Lactoferrin is a natural inhibitor of plasminogen activation

The plasminogen system is essential for dissolution of fibrin clots, and in addition, it is involved in a wide variety of other physiological processes, including proteolytic activation of growth factors, cell migration, and removal of protein aggregates. On the other hand, uncontrolled plasminogen activation contributes to many pathological processes (e.g. tumor cells' invasion in cancer progression). Moreover, some virulent bacterial species (e.g. Streptococci or Borrelia) bind human plasminogen and hijack the host's plasminogen system to penetrate tissue barriers. Thus, the conversion of plasminogen to the active serine protease plasmin must be tightly regulated. Here, we show that human lactoferrin, an iron-binding milk glycoprotein, blocks plasminogen activation on the cell surface by direct binding to human plasminogen. We mapped the mutual binding sites to the N-terminal region of lactoferrin, encompassed also in the bioactive peptide lactoferricin, and kringle 5 of plasminogen. Finally, lactoferrin blocked tumor cell invasion in vitro and also plasminogen activation driven by Borrelia. Our results explain many diverse biological properties of lactoferrin and also suggest that lactoferrin may be useful as a potential tool for therapeutic interventions to prevent both invasive malignant cells and virulent bacteria from penetrating host tissues.

serum and exocrine secretions. Antimicrobial, antitumor, and immunomodulatory activities have been attributed to LF (7,8). Here, we demonstrate a so far unknown function of LF in regulation of fibrinolysis and unravel the molecular determinants and mechanism whereby lactoferrin blocks plasminogen activation.
The Western blot analysis confirmed the specificity of LF (Fig. 1B). We verified the interaction further by co-immunoprecipitating LF with M6P/IGF2R from human granulocytes, which are known to produce high amounts of LF (Fig. 1C). M6P, which was found in LF (14), did not significantly influence the interaction between LF and M6P/IGF2R, suggesting that the M6P-binding sites in M6P/IGF2R were not required for the interaction (Fig. 1C). These findings were proven by an in vitro binding assay wherein M6P did not block the interaction, similarly to soluble uPAR ( Fig. 2A). In contrast, the presence of Plg, a M6P/IGF2R ligand (9), enhanced the weak interaction between LF and M6P/IGF2R ( Fig. 2A). Based on the latter finding, we tested the possibility of a direct LF-Plg binding. Indeed, an in vitro binding assay revealed the direct binding of LF and Plg (Fig. 2B).
To map the determinants responsible for the LF-Plg binding, we designed three peptides: pLF1 derived from the N-terminal end of LF and encompassed within lactoferricin (LFC), a small bioactive fragment with anti-pathogenic activity (15); pLF2 derived from the helix linker region of LF encompassing a stretch of lactoferrampin, another bioactive peptide (16); and pLF3 made up of the C-terminal part containing a C-terminal lysine (Table 3), because C-terminal lysines are generally known to be involved in Plg binding (17). We found that pLF1 blocked the binding of LF and Plg with an IC 50 of about 6 g/ml (2.8 M) (Fig. 2, C-E). The lysine analogue tranexamic acid (TA), which is known to block the lysine-dependent Plg binding, did not have significant effects on the binding as well as the other peptides. This indicates that the Plg-binding site is encompassed within the N-terminal part of LF. Coomassie-stained gel band co-purified with M6P/IGF2R was cut for LC-MS/MS analysis. B, the fractions were analyzed by immunoblotting (IB) using mAbs to M6P/IGF2R (MEM-238), LF (4C5), and uPAR (H2). C, the human granulocyte cell lysate, obtained with 1% Triton X-100 together with 0.5% dodecylmaltoside as detergents, was subjected to immunoprecipitation on a 96-well plastic plate (50 l/well) coated with goat anti-mouse IgG (5 g/ml) and specific antibodies to M6P/IGF2R (mAb clone MEM-238), LF (mAb clone 4C5), and uPAR (mAb clone H2); as isotype control, mAb AFP-01 was used (ctr) (all 5 g/ml). The incubation was performed for 4 h at 4°C in the presence or absence of mannose 1-phosphate (M1P) and M6P (both 10 mmol/liter). The precipitates were analyzed by SDS-PAGE followed by immunoblotting using the specific mAb to LF (LF95-4C5).

Table 1 MS analysis of the 80-kDa protein band co-purified with M6P/IGF2R
Peptides matched to LF (Mascot protein score ϭ 613). M6P/IGF2R was purified with the anti-M6P/IGF2R mAb MEM-240 from lysates of THP-1 cells. The 80 kDa band co-purified with M6P/IGF2R was excised from the gel for MS/MS analysis. Identification was performed by MASCOT and the NCBI non-redundant protein database. Peptides identified to match to LF are listed.

