Pleiotrophin, a multifunctional cytokine and growth factor, induces leukocyte responses through the integrin Mac-1

Pleiotrophin (PTN) is a multifunctional, cationic, glycosaminoglycan-binding cytokine and growth factor involved in numerous physiological and pathological processes, including tissue repair and inflammation-related diseases. PTN has been shown to promote leukocyte responses by inducing their migration and expression of inflammatory cytokines. However, the mechanisms through which PTN mediates these responses remain unclear. Here, we identified the integrin Mac-1 (αMβ2, CD11b/CD18) as the receptor mediating macrophage adhesion and migration to PTN. We also found that expression of Mac-1 on the surface of human embryonic kidney (HEK) 293 cells induced their adhesion and migration to PTN. Accordingly, PTN promoted Mac-1–dependent cell spreading and initiated intracellular signaling manifested in phosphorylation of Erk1/2. While binding to PTN, Mac-1 on Mac-1–expressing HEK293 cells appears to cooperate with cell-surface proteoglycans because both anti-Mac-1 function-blocking mAb and heparin were required to block adhesion. Moreover, biolayer interferometry and NMR indicated a direct interaction between the αMI domain, the major ligand-binding region of Mac-1, and PTN. Using peptide libraries, we found that in PTN the αMI domain bound sequences enriched in basic and hydrophobic residues, indicating that PTN conforms to the general principle of ligand-recognition specificity of the αMI domain toward cationic proteins/peptides. Finally, using recombinant PTN-derived fragments, we show that PTN contains two distinct Mac-1–binding sites in each of its constitutive domains. Collectively, these results identify PTN as a ligand for the integrin Mac-1 on the surface of leukocytes and suggest that this interaction may play a role in inflammatory responses.


Pleiotrophin (PTN)
is a glycosaminoglycan-binding cytokine and growth factor with potent mitogenic and angiogenic activities. Together with the related protein midkine, it forms a unique family of cytokines that are normally expressed during embryogenesis and neonatal development but are also produced by injured tissues during repair and regeneration (1,2). PTN has also been identified as crucial in the maintenance of hematopoietic stem cells (3,4). Moreover, PTN levels have been shown to be highly elevated in a large number of cancer cell lines as well as being correlated with their metastatic abilities (5)(6)(7)(8)(9)(10)(11)(12)(13). Several studies have also associated PTN with inflammation and leukocyte recruitment (13,14). There are now efforts both to develop PTN inhibitors to treat cancer and to use PTN itself for prevention of tissue injury during ischemia (15)(16)(17)(18). Several proteins have been proposed as receptors of PTN. The most studied among these include the heparan sulfate proteoglycan (HSPG) N-syndecan and the chondroitin sulfate proteoglycan receptor-type protein tyrosine phosphatase (19 -21). Several non-proteoglycan receptors are also known to bind PTN, including nucleolin (22,23) and the integrin ␣ V ␤ 3 (24).
Structurally, PTN is made up of two thrombospondin type-1 repeat (TSR) domains flanked by unstructured termini, both of which are highly basic (25,26). The TSR domains also contain a number of basic amino acid clusters with two clusters in the C-terminal TSR domain having the highest affinity for glycosaminoglycans (25,27). It is thought that the basic nature of PTN is crucial to its ability to interact with receptors and initiate signaling.
Although the role of PTN in mitogenesis and angiogenesis has been extensively researched, its function in inflammation has not been widely explored. Given the fact that injured tissues express PTN, it is likely that leukocytes are exposed to it at sites of inflammation. Indeed, some studies have suggested that PTN plays a role in recruitment of leukocytes during inflammation (13,14). However, the identity of the PTN receptor(s) on leukocytes that mediate this response and the role of PTN signaling in these cells remain unknown. We hypothesized that integrin ␣ M ␤ 2 (Mac-1, CD11b/CD18, CR3) can serve as a receptor for PTN. This hypothesis was based on our recent finding that Mac-1 binds cationic proteins, many of which, similar to PTN, interact with heparin (28 -30).
Mac-1 is an adhesion receptor expressed on the surface of myeloid cells such as monocyte/macrophages and neutrophils. As an adhesion receptor, Mac-1 is crucial to numerous leukocyte responses, including migration, phagocytosis, degranulation, adherence to microorganisms, and others (31,32). Among integrins, which are in general notorious for their capacity to recognize multiple ligands, Mac-1 is its most promiscuous member. In contrast to many other integrins that recognize the RGD motif, Mac-1 does not bind this sequence. Instead, as we have recently revealed, Mac-1 has preferences for sequences enriched in basic and hydrophobic amino acid residues (29,30,33,34). The identification of Mac-1 recognition motifs has led to the prediction that other cationic proteins and peptides released from leukocytes and other cells during tissue injury can serve as Mac-1 ligands (29). Most of Mac-1's ligand binding and thus its broad ligand binding specificity can be attributed to the ␣ M I domain, a region of ϳ200 amino acid residues inserted into the ␣ M subunit, inasmuch as its removal in Mac-1 eliminates ϳ85% of ligand binding (35).
In this study, we present biochemical and cell biology evidence that PTN is a ligand for Mac-1. Using various Mac-1expressing cells, we demonstrate that Mac-1 supports adhesion to PTN, promotes spreading, and induces activation of the MAP kinase Erk1/2. Furthermore, PTN induces a potent migratory response in Mac-1-expressing human embryonic kidney 293 (Mac-1 HEK293) cells and in isolated murine macrophages. Direct interaction between PTN and the ␣ M I domain, the major ligand-binding region of Mac-1, has been shown using biolayer interferometry analyses and confirmed by solution NMR spectroscopy. Finally, using peptide array analysis, we identified the putative ␣ M I domain-binding sites in both TSR domains of PTN and validated their importance in cell adhesion and signaling assays with individual recombinant PTN fragments.

