Recombinant toxins that bind to the urokinase receptor are cytotoxic without requiring binding to the alpha(2)-macroglobulin receptor.

The alpha(2-)macroglobulin receptor (alpha(2)MR) has been reported to mediate the internalization of the urokinase plasminogen activator receptor (uPAR) via ligand binding to both receptors. To target malignant uPAR-expressing cells and to determine whether uPAR can internalize without ligand binding to alpha(2)MR, we engineered two recombinant toxins, ATF-PE38 and ATF-PE38KDEL. Each consists of the amino-terminal fragment (ATF) of human urokinase and a truncated form of Pseudomonas exotoxin (PE) devoid of domain Ia, which binds alpha(2)MR. ATF-PE38 and ATF-PE38KDEL were cytotoxic toward malignant uPAR-bearing cells, with IC(50) values as low as 0.02 ng/ml (0.3 pM). Cytotoxicity could be blocked using either recombinant urokinase or free ATF, indicating that the cytotoxicity of the recombinant toxins was specific. Radiolabeled ATF-PE38 had high affinity for uPAR (K(d) = 0.4-8 nM) on a variety of different malignant cell types and internalized at a rate similar to that of ATF. The cytotoxicity was not diminished by receptor-associated protein, which binds and shields the alpha(2)MR from other proteins, or by incubation with phorbol myristate acetate, which is known to decrease the number of alpha(2)MRs in U937 cells or by antibodies to alpha(2)MR. Therefore, these recombinant toxins appear to internalize via uPAR without association with the alpha(2)MR.

The ␣ 2-macroglobulin receptor (␣ 2 MR) has been reported to mediate the internalization of the urokinase plasminogen activator receptor (uPAR) via ligand binding to both receptors. To target malignant uPAR-expressing cells and to determine whether uPAR can internalize without ligand binding to ␣ 2 MR, we engineered two recombinant toxins, ATF-PE38 and ATF-PE38KDEL. Each consists of the amino-terminal fragment (ATF) of human urokinase and a truncated form of Pseudomonas exotoxin (PE) devoid of domain Ia, which binds ␣ 2 MR. ATF-PE38 and ATF-PE38KDEL were cytotoxic toward malignant uPAR-bearing cells, with IC 50 values as low as 0.02 ng/ml (0.3 pM). Cytotoxicity could be blocked using either recombinant urokinase or free ATF, indicating that the cytotoxicity of the recombinant toxins was specific. Radiolabeled ATF-PE38 had high affinity for uPAR (K d ‫؍‬ 0.4 -8 nM) on a variety of different malignant cell types and internalized at a rate similar to that of ATF. The cytotoxicity was not diminished by receptor-associated protein, which binds and shields the ␣ 2 MR from other proteins, or by incubation with phorbol myristate acetate, which is known to decrease the number of ␣ 2 MRs in U937 cells or by antibodies to ␣ 2 MR. Therefore, these recombinant toxins appear to internalize via uPAR without association with the ␣ 2 MR.
Urokinase plasminogen activator (uPA) 1 is a serine protease that activates plasminogen to plasmin, which in turn degrades fibrin and other extracellular matrix proteins (1). Human uPA is 411 amino acids in length (2) and is produced by kidney cells or fibroblasts as a 55-kDa protein glycosylated at several residues including Thr 18 and Asn 302 (3,4). The binding domain is the epidermal growth factor-like amino-terminal fragment (ATF; amino acids 1-135, 15 kDa) that binds with high affinity (K d ϭ 0.5 nM) to its receptor uPAR (5). uPA is cleaved between Lys 158 and Ile 159 by plasmin or kallikrein to the active twochain protease. The catalytic domain of uPA can be inactivated upon interaction with one of the plasminogen activator inhibitors (PAI-1, PAI-2, or protease nexin-1) (6,7). Whereas active uPA bound to uPAR is stable on the cell surface, the uPA⅐PAI complex internalizes, preventing uPA from asserting its biologic functions (5). uPAR also can bind vitronectin, an important adhesion protein in the plasma and extracellular matrix (8). In this way, uPAR is believed to play a role in the movement or invasion of malignant cells through the extracellular matrix. uPAR is overexpressed on a variety of tumors, including monocytic and myelogenous leukemias (9,10) and cancers of the breast (11), bladder (12), thyroid (13), stomach (14), liver (15), pleura (16), lung (17), pancreas (18), and ovaries (19).
