labeling of PTP-S-PTP-S was purified from rat brain and labeled as described previously (10,24). Briefly, dried 125 I-Bolton-Hunter reagent (100 Ci) was solubilized with samples (10 g of protein in 100 l of 100 mM sodium phosphate buffer, pH 8.0), followed by incubation for 3 h on ice and then mixed with 30 l of 1 M glycine, pH 7.5. After a 2-h incubation at 4°C, free 125 I-Bolton-Hunter reagent was removed by passing through a Chroma Spin 30 column equilibrated with 0.05% Triton X-100, 0.5 mg/ml BSA, 0.15 M NaCl, 10 mM sodium phosphate, pH 7.2. The specific radioactivity of the sample thus prepared was 3.3 ϫ 10 6 cpm/g.
Binding Assay-Wells of Nunc Maxisorp Immunoplates were coated with 35 l of 1ϳ5 g/ml midkine or pleiotrophin in 5 mM Tris-HCl, pH 8.0, at 4°C overnight. The wells were washed three times with phosphate-buffered saline and then blocked with 1% BSA/phosphate-buffered saline for 1 h at room temperature. 125 I-PTP-S diluted in 0.5% BSA, 2 mM CaCl 2 , 2 mM MgCl 2 , 0.1% CHAPS, 0.15 M NaCl, 10 mM sodium phosphate, pH 7.2, was added to the coated wells. When inhibition experiments were performed, inhibitors (pleiotrophin or glycosaminoglycans) were premixed with 125 I-PTP-S before addition to the wells. The plates were incubated for 5 h at room temperature and then the wells were washed three times with 1 mM CaCl 2 , 1 mM MgCl 2 , 0.15 M NaCl, 10 mM Tris-HCl, pH 7.2. The bound materials were released by adding 200 l of 0.1 M NaOH, 0.2% SDS to the wells. The plates were shaken for 15 min at room temperature and then the eluted radioactivity was measured using a ␥ counter. 125 I-Labeled PTP-S was digested with chondroitinase ABC as described previously (10). Briefly, 125 I-PTP-S was diluted with 100 l of 0.5% BSA, 2 mM MgCl 2 , 2 mM CaCl 2 , 0.15 M NaCl, 10 mM sodium acetate, 10 mM Tris-HCl, pH 7.5, to a final concentration of 2 g/ml. Aliquots (5 milliunits) of protease-free chondroitinase ABC was added to the samples, and the solutions were incubated for 30 min at 30°C for use in binding assays.
Other Methods-Boyden chamber cell migration assays were performed using cortical neurons from embryonic day-17 Sprague-Dawley rats as described previously (11). Protein concentration was determined using a Micro BCA kit using BSA as a standard. Fig. 1 shows the binding profile of 125 I-labeled PTP-S to human midkine-coated ELISA plates. Scatchard analyses of the binding of PTP-S to midkine showed low (K d ϭ 3.0 nM) and high (K d ϭ 0.58 nM) affinity binding sites (Fig. 1B), which were similar to those of pleiotrophin-PTP binding (10). As shown in Fig. 1, the C-terminal half of midkine exhibited exactly the same binding properties to PTP-S as native midkine. On the other hand, the N-terminal half of midkine showed no binding activity to PTP-S (Fig. 1). Soluble pleiotrophin premixed with PTP-S inhibited the binding of PTP-S to pleiotrophin-coated ELISA plates (Fig. 2). In a similar dose-dependent manner, soluble pleiotrophin also inhibited the binding of PTP-S to midkine on the plates (Fig. 2), suggesting that pleiotrophin and midkine bind to the same binding site on PTP-S with a similar affinity. However, fairly high concentrations of pleiotrophin were required for inhibition (IC 50 ϭ ϳ600 nM) compared with the K d values of midkine-or pleiotrophin-PTP-S binding obtained by solid-phase binding assay. These observations suggested that substrate-bound forms of midkine and pleiotrophin exhibit orders of stronger affinity to PTP-S than the soluble forms.

Binding of PTP-S to Midkine-
Midkine has two clusters of basic amino acids (Clusters I and II) located at the surface on one side of the C-terminal half domain, which are considered to be heparin binding sites (23). Cluster I contains Lys 76 , Arg 78 , and Lys 99 , and Cluster II contains Lys 83 , Lys 84 , and Arg 86 ; amino acids were numbered according to mouse midkine. Among these, Lys 76 , Arg 78 , Lys 83 , and Lys 99 are conserved in midkine and pleiotrophin of all species examined to date. On the other hand, Lys 84 is conserved only in midkine of various species but is changed to Arg in pleiotrophin, and Arg 86 is changed to Leu in pleiotrophin of various species and midkine of some species (23).
Five mouse midkine mutants were prepared, in which some of the basic amino acids in the Cluster I and/or II were changed to glutamine: R78Q, K83Q, K84Q, K83Q/K84Q, and R78Q/ K83Q/K84Q (21,22). As shown in Fig. 3, K83Q, K84Q, and K83Q/K84Q exhibited essentially the same binding activities to PTP-S as the native midkine, suggesting that Cluster II is not essential for midkine-PTP binding. In contrast, R78Q and R78Q/K83Q/K84Q exhibited only low affinity binding to PTP-S, suggesting that Cluster I plays an important role in the high affinity binding between PTP and midkine ( Fig. 3 and Table I).
Effects of Chondroitinase ABC Digestion of PTP-S on the PTP-midkine Binding-Chondroitin sulfate chains of PTP play an essential role in its high affinity binding to pleiotrophin (10). Chondroitinase ABC digestion of PTP-S reduced its affinity also to midkine (Fig. 4). In contrast to the intact PTP-S showing high (K d ϭ ϳ0.5 nM) and low (K d ϭ ϳ3 nM) affinity binding sites, chondroitinase ABC-digested PTP-S exhibited only a low affinity binding site (K d ϭ 8.8 nM) (Fig. 4, A and B). In addition, R78Q (Fig. 4, C and D) and R78Q/K83Q/K84Q (Table I), which have a mutation at Arg 78 , showed a single binding site to intact PTP-S with a K d value of 2.8 nM, in a similar affinity range to the chondroitinase ABC-digested PTP-S (ϳ8 nM). This suggested that Arg 78 is involved in binding to chondroitin sulfate to make up the high affinity binding site.
Influence of Glycosaminoglycans on PTP-midkine Binding-Previously, we reported that pleiotrophin-PTP-S binding is inhibited strongly by heparin, moderately by heparan sulfate and chondroitin sulfate C, and very weakly by chondroitin sulfate A (10). Glycosaminoglycans inhibited midkine-PTP-S interactions similarly (Fig. 5). Heparin strongly inhibited binding of PTP-S to midkine (IC 50 ϭ 10 ng/ml), heparan sulfate showed moderate inhibition (IC 50 ϭ 100 ng/ml), and keratan sulfate exerted almost no effect. On the other hand, various types of chondroitin sulfate exerted diverse influences on midkine-PTP-S binding. Chondroitin sulfate D and chondroitin sulfate E strongly inhibited binding (IC 50 ϭ ϳ70 ng/ml for both types of chondroitin sulfate). Chondroitin sulfate B and chondroitin sulfate C showed moderate inhibitory effects (IC 50 ϭ 500 ng/ml and 1000 ng/ml, respectively), but chondroitin sulfate A exerted almost no effect (IC 50 Ͼ 100 g/ml). Similar sensitivities to the various chondroitin sulfates were observed for pleiotrophin-PTP binding (data not shown; data partly shown in Ref. 10).
Cell Migration-inducing Activity of Midkine-We reported previously that pleiotrophin induced cell migration of cortical neurons (11). Midkine also induced neuronal migration in Boyden chamber cell migration assay with essentially the same dose dependence profile as that of pleiotrophin (data not shown; see Fig. 3A of Ref. 11). Boyden chamber cell migration assay indicated that the C-terminal half of midkine exhibited full cell migration-inducing activity but the N-terminal half was devoid of activity (Fig. 6A). Midkine mutants, K83Q, K84Q, and K83Q/K84Q, which have amino acid replacements in Cluster II, showed normal levels of activity. In contrast, R78Q and R78Q/K83Q/K84Q exhibited low cell migration-inducing activity (Fig. 6B). These results suggested that Cluster I is sufficient for the neuronal migration-inducing activity of midkine.
Influence of Glycosaminoglycans on Midkine-induced Neuronal Migration-Midkine-induced neuronal migration was inhibited strongly by heparin, moderately by heparan sulfate, but  Table I. not by keratan sulfate (Fig. 7). As in the case of midkine-PTP-S binding, various types of chondroitin sulfate exerted diverse effects on midkine-induced neuronal migration. Chondroitin sulfate A exhibited almost no effect (Fig. 7). On the other hand, midkine-induced neuronal migration was inhibited strongly by chondroitin sulfate E and moderately by chondroitin sulfates B, C, and D (Fig. 7). Similar inhibitory effects by chondroitin sulfates were observed for pleiotrophin-induced neuronal migration (data not shown; data partly shown in Ref. 11).

DISCUSSION
In this study, we demonstrated that midkine binds to PTP. The characteristics of binding of midkine to PTP were indistinguishable from those of pleiotrophin (10), suggesting that PTP is a common receptor of midkine and pleiotrophin. Here, the C-terminal half of midkine was revealed to be sufficient for the binding. The C-terminal half domain of midkine exhibits various activities: strong heparin-binding activity, neurite promoting activity, and tissue plasminogen activator enhancing activity (21,22). On the other hand, specific functions have not been found for the N-terminal half of midkine, although it weakly binds to heparin (21,22,26).
NMR spectroscopy indicated that there are two heparinbinding sites in the C-terminal half domain: Cluster I, which is composed of Lys 76 , Arg 78 , and Lys 99 , and Cluster II, which is composed of Lys 83 , Lys 84 , and Arg 86 (23). On the other hand, in the N-terminal half domain, the basic amino acids do not form clusters which are expected to interact with the sulfate groups on heparin (23,26). Our data showed that PTP-midkine binding was significantly affected by the mutation of Arg 78 , but not by mutations of Lys 83 , Lys 84 , or Lys 83 ϩ Lys 84 . Here, mutation of Arg 78 resulted in loss of high affinity binding between midkine and PTP (Fig. 3), and the chondroitin sulfate portion of PTP plays an essential role in formation of the high affinity binding site (Fig. 4). Therefore, it seems that Arg 78 of midkine is involved in binding to chondroitin sulfate on PTP. In support of this idea, various chondroitin sulfate preparations differentially affected midkine-PTP binding (Fig. 5). Among various chondroitin sulfate species, there was a significant difference in the inhibiting activity. This finding suggested that there must be a specific structural motif of chondroitin sulfate that strongly inhibits midkine-PTP binding. However, the nature of this structure is not known at present because commercially available chondroitin sulfate samples contain considerable heterogeneity. Nevertheless, it is possible to speculate that Arg 78 of midkine recognizes a specific structure of chondroitin sulfate on the PTP molecule, which is also present in chondroitin sulfates C, D, and E, but not in chondroitin sulfate A. An oversulfated structure is one of the candidates; however, the fine structure of chondroitin sulfate chains of PTP must be determined to further clarify this point. A similar finding was reported for DSD-1-PG, a chondroitin sulfate proteoglycan expressed in the rodent central nervous system, that is recognized by a monoclonal antibody 473HD (27). DSD-1-PG exhibited neurite outgrowth-promoting activity, which was blocked by 473HD or by chondroitinase ABC digestion of this proteoglycan (27). The binding of 473HD to DSD-1-PG was inhibited by chondroitin sulfates C and D, but not by chondroitin sulfates A or B (27,28), suggesting that a specific structural motif of chondroitin sulfate plays an important physiological function in the brain.
Chondroitinase ABC-treated PTP showed markedly reduced binding affinity to midkine. Mutations of midkine at Arg 78 , Lys 83 , and Lys 84 did not influence binding to the chondroitinase ABC-treated PTP, suggesting that these amino acids do not play an essential role in binding to the core glycoprotein portion of PTP. In summary, there seems to be a hierarchy with three steps in the binding between PTP and midkine: 1) low affinity binding between midkine and core glycoprotein portion of PTP (K d ϭ ϳ8 nM); 2) medium affinity binding between midkine and PTP bearing general structure of chondroitin sulfate (K d ϭ ϳ3 nM); and 3) high affinity binding between midkine and PTP bearing a specific structural motif of chondroitin sulfate (K d ϭ ϳ0.6 nM), which involves a specific contribution of Arg 78 of midkine.
Boyden chamber cell migration assay indicated that the mutation of Arg 78 of midkine significantly reduced the neuronal migration-inducing activity of this factor (Fig. 6). In contrast, mutations of Lys 83 and Lys 84 did not influence this activity. These observations suggested that the high affinity binding of midkine and PTP is important for the neuronal migrationinducing activity. Here, heparin strongly inhibited midkineand pleiotrophin-induced neuronal migration, and only the substrate-bound forms of these factors exhibit this activity (11,17), which is consistent with the finding that PTP exhibits very low affinity to soluble pleiotrophin (Fig. 2). In contrast, plasminogen activator-enhancing activity of midkine was markedly reduced by double mutation of Lys 83 and Lys 84 , but not by the single mutation of Arg 78 , Lys 83 , or Lys 84 (22). The soluble forms of midkine and pleiotrophin enhance plasminogen activator activity. However, it has been suggested that enzymatic dimerization of midkine and pleiotrophin induced by heparin-like oligosaccharides (presumably endogenous heparan sulfate) is required for plasminogen activator-enhancing activity (17). Here, exogenously added heparin could substitute for endogenous heparan sulfate (17). Taken together, these two activities of midkine and pleiotrophin are thought to be mediated by distinct receptors. Neurite-promoting activity of midkine was also markedly reduced by mutation of Arg 78 , while mutations of Lys 83 and Lys 84 were less effective (22). These observations suggested that the neurite-promoting and the neuronal migration-inducing activities of midkine are mediated at least partly by the same or similar receptor(s).
Midkine binds to a syndecan family heparan sulfate proteoglycan, ryudocan, with high affinity (29). Pleiotrophin/heparinbinding growth-associated molecule binds to N-syndecan, which is thought to be another pleiotrophin receptor involved in pleiotrophin-induced neurite extension (30). It would be helpful to examine the binding of syndecan family proteoglycans with midkine mutants to determine the physiological significance of these interactions.