Mac-1 is involved in adhesion of Mac-1-expressing cells to PTN
To assess whether PTN can bind Mac-1, we initially examined the ability of Mac-1 HEK293 to adhere to immobilized PTN. As shown in Fig. 1A, Mac-1 HEK293 cells adhered to immobilized PTN in a concentration-dependent manner with saturable adhesion achieved with a PTN coating concentration of ϳ100 nM. Wild-type HEK293 cells also strongly adhered to PTN. However, the presence of Mac-1 significantly augmented adhesion (Fig. 1A). Because PTN has high affinity for glycosaminoglycans and HSPG N-syndecan binds PTN, we examined the possibility that HSPGs on wild-type HEK293 cells mediate adhesion to PTN. Preincubation of cells with heparin (1 g/ml) almost completely eliminated adhesion (Fig. 1B), indicating that on the surface of these cells HSPGs are entirely responsible for PTN binding. However, adhesion of Mac-1 HEK293 cells was reduced by only ϳ60%, suggesting that Mac-1 on these cells may be involved in binding PTN. To investigate this possibility, Mac-1 HEK293 cells were treated with the function-blocking mAb 44a (directed against the ␣ M I domain of Mac-1). Although adhesion tended to decrease, it was not significantly altered with the maximal blocking effect (19 Ϯ 10%) observed at 10 g/ml of mAb. Likewise, although heparin inhibited adhesion in a dose-dependent manner, the maximal blocking effect attained did not exceed 59 Ϯ 6%. However, when Mac-1 HEK293 cells were treated with both mAb 44a and heparin, cell adhesion was inhibited by Ͼ95%. These results suggest both HSPG and Mac-1 can act as receptors for PTN.
Consistent with the role of Mac-1 in adhesion to PTN, Mac-1 HEK293 cells spread with the formation of actin filaments as detected by staining with Alexa Fluor 546-conjugated phalloidin (Fig. 1C, upper panels). By contrast, spreading of wild-type HEK293 cells was visibly less even after 24 h (Fig. 1C, bottom panels). Quantification of cell spreading confirmed this observation (Fig. 1D).
To investigate the relevance of PTN-Mac-1 interaction in immune cells, we performed adhesion assays using IC-21 cells, a murine macrophage cell line naturally expressing Mac-1. As shown in Fig. 1E, IC-21 cells adhered to immobilized PTN, and ϳ65% of adhesion was inhibited by mAb M1/70 against the murine ␣ M integrin subunit. This is in contrast to Mac-1 HEK293 cells whose adhesion to PTN was reduced modestly by mAb 44a (Fig. 1B). Moreover, heparin, which greatly decreased adhesion of Mac-1 HEK293 cells to PTN, had little effect on adhesion of IC-21, suggesting that Mac-1 on the surface of IC-21 cells largely contributes to the interaction with PTN. Together, these results identify PTN as an adhesive ligand for Mac-1 and suggest that on the surface of different cells both Mac-1 and HSPGs are involved in PTN adhesion, albeit to a different extent.

PTN bound to extracellular matrix (ECM) proteoglycans supports adhesion of Mac-1 HEK293 cells
Because PTN is secreted from cells and deposited into the ECM, we explored whether PTN bound to the ECM components is capable of supporting adhesion. Because Mac-1 is known to interact with a number of ECM proteins, including fibronectin, vitronectin, collagens, Cyr61, and others (36), a heterogeneous mixture of these molecules such as those found in Matrigel was not suitable as a substrate for investigations of PTN-Mac-1 interactions. PTN is only known to bind proteoglycans; therefore, the most relevant form of PTN encountered by immune cells in vivo is likely to be PTN anchored to ECM proteoglycans. To simulate such an environment, we studied cell adhesion to PTN prebound to aggrecan, a common proteoglycan found in the ECM. Fig. 2 shows that at two concentrations of PTN tested both Mac-1 HEK293 and wild-type HEK293 (HEK293) cells adhered to aggrecan-bound PTN with Mac-1 HEK293 cells adhering at a significantly higher level than HEK293 cells (p Ͻ 0.001). Neither type of cells had affinity for aggrecan itself. Notably, binding of PTN to aggrecan did not reduce cell adhesion, indicating that PTN in its aggrecanbound form remains an efficient Mac-1 ligand.

PTN induces migration of Mac-1-expressing cells
PTN is known to induce cell migration, and in some instances, this effect has been shown to be integrin-dependent (24,37). Therefore, we investigated whether Mac-1 can support PTN-induced migration. In particular, using a Transwell sys-PTN is a ligand for integrin Mac-1 tem, we compared the ability of wild-type and Mac-1 HEK293 cells to migrate toward PTN. Previous studies reported that these cell lines are a useful system for assessing the role of Mac-1 in migration (38). PTN induced a potent migratory response (Fig. 3, A and B). In contrast, wild-type HEK293 migrated only slightly. Direct evidence that migration of Mac-1 HEK293 cells was dependent on Mac-1 was obtained in the experiments in which anti-Mac-1 mAb 44a was tested. Although mAb 44a eliminated cell migration, an IgG1 isotype control for this mAb had no significant effect (Fig. 3B). Heparin (1 g/ml) also inhibited migration to a modest but significant degree.
In a separate set of experiments, we tested whether PTN can induce migration of mouse macrophages isolated from the peritoneum of wild-type and Mac-1-deficient mice. Macrophages were purified from a total population of peritoneal cells, and their migration was examined in a Transwell system. As shown in Fig. 3, C and D, PTN induced a strong migratory response in wild-type macrophages. PTN also induced migration of Mac-1-deficient macrophages. However, compared with wildtype cells, migration of Mac-1-deficient macrophages was significantly impaired (by 76 Ϯ 21%). These results indicate that Mac-1 is a crucial mediator of PTN-induced cell migration. Furthermore, because macrophages lacking Mac-1 can