It has been shown that internalization of the uPA⅐PAI complex via uPAR requires both portions of the uPA⅐PAI complex to make contact with a different receptor, the ␣ 2 -macroglobulin receptor (␣ 2 MR, also termed low density lipoprotein receptorrelated protein (LRP)) (20 -23). LRP/␣ 2 MR binds a variety of other ligands and ligand complexes, including tissue plasminogen activator and tissue plasminogen activator⅐PAI (20), lipoprotein lipase (24), lactoferrin (25), and very low density lipoprotein (26). LRP/␣ 2 MR also binds and internalizes protein toxins, including Pseudomonas exotoxin (PE) (27) and saporin (SAP) (28). The 40-kDa receptor-associated protein (RAP) also binds to LRP/␣ 2 MR and can displace other known ligands from binding (29). uPAR itself internalizes along with the uPA⅐PAI complex and LRP/␣ 2 MR (30). ATF, the binding domain of uPA, has been reported not to internalize, since the catalytic domain of uPA is needed to bind PAI (31).
To study the internalization of uPAR, Cavallaro et al. (32,33) chemically conjugated uPA to the plant toxin SAP, which like most protein toxins requires internalization in order to be cytotoxic, and uPA-SAP was selectively cytotoxic toward uPARexpressing cells. LRP/␣ 2 MR was found to be involved in the internalization of uPA-SAP due to the binding of SAP to LRP/ ␣ 2 MR, because 1) cytotoxicity could not be competed by an excess of the uPA catalytic domain, indicating that formation of the uPA-PAI complex was not necessary for internalization of uPA (32), 2) an excess of uPA-SAP could displace the binding of RAP to LRP/␣ 2 MR, and 3) decreasing LRP/␣ 2 MR expression on cells using phorbol 12-myristate 13-acetate (PMA) resulted in resistance to uPA-SAP (28). More recently, a fusion toxin was made containing ATF and a saporin isoform (SAP-3), and ATF-SAP-3 was also cytotoxic due to the binding of saporin to LRP/␣ 2 MR (34). These studies were consistent with the conclusion that uPAR requires LRP/␣ 2 MR for internalization. To test this hypothesis directly, we decided to produce a chimeric toxin that would bind to uPAR but not to PAI or to LRP/␣ 2 MR. This was accomplished by producing a fusion of ATF with PE38, a truncated form of PE that is known not to bind to LRP/␣ 2 MR.
PE is a 613-amino acid single-chain bacterial toxin composed of domains Ia, II, Ib, and III (amino acids 1-252, 252-364, 365-399, and 400 -613, respectively), which carry out functions required for intoxication, as elucidated by structural and functional studies (35,36). A current model of how PE kills cells contains the following steps. 1) The C-terminal residue (lysine 613) is removed by a carboxypeptidase in the plasma or culture medium (37). 2) Domain Ia binds to LRP/␣ 2 MR and is internalized via endosomes to the transreticular Golgi (27). 3) After internalization, domain II is proteolytically cleaved between amino acids 279 and 280 by furin (38 -40). 4) The disulfide bond between cysteines 265 and 287, which joins the two fragments generated by proteolysis, is reduced. 5) Amino acids 609 -612 (REDL) bind to the intracellular KDEL receptor, which transports the 37-kDa carboxyl-terminal fragment from the transreticular Golgi apparatus to the endoplasmic reticulum (41,42). 6) Amino acids 280 -313 mediate translocation of the toxin to the cytosol (43,44). 7) The ADP-ribosylating enzyme within amino acids 400 -602 inactivates EF-2 (45), leading to cell death, which is facilitated by apoptosis (46). Replacement of the native carboxyl-terminal sequence of PEcontaining toxins, REDLK, with the amino acids KDEL results in increased cytotoxic activity associated with improved KDEL receptor binding (42,47,48). The most common truncated form of PE for the production of fusion toxins is PE38, which is missing all of domain Ia and part of domain Ib (amino acids 365-380) (47,49). The fusion toxins produced for the present study were ATF-PE38 and the more active version ATF-PE38KDEL.