LF residues Peptide
Observed mass

Lactoferrin controls plasminogen activation
To map the determinants responsible for binding of Plg to LF, we applied fragments derived from Plg generated as described previously (18): angiostatin, consisting of kringle domains 1-4; mini-Plg containing kringle 5 and the proteolytic domain; and micro-Plg encompassing just the proteolytic domain. LF bound strongest to mini-Plg, although we detected a weak binding to all fragments (Fig. 2, F and G). Upon a higher stringency (i.e. in the presence of 0.1% detergent Triton X-100), LF bound only to mini-Plg (Fig. 2F). These findings indicate that the major LF-binding site is encompassed within kringle 5 of Plg.
In the next steps, we quantified the binding of LF and Plg by surface plasmon resonance (SPR). To preserve the surface of LF as much as possible, we opted for its immobilization on a streptavidin sensor chip, which facilitated a flexible attachment following biotinylation of the oligosaccharides present at Asn 147 and Asn 478 in the N-and C-terminal lobe of LF, respectively (19). Kinetic analysis revealed that LF and Plg formed a high-affinity complex with the dissociation constant equal to 69 Ϯ 24 nM (Table 4 and Fig. 3A).
Interestingly, the SPR profile of Plg association and dissociation gives the best fit to a two-state reaction model (Fig. 3A). According to this model, the proteins initially form a metastable complex 1, which is rearranged on a millisecond time scale to a final, slowly dissociating complex 2 (Table 4). To verify the two-state reaction model, we examined dissociation of the LF-Plg complex in two separate SPR experiments, comparing short (60-s) and prolonged (300-s) Plg association time. Indeed, in the experiment with prolonged association time, we detected the formation of a higher amount of the slowly dissociating complex 2 (data not shown), which ultimately confirmed the twostep interaction mechanism.
A freshly prepared LF sensor chip had a very low capacity to bind Plg and required activation by three consecutive 6-s pulses of 100 mM HCl. The sensor chip response decreased by 100 response units during the treatment, but the LF surface became fully competent to interact with Plg and maintained an unchanged reactivity over 1 month of running experiments.
This surprising observation prompted us to investigate the underlying reasons. As the LF-bound iron might be depleted during activation with the acid (20), at first, we tested whether the resaturation of LF with iron could influence Plg binding. However, no change in complex formation was found; therefore, iron presence had no direct influence on the binding.
Second, we examined the oligomeric state of biotinylated LF used in the sensor chip preparation, as LF is able to form tightly bound dimers in diluted solutions at physiologic pH (21), which might influence the Plg binding. We measured molecular radius by dynamic light scattering (DLS) and found that biotinylated LF preparation exhibited a higher average particle size than the initial protein (Table 5), with the weight-averaged molar mass of biotinylated LF corresponding prevalently to the dimeric form (177 kDa versus 106 kDa of native LF). The amount of high-molecular weight aggregates was negligible. It suggests that the unresponsiveness of the fresh LF surface might reflect the LF dimerization during labeling, which had occluded surface patches needed for the interaction with Plg. The dimerization of LF occurs also at physiological conditions (22) and may represent a mechanism to control its interaction with Plg.
Each of the three proteins, sM6P/IGF2R, Plg, and LF, had been detected in human serum (23)(24)(25)(26). We found that LF coimmunoprecipitated with both sM6P/IGF2R and Plg in human serum (Fig. 3B). In addition, blue native electrophoresis analysis (BN-PAGE) implied the presence of several complexes associated with LF in serum (Fig. 3C); the low molecular weight species seemed to contain exclusively LF and very likely corresponded to dimers that did not bind Plg as described above; the intermediate diffused complex(es) Ͻ272 kDa contained apparently both LF and Plg; the intermediate diffused complex(es) of ϳ500 kDa contained apparently predominantly sM6P/IGF2R, as shown previously (12); and the high molecular weight complex(es) of Ͼ880 kDa appeared to contain Plg, LF, and sM6P/ IGF2R (Fig. 3C, arrowheads). These data suggest that LF may co-exist with Plg in serum in native complexes.

Lactoferrin controls plasminogen activation
Next, we tested whether binding of LF to Plg could interfere with Plg activation. First, we found that pre-incubation with LF abrogated AF647-labeled Plg binding to THP-1 cells (known to express high levels of uPAR and uPA (11)) in a concentrationdependent manner with the maximal inhibition (52%) at the concentration of 50 g/ml (Fig. 4A). We analyzed only the binding to viable cells. Second, when the cells were subjected to a Plg activation assay, we found a concentration-dependent inhibitory effect of LF with maximal inhibition at a concentration of 10 g/ml (Fig. 4B). This effect was independent of M6P/ IGF2R, previously identified by us to bind and internalize Plg (9,11), because in both control and M6P/IGF2R-silenced THP-1 cells, generated previously by us (11), a similar effect was A, LF, Plg, and BSA (as a negative control), coated on wells of a plastic plate (5 g/ml), were incubated for 4 h on ice with purified M6P/IGF2R (5 g/ml) in the absence or presence of additional molecules: mannose 1-phosphate (M1P) and M6P (both 10 mmol/liter), soluble uPAR (5 g/ml), Plg (50 and 100 nmol/liter), or LF (20 and 40 g/ml). Bound material was analyzed by immunoblotting (IB) using the specific mAb MEM-238 to M6P/IGF2R. B, coated Plg or BSA (5 g/ml) was incubated for 4 h on ice with LF (20 g/ml). Before LF was added, some wells were pre-incubated for 2 h on ice with 20 g/ml LF-derived peptides (pLF1, pLF2, or pLF3) and scrambled peptide (pSCR) or 5 mmol/liter tranexamic acid (TA). Bound material was analyzed by immunoblotting using human LF-specific mAb 4E2. C, the binding assay was performed as in B but with the pre-incubation with increasing concentrations of pLF1 or scrambled peptide pSCR as indicated. D and E, densitometric quantifications of LF binding assays. A representative immunoblot is shown in B and in C and D, respectively. Bars, relative means with S.D. (error bars) of at least three independent experiments. The scrambled peptide pSCR served as a control, and the corresponding bands were set as a relative maximum of 100%. Immunoblots were quantified with Fuji MultiGauge software. Bands corresponding to LF binding to BSA were subtracted from the bands of LF binding to Plg bands; in addition, a background correction for each individual lane was applied. The relative IC 50 value was determined by a 4PL nonlinear regression curve fit, in accordance with National Institutes of Health assay guidance (64). Curve fitting was performed with MasterPlex ReaderFit, published by the MiraiBio Group of Hitachi Software Engineering America, Ltd. F, the binding assay was performed as in B, but Plg and the Plg fragments angiostatin, mini-Plg, and micro-Plg, or BSA, were coated onto plastic wells. Optionally, the binding assay was performed in the presence of Triton X-100 (0.1%). G, densitometric quantification was performed as in D and E. A representative immunoblot is shown in F, without the detergent; *, p Ͻ 0.05; **, p Ͻ 0.005.