PTN is a ligand for integrin Mac-1
migrate, albeit at a significantly reduced rate, other integrins (most likely ␣ V ␤ 3 ) may contribute to migration.

PTN induces activation of MAP kinase (MAPK) Erk1/2
The interaction of integrins with ligands is known to induce activation of MAP kinases. To determine whether PTN activates MAPKs via binding to Mac-1, adherent Mac-1 HEK293 cells were treated with different concentrations of PTN, and phosphorylation of Erk1/2 was examined. A PTN concentration of 0.7 M induced a significant increase in phosphorylation of Erk1/2, and this response was blocked by mAb 44a and heparin (Fig. 4, A and B). The simultaneous use of two reagents increased the inhibitory effect; however, the cumulative effect did not reach a statistical significance. Treatment of control HEK293 cells with PTN did not induce phosphorylation of Erk1/2 (Fig. 4A). These data suggest that in this system initiation of intracellular signaling by PTN requires both Mac-1 and HSPGs.

Biochemical analyses of the interaction between PTN and ␣ M I domain
To further characterize the Mac-1-PTN interactions and determine domains of Mac-1 responsible for PTN binding, we analyzed the binding parameters of the interaction between the ␣ M I domain and PTN. We focused on the ␣ M I domain because this domain is the major ligand-binding region in Mac-1, and previous studies have shown that several basic proteins and peptides interact with it (28 -30, 34, 39). To measure the affinity of the ␣ M I domain-PTN interaction, we used biolayer interferometry (BLI) in which PTN was coupled to the matrix coating the biosensor via lysines. The interaction between PTN with both active and nonactive forms of the ␣ M I domain was measured in 20 mM HEPES buffer containing 150 mM NaCl, 1 mM MgCl 2 , and 0.05% Tween 20, pH 7.5. When PTN was allowed to interact with various concentrations of soluble active ␣ M I domain, a dose-dependent binding was observed (Fig. 5, A and  B). The dissociation rate constant (K d ) value determined from the maximal responses achieved in the equilibrium portion of the BLI sensorgrams was 1.2 Ϯ 0.2 M. The interaction of PTN with nonactive ␣ M I domain was also detected (Fig. 5C); however, binding was lower than that of the active form, which precluded an accurate assessment of the K d value. In contrast to the ␣ M I domains, the active form of the I domain of integrin ␣ L ␤ 2 , which has narrow ligand specificity and binds only intercellular adhesion molecule and JAM-1 molecules, did not bind PTN (Fig. 5C). To assess how divalent cations might affect the interaction of PTN with ␣ M I domain, the binding of active ␣ M I domain was measured in the presence of 5 mM EDTA. As shown in Fig. 5C, EDTA only modestly decreased binding of active ␣ M I domain. The latter result indicates that the ␣ M I domain binding to PTN is not dependent on the presence of Mg 2ϩ .
Because recognition by the ␣ M I domain of many of its ligands depends on basic amino acids, we sought to determine whether immobilization of PTN onto the BLI sensor using lysine side chains may potentially reduce PTN's binding to ␣ M I domain. Therefore, we repeated the experiments by randomly biotinylating glutamate and aspartate side chains of PTN and attaching the biotinylated PTN to streptavidin-coated sensors. As determined from the sensorgrams shown in Fig. 5D, the alternative mode of immobilization did not change considerably the binding affinity of the two proteins (K d ϭ 3.9 Ϯ 0.1 M).
We also examined the interaction of PTN with the active ␣ M I domain using solution NMR. Fig. 6A shows the 1 H-15 N HSQC spectrum of PTN in the presence and absence of 1 molar eq of active ␣ M I domain. Addition of the ␣ M I domain decreased the signal intensities of residues in the TSR domains of PTN by ϳ40%, whereas PTN residues in the unstructured N and C termini were not affected by ␣ M I domain (Fig. 6B). A signal intensity decrease often takes place when a small protein binds a larger protein or if the interactions between the two is dynamic and the kinetics of the interaction is on the intermediate NMR time scale (40). The observation of a signal intensity decrease therefore indicates that the TSR domains of PTN are involved in binding ␣ M I domain. Together, these data indicate that PTN has a significant affinity for the active form of the ␣ M I domain.