EXPERIMENTAL PROCEDURES
Plasmid Preparation-Schematic structures for uPA and the recombinant proteins are shown in Fig. 1. Plasmids contained DNA encoding recombinant toxins or ATF under control of the T7 promoter for expression in Escherichia coli BL21/DE3. DNA encoding ATF was amplified from a human spleen cDNA library (CLONTECH, Palo Alto, CA) using the primers AA1 (5Ј-GCG ACT CCC ATA TGA GCA ATG AAC TTC ATC AAG-3Ј) and AA135 (5Ј-AGG AGA GGA GGA AGC TTT TCC ATC TGC GCA GTC-3Ј). The 430-base pair fragment obtained encoded the ATF sequence and contained a 5Ј NdeI and a 3Ј HindIII site. The two plasmids pVRU9, encoding ATF-PE38, and pVRU9K, encoding ATF-PE38KDEL, were produced by ligating the 410-base pair NdeI-HindIII polymerase chain reaction fragment into either the 4.1-kb NdeI-HindIII fragment of pRK29 or the 4.0-kb NdeI-HindIII fragment of pRK29K (50), respectively. The vector pRK29, which encodes Mik-␤1(Fv)-PE38 and contains NdeI and HindIII restriction sites 5Ј and 3Ј of the Mik-␤1(Fv)-encoding region, was produced by ligating the 770base pair XbaI-HindIII fragment of pRK28 (51) with the 4.1-kb XbaI-HindIII fragment of pRK79 (50). To make pRKHB9, encoding the antitransferrin receptor recombinant immunotoxin anti-TFR(Fv)-PE38, the 0.35-kb NdeI-BamHI and 0.35-kb BamHI-HindIII fragments of plasmid pJBDT1-anti-TFR(Fv) (52) were ligated to the 4.1-kb NdeI-HindIII fragment of pRK79. pVRU, encoding ATF, was constructed by ligating the 410-base pair NdeI-HindIII fragment from pVRU9 to the 3.0-kb NdeI-HindIII fragment of pRKGC. The intermediate vector pRKGC, which encodes human granulocyte colony-stimulating factor, was constructed from an NdeI-HindIII fragment of a polymerase chain reaction product ligated into the 3.0-kb NdeI-HindIII fragment of pRKL4 (53). Plasmid sequences were verified using an automated DNA sequencer from Applied Biosystems (Perkin-Elmer) to rule out polymerase chain reaction or oligonucleotide construction errors.
Production of Recombinant Proteins-Plasmids encoding ATF-PE38 and ATF-PE38KDEL were expressed and purified as described previously for other single-chain immunotoxins (54). The BL21/DE3 used for transformation contained the plasmid pUBS500 to facilitate the expression of plasmids containing the codons AGA and AGG (55). Inclusion bodies were obtained from E. coli, washed in detergent, and then denatured and reduced in buffer containing 7 M guanidine and 10 mg/ml dithioerythritol. The reduced denatured protein at 10 mg/ml was renatured in a redox buffer containing 0.1 M Tris-HCl, pH 8.0, 0.5 M L-arginine-HCl, 0.9 mM oxidized glutathione, and 10 mM EDTA for 80 h at 10°C. The renatured protein was dialyzed and then purified by ion-exchange chromatography (Q Sepharose followed by MonoQ; Amersham Pharmacia Biotech) followed by sizing chromatography (TSK G3000SW, TosoHaas, Philadelphia, PA). Free recombinant ATF was produced from inclusion bodies in E. coli. using the methods described for anti-Tac(scdsFv) (56). Proteins were Ͼ95% pure as assessed by SDS-polyacrylamide gel electrophoresis (data not shown). The yields of insoluble inclusion body protein were 50 -150 mg/liter of E. coli culture induced in a shake flask at an A 650 of 2-3. The yield of purified monomeric recombinant protein from total insoluble inclusion body protein was about 10% for both ATF-PE38 and ATF-PE38KDEL and 20% for ATF. The recombinant urokinase used for competition studies was pharmaceutical grade (Abbot Laboratories, North Chicago, IL). RAP-GST and blocking antibodies to LRP/␣ 2 MR were kindly provided by Dr. Dudley Strickland. Saporin was obtained from Advanced Targeting Systems (Carlsbad, CA). ␤-Glucuronidase was purchased from Sigma. The uPAR-binding peptide SLNFSQYLWS and the negative control peptide SLNASQYLWS (57) were synthesized by Genosys (The Woodlands, TX).