Lactoferrin controls plasminogen activation
observed: 75 and 82% inhibition, respectively. However, M6P/ IGF2R-silenced cells displayed a 30% higher capacity to activate Plg compared with control cells (Fig. 4B), which was in line with our previous work showing the M6P/IGF2R-dependent negative regulation of Plg activation by means of the stable genetic knockout and knockdown experiments (10,11). Third, by a cell-free system, we found that LF hampered Plg activation on coated uPA in a concentration-dependent manner with an IC 50 value of 3 g/ml (43 nM), similarly to the plasmin inhibitor aprotinin (Fig. 4C). An in vitro binding assay revealed that LF completely blocked the binding of Plg to uPA at a concentration of 10 g/ml (Fig. 4D), which is in agreement with its effect on Plg activation. Finally, we performed an in vitro binding assay in the presence of mAbs specific to both LF and Plg, and we found that the anti-LF mAb 1D2 blocked the interaction between LF and Plg (Fig. 5A). A pre-incubation of LF with mAb 1D2 abrogated the inhibition of Plg activation by LF at a concentration of 20 g/ml (Fig. 5, B and C).
We also tested whether LF was able to block Plg activation by purified SK, a protein secreted by Streptoccoci that binds and activates Plg independently of uPA (27,28). We found that LF Table 4 Kinetics and affinity of LF-Plg binding The data are averaged from four independent experiments and expressed as mean Ϯ S.D. k ON , association rate constant of the initial complex formation; k OFF1 , dissociation rate constant of the initial complex 1 breakup; k ONconf , rearrangement rate constant; k OFF2 , dissociation rate constant of the final complex 2; K D , equilibrium dissociation constant.

Lactoferrin controls plasminogen activation
did not have any inhibitory effect on Plg activation by SK (Fig.  4E). Together, these results suggest that LF inhibits plasmin generation on the surface of cells through a blockade of Plg binding to uPA. LF is an iron-binding glycoprotein and a member of the lactotransferrin family. However, we did not observe any inhibition of Plg activation by human transferrin (TF) (Fig. 4F). Although LF is structurally highly similar to TF (29), the superimposed structure comparison of LF and TF showed that their N-terminal parts possessed several distinctions very likely caused by the presence of several positively charged amino acid residues within LF not present in TF (e.g. an N-terminal stretch of arginines) (Fig. 6, A (arrow), B, and C). Because we mapped the Plg-binding site into this N-terminal part of LF (Fig. 2, B-E), these differences might be decisive in a contrasting capability of LF and TF to block Plg activation. This finding further underlines the specificity of the LF-mediated inhibition of Plg activation.
In the last series of experiments, we tested the effect of LF on patho/physiological processes associated with the Plg system. The Plg system is utilized by migratory cells to penetrate tissues (e.g. endothelial cells during angiogenesis harness uPA-dependent Plg activation) (30,31). However, tumor cells also activate Plg by uPA, which contributes to their dissemination (32). Accordingly, we performed in vitro Matrigel cell invasion assays by using the kidney carcinoma cell line TCL-598 and primary human umbilical vein endothelial cells (HUVECs), both shown by us previously to migrate in a uPA/Plg-dependent manner (10,12). We found that LF reduced the invasiveness of both TCL-598 cells and HUVECs (Fig. 7, A and B). Thus, LF can exert negative regulatory effects on cell migration via inhibition of Plg activation.
Also virulent bacteria hijack the Plg system to increase their invasiveness. For instance, Streptococci secret SK, which on the one hand binds one "enzyme" Plg molecule and on the other hand activates another "substrate" Plg molecule in a uPA-independent manner (27,33). In contrast, Borrelia do not produce any intrinsic Plg activator; instead, they bind and employ components of the host's fibrinolytic system, namely Plg and uPA, via outer surface proteins (4,5,34,35) to penetrate the host. Thus, we tested the ability of LF to inhibit Plg activation by B. burgdorferi and S. pyogenes. When we incubated human serum with B. burgdorferi and the plasmin-specific chromogenic substrate, we observed a blockade of Plg activation by LF (Fig. 7C). In contrast, in the presence of S. pyogenes, LF did not reduce Plg activation (Fig. 7D). This observation was in line with the above-mentioned experiments with purified enzymes showing that LF was able to inhibit Plg activation by uPA but not SK (Fig. 4). Taken together, these results show that LF can modulate patho/physiological processes of Plg, such as tumor cell invasion and invasion of certain bacteria.