Screening of the PTN-derived peptide libraries for ␣ M I domain binding
To localize the putative ␣ M I domain-binding sites in PTN, we screened the cellulose-bound peptide library representing the complete sequence of PTN. The library, consisting of 9-mer peptides with a 3-residue offset (Fig. 7A), was probed with 125 Ilabeled recombinant ␣ M I domain. The screening identified four peptide clusters that strongly bound ␣ M I domain (Fig. 7B). These clusters correspond to the following PTN sequences: 1-15 (cluster 1; spots 1-3), 46 -69 (cluster 2; spots 16 -21), 82-99 (cluster 3; spots 28 -31), and 106 -132 (cluster 4; spots 36 -42). Control spots containing only ␤-Ala spacer (spots 49 and 50) were negative as were many other peptides. The ␣ M I domain-binding peptides were analyzed by a computer pro-

PTN is a ligand for integrin Mac-1
gram designed to determine the capacity of peptides to interact with the ␣ M I domain (29). The program assigns each peptide an energy value, which serves as a measure of probability that the ␣ M I domain binds this sequence: the lower the energy, the higher the likelihood that the sequence binds the ␣ M I domain. As previously determined, strong ␣ M I domain-binding peptides derived from various Mac-1 ligands have energy values in the range of Ϫ20 to 2 kJ/mol. The analyses showed a good relationship between the energy scores and the ␣ M I-binding activity of PTN-derived peptides revealed in peptide scans (Fig. 7A). In agreement with previous findings (29), peptides enriched in positively charged and hydrophobic residues had the highest affinities for the ␣ M I domain (spots 2, 16 -19, 29, 38, and 42). For example, the stretch of overlapping peptides KQT-MKTQRCKIPCNWKKQ (spots 16 -19) and peptide KTRTG-SLKR (spot 29) contain several short motifs, HyB and BHy where Hy represents any hydrophobic residue and B (basic) is either arginine or lysine, that have been shown to be abundant in the ␣ M I domain binders (29). When mapped onto the threedimensional structure, the ␣ M I domain-binding sequences were found in both TSR domains as well as the N-and C-terminal tails (Fig. 7C). These results indicate that, in the case of PTN, the ␣ M I domain has affinity for its positively charged regions, and thus recognition specificity toward PTN conforms to the general principles of ligand recognition exhibited by Mac-1 toward cationic proteins (29).

Individual TSR domains of PTN support Mac-1-dependent cell adhesion and induce activation of Erk1/2
The above data from peptide library screening suggest that PTN may contain several ␣ M I domain-binding sites. These sites have been mapped to the N-and C-terminal (NTD and CTD) TSR domains of PTN as well as the C-terminal tail. To investigate the contribution of each domain to Mac-1 binding, we prepared truncated PTN fragments corresponding to the NTD (residues 1-57), CTD (residues 58 -114), and PTN-short (residues 1-114) with the C-terminal tail deleted (Fig. 8A) and examined their ability to support cell adhesion and induce MAPK activation (Fig. 8, B-F). The data showed that removing the C-terminal tail of PTN did not reduce significantly the ability of PTN-short to support adhesion of Mac-1 HEK293 cells compared with intact PTN (Figs. 8B and 1A). Furthermore, the pattern of adhesion of HEK293 cells to PTN-short (Fig. 8B) was similar to that of wild-type PTN (Fig. 1A). Both NTD and CTD were also able to support adhesion (Fig. 8, C and D). However, removing either one of the structural domains of PTN resulted in the reduction of their adhesion-promoting activity (Fig. 8, C and D). In particular, although a 100 nM coating concentration of wild-type PTN was sufficient to achieve a saturable level of adhesion, ϳ5and ϳ10-fold higher concentrations of CTD and NTD, respectively, were required to reach saturation. Furthermore, the maximal level of adhesion mediated by these frag-

PTN is a ligand for integrin Mac-1
ments was lower than that mediated by intact PTN. The effect of the removal of either structural domain of PTN on proteoglycan binding was even more severe. Specifically, wild-type HEK293 cells adhered poorly to NTD and CTD even at a concentration of 1 M (Fig. 8, C and D). To rule out the possibility that differences in the adhesion-promoting activity of PTNderived fragments were due to their different coating efficiencies, we examined the ability of soluble fragments to inhibit adhesion. Wild-type PTN and all fragments blocked adhesion in a dose-dependent manner (supplemental Fig. S1). At 5 M, wild-type PTN was the most potent inhibitor (78 Ϯ 13% inhibition of adhesion) followed by PTN-short (68 Ϯ 15%), CTD (57 Ϯ 10%), and NTD (38 Ϯ 12%) (Fig. 8E), thus corroborating the results of adhesion assays. Consistent with the presence of the Mac-1-binding sites in both NTD and CTD domains, treatment of Mac-1 HEK293 cells with each fragment resulted in phosphorylation of Erk1/2, whereas only basal phosphorylation was detected in untreated cells ( Fig. 8F and supplemental Fig. S2).