Cytotoxicity Assays-The glioma line 897 was kindly provided by Dr. Bigner at Duke University, and the remaining cell lines were available from the ATCC. The cell lines U937, CA46, HUT-102, Raji, and Daudi, which grew in suspension, were plated at 4.0 ϫ 10 4 cells/well in a 96-well plate and immediately incubated with toxins in 100-l aliquots for 24 h (unless otherwise specified) at 37°C. The remaining adherent cell lines were plated the day before toxin addition at 1.5 ϫ 10 4 /well and incubated with toxins in 200-l aliquots. The cells were pulsed with 1 Ci/well of [ 3 H]leucine and then incubated for 6 -8 h at 37°C. The protein was then harvested onto glass fiber filters, which were read in a scintillation counter to determine inhibition of protein synthesis. Adherent cell lines required a freeze and thaw step prior to harvesting. To induce a reduction in surface LRP/␣ 2 MR expression, U937 cells were resuspended in RPMI plus 10% fetal bovine serum containing 150 nM PMA (Sigma) and plated at 1.0 ϫ 10 4 cells/well in a 96-well plate. After 72 h, the activated cells had adhered and were washed twice with 200 l of PMA-free RPMI plus 10% fetal bovine serum and then incubated with 200 l of PMA-free RPMI plus 10% fetal bovine serum containing various dilutions of toxins for 48 h at 37°C. The cells were then pulsed with [ 3 H]leucine and harvested as above.
Binding Studies-ATF-PE38, ATF-PE38KDEL, or ATF (150 g/100 l) were each radiolabeled with 1 mCi of Na 125 I in the presence of 10 g of chloramine T and 0.15 M sodium phosphate, pH 7.5, and after adding 83 g of sodium metabisulfite purified on a PD-10 column (Amersham Pharmacia Biotech) equilibrated and eluted with 0.2% human serum albumin in phosphate-buffered saline. Specific activities were typically 3.5-4 Ci/g. For binding studies, 200-l aliquots of 5 ϫ 10 5 cells in binding buffer (Dulbecco's modified Eagle's medium containing 0.1% bovine serum albumin plus 0.2% sodium azide) were incubated with 0.5, 1.0, 2.0, 4.0, 8.0, or 16.0 nM 125 I-ATF-PE38 in the presence or absence of an excess (up to 1000-fold) of urokinase or ATF. After 90 -120 min at 4°C, the cells were washed by centrifugation twice with cold binding buffer and counted. Adherent cells were liberated for binding assays using 0.2% EDTA in phosphate-buffered saline. For displacement assays, 200-l aliquots of 5 ϫ 10 5 U937 cells in binding buffer were treated with 0.25-0.5 nM 125 I-ATF-PE38 or 125 I-ATF-PE38KDEL and differing concentrations of recombinant proteins or peptides. After 90 -120 min, the cells were washed and counted as described above. In the same manner, 125 I-RAP-GST, radioiodinated like 125 I-ATF-PE38, was displaced from U937 cells using unlabeled RAP-GST, ATF-PE38KDEL, or Saporin. Displacement assays on the adherent cell line A172 were performed similarly except that the cells were plated in 24-well plates at 10 4 /well the day before use, and after incubating with 200-l aliquots of unlabeled proteins on a rocker at 4°C, the cells were washed, and the bound radiolabel was quantitated by dissolving cells in NaOH.
Internalization Assay-U937 cells in 200-l aliquots of 2 ϫ 10 6 cells in media (RPMI containing 10% fetal bovine serum) were incubated with 8 mM concentrations of either 125 I-ATF-PE38KDEL or 125 I-ATF. After incubating for either 1 h at 4°C or 1, 2, or 3 h at 37°C, the cells were washed twice by centrifugation with either medium (to detect bound plus internalized radiolabel) or with medium adjusted to pH 3.0 with HCl (to detect internalized radiolabel).