Discussion
LF is known to display a plethora of antimicrobial properties and antitumor activities (7). Its iron binding ability and the unique positively charged N-terminal region are considered to explain most of its biological activities (8). Here, we identified a so far unknown function of LF; it binds Plg and blocks its activation by uPA but not by bacterial SK.
We identified the interaction by means of MS analysis of the proteins co-immunoprecipitated with M6P/IGF2R. This suggests that M6P/IGF2R, a Plg receptor, might be involved in effects of LF. In this respect, it is of interest that M6P/IGF2R is required for herpes simplex virus type 1 infection (36), and LF inhibits herpes simplex infection and trafficking (37). Moreover, also nucleolin, another protein identified by us as a partner of M6P/IGF2R (Table 2), seems to be involved in herpes simplex virus type 1 infection (38). The putative functional network among M6P/IGF2R, LF, and herpes simplex virus remains a topic for future study.
Here, we mapped the mutual binding sites of LF and Plg within the N-terminal region of LF and kringle 5 of Plg, respectively. We show that through this interaction, LF controls Plg activation. We propose that LF inhibits Plg activation on cells via the blockade of the Plg-uPA interaction, which is in agreement with the observation that the majority of lysine-dependent Plg binding to the cell surface depends on the presence of uPA (39). However, it is also possible that LF via binding to Plg blocks a conformational conversion of Plg (40) required for its activation.
LF shares significant sequence and structural homology with TF, a further member of the iron-binding transferrin family. However, TF does not possess the inhibitory activity of LF toward Plg, which might be due to the different composition within their N termini.
Plg is an inactive zymogen circulating in plasma at relatively high concentration (2 M) (23,41). Plg binding to the matrix protein fibrinogen (K D ϭ 0.32 M) is required for conversion of Plg to active plasmin during blood clot resolution (42). In addition, Plg binds with K D values ranging from 0.1 to 2 M also to various cell surface receptors, which is indispensable for cell migration (17,39,43). Both pathways (i.e. Plg activation on the clots and on the cellular surface) are mediated by Plg activators, tissue-type plasminogen activator and uPA, respectively (32). Strikingly, Plg binds with higher affinity (K D ϭ 420 pM) to bacterial SK (44 -46). SK is a Plg activator employed by invasive human pathogens to facilitate their spreading through host barriers (27,28). Plg activators are susceptible to inhibition by specific plasminogen activator inhibitors. Also, direct plasmin inhibitors are present in plasma and rapidly inactivate free unbound plasmin (e.g. ␣2-antiplasmin binds to various Plg types with K D values ranging from 0.2 to 1.9 M) (47-50).

Lactoferrin controls plasminogen activation
By using SPR, we determined the dissociation constant of the Plg-LF interaction to be 69 Ϯ 24 nM and the IC 50 of LF in inhibiting Plg activation to be 43 nM. These measures are similar yet somewhat higher than the concentrations of LF in human serum, which range from 5 to 30 nM in healthy donors (51,52). 3 This suggests that under normal conditions, Plg activation would not be blocked by LF. However, LF concentrations might be up-regulated in site-specific microenvironments (e.g. upon degranulation of neutrophils) (7,53). Further, high levels of LF were reported to be secreted by apoptotic cells (54). In apoptosis, LF might be necessary to down-regulate extracellular plasmin activity to avoid unwanted degradation of the surrounding tissue, inappropriate cell migration, or activation of proenzymes. The serum levels of LF might further increase during pregnancy (55). Thus, an anti-fibrinolytic property of LF may provide an intriguing clue to the reported higher risk of thromboembolism during pregnancy (56,57).
The physiological relevance of the LF-Plg complex is indicated by the fact that we were able to coimmunoprecipitate both proteins from human serum. The SPR profile of the LF-Plg interaction suggests a two-state model with metastable initial complex 1 and stable complex 2. Time-dependent stabilization of the LF-Plg interaction may have important physiological . LF blocks Plg activation by uPA. A, to measure Plg binding to cells, Plg-AF647 (10 g/ml) was pre-incubated on ice for 30 min with or without LF at the indicated concentrations. Then THP-1 cells were incubated on ice for 20 min with the Plg preparations, and the cell surface-bound Plg was analyzed by flow cytometry. Histograms are representative of three experiments. The values show the percentage of differential mean fluorescence intensity. The anti-mouse IgG-AF647 conjugate was used as a negative control. B, both control and M6P/IGF2R-silenced THP-1 cells were incubated at 37°C in serum-free medium together with Plg (50 nmol/liter) and the chromogenic plasmin substrate S-2251 (0.8 mmol/liter) in a 96-well plate (5 ϫ 10 5 cells/well). The reactions were performed with or without LF or aprotinin at the indicated concentrations. The absorbance change at 405 nm was monitored by using an ELISA reader. C, in a cell-free system, coated uPA (20 nmol/liter) was incubated at 37°C with Plg (50 nmol/liter) and substrate S-2251 (0.8 mmol/liter) in a 96-well plate. The reactions were performed as in B. The experiments in B and C were performed at least in triplicates, and results are representative of two independent experiments. D, uPA, coated as in C, was incubated with purified Plg (50 nmol) in the presence of aprotinin (2.5 g/ml) and the indicated concentrations of LF. Bound material was analyzed by immunoblotting (IB). Results are representative of three independent experiments. E, the Plg activation assays were performed as in C, but SK was coated instead of uPA (20 nmol/liter). The reactions were performed in the presence of LF as indicated. F, the Plg activation assays were performed as in C in the presence of TF at the indicated concentrations. The experiments were performed at least in triplicates, and results are representative of three independent experiments: *, p Ͻ 0.05; **, p Ͻ 0.005. Error bars, S.D.

Lactoferrin controls plasminogen activation
consequences in situations where prolonged elevated levels of interacting proteins occur. Our data further suggest that the LF dimerization status, rather than its iron saturation, might modulate its capacity to interact with Plg.
In summary, LF intrinsically inhibits Plg activation through blocking the interaction between Plg and uPA. Thus, LF represents a novel pathway and tool to regulate plasmin-dependent functions.

Materials
Ammonium persulfate, TEMED, SDS, acrylamide, and N,NЈ-methylenebis-acrylamide were purchased from SERVA (Heidelberg, Germany). Human uPA and Glu-Plg were products of Calbiochem (Darmstadt, Germany). BSA was from Carl Roth (Karlsruhe, Germany). Lactoferrin, streptokinase, aprotinin (inhibitor of serine proteases), ␣2-antiplasmin (inhibitor of plasmin), Polybrene, mannose 1-phosphate, mannose 6-phosphate, all protease inhibitors, puromycin, tranexamic acid, and monensin were from Sigma-Aldrich. Beriglobin was from CSL Behring (King of Prussia, PA). Nonidet P-40 and Triton X-100 were obtained from Pierce. Plasminspecific substrate S-2251 was a product of CoaChrom Diagnostica (Vienna, Austria). The streptavidin-HRP conjugate was supplied by GE Healthcare (Uppsala, Sweden). The LF-derived peptides were synthesized by Peptide 2.0 (Chantilly, VA). Human serum samples were obtained from healthy donors after informed consent that a part of the routine blood samples would be used for research purposes (approved by the Ethical Committee of the Medical University of Vienna). The study abides by the Declaration of Helsinki principles. All samples were stored at Ϫ20°C before use.

Antibodies
The mAb MEM-238 to M6P/IGF2R was generated by us and produced by and purchased from EXBIO (Prague, Czech Republic); it was biotinylated or fluorescently conjugated by us. The AFP-01 mAb was generously provided by Dr. Václav Hoøejší (Institute of Molecular Genetics, Academy of Sciences of the Czech Republic, Prague, Czech Republic). mAbs to human LF (LF5-1D2, LF65-3D5, LF95-4C5, and LF124-5E2) were generated in the laboratory of Otto Majdic. Both biotinylated and FITC-conjugated rabbit polyclonal Ab to LF were purchased from Abcam (Cambridge, UK), and the mAbs to Plg (7Pg and 4Pg) were from Technoclone (Vienna, Austria). The goat anti-mouse IgG secondary antibody, both unlabeled and HRP-conjugated, was from Sigma-Aldrich. Goat anti-mouse IgG ϩ IgM (HϩL)-FITC conjugate was from An der Grub (Kaumberg, Austria); the anti-mouse IgG conjugates labeled fluorescently with Alexa Fluor 488 or Alexa Fluor 647 (AF647) were from Molecular Probes (Invitrogen).  Fig. 2B, but in the presence of various anti-LF mAbs (3D5, 4C5, 5E2, and 1D2) and anti-Plg mAbs (4Pg and 7Pg), all at 10 g/ml. As an isotype control (CTR) mAb, we used AFP-01 against ␣-fetoprotein (10 g/ml). B and C, plasminogen activation assays were performed as in Fig. 4B, but Plg was pre-incubated for 30 min on ice with or without the indicated concentrations of anti-LF mAb. Error bars, S.D.

Cells
The human monocytic cell line THP-1 was from ATCC, and the human kidney epithelial tumor cell line TCL-598 was obtained from the Novartis Research Institute (Vienna, Austria). The cell lines were cultured in RPMI 1640 medium (Gibco/Invitrogen) supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, 2 mM L-glutamine and 10% heat-inactivated FCS (Sigma). All the cells were grown in a humidified atmosphere at 37°C and 5% CO 2 and passaged suitably using trypsin-EDTA solution. Peripheral blood was separated into plasma, peripheral blood mononuclear cells, and granulocytes with red blood cells using the Lymphoprep gradient centrifugation (Axis Shield, Oslo, Norway), the latter followed by red cell lysis using aqua test.