Discussion
PTN is a 15-kDa basic heparin-binding protein that induces proliferation, cell growth, and angiogenesis in a wide variety of cells by interacting with specific receptors, including receptortype protein-tyrosine phosphatase (RPTP) ␤/, anaplastic lymphoma kinase, N-syndecan, and ␣ V ␤ 3 (3-5, 9, 24). PTN has also been implicated in mediating inflammation based on its ability to trigger in vivo migration of neutrophils and monocyte/ macrophages (13,14). In this study, we demonstrated that PTN induces a potent migratory response in macrophages in vitro and identified integrin Mac-1 as the major receptor mediating cell migration. In support of this finding, we showed that PTN induces migration of Mac-1-expressing HEK293 cells and wild-type mouse macrophages but not wild-type HEK293 cells or Mac-1-deficient macrophages. Mac-1 expressed on HEK293 cells also augmented adhesion to PTN compared with wild-type HEK293 cells and specifically promoted cell spreading. Similarly, PTN also supported adhesion of IC-21 mouse macrophages in a Mac-1-dependent manner. In addition, Mac-1 was essential for PTN-induced activation of Erk1/2. Finally, we showed direct interaction between ␣ M I domain, the ligand-binding region of Mac-1, and PTN using BLI and NMR. These observations establish PTN as a novel ligand for integrin Mac-1 and suggest a role for this protein in the pathophysiologic functions of myeloid leukocytes, which specifically express this receptor.
Integrin Mac-1 is a member of the ␤ 2 subfamily of integrin adhesion receptors and is the major receptor on the surface of neutrophils and monocytes/macrophages. This receptor contributes to leukocyte adhesion to and diapedesis through the inflamed endothelium and controls leukocyte migration to sites of inflammation. Moreover, ligand engagement by Mac-1 initiates a variety of cellular responses, including phagocytosis, neutrophil degranulation and aggregation, and expression of cytokines/chemokines and many other pro-and anti-inflammatory molecules (31,32,41). Innumerable roles played by Mac-1 in leukocyte biology arise from its multiligand binding properties. Indeed, this receptor exhibits broad recognition specificity and is capable of binding an extremely diverse group of protein and nonprotein ligands. We recently showed that, within its many ligands, Mac-1 binds not to a specific amino acid sequence but rather has a preference for the sequence patterns consisting of a core of positively charged residues flanked by hydrophobic residues (29). In particular, the binding motifs for Mac-1 can be coded as HyBHy, HyHyBHy, HyBHyHy, and HyHyBHyHy. Other amino acids can also be found, but in general, their proportion within the Mac-1-binding motifs is very small, and negatively charged (acidic) residues are largely omitted.
Analyses of the PTN sequence by a previously developed program (29) that predicts the Mac-1-binding sites in its ligands showed that PTN contains several potential ␣ M I domain-recognition sequences, and this prediction was confirmed experimentally (Fig. 7). Screening of the peptide library spanning the sequence of PTN demonstrated the presence of four clusters composed of overlapping ␣ M I domain-binding peptides (Fig. 7,  A and B). The ␣ M I domain-binding peptides in clusters 2 and 3 correspond to the segments that form the structural NTD and CTD of PTN. Sequences in the C-terminal part of cluster 2 correspond to the segment in the hinge region (residues 58 -66) connecting the TSR domains (Fig. 7, A and C). In addition, the ␣ M I domain-binding sequences identified within clusters 1 and 4 are present in the N-and C-terminal tails, respectively (Fig. 7C). The PTN structure shows that each PTN domain possesses large basic surfaces formed by segments

PTN is a ligand for integrin Mac-1
enriched in lysine and arginine that are brought into close proximity by PTN folding (Fig. 9) (25). Within CTD, Lys 68 , Lys 84 , Arg 86 , Lys 91 , Arg 92 , and Lys 107 contribute to the formation of an extended basic surface (Fig. 9B). Within NTD, Arg 35 , Arg 39 , and Lys 49 contribute to the basic surface on one side of the ␤-sheet forming the domain (Fig. 9A). Side chains of Tyr 69 , Leu 90 , Val 103 , and Ile 105 in CTD form a small hydrophobic cluster in the middle of CTD (25). Furthermore, Phe 63 in the hinge segment that connects NTD and CTD has significant contacts with side chains of residues in the hydrophobic cluster. Because of these interactions, the bend formed in the hinge places Lys 60 and Lys 61 close to the basic surface in CTD (Fig. 7C). Interestingly, all of the residues forming the basic surfaces in CTD and NTD as well as those found in the hydrophobic cluster and the hinge were found in the ␣ M I domain-binding clusters 2 and 3, and two residues (Val 103 and Lys 105 ) are in the beginning of cluster 4. Because basic and hydrophobic amino acid residues were shown to be the main contributors to the interaction between the ␣ M I domain and peptides in the previously screened libraries of Mac-1 ligands (29,30,34), it is possible that the basic surfaces in CTD and NTD form the binding sites for the ␣ M I domain. The identification of the majority of the ␣ M I domain-binding residues in NTD and CTD is in agreement with the NMR data showing that many residues in these domains were perturbed in the presence of ␣ M I domain (Fig. 6).
To validate the relevance of the proposed Mac-1-binding sites, we investigated adhesion of Mac-1-expressing HEK293 cells to truncated PTN fragments. The results showed that immobilized NTD and CTD can independently support Mac-1-dependent adhesion, and soluble NTD and CTD can inhibit cell adhesion to intact PTN, albeit to a lower extent than soluble intact PTN (Fig. 8, B-E). Interestingly, removal of each NTD or CTD reduced PTN's affinity for cell-surface proteoglycans to a greater extent than its affinity for Mac-1 (Fig. 8, C and D), suggesting that proteoglycans are involved but not absolutely required for adhesion. Based on these data, it is reasonable to propose that each of the two structural domains of PTN can bind Mac-1 independently. Furthermore, it appears that both domains of PTN are required for the optimal proteoglycan binding, and thus the enhanced adhesion of Mac-1-expressing cells to intact PTN may result from the cooperative binding of Mac-1 and proteoglycans.
Consistent with the role of proteoglycans in Mac-1mediated cell adhesion, previous studies demonstrated that soluble heparin partially blocks adhesion of Mac-1-expressing cells, including human monocytes to various ligands (28,30,34). Moreover, the inhibitory effect of anti-Mac-1 functionblocking reagents has been potentiated by heparin. We observed similar effects of heparin and anti-Mac-1 functionblocking mAb 44a on adhesion of Mac-1 HEK293 cells to PTN (Fig. 1). Heparin also partially inhibited migration of Mac-1 HEK293 cells. However, in contrast to adhesion, mAb 44a completely inhibited migration even in the absence of heparin, indicating that Mac-1 was absolutely required for migration (Fig. 3,  C and D). These findings suggest that, although HSPGs may influence cell migration in this system, PTN engagement by Mac-1 is indispensable for triggering intracellular signaling that initiates migration. Heparin may exert its inhibitory effect by binding to the heparin-binding sites in PTN or by interacting with Mac-1. Indeed, both NTD and CTD of PTN contain sites that bind heparin with high affinity (27,42). Furthermore, heparin has also been shown to interact with the ␣ M I domain of

PTN is a ligand for integrin Mac-1
Mac-1 (43), although the binding site(s) remains uncertain. These findings suggest that, if present, HSPGs on leukocytes may participate in optimal adhesion and migration to PTN. Moreover, PTN has high affinity for chondroitin sulfate (CS) A and E with a K d in the low nM range (27,42). This suggests that CS, which (similar to heparin) is covalently attached to several core proteins, creating a variety of CSPGs on the cell surface, can act cooperatively with Mac-1. The complex relationship among PTN, Mac-1, and cell-surface HS/CS proteoglycans remains poorly understood. Because PTN contains two glycosaminoglycan-binding sites in NTD and CTD and can bind the ␣ M I domain through the same basic surfaces, it may potentially bridge Mac-1 and proteoglycans, inducing their clusterization. The additional binding site for glycosaminoglycans in the C-terminal tail of PTN may also participate in the formation of complexes between Mac-1 and HS/CSPG, further increasing the complexity of multimolecular clusters. Given noticeable differences in the affinity of PTN domains for glycosaminoglycans, it is reasonable to propose that this variability may influence the interactions of PTN with Mac-1 on leukocytes and thus affect cell adhesion, migration, and other Mac-1-dependent leukocyte responses. Interestingly, although HSPGs on Mac-1-expressing HEK293 cells and human monocytes (28) cooperate with Mac-1 in PTN binding, their role in adhesion of IC-21 mouse macrophages appears to be less important because heparin alone does not inhibit adhesion, whereas anti-Mac-1 mAb produces a strong inhibitory effect (Fig. 1E). The reason for the differential involvement of proteoglycans in PTN recognition is unclear but may stem from differences in the composition of glycosaminoglycans expressed on various cells. In addition to cell-surface proteoglycans, ECM proteoglycans may also play a role in storage and presentation of PTN to migrating cells. In this regard, we showed that PTN bound to aggrecan, a well-known CSPG in the ECM, supports adhesion, suggesting that PTN deposited in the ECM remains an active Mac-1 ligand. Binding of several ligands to Mac-1 is controlled by the activation state of the receptor. Previous studies demonstrated that the ␣ M I domain exists in two different conformations, active and nonactive, with the position of the C-terminal ␣7 helix regulating its activation state (44). In contrast to such Mac-1 ligands as C3bi, recognition of which requires the active state of the ␣ M I domain (45), the nonactive form of the ␣ M I domain bound PTN, albeit to a lower extent than active ␣ M I domain (Fig. 5C). This finding suggests that the ␣ M I domain binding to PTN is not strictly activation-dependent. Furthermore, in contrast to other integrin ligands, binding of PTN to the active ␣ M I domain still occurred in the absence of Mg 2ϩ or in the presence of EDTA (Fig. 5C). Similar characteristics, i.e. the lack of dependence on the activation state of the ␣ M I domain and divalent cations, have been noted with some Mac-1 ligands, including opioid peptide dynorphin A (30), fibrinogen peptide P2-C (39), and cationic proteins myeloperoxidase and elastase. 4 Because all these molecules are highly positively charged and their binding to the ␣ M I domain is mediated by sequences enriched in basic residues (29), it is tempting to speculate that cationic peptides/proteins represent a unique group of Mac-1 ligands that bind the ␣ M I domain through activation-and cation-insensitive mechanisms. Further structural studies of the ␣ M I domain-PTN complex may help to define the mechanism of this interaction.
Although the biological significance of leukocyte interaction with PTN remains to be established, the expression of PTN in different cells during regeneration after injury (2,46) suggests that it may play a role in inflammatory responses. In this regard, PTN has been shown to promote neutrophil and mononocyte/ macrophage recruitment during liver regeneration and peritoneal fibrosis (13,14). Our in vitro studies showing that Mac-1 can mediate migration of macrophages and Mac-1-expressing HEK293 cells to PTN suggest that leukocyte migration in vivo may also be driven by this receptor. Furthermore, the only known PTN homolog, midkine (MK), which shares 50%

Residue Number
Intensity Ratio

PTN is a ligand for integrin Mac-1
sequence identity and a similar tertiary structure with PTN, is expressed in damaged tissues and has a comparable proinflammatory profile (47). It has been reported that MK enhances migration of inflammatory leukocytes in a number of pathological conditions induced in experimental animals (14, 48 -52).
MK receptors responsible for triggering these responses are not known, but MK has been shown to activate both ␤ 1 and ␤ 2 integrins (53,54). Furthermore, adhesion to immobilized MK was shown to be mediated by ␤ 2 integrins, although the nature of ␣ subunit pairing with the ␤ 2 subunit has not been deter-

PTN is a ligand for integrin Mac-1
mined. Both PTN and MK have also been shown to induce expression of inflammatory cytokines in peripheral blood monocytes (47,55). Because ligand engagement by Mac-1 is known to initiate intracellular signaling that regulates numerous leukocyte responses, including expression of cytokines, a receptor that mediates this effect upon PTN and MK binding is most likely to be Mac-1. The potential role of PTN in inducing other reactions of monocyte/macrophages that are known to be mediated by Mac-1 merits further investigation.