RESULTS
To determine whether uPAR alone can internalize uPA, recombinant toxins were constructed that contained the binding domain of uPA fused to truncated forms of PE known not to bind to LRP/␣ 2 MR. The cytotoxicity of the recombinant toxins toward malignant cells expressing different levels of uPAR was then measured.
Expression and Purification of Recombinant Immunotox-ins-The recombinant fusion toxins ATF-PE38 and ATF-PE38KDEL, shown schematically in Fig. 1, were produced in E. coli. Recombinant inclusion body protein was denatured and reduced, refolded to active monomer in a redox buffer, and then purified by anion exchange and sizing chromatography to Ͼ95% purity by SDS-polyacrylamide gel electrophoresis (data not shown). The yield of pure 53-kDa monomer was about 10% of the total denatured protein, which provided about 10 mg of pure monomeric protein per liter of refolding solution. On SDSpolyacrylamide gel electrophoresis analysis, both proteins displayed one prominent band at the expected molecular mass of 53 kDa. Thus, using a production protocol that is now standard for PE38-containing immunotoxins, these chimeric toxins could be expressed and purified easily, in high yield and purity. Cytotoxicity of Recombinant Toxins on U937 Cells-To determine whether recombinant ATF-containing toxins would be cytotoxic to uPAR-expressing cells, ATF-PE38 and ATF-PE38KDEL were incubated with the monocytic leukemia line U937 for 20 h, followed by incubation with [ 3 H]leucine to determine inhibition of protein synthesis. Fig. 2A shows that both recombinant ATF-toxins were cytotoxic to U937 cells in a concentration-dependent manner. The recombinant toxin concentration required for 50% inhibition of protein synthesis (IC 50 ) was 7.5 pM for ATF-PE38 and 3.8 pM for the more active ATF-PE38KDEL. Cytotoxicity was Ͼ90% with a 1.9 nM concentration of either toxin. Thus, ATF-toxins that were not expected to interact with PAI or LRP/␣ 2 MR were still cytotoxic toward uPAR-expressing U937 cells.
Specificity of the Cytotoxic Activity of ATF Toxins-To determine whether the cytotoxicity toward U937 cells required the ATF-toxins to bind to uPAR or was instead due to nonspecific internalization, several control experiments were performed. First, as shown in Fig. 2A, U937 cells were coincubated with ATF-PE38 or ATF-PE38KDEL and a 300-fold molar excess of uPA. Excess uPA completely prevented the cytotoxic activity of either recombinant toxin, indicating that their binding to uPAR was required for their cytotoxicity. Second, the recombinant toxin anti-Tac(Fv)-PE38 (50), which binds to CD25 instead of uPAR but contains the same toxin domains as ATF-PE38, was tested against U937 cells. As shown in Fig. 2A, anti-Tac(Fv)-PE38 was not cytotoxic toward CD25-negative U937 cells, indicating that the cytotoxic activity of ATF-PE38 was not due to nonspecific internalization.
Several additional specificity experiments were performed to determine whether the addition of excess urokinase prevented the cytotoxicity of ATF-toxins by proteolytically destroying the toxin rather than by blocking the binding of ATF-toxin to the urokinase receptor. In Fig. 2B, U937 cells were incubated with increasing concentrations of ATF-PE38KDEL (0, 1.9, 19, 190, and 1900 pM) and a constant concentration of urokinase (400 nM), ATF (1.3 M), or the anti-transferrin receptor (anti-TFR) monoclonal antibody HB21 (800 nM) (58). Protein synthesis in the absence of ATF-PE38KDEL, as depicted in Fig. 2B by points along the y axis, was slightly (15%) higher in the presence of ATF and urokinase. The cytotoxicity of ATF-PE38KDEL (IC 50 ϭ 32 Ϯ 2 pM) was prevented by ATF or urokinase (IC 50 Ͼ 1900 pM) but not by anti-TFR (IC 50 ϭ 26 Ϯ 5 pM). Large excesses of competitor over recombinant toxin are necessary to prevent cytotoxicity, since binding of competitors is reversible but events following toxin internalization are not. To determine whether urokinase inhibited the activity of ATF-PE38KDEL simply by proteolytically inactivating PE38KDEL, the positive control immunotoxin anti-TFR(Fv)-PE38KDEL (54) (0, 0.16, 1.6, 16, and 160 pM) was tested in place of ATF-PE38KDEL combined with ATF, urokinase, or anti-TFR. As shown in Fig. 2B, the cytotoxicity of anti-TFR(Fv)-PE38KDEL (IC 50 ϭ 3.3 Ϯ 0.3 pM) was prevented by anti-TFR (IC 50 Ͼ 160 ng/ml) but not by 400 nM urokinase (IC 50 ϭ 4 Ϯ 0.5 pM) or 1.3 M ATF (IC 50 ϭ 3 Ϯ 0.3 pM). Thus, urokinase appears to block the cytotoxicity of ATF-PE38KDEL by preventing its binding to the urokinase receptor rather than by proteolytically inactivating PE38KDEL or by stimulating cellular protein synthesis. Similar results were obtained when the recombinant toxins ATF-PE38 and anti-TFR(Fv)-PE38 were substituted for ATF-PE38KDEL and anti-TFR(Fv)-PE38KDEL, respectively (data not shown).
To explore the unlikely possibility that proteolytic inactivation by urokinase of ATF-PE38KDEL could occur despite the lack of inactivation by urokinase of anti-TFR(Fv)-PE38KDEL, the experiments shown in Fig. 2B were repeated at 4°C. U937 cells were exposed to recombinant toxins in the presence or absence of competitors for 4 h at 4°C, and the washed cells were incubated at 37°C for a total of 24 h prior to pulsing with [ 3 H]leucine. As shown in Fig. 2C, with this 4-h toxin exposure, the cytotoxicity of ATF-PE38KDEL (IC 50 ϭ 10 Ϯ 4 ng/ml) was still prevented by urokinase or ATF (IC 50 Ͼ 100 ng/ml) but not by anti-TFR (IC 50 ϭ 11 Ϯ 2 ng/ml). As in Fig. 2B, it is evident in Fig. 2C that in the absence of toxin (y axis), the high concentration of urokinase (400 nM) or ATF (1.3 M) resulted in a moderate (ϳ25%) increase in protein synthesis. Parallel experiments using anti-TFR(Fv)-PE38 or anti-TFR(Fv)-PE38 as in Fig. 2B confirmed these findings (data not shown). These specificity experiments therefore support the hypothesis that ATFtoxins kill cells after binding specifically to the urokinase receptor.
Cytotoxicity and Specificity of ATF-Toxins toward a Variety of Malignant Cells-To determine whether uPAR expressed on cells other than U937 would constitute a target sufficient for ATF internalization, the ATF-toxins were incubated with different types of malignant cells. As shown in Fig. 2, D-F, ATF-PE38 and ATF-PE38KDEL were very cytotoxic toward HT-29 colon carcinoma cells and A172 and SN19 glioblastoma cells. In all three cases, an excess of ATF prevented the cytotoxic activity of ATF-toxins, indicating that uPAR was involved in the internalization of ATF in these cells. In Fig. 2D, the cytotoxicity of ATF-PE38KDEL (IC 50 ϭ 0.03 ng/ml, 0.5 pM) could not be reproduced by PE38KDEL, which lacks a binding domain, indicating that HT-29 cells also do not nonspecifically internalize ATF-PE38KDEL. Fig. 2D shows that an excess of ATF was incapable of preventing the cytotoxic activity of anti-TFR(Fv)-PE38, indicating that the prevention of the cytotoxic activity of ATF-PE38KDEL by excess ATF was due to its blocking of uPAR and not due to nonspecific effects. Table I, the ATFtoxins were cytotoxic toward a variety of malignant cells, including breast cancer, colon carcinoma, epidermoid carcinoma, gastric carcinoma, lung carcinoma, medulloblastoma, and glioblastoma. The difference between the cytotoxic activities of ATF-PE38 and ATF-PE38KDEL was less than 2-fold on MCF7 breast cancer, A431 epidermoid carcinoma, A172 glioblastoma, A549 lung carcinoma, and U937 cells. This difference was over 10-fold for MDA-MB231 breast cancer, HT-29 colon cancer, HTB-103 gastric carcinoma, and on DAOY medulloblastoma cells. U373 glioma cells were nearly 25-fold more sensitive to ATF-PE38KDEL than to ATF-PE38. Thus, as observed with other recombinant toxins that require internalization and intracellular trafficking mediