The Binding of Chondroitin Sulfate to Pleiotrophin/Heparin-binding Growth-associated Molecule Is Regulated by Chain Length and Oversulfated Structures*

Pleiotrophin is an 18-kDa heparin-binding growth factor, which uses chondroitin sulfate (CS) proteoglycan, PTPζ as a receptor. It has been suggested that the D-type structure (GlcA(2S)β1-3GalNAc(6S)) in CS contributes to the high affinity binding between PTPζ and pleiotrophin. Here, we analyzed the interaction of shark cartilage CS-D with pleiotrophin using a surface plasmon resonance biosensor to reveal the importance of D-type structure. CS-D was partially digested with chondroitinase ABC, and fractionated using a Superdex 75pg column. The ≥18-mer CS fractions showed significant binding to pleiotrophin, and the longer fractions had stronger affinity for pleiotrophin than the shorter ones. The ∼46-mer CS fraction bound to densely immobilized pleiotrophin with high affinity (KD = ∼30 nm), and the binding reactions fitted the bivalent analyte model. However, when the density of the immobilized pleiotrophin was lowered, the strength of affinity remarkably decreased (KD = ∼2.5 μm), and the reactions no longer fitted the model and were considered to be monovalent binding. The 20∼24-mer fractions showed low affinity binding to densely immobilized pleiotrophin (KD = 3∼20 μm), which seemed to be monovalent. When ∼22-mer CS oligosaccharides were fractionated by strong anion exchange HPLC, each fraction differed in affinity for pleiotrophin (KD = 0.36 ∼ >10 μm), and the affinity correlated with the amounts of D- and E- (GlcAβ1-3GalNAc(4S,6S)) type oversulfated structures. These results suggest that the binding of pleiotrophin to CS is regulated by multivalency with CS ∼20 mer as a unit and by the amounts of oversulfated structures.

The binding of phosphacan to pleiotrophin and midkine depends on the CS portion of this proteoglycan, and the removal of CS resulted in a remarkable decrease in the binding affinity (9,10). It was revealed that pleiotrophin inactivated the tyrosine phosphatase activity of this receptor leading to an increase in the tyrosine phosphorylation levels of specific substrates such as cat-1 and ␤-catenin (4,11). Several researchers suggested that pleiotrophin induces the dimerization of PTP, which results in the inactivation of its enzymatic activity (4,11).
On the other hand, midkine and pleiotrophin easily formed noncovalently bound multimers, and it has been suggested that multimers larger than dimers are the active forms (12)(13)(14)(15). Furthermore, both growth factors were cross-linked by transglutaminases forming covalently bound multimers (12)(13)(14)(15), and this multimerization process was remarkably promoted by heparin and CS (13,15). These observations raise the possibility that multimers of pleiotrophin and midkine bind with PTP through CS inducing the dimerization of this receptor.
Previously, we demonstrated that the structure of CS on phosphacan changed during development of rat brain, and even at the same developmental stage, the CS structure varied depending on the region of the brain (16). An analysis using a surface plasmon resonance biosensor indicated that phosphacan bearing the CS with D unit (GlcA(2S)␤1-3GalNAc(6S)) had higher affinity for pleiotrophin than that without this structure (16). The binding of phosphacan with pleiotrophin and midkine was inhibited by various CS preparations (9,10). The binding was inhibited strongly by squid cartilage CS-E and shark cartilage CS-D, moderately by shark cartilage CS-C and very weakly by whale cartilage CS-A (9, 10). CS-E and CS-D contain large amounts of E-type (GlcA␤1-3GalNAc(4S,6S)) and D-type structures, respectively. C unit (GlcA␤1-3GalNAc(6S)) is the major component of CS-C, which additionally contains moderate amounts of D unit. On the other hand, the amounts of the oversulfated structures are very small in CS-A, in which A unit (GlcA␤1-3GalNAc(4S)) is the major component. These findings suggested that D-and E-type oversulfated structures in CS chains play important roles in the determination of the affinity of pleiotrophin and midkine for PTP.
However, the contribution of the D unit in CS to the binding with these growth factors is yet to be clearly demonstrated. Some researchers reported that CS-E but not CS-D strongly bound with midkine and suggested that the E-type structure was the critical determinant of the binding affinity for midkine (17)(18)(19). Other studies, however, indicated that CS-E strongly bound with pleiotrophin, but CS-D also bound to this growth factor with high affinity (16,20,21). In the present study, we analyzed the interaction of CS-D with pleiotrophin to reveal the structural determinants involved in the pleiotrophin-CS interaction. We suggest that three parameters: chain length, amounts of oversulfated structures, and pleiotrophin multimerization, play important roles in determination of the binding affinity between CS and pleiotrophin.
Preparation and Fractionation of CS Oligosaccharides-First, 120 mg of shark cartilage CS-D (O unit (GlcA␤1-3GalNAc): A unit:C unit:D unit:E unit is 1.5:25.9:50.4:18.8:3.4; average molecular mass ϳ20 kDa) was dissolved in 1.5 ml of 30 mM sodium acetate, 0.1 M Tris-HCl, pH 7.8. The solution was incubated with 30 milliunits of protease-free chondroitinase ABC at 30°C until the absorbance at 232 nm of the sample reached 8.5% of the value obtained after complete digestion. The sample was heated at 100°C for 5 min and then applied to a column of Bio-Gel P-10 (1.6 ϫ 90 cm) equilibrated with 0.4 M ammonium acetate. The fractions corresponding to K av values of 0.16ϳ0.61 and 0.45ϳ0.77 were pooled. After lyophilization, the samples were applied to a Superdex 75pg column (1.6 ϫ 60 cm) and eluted with 0.2 M NH 4 HCO 3 at a flow rate of 1 ml/min. The fractions were lyophilized and then dissolved in distilled water. The 22-mer fraction was further chromatographed using a Spherisorb S5 SAX column (4.6 ϫ 250 mm) with a linear gradient of NaCl from 0.8 M to 1.35 M at pH 3.5 at a flow rate of 0.5 ml/min. The fractions were desalted using a HiTrap desalting column equilibrated with 0.2 M NH 4 HCO 3 and lyophilized.
Characterization of CS-The disaccharide composition of CS was determined as follows. CS (2 g) was diluted with 20 l of 50 mM Tris-HCl, pH 7.4, and 15 mM sodium acetate. The samples were incubated with 50 milliunits of chondroitinase ABC for 2 h at 37°C, and then heated at 100°C for 1 min. The samples were applied to a YMC PA-03 column (4.6 ϫ 250 mm) and eluted with a linear gradient of NaH 2 PO 4 from 16 to 500 mM over 60 min at a flow rate of 1 ml/min. The eluates were monitored by absorbance at 232 nm. Chain lengths of CS (up to 24 mer) were determined using a Voyager DE-STR MALDI-TOF MS spectrometer. The chain lengths of CS larger than 24 mer were estimated from the K av value of each fraction on the Superdex 75pg column with a standard curve generated using 6 -20 mer of CS oligosaccharides.
Solid Phase Binding Assay-Wells of Nunc Maxisorp immunoplates were coated with 2 g/ml of pleiotrophin in 30 l of 5 mM Tris-HCl, pH 7.8 at 4°C overnight. The wells were washed three times with phosphate-buffered saline (PBS) and then blocked with 1% bovine serum albumin/PBS for 1 h at room temperature. CS oligosaccharides diluted with 1 mM CaCl 2 , 1 mM MgCl 2 , and 0.05% Tween 20 in PBS (reaction buffer) were added to the wells (30 l/well), and the plates were incubated for 2 h at room temperature. Then, 5 l of 1 g/ml of phosphacan/ 6B4 proteoglycan were added to each well, and the plates were incubated for 1 h at room temperature. After three washes with the reaction buffer, 35 l of anti-6B4 proteoglycan (1:500) diluted in the reaction buffer were added to each well, and the plates were incubated for 30 min at room temperature. After three more washes, 35 l of horseradish peroxidase-conjugated anti-rabbit IgG (1:500) diluted in the reaction buffer were added to each well, and the plates were again incubated for 30 min at room temperature. The wells were washed three times with the reaction buffer, and 200 l of 0.1 mg/ml of o-phenylenediamine, 0.006% H 2 O 2 were added to each well. After ϳ10 min, the reaction was stopped by adding 10 l of 1 M HCl, and the absorbance at 450 nm was measured.
Immobilization of Pleiotrophin on the Sensor Surface-Pleiotrophin was immobilized on the surface of a CM5 sensor chip by amine coupling, where the primary amino groups on the protein were coupled to the carboxymethylated dextran on a sensor chip surface. The carboxymethylated dextran on the sensor chip was activated by injection of 35 l of N-ethyl-NЈ-(3-dimethylaminopropyl)-carbodiimide hydrochloride/N-hydroxysuccinimide (0.2/0.05 M). Then, 20 g/ml of pleiotrophin in 10 mM maleate buffer, pH 6.0, was injected onto the activated sensor surface. The remaining unreacted sites were blocked by injecting 35 l of 1 M ethanolamine, pH 8.5. The amounts of pleiotrophin immobilized on the sensor surface were controlled within the range of 2,900 and 6,300 resonance units (RU) by changing the injection time. All steps were carried out in a continuous flow of a solution containing 10 mM HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, and 0.005% Tween 20 (HBS running buffer) at 5 l/min. Surface Plasmon Resonance Analysis-All experiments were carried out at a flow rate of 20 l/min at 25°C. Each sample was diluted in HBS running buffer and injected onto the sensor surface. The sensor surface was regenerated with 10 l of 2 M NaCl after a dissociation phase. To correct for the bulk effects and the nonspecific binding of samples, the equivalent sample solutions were injected onto an untreated sensor surface, and the responses obtained were subtracted from the pleiotrophin-immobilized sensor surface data. The kinetic parameters were evaluated with BIAevaluation software 4.1 using a bivalent analyte model and Scatchard plot. The data are expressed as the mean Ϯ mean absolute deviation of at least two independent measurements.
Cross-linking Experiments-Pleiotrophin (0.3 g) was incubated at room temperature with various concentrations of CS fractions in a final volume of 10 l in 0.05% Tween 20/PBS. Then, 0.5 l of 25 mM DSS was added to the each sample, which was incubated at room temperature for 30 min. The reactions were quenched for 30 min with 100 mM Tris-HCl, pH 7.8, and the samples were analyzed by 12.5% SDS-PAGE and Western blotting. Cross-linked pleiotrophin was detected using anti-human pleiotrophin antibody and ECL kit according to the supplier's protocol.

RESULTS
Binding of CS Oligosaccharides to Pleiotrophin-Previously, we demonstrated that shark cartilage CS-D bound to pleiotrophin with high affinity using the BIAcore system (16). Many of the chondroitin sulfatedependent activities of pleiotrophin such as promotion of neurite exten-sion and migration of embryonic neurons are mediated by substratebound form of this growth factor (1-3). Thus we immobilized pleiotrophin on the CM5 sensor chips by the amine coupling method because it seems that this closely models the active form of pleiotrophin. However, there was a possibility that this coupling method could destroy lysine residues that are essential for binding with CS. So we firstly immobilized pleiotrophin preincubated with excess amounts of CS oligosaccharides to protect CS binding basic amino acids. However, it became apparent that the sensorgrams of the binding of various CS preparations to pleiotrophin were essentially the same, whether or not pleiotrophin was preincubated with CS oligosaccharides (data not shown). Thus in the following experiments, pleiotrophin was coupled to the sensor chips without preincubation with CS oligosaccharides.
To characterize the CS structure required for binding to pleiotrophin, we fractionated CS-D partially digested with protease-free chondroitinase ABC by Superdex 75 gel permeation chromatography. Each fraction (10 g/ml as hexuronate) was applied to a CM5 sensor chip containing immobilized pleiotrophin (Fig. 1). In this assay, a saturated amount of pleiotrophin was immobilized on the sensor chip (5,942 RU) to maximize the sensitivity. As shown in Fig. 1A, CS fractions bound to pleiotrophin, but the binding decreased sharply as the chain length shortened. Although the 18-mer fraction showed slight binding to pleiotrophin, we could not detect any binding for Յ16-mer oligosaccharides in this assay system (Fig. 1B).
In a previous study, we demonstrated that the binding of phosphacan to pleiotrophin was strongly inhibited by CS-D (10). Using a solid phase binding assay, we next examined the inhibitory effects of CS oligosaccharides on the binding of phosphacan to pleiotrophin. The binding was inhibited efficiently by Ն20-mer oligosaccharides, moderately by 16-mer and 18-mer oligosaccharides, and not by 14-mer oligosaccharides (Fig. 2). These findings suggest that 16ϳ18 mer was the basic functional unit required for the binding to pleiotrophin.
Multivalent Binding of CS to Pleiotrophin-The above experiments indicated that longer CS bound more efficiently to pleiotrophin than shorter ones. This suggests that the affinity of CS for pleiotrophin was increased by multivalent binding of long polysaccharides to multiple pleiotrophin immobilized on the sensor chip. To test this possibility, the ϳ46-mer CS fraction (F-a, Fig. 1A) was applied to sensor chips immobilized with various densities of pleiotrophin. Three sensor chips were prepared, on which low (2,461 RU; low density condition), medium (2,905 RU; medium density condition), and high (4,753 RU; high density condition) amounts of pleiotrophin were immobilized. During the immobilization process, the amounts of immobilized pleiotrophin increased rapidly to ϳ4,000 RU (ϳ4 ng/mm 2 ). Then, the reaction became very slow with a maximum immobilization of ϳ6,000 RU (ϳ6 ng/mm 2 ) (data not shown). This suggests that pleiotrophin was immobilized sparsely on the sensor chips under the low and medium density conditions, and the density of immobilized pleiotrophin was semisaturated under the high density condition. Fig. 3 shows the binding of fraction F-a to the sensor chips immobilized with three different densities of pleiotrophin. Under the high density condition (Fig. 3A), CS associated relatively slowly with pleiotro-FIGURE 1. Gel permeation column chromatography of chondroitin sulfate D partially digested with chondroitinase ABC. Shark cartilage CS-D was partially digested with protease-free chondroitinase ABC, and then applied to a Bio-Gel P10 column. The fractions corresponding to the K av values of 0.16ϳ0.61 (A) and 0.45ϳ0.77 (B) were next applied to a Superdex 75pg column. The fractions indicated by rectangular columns were collected (F-a ϳl). Each fraction (10 g/ml) was applied to a CM5 sensor chip immobilized with 5,942 RU of pleiotrophin (ϳ6 ng/mm 2 ). The steady state binding responses are indicated by the height of the columns. Numbers 6 -22 above or below the peaks indicate the elution positions of 6 -22-mer oligosaccharides, respectively. The shaded column in A indicates the F-h fraction, which was next subjected to strong anion exchange HPLC. Vo, void volume; Vt, total volume. phin, and the dissociation also proceeded gradually. In contrast, under the low density condition (Fig. 3E), fraction F-a associated with pleiotrophin very quickly and also dissociated quickly, showing "box-shaped" sensorgrams. The sensorgrams were intermediate between the two under the medium density condition (Fig. 3C).
We analyzed these sensorgrams using various binding models in BIAevaluation software 4.1. Although the 1:1 (Langmuir) binding model is most commonly used, these sensorgrams did not fit this model well. On the other hand, the sensorgrams shown in Fig. 3, A and C fitted the bivalent analyte model well. The bivalent analyte model is schematically represented in Fig. 9A, in which one CS chain (C) binds with two pleiotrophin units (P) immobilized on the sensor chip. In this model, CS binds with one pleiotrophin unit (reaction 1) and then with another (reaction 2), and the kinetic parameters for each reaction can be calcu-lated using the BIAevaluation software, in which the dissociation constant of the reaction 1 (K D1 ) characterizes the binding. The K D1 values for the interaction under the medium and high density conditions were calculated to be 2.2 and 0.5 M, respectively (Table 1).
Although K D1 represents the strength of the interaction between CS and one pleiotrophin unit, we cannot know from this value the overall strength of the binding between these molecules in a bivalent reaction. However, using a steady state analysis (Scatchard plot), we can reveal overall binding strength between CS and the immobilized pleiotrophin (K D , avidity), in which the sum of the binding strength of reaction 1 and reaction 2 is calculated. So, the sensorgrams were then analyzed by plotting steady state binding responses against concentrations of CS (Fig. 3, B, D, F). As shown in Table 1, the K D values for the interaction under the high, medium, and low density conditions were calculated to

TABLE 1 Kinetic parameters for the interaction between the CS 46-mer fraction and various densities of PTN
The K D , k a1 , k dl , K D1 , k a2 , and k d2 values were calculated from the sensorgrams using six or more concentrations of analyte (CS) in two or three independent experiments.

PTN
Scatchard analysis, K D Bivalent analyte model The values could not be determined because the sensorgrams did not fit the bivalent analyte model.  FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 be 62 nM, 0.4 M, and 2.5 M, respectively. The K D values for the interactions under high and medium density conditions were 6ϳ8-fold smaller than the K D1 values, suggesting that bivalent binding of CS to two pleiotrophin units led to the stronger interaction between these molecules. The sensorgrams of the low density condition could not be analyzed with the bivalent model, and only the results of the Scatchard analysis were available (Fig. 3F).

Binding of Chondroitin Sulfate to Pleiotrophin
Chain Length-dependent Binding of CS to Pleiotrophin-The CS oligosaccharides of various chain lengths were applied to sensor chips immobilized with 5,942 RU of pleiotrophin. In this experiment, a saturated level of pleiotrophin was immobilized on the sensor chips to improve sensitivity for the binding of short oligosaccharides. The sensorgrams of the long CS (Fig. 4, A and B) were similar to that shown in Fig. 3A. On the other hand, shorter CS showed box-shaped sensorgrams (Fig. 4C). Scatchard analysis indicated that the affinity of CS fractions decreased as the chain length shortened (Table 2), and F-a (ϳ 46 mer), F-c (ϳ 34 mer), and F-e (ϳ28 mer) CS fractions showed K D values of 30, 170, and 950 nM, respectively.
Under this assay condition, the sensorgrams of CS fractions longer than 34 mer fit the bivalent analyte model well (Table 2). K D values were again 6ϳ13-fold smaller than K D1 values, indicating that bivalent binding increased the affinity (avidity). The sensorgrams of CS fractions shorter than 28 mer were a typical box shape and could not be analyzed with the bivalent analyte model, suggesting that these short CS oligosaccharides interacted with only one pleiotrophin unit.
Next, we cross-linked pleiotrophin with DSS in the presence of ϳ46mer (F-a), ϳ24-mer (F-g), and 10-mer fractions (Fig. 5). In the presence of cross-linker alone, almost no dimers were observed (Fig. 5g). Also in the presence of the 10-mer fraction, almost no dimers were observed (Fig. 5, e and f). In contrast, the F-g fraction induced dimer formation of pleiotrophin (Fig. 5, c and d), and in the presence of the F-a fraction, trimers and tetramers were additionally observed (Fig. 5, a and b). These results indicated that ϳ46 -and ϳ24-mer fractions accommodate four and two pleiotrophin molecules, respectively and further suggested that pleiotrophin dimer was the basic functional unit.
Oversulfated Structures in CS Regulate the Affinity for Pleiotrophin-To examine the significance of structural variations of CS in pleiotrophin-CS interaction, we fractionated F-h oligosaccharides (ϳ 22 mer) with a Spherisorb S5 SAX column (Fig. 6). The separated fractions (F-h-1ϳ11) were applied to a sensor chip immobilized with 6,300 RU of pleiotrophin, and the steady state binding of CS oligosaccharides was measured (Fig. 7). Whereas F-h-1 ϳ4 showed very weak binding to pleiotrophin, the binding responses increased from F-h-5 to F-h-11. Because these fractions were all 20ϳ24-mer oligosaccharides (Table 3), it was concluded that the differences in the binding response were caused by the structural variation of CS. F-h-1ϳ6 contained 10ϳ21% D unit with almost no E unit (Table 3). On the other hand, F-h-7ϳ11 contained 1ϳ6% E unit with 22ϳ25% D unit. Several reports suggested that D-and E-type structures are the major determinants of the affinity of CS for pleiotrophin (16 -19). So, we plotted amounts of these structures in Fig. 7, revealing that the affinity of CS oligosaccharides for pleiotrophin correlated well with the contents of these oversulfated structures. It should be noted that F-h-10 and F-h-11 showed markedly enhanced binding to pleiotrophin compared with F-h-8 and F-h-9 (Fig.  7). Although these fractions contained comparable amounts of D unit, F-h-10 and F-h-11 contained higher amounts of E unit than F-h-8 and F-h-9. Bao et al. (21) reported that CS oligosaccharides containing E unit showed stronger affinity for pleiotrophin than CS oligosaccharides containing D unit but without E unit. Thus, the differences in the binding efficiency of F-h-8ϳ11 might be explained by the difference in the

TABLE 2 Kinetic parameters for the interaction between pleiotrophin and CSs of various chain length
The K D , k a1 , k d1 , K D1 , k a2 , and k d2 values were calculated from the sensorgrams using six or more concentrations of analyte (CS) in two or three independent experiments.

Fraction
The values could not be determined because the sensorgrams did not fit the bivalent analyte model.
amounts of E unit. However, there also is a possibility that this was caused by the difference in the CS sequences. The sensorgrams of these oligosaccharides were all box-shaped, and the K D values were calculated using a Scatchard analysis (Fig. 8). F-h-6 with 21% D and no E structures showed a K D value of 2.7 M, and F-h-12 with 25% D and 5% E structures had a K D value of 0.36 M (Table 3). These results suggested that the amounts of D-and E-type structures contribute to the determination of the affinity of CS for pleiotrophin.

DISCUSSION
In this study, we demonstrated that the interaction between CS and pleiotrophin is regulated by three parameters: 1) chain length, 2) amount of oversulfated structures of CS, and (3) multivalency of the interaction. This suggests that the structural heterogeneity of CS plays important roles in the signal transduction of pleiotrophin.
Multivalent Interaction of CS with Pleiotrophin-Pleiotrophin and midkine are composed of two domains: N-and C-terminal halves (1, 2). The C-terminal half of these growth factors contains two clusters of basic amino acids (Clusters I and II), which have been shown to function as heparin binding sites for midkine (1,14). The basic amino acid residues in Clusters I and II are conserved or type-conserved in pleiotrophin and midkine except for one arginine residue in Cluster II (14). We observed no difference between the bindings of pleiotrophin and midkine to phosphacan probably because of such common structural features (9, 10).
Based on an NMR analysis of midkine in solution, Iwasaki et al. (14) suggested that midkine forms a head-to-head dimer in the presence of heparin oligosaccharides. In this dimeric structure, two Cluster IIs from each midkine fused to form a heparin binding site at the dimer interface, and the Cluster Is work as separate heparin binding sites on the distal sides of the dimer (Fig. 9B). These heparin binding sites of the midkine dimer were considered to fit the three sulfate group clusters formed in heparin 20 mer (14). Furthermore, Kaneda et al. (23) indicated that heparin 22 mer but not 12 mer inhibited midkine-induced neurite outgrowth. These observations suggest that the midkine dimer is the basic functional unit, which interacts with heparin ϳ20 mer. Previously, we indicated that a mutation at Arg 78 in Cluster I of midkine resulted in the loss of CS-dependent high affinity binding to phosphacan, suggesting that heparin binding sites in midkine serve as CS binding sites (10).
On the other hand, the heparin binding site of pleiotrophin was not as clearly identified as that of midkine. Despite the highly conserved structure of pleiotrophin and midkine, it has been reported that both N-terminal and C-terminal halves of HB-GAM (pleiotrophin) strongly bound with heparin (24). Whereas Cluster II of midkine contains 3 basic amino acid residues, that of HB-GAM contains only 2 basic amino acids (14). This structural difference might result in the different heparin binding mechanism between these molecules. Although the heparin binding sites of pleiotrophin and midkine might be somewhat different, the HB-GAM monomer was considered to bind with the heparin decasaccharide like midkine (25,26). Quite recently, Bao et al. (21) demonstrated that CS octasaccharides containing D unit weakly bound to a pleiotrophin-conjugated affinity column, suggesting that pleiotrophin monomer accommodate CS ϳ8 mer. Based on these observations, we assume that the pleiotrophin monomer has two heparin binding sites, which also serve as CS binding sites (Fig. 9B). Fig. 1 indicated that CS oligosaccharides longer than 18 mer bound significantly to pleiotrophin immobilized on the sensorchips. We could not detect the binding of CS oligosaccharides shorter than 16 mer, although CS 8ϳ10 mer could bind to pleiotrophin-conjugated affinity column (21). Because affinity chromatography can separate very weak affinity substances, it seems that the affinities of CS oligosaccharides shorter than 16 mer were too weak to be detected using BIAcore system.   On the other hand, the binding of phosphacan to pleiotrophin was inhibited by Ն16-mer CS oligosaccharides (Fig. 2), suggesting that the CS 16ϳ18 mer was the basic functional unit, which could presumably interact with the pleiotrophin dimer. Because pleiotrophin was immobilized on the sensor chip through a flexible 100-nm carboxymethylated dextran, it would move fairly freely, and the pleiotrophin molecules would be able to associate with each other. Thus, it was expected that CS could interact with pleiotrophin multimers on the sensor surfaces. Although the 18-mer fraction showed significant binding to pleiotrophin, its binding affinity was very low (K D Ͼ20 M). On the other hand, the 46-mer fraction showed high affinity binding to pleiotrophin (K D ϭ 30 nM) under the same condition (Table 2), in which a saturated amount of pleiotrophin was immobilized on the sensor chips (ϳ6,000 RU). Whereas the sensorgrams of the interaction of pleiotrophin with 34ϳ46-mer CS polysaccharides fitted the bivalent analyte model well, CS fractions shorter than 28 mer did not fit this model (Fig. 4 and Table  2). This suggested that Ն34-mer CS polysaccharides could bind with two pleiotrophin units.
When a semisaturated amount of pleiotrophin was immobilized on the sensor chips (4,753 RU; high density condition), the 46-mer fraction showed high affinity binding to pleiotrophin (K D ϭ 62 nM) (Fig. 3). The affinity of the 46-mer fraction for pleiotrophin remarkably decreased (K D ϭ 2.5 M), when the density of immobilized pleiotrophin was lowered (2,461 RU; low density condition). When a moderate amount of pleiotrophin was immobilized on the sensor chip (2,905 RU; medium density condition), the affinity was also intermediate between the two conditions (K D ϭ 0.4 M). Analysis of the sensorgrams shown in Fig. 3A using the bivalent analyte model indicated that K D1 was 0.5 M, the value of which was ϳ8-fold larger than that of K D (Table 1). Based on the observation that CS 46 mer accommodated 4 pleiotrophin molecules (Fig. 5), we suggest that the 46-mer CS fraction bound first to the pleiotrophin dimer with a K D1 of 0.5 M, and then to a second pleiotrophin dimer, resulting in a stronger overall binding (K D ϭ 62 nM) (Fig. 9B,  (a)). Under the medium density condition, the K D and K D1 values were calculated to be 0.4 and 2.2 M, respectively (Table 1). This K D value was close to the K D1 value in the high density condition, suggesting that CS interacted with two pleiotrophin molecules under the medium density condition (Fig. 9B, (b)). On the other hand, the sensorgrams under the low density condition did not fit the bivalent analyte model, suggesting that CS bound with one molecule of pleiotrophin. The K D value was calculated to be 2.5 M, which was close to the K D1 value under the medium density condition (K D ϭ 2.2 M). These results suggested that the K D values for the interaction of the 46-mer CS fraction with the monomer, dimer, and tetramer were ϳ2,500, ϳ500, and ϳ60 nM, respectively (Fig. 9B). However, we cannot exclude the possibility that the reaction 1 under the medium density condition (Fig. 9B, (b)) and the binding under the low density condition (Fig. 9B, (c)) were mediated by pleiotrophin dimers.
Importance of Oversulfated Structures in the Interaction of CS with Pleiotrophin-When the 22-mer CS fraction was subjected to strong anion exchange HPLC, CS oligosaccharides were separated depending on the amounts of D-and E-type structures (Table 3). Fractions F-h-1 and -2, which were estimated to contain only one D unit per chain, showed no binding to pleiotrophin. Fractions F-h-3 and -4, which were considered to be composed of a mixture of CS oligosaccharides containing 1 or 2 D units per chain, slightly bound to pleiotrophin, but the affinity was very low. On the other hand, F-h-6, which was estimated to

TABLE 3 Interaction between pleiotrophin and CS 22-mer fractions with various contents of D and E units
The K D values were calculated from the sensorgrams using six or more concentrations of analyte (CS) in two or three independent experiments. contain 2 D units per chain, showed significant affinity for pleiotrophin (K D ϭ 2.7 M) (Fig. 9C, (b)). F-h-12 was considered to contain 3 D and/or E units per chain, and showed relatively high affinity for pleiotrophin (K D ϭ 0.36 M) (Fig. 9C, (a)). This suggests that the strength of binding between CS and pleiotrophin depended on the number of oversulfated units that could presumably interact with the heparin binding sites of pleiotrophin (Fig. 9C).
Because the CS fractions used in this study were still a mixture of oligosaccharides with various sequences, we do not know the precise CS sequence involved in binding with pleiotrophin. However, the structure of shark cartilage CS-D was characterized by a high frequency of A-Dcontaining sequences such as A-D-A, A-D-C, and D-A-D-A, and no D-D tetrasaccharide sequence was found (27,28). These A-D-containing regions seemed to be separated by sequences without a D unit such as C-C, C-A, C-C-C, C-C-A, C-C-C-C, and C-C-C-A (27,28). Combinations of these sequences would generate CS 22 mer containing 1, 2, or 3 D units. Because of this structural limitation, CS-D would not contain sequences with densely clustered D units. On the other hand, E units could be highly clustered in the CS sequences. Zou et al. (18) demonstrated that an artificial CS-E structure, of which up to 95% was E units, could be formed in various CS preparations using squid or human N-acetylgalactosamine 4-sulfate 6-O-sulfotransferase. The affinity of these artificial CS-E for midkine correlated with the amounts of E unit, and the CS-E with a dense E unit cluster bound quite strongly with midkine (18). The clustered E units in CS-E might interact with each of Cluster Is and IIs on the midkine dimer, leading to strong interaction. Such different features of D-and E-type structures might play important roles in the determination of the binding affinity of CS for pleiotrophin and midkine, leading to the different functions of D unit-and E unit-rich CS.
Hikino et al. (29) demonstrated that shark cartilage CS-D and squid cartilage CS-E exerted distinct effects on cultured embryonic hippocampal neurons. Whereas CS-D promoted the outgrowth of short dendrite-like neurites, CS-E stimulated the outgrowth of long axon-like processes (29). Using various CS/dermatan sulfate (DS) preparations, they found that D-and/or iD-(IdoA(2S)␣1-3GalNAc(6S)) type structures were involved in the promotion of the outgrowth of dendrite-like processes. On the other hand, E and/or iE (IdoA␣1-3GalNAc(4S,6S))type structures were suggested to play roles in the promotion of the outgrowth of axon-like neurites (29). Bao et al. (20) also reported that CS/DS hybrid chains purified from embryonic pig brain efficiently bound with pleiotrophin. These CS/DS hybrid chains were separated on a pleiotrophin-conjugated column into unbound, low affinity, and high affinity fractions. The latter two fractions promoted the outgrowth of dendrite-and axon-like neurites by cultured embryonic hippocampal neurons, respectively (20). Pleiotrophin expressed by glial cells mediated the dendrite-promoting activity of the low affinity CS/DS fraction, but this growth factor was not involved in the axon-promoting activity of the high affinity fraction (20). These low and high affinity CS/DS fractions showed high contents of D/iD and E/iE units, but the latter fraction was more heavily sulfated and contained a larger amount of the E-type structure than the former one (20). The low affinity binding of D unit-rich CS with pleiotrophin might play important roles in the formation of dendrites.
CS and PTP Signaling-Midkine is multimerized by tissue transglutaminase, and this process is highly promoted by glycosaminoglycans such as heparin (13). Pleiotrophin is also easily multimerized by this enzyme under similar conditions. 3 Kojima et al. (13) indicated that multimerization of midkine is required for its enhancing effects on plasminogen activator activity in bovine aortic endothelial cells. These observations suggest that pleiorophin/midkine multimers, but not monomers activated the receptors. At present, N-syndecan, anaplastic lymphoma kinase (ALK) and PTP have been identified as receptors for pleiotrophin/midkine (3,4,9,30,31). Among them, the signal transduction mechanism of PTP has begun to be elucidated, and it has been proposed that tyrosine phosphatase activity of PTP is inactivated by pleiotrophin probably through dimerization of this receptor (4,11). Recently, we demonstrated that PTP-pleiotrophin signaling is involved in the morphogenesis of cerebellar Purkinje cells (32). In the postnatal 3 N. Maeda, unpublished observation. FIGURE 9. Models of the binding of CS oligosaccharides with pleiotrophin. A, bivalent analyte model is composed of two binding reactions (reaction 1 and reaction 2). In reaction 1, C (CS) binds with the first P (pleiotrophin) to form the complex CP. In reaction 2, CP binds with a second P to form the complex CP2. The rate constants of each reaction (k a1 , k d1 , k a2 , and k d2 ) can be calculated using BIAevaluation software 4.1. The affinity of C for P is calculated by the equation, K D1 ϭ k d1 /k a1 . B, interaction of fraction F-a (ϳ46 mer) with various amounts of pleiotrophin shown in Fig. 3 was schematically modeled. Because this fraction contained ϳ20% D unit, 5 D units (yellow circle) were estimated to be present in this polysaccharide. The pleiotrophin dimer (blue ellipse) was assumed to have four heparin binding sites (green and red rectangles). Under the high density condition (a), the 5 D units in 46 mer could interact with 5 heparin binding sites on two pleiotrophin dimers. The reaction fitted the bivalent analyte model, and K D1 was calculated to be 0.5 M. On the other hand, Scatchard analysis indicated that the overall affinity of F-a polysaccharide for pleiotrophin was ϳ8-fold stronger (K D ϭ 60 nM) than the affinity of reaction 1. Under the medium density condition (b), K D1 and K D were 2.2 and 0.4 M, respectively. This K D value was close to the K D1 value of reaction 1 under the high density condition, suggesting that the overall reaction was the binding of CS with two pleiotrophin monomers (b). However, there also is a possibility that CS bound with pleiotrophin dimer in the reaction 1 under this condition. The reaction under the low density condition (c) did not fit the bivalent analyte model, and K D was calculated to be 2.5 M. This K D value was close to the K D1 value of reaction 1 under the medium density condition, suggesting that CS bound with the pleiotrophin monomer (c). However, we again cannot exclude the possibility that CS bound with pleiotrophin dimer under this condition. C, as summarized in Table 3, the affinity of ϳ22-mer oligosaccharides for pleiotrophin depended on the CS structure. CS oligosaccharides containing 3 oversul- cerebellar cortex, PTP, pleiotrophin and D unit-rich CS were accumulated in the molecular layer, and it has been considered that these molecules cooperatively played roles in the dendrite formation of Purkinje cells (32,33). Transglutaminase type 2 was also reported to be expressed in the molecular layer (12), suggesting that these components constitute the extracellular signaling complex involved in the development of Purkinje cell dendrites.
In the postnatally developing rodent cerebellum, phosphacan-bearing CS rich in D unit was abundantly expressed (33). This CS proteoglycan might trap pleiotrophin through D unit-rich CS and stimulate the multimerization of pleiotrophin by transglutaminase. The multimerized pleiotrophin might move to CS of PTP expressed on the surface of Purkinje cells, leading to the cross-linking of the receptor molecules. The efficiency of this cross-linking process could be highly dependent on the CS structure of PTP and pleiotrophin multimerization. If CS of PTP contained a cluster of oversulfated disaccharide units and pleiotrophin multimers were present, pleiotrophin would bind CS chains of two PTP and stably cross-link the receptor molecules leading to strong signaling. In this case, a higher pleiotrophin multimer should cause more stable cross-linking of PTP. If the level of oversulfated disaccharide units was low and/or only a lower pleiotrophin multimer such as a dimer was present, the dimerization of PTP, and the signaling might be transient and weak.
We previously indicated that phosphacan with the CS-containing D unit showed ϳ5-fold stronger affinity for pleiotrophin (K D ϭ 0.14 nM) than phosphacan without D unit (16). This difference in the affinity disappeared after chondroitinase ABC treatment of phosphacan, and both preparations showed a similar lower affinity binding to pleiotrophin. This is consistent with the present results that the affinity of CS for pleiotrophin depends on the D unit content. However, it should be noted that the core protein portion of phosphacan could also contribute to the binding to pleiotrophin (9,16). Although we do not know the mechanism of the binding between pleiotrophin and phosphacan core protein, the affinity of phosphacan core protein for pleiotrophin was relatively high (K D ϭ 1.5ϳ13 nM) (9,16). This suggests that there is cooperative interaction among the core protein, CS, and pleiotrophin. In this case, a lower pleiotrophin multimer such as a dimer and CS with few D units might be sufficient to cause stable cross-linking of PTP.
In this study, we demonstrated that the binding of pleiotrophin with CS was dependent on the chain length and amount of oversulfated structure in CS and on the pleiotrophin multimerization. At present, it is unknown whether specific CS sequences are required for the binding with pleiotrophin. Future study is required to study this problem.