Reagents
The mouse mAb 44a directed against the human ␣ M integrin subunit was purified from conditioned media of hybridoma cells obtained from the American Type Culture Collection (Manassas, VA) using protein A-agarose. Alexa Fluor 546-conjugated phalloidin was purchased from Life Technologies. The mouse mAb G3A1, an IgG1 isotype control for mAb 44a, was obtained from Cell Signaling Technology (Beverly, MA). BSA, polyvinylpyrrolidone (PVP), EDTA, heparin, and aggrecan were purchased from Sigma. Calcein-AM was purchased from Molecular Probes (Eugene, OR). mAbs directed to total Erk1/2 and the phosphorylated form of Erk1/2 were from Cell Signaling Technology. Goat anti-rabbit IgG (heavy ϩ light)-HRP conjugated antibody was from Bio-Rad. Protease and phosphatase inhibitor mixture was from Thermo Scientific (Rockford, IL).

Expression and purification of recombinant proteins
Expression and purification of PTN were performed as described previously (25). Briefly, human PTN ORF cloned into the pET-15b vector was transformed into Origami B (DE3) cells (Novagen, Madison, WI). The transformed cells were grown in M9 medium at 37°C until induction with 0.25 mM isopropyl 1-thio-␤-D-galactopyranoside when A 600 reached 0.8. The culture was then incubated with shaking overnight at room temperature. The PTN was purified from supernatant using hepa-rin-affinity chromatography with a 5-ml HiTrap heparin column (GE Healthcare). Truncated PTN fragments, PTNshort (residues Gly 1 -Lys 114 ), NTD (residues Gly 1 -Cys 57 ), and CTD (residues Asn 58 -Lys 114 ) were cloned into the pET-15b vector and expressed as described above. Expression of recombinant active (residues ␣ M Glu 123 -Lys 315 ) and nonactive (residues ␣ M Gln 119 -Glu 333 ) ␣ M I domains as well as active ␣ L I domain (residues ␣ L Gly 127 -Tyr 307 ; K287C,K294C) were described previously (39). Alternatively, the ␣ M I domain ORF was cloned into pHUE vector (56) as a fusion protein with ubiquitin and transformed into BL21 (DE3) cells (Bioline). The ␣ M I domains were purified with 5-ml HisTrap HP nickel columns (GE Healthcare). To remove the His-tagged ubiquitin, the fusion protein was incubated with the ubiquitinase USP2 overnight at a protein-to-enzyme ratio of 30:1. The pure ␣ M I domain was separated from other impurities by size-exclusion chromatography using a Superdex 75 column (GE Healthcare). The ␣ M I domain was biotinylated with EZ-Link Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific, Inc., San Jose, CA). The efficiency of biotinylation was tested using a biotin quantitation assay kit (Thermo Fisher Scientific, Inc.) with biotinylated horseradish peroxidase as a standard.

Screening of peptide libraries for ␣ M I domain binding
A peptide library was prepared by parallel spot synthesis on cellulose membranes as described previously (57,58). Purified recombinant ␣ M I domain was labeled with 125 I using Iodo-Gen (Pierce). The library spanning the sequence of PTN was synthesized as 9-mer overlapping peptides with a 3-amino acid offset. Peptides were C-terminally attached to the cellulose via a (␤-Ala) 2 spacer and were acetylated N-terminally. The membrane with attached peptides was blocked with 1% BSA and then incubated with 10 g/ml 125 I-labeled ␣ M I domain in TBS containing 1 mM MgCl 2 , 0.05% Tween 20, and 1 mM dithiothreitol. After washing, the membrane was dried, and the ␣ M I domain binding was visualized by autoradiography and analyzed by densitometry.

NMR
PTN 1 H-15 N HSQC NMR data were collected using a Bruker 600-MHz Avance III HD spectrometer (Bruker Corp., Billerica, MA) equipped with a Prodigy probe. The NMR samples contained 40 M PTN in PBS buffer at pH 7.5 and either 0 or 40 M unlabeled ␣ M I domain. Data were processed using NMRPipe (59) and analyzed with NMRView (60).

PTN is a ligand for integrin Mac-1 Cell adhesion assays
Cell adhesion assays were performed according to methods described previously (36,61). Briefly, 96-well microtiter plates (Immulon 4HBX, Thermo) were coated with different concentrations of PTN and PTN fragments (PTN, PTN-short, CTD, and NTD) for 3 h at 37°C and blocked with 1% PVP in PBS for 1 h at 37°C. The cells were labeled with 7.5 M calcein-AM (Thermo) for 30 min at 37°C. The labeled cells were washed with Hanks' balanced salt solution containing 0.1% BSA and resuspended in the same buffer at a concentration of 5 ϫ 10 5 / ml. Aliquots (100 l) of labeled cells were added to each well and incubated for 30 min at 37°C. Non-adherent cells were removed with two washes with PBS, and fluorescence was measured with a fluorescence plate reader (Perseptive Biosystems, Framingham, MA). For inhibition experiments, labeled cells were treated with 5 g/ml anti-Mac-1 antibodies (anti-␣ M mAb 44a and M1/70 for Mac-1 HEK293 and IC-21 macrophages, respectively) or heparin (1 g/ml) for 15 min at 22°C before they were added to the wells. For PTN inhibition experiments, cells were treated with either 2 or 5 M soluble PTN or PTN fragments for 20 min at 22°C before they were added to the wells. To test the effect of aggrecan-bound PTN, 10 g/ml aggrecan in PBS was used to coat 96-well microtiter plates overnight at 4°C as described (62). Then PTN was added to aggrecan-coated wells for 3 h at 37°C. The remaining steps were the same as described above. The coating efficiency of aggrecan was determined using biotinylated aggrecan.