Bacteria
S. pyogenes type strain was obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (DSMZ number 20565) and grown in brain-heart infusion broth (Carl Roth, Karlsruhe, Germany) at 37°C. Bacterial growth was observed by absorbance at 600 nm (A 600 ) and was correlated to cfu/ml by plating on Columbia blood agar plates (bioMérieux, Vienna, Austria). An overnight culture of the bacteria was inoculated to fresh brain-heart infusion broth at an A 600 of 0.05 and grown at 37°C until it reached an A 600 of 0.47 (ϳ1 ϫ 10 8 cfu/ml). Bacteria were pelleted at 6,000 ϫ g for 10 min at room temperature and resuspended in sterile PBS for use in the experiment.
B. burgdorferi sensu stricto strain B31 was obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (DSMZ number 4680). Borrelia were grown in a modified BSK II medium as described (58). They were grown to logarithmic phase and enumerated using a counting chamber. The respective amount of culture was centrifuged at 6,000 ϫ g for 10 min at room temperature, and the resulting pellet was washed once with 1 ml of sterile PBS. Sub-

Lactoferrin controls plasminogen activation
sequently, the bacteria were brought up to a density of 4 ϫ 10 5 /l in fresh PBS, and 25 l (containing 10 7 spirochetes) were added to each reaction.

Flow cytometry and cell surface-binding assay
Cells were harvested, washed with PBS containing 1% BSA and 0.02% sodium azide, blocked with 2% beriglobin for 15 min on ice, and afterward incubated for 30 min on ice with specific mAb, either fluorescently labeled or unlabeled. For the latter, a second step staining was done with FITC-conjugated F(abЈ)2 anti-mouse IgGϩIgM antibodies. Before analysis, the cells were washed again. Apoptotic and necrotic cells were discriminated by staining with 7-aminoactinomycin D and the annexin V-PE conjugate (BD Biosciences). Then the analysis was carried on as described above. Flow cytometry was performed with an LSRII flow cytometer (BD Biosciences). Data acquisition was executed with the FACS DIVA software. Data analysis was accomplished with the FlowJo software (Treestar Inc., Ashland, OR).
To measure Plg binding to cells, the Plg molecule labeled with either Alexa Fluor 488 or AF647 (10 g/ml), prepared in our laboratory, was pre-incubated on ice for 30 min with or without LF in PBS at the indicated concentrations. In parallel, THP-1 cells were harvested, washed, and then incubated on ice for 20 min with the Plg preparations at the indicated concentrations. Afterward, the cells were washed with PBS, and cell surface-bound Plg was analyzed by flow cytometry.

RNAi
The stable knockdown for M6P/IGF2R in THP-1 was generated by the delivery of a short hairpin RNA expression cassette as described in detail (10,11).

Trypsin in-gel digestion and LC-MS/MS analysis
The 80 kDa Coomassie-stained gel band containing proteins co-purified with M6P/IGF2R was cut and destained with a mixture of methanol and 50 mM ammonium hydrogen carbonate. Proteins were reduced by DTT, alkylated by iodoacetamide, and treated with trypsin in cleavage buffer (100 mM Tris-HCl, pH 8.0, 1 mM CaCl 2 , 10% acetonitrile). The digestion was carried out overnight at 37°C and stopped by formic acid (1%). Peptide extraction was performed using TFA (0.1%) for the analysis. The HPLC system used was from the UltiMate TM system (Dionex Corp., Sunnyvale, CA) equipped with a PepMap C18 purification column (300 m ϫ 5 mm) and a 75 m ϫ 150-mm analytical column of the same material. The peptides were loaded on LC Packings "Switchos" module (Dionex) and a linear gradient (5-80%) of acetonitrile in formic acid (0.1%) was used for the elution. The LC-MS/MS analyses were carried out with the UltiMate TM system interfaced to an LTQ linear ion trap mass spectrometer (Thermo, San Jose, CA). The nanospray source of Proxeon (Odense, Denmark) was used with the distal coated silica capillaries of New Objective (Woburn, MA) at a voltage of up to 1500 V. Peptide spectra were recorded over the mass range of m/z 450 -1600, and MS/MS spectra were recorded in information-dependent data acquisition over the mass range of m/z 200 -2000. One full spectrum was recorded, followed by 4 MS/MS spectra, automatic gain control was applied, and the collision energy was set to the arbitrary value of 35 as collision gas. MS/MS spectra were interpreted by Mascot (Matrix Science Ltd., London, UK), peptide tolerance was set to Ϯ2 Da, and MS/MS tolerance was set to Ϯ0.8 Da, by using the nr (nonredundant) protein database (NCBI Resources, National Institutes of Health, Bethesda, MD). Proteins identified by the MASCOT search were inspected to create a protein list. The whole analysis was performed by the Mass Spectrometry Facility, Medical University of Vienna, Vienna, Austria.

Protein purification
M6P/IGF2R was isolated from THP-1 cells on cyanogen bromide-activated Sepharose 4B (Uppsala, Sweden) conjugated with the anti-M6P/IGF2R mAb MEM-238. The cells were lysed in lysis buffer containing 20 mM Tris-HCl (pH 8.2), 140 mM NaCl, 1% Nonidet P-40 detergent, and a mixture of protease inhibitors (10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 100 nM N ␣ -p-tosyl-L-lysine chlormetylketone, 100 nM N-CBZ-L-phenylalanine chloromethyl ketone, 1 mM pepstatin A) for 30 min on ice. The lysate was clarified by centrifugation at 2,000 ϫ g for 30 s at 4°C and loaded onto the column. The column was washed with lysis buffer (3ϫ volume). The elution was performed using the lysis buffer under stepwise increase of the pH (9.0, 10.0, and 11.7). After elution, the samples were immediately neutralized by adjusting the pH to 7.0 using HCl. Afterward, the fractions were analyzed by SDS-PAGE followed by immunoblotting or Coomassie Blue gel staining. By using a uPA-conjugated agarose column, soluble uPAR was isolated as described previously (59) from the conditioned medium of CHO cells expressing recombinant soluble uPAR. Purified proteins were optionally used in in vitro binding assays.

In vitro binding assay
For the binding assay, various molecules solubilized in PBS (pH 8.7) at a concentration of ϳ5 g/ml were coated on wells of a 96-well BD Falcon TM plate for 2 h at 37°C. Then the wells were blocked with 1% BSA in PBS for 1 h at room temperature and washed two times with binding buffer (20 mM Tris-HCl, 140 mM NaCl, pH 7.5). The wells were incubated for 4 h on ice with binding buffer supplemented with purified assayed proteins (ϳ5 g/ml) in the absence or presence of additional molecules and afterward washed two times with ice-cold binding buffer. Bound material was analyzed by SDS-PAGE and immunoblotting.

Immunoprecipitation
For immunoprecipitation, a 96-well BD Falcon TM plastic U-bottom plate was coated with goat anti-mouse IgG (5 g/ml in PBS, pH 8.7) for 2 h at 37°C. The plate was washed twice with PBS, incubated with specific antibodies (10 g/ml in PBS) for 2 h at 37°C, blocked with 1% BSA in PBS for 1 h at 37°C, and washed twice with cold lysis buffer. Granulocytes were lysed for 30 min on ice with 1% Triton X-100 together with 0.5% dodecyl maltoside in 20 mM sodium phosphate buffer (pH 7.4) with a mixture of protease inhibitors and benzonase (Merck, Darmstadt, Germany; 50 units/ Lactoferrin controls plasminogen activation ml). Afterward, the lysate was passed twice through a needle (25 gauge). The lysate was clarified by centrifugation at 2,000 ϫ g for 30 s at 4°C. The supernatant, or in some experiments human serum, was loaded into the wells (50 l/well) and incubated for 4 h at 4°C. Then the wells were washed four times with cold lysis buffer, and the precipitates were collected with sample buffer and analyzed using SDS-PAGE followed by immunoblotting.

Electrophoresis and immunoblotting
Samples obtained from cell lysis, protein purification, in vitro binding assays, in vitro proteolysis assay, or immunoprecipitation analysis were analyzed by electrophoresis on an appropriate SDS-polyacrylamide gel, followed by a transfer at constant voltage (15 V) to Immobilon polyvinylidene difluoride membranes (Millipore Co., Bedford, MA). The membranes were blocked using 4% nonfat milk and immunostained with specific mAb. For visualization of proteins, the chemiluminescence image analyzer FUJIFILM/LAS-4000 was used.

BN-PAGE
As described in detail elsewhere (12), in the first dimension, a native separation gel was used. Ferritin (monomer (440 kDa) and dimer (880 kDa)), jack bean urease (monomer (91 kDa), trimer (272 kDa), and hexamer (545 kDa)) and BSA (all from Sigma) were used as markers. The serum samples were loaded as described (12), and electrophoresis was performed at 80 V and room temperature overnight. The following day, the lanes were cut from the gel, put on the top of a second-dimension SDS-polyacrylamide gel, and run at room temperature overnight. Afterward, the gel was blotted onto a Immobilon polyvinylidene difluoride membrane, and the membrane was blocked with 4% milk and then incubated with specific antibodies for visualization.

Plasminogen activation assay
For the analysis of Plg activation in a cell-free system, uPA or SK (10 nM) in PBS (pH 8.7) was coated on a 96-well Falcon plate for 2 h at 37°C. The wells were blocked with 1% BSA in PBS for 1 h at room temperature, washed two times with PBS, and afterward incubated at 37°C in PBS with human Glu-Plg (50 nM) and the chromogenic plasmin substrate (S-2251; 0.8 mM).
For the analysis of Plg activation at the cell surface, the cells were washed twice with serum-free RPMI 1640 medium and afterward incubated at 37°C in serum-free medium together with human Glu-Plg (50 nM) and S-2251 in a 96-well plate (5 ϫ 10 5 cells/well). The absorbance change at 405 nm was monitored at different time points by using an ELISA reader (SpectraMax M5, Molecular Devices). To block the activity of unbound plasmin and to study Plg activation at the cell surface only, the assays were performed in the presence of ␣2-antiplasmin (2.5 g/ml). Optionally, various substances were added to the reaction.
For the analysis of Plg activation by bacteria, a 2ϫ master mix containing BSA (final concentration, 2 g/ml), Plg (50 nM), and the chromogenic substrate S-2251 (0.8 mM) diluted in PBS was prepared. Bacteria were added to the master mix (5 ϫ 10 7 cfu/ml in the case of S. pyogenes and 1 ϫ 10 7 cfu/ml for Borrelia). Aliquots without bacteria were used as controls. Finally, the reaction volume was made up to 100 l with PBS. Tubes were shaken at 37°C overnight, and the color change was measured in triplicate using an EL 808 microplate reader (BioTek, Bad Friedrichshall, Germany). The basal level of Plg activation measured only with human serum without bacteria was subtracted.

Cell invasion assay
The invasion of both TCL-598 cells and HUVECs was measured by an assay based on the Boyden chamber principle with filter inserts containing 8-m pore size polycarbonate membranes coated with basement membrane Matrigel matrix (BD Biosciences). Optionally, the Matrigel was pre-incubated with inhibitors (LF or aprotinin) diluted in serum-free medium for 1 h at 37°C. Either epidermal growth factor (10 ng/ml) or vascular endothelial growth factor (50 ng/ml) in RPMI 1640 medium (10% FCS) was added into the lower chambers to induce migration of the TCL-598 cells or HUVECs, respectively. The cells were seeded into the upper part of the filter (1 ϫ 10 5 cells/filter) and incubated for 8 h at 37°C. Then the filters were removed, the upper surface of the membrane was scraped free of cells, and the number of cells invaded onto the lower side of the filters was evaluated after fixing and staining with crystal violet by counting.

Structure comparisons
All three-dimensional structures were retrieved from the Protein Data Bank (PDB) (60). Comparison of structures was carried out by Multiprot (61), and the structures were visualized by WebLabViewerLite (Molecular Simulations, Inc.). Superimposed structure of LF (real structure; PDB code 1LFG (19)) was compared with TF (real structure; PDB code 3QYT (62)). The superimposed part covered 441 C␣ atoms with a 1.2-Å root mean square deviation. The N-terminal part of LF corresponding to the antimicrobial peptide LFC (partial structure; PDB code 1LFG (19), residues Gly 1 -Ala 48 ) was compared with the corresponding part of transferrin (partial structure; PDB code 3QYT (62); residues Val 1 -Ala 51 ). The superimposed part covers 45 C␣ atoms with a 0.88 Å root mean square deviation.

Determination of binding by SPR
A Biacore 3000 instrument with a streptavidin Sensor Chip SA (GE Healthcare) was used. All experiments were performed at 25°C in PBS (10 mM phosphate buffer, pH 7.2, 150 mM NaCl) with 0.005% Tween 20 as running buffer. For biotinylation, iron-saturated recombinant human LF (Sigma, L1294), 1.8 mg/ml in 0.1 M sodium acetate, pH 5.5, was treated with 10 mM sodium meta-periodate for 15 min on ice to introduce carbonyl groups on LF's glycans. Oxidized LF was buffer-exchanged using a desalting column equilibrated with PBS, giving 0.7 mg/ml LF solution. A DMSO stock solution of EZ-link hydrazide biotin (Thermo Scientific, 21339) was added to a 5 mM final concentration, and the mixture was incubated for 45 min at 25°C. Unreacted ligand was removed on a desalting column Lactoferrin controls plasminogen activation equilibrated with PBS. Biotinylated LF (ϳ1 g/ml) was immobilized on the SA sensor chip to a final level of 1400 response units. After sensor chip pretreatment with 100 mM HCl (see "Results" for details), serially diluted plasminogen (10 -160 nM) was injected at a 30 l/min flow rate, allowing an association and dissociation time of 60 and 120 s, respectively. Regeneration of the LF surface was accomplished by a 13-s injection of 10 mM HCl. Binding data were double-referenced (63), and the kinetic constants and affinity were derived from fitting a 1:1, two-state reaction model as implemented in BIAevaluation software version 4.1.1. Rate constants were approximated globally, maximal responses were fitted locally, and the bulk response was set to zero. The influence of the LF iron content on the binding of plasminogen was tested after iron resaturation with 10 M iron(III) sulfate in 100 M imidazole, pH 7.0, and 10 M sodium bicarbonate.

Light-scattering measurements
LF solutions were centrifuged at 10,000 ϫ g for 10 min, transferred into a 1-l cuvette, and measured in a Dynapro NanoStar instrument controlled by Dynamics software version 7.7.0.125 (Wyatt Technology). Both DLS and static light-scattering data were collected from sample. Evaluation of data was performed by the Dynamics software. Adopting a globular protein model, DLS autocorrelation data were analyzed by a cumulant analysis with polydispersity calculation and a regularization analysis by the Dynals algorithm for characterization of average particle radius and high-molecular weight species, respectively. Static light scattering averaged intensity was used for molar mass calculation using a sphere model. To cull the acquisitions influenced by dust or irregular particles, an automatic filtering of autocorrelation functions was applied with a baseline threshold of 0.997-1.003 and a maximal allowed sum-of-squares error for cumulant fit equal to 10.

Statistical analysis
All experiments were performed at least three times in at least triplicates. The data were expressed as mean values with S.D. Statistical significance was evaluated by using Student's t test; values of p Ͻ 0.05 (*) and p Ͻ 0.005 (**) (as indicated) were considered to be significant or highly significant, respectively.