Cell migration assays
Cell migration assays were performed under sterile conditions using Transwell inserts (Costar, Corning, NY) with a pore size of 8 (HEK293 cells) or 5 m (macrophages) as described (30,38). Briefly, after coating PTN on the Transwell membrane at 37°C for 3 h, cells (100 l) were loaded in the upper chamber of the Transwell system at a concentration of 3 ϫ 10 6 /ml. For inhibition experiments, cells were pretreated with 20 g/ml anti-␣ M mAb 44a or 1 g/ml heparin for 20 min at 22°C before loading into the upper chamber of the Transwell system. The lower chamber contained 600 l of DMEM. Migration took place at 37°C in a 5% CO 2 humidified atmosphere for 16 h. For detection, the HEK293 cells were labeled with calcein-AM for 30 min at 37°C, and macrophages were labeled with FITC during isolation. Cells from the upper chamber of the Transwells were removed by wiping with a cotton-tipped applicator. Images of the cells on the underside of the Transwell filter were taken with a Leica DM 4000B microscope camera (Leica, Buffalo Grove, IL).

Cell spreading assays
For cell spreading assays, glass coverslips coated with 1.5 g/ml PTN were placed in 6-well plates (Costar), and 4 ϫ 10 5 cells were added to each well. After incubation at 37°C for different times (30 min, 1, 2, 8, and 24 h), cells were washed with PBS, fixed with 2% paraformaldehyde for 10 min at room temperature, and then treated with 0.1% Triton X-100. For actin staining, the cells were treated with 200 l of 165 mM phalloidin solution for 20 min at room temperature. DAPI was added for DNA staining. The specimens were imaged with a Leica SP5 laser-scanning confocal microscope using 63ϫ/1.4 objective. The diameters of 20 -50 cells for each time point were counted using ImageJ software from images taken with a 20ϫ/0.5 objective. The diameter was defined as the largest distance between two opposite sides of the cell.

Biolayer interferometry
The experiments were performed using an Octet K2 instrument (ForteBio, Pall Corp.). Purified PTN was immobilized on the amine-reactive second-generation (AR2G) biosensor using the amine coupling kit according to the manufacturer's protocol. Different concentrations of the active and nonactive forms of ␣ M I domain and active ␣ L I domain were applied in the mobile phase, and the association between the immobilized and flowing proteins was detected. Experiments were performed in 20 mM HEPES, 150 mM NaCl, 1 mM MgCl 2 , and 0.05% (v/v) Tween 20, pH 7.5, or in 20 mM HEPES, 150 mM NaCl, 5 mM EDTA, and 0.05% (v/v) Tween 20, pH 7.5. The PTN-coated surface was regenerated with 25 mM NaOH. Analyses of the binding kinetics were performed using ForteBio Data Analysis 9.0 software. The K d was obtained by curve fitting of the association and dissociation phases of sensorgrams using a heterogeneous ligand model. BLI experiments were also carried out using PTN biotinylated through glutamate or aspartate side chains with amine-PEG 2 -biotin (Thermo Fisher Scientific). Biotinylated PTN at a concentration 50 g/ml was immobilized on the streptavidin biosensor in 20 mM HEPES buffer supplemented with 150 mM NaCl, 0.1% BSA, 0.02% Tween 20, 1 mM MgCl 2 , and 1 mM CaCl 2 for 15 min, which generated a saturated level of PTN on the biosensor. After immobilization, the biosensor was washed with HEPES buffer, and different concentrations of ␣ M I domain were added to the immobilized PTN. After each association-dissociation cycle, the biosensor was regenerated using 20 mM NaOH.

Western blotting
HEK293 cells and Mac-1 HEK293 cells were added to the wells of 6-well culture plates at 2 ϫ 10 6 /well and cultured for 2 h in DMEM/F-12 ϩ 10% FBS. Adherent cells were treated with PTN and PTN fragments (0.7 M). After 30 min at 37°C in a humidified 5% CO 2 atmosphere, cells were washed with icecold PBS containing a protease and phosphatase inhibitor mixture. Next, 200 l of the lysis buffer (50 mM Tris-HCl, pH 7.4, 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, and protease and phosphatase inhibitor mixture) was added to each well, and cells were incubated on ice for 1 h. Protein concentration in the lysates was quantified by a BCA assay (Thermo Scientific), and equal amounts of protein (20 g) were loaded onto 12.5% SDSpolyacrylamide gels. After the transfer of proteins onto nitrocellulose membranes, the membranes were blocked with skim milk and incubated with anti-Erk1/2 and anti-phospho-Erk1/2 (Tyr 202 and Tyr 204 ) rabbit mAbs. The mAb binding was detected using HRP-conjugated goat anti-rabbit antibodies and chemiluminescence (Thermo Scientific).

Statistical analysis
All data are presented as the mean Ϯ S.E. The statistical differences were determined using one-way analysis of variance PTN is a ligand for integrin Mac-1 using SigmaPlot 11.0 software (Systat Software, San Jose, CA). For multiple comparisons, the Bonferroni correction method was used. Differences were considered significant if the p value was less than 0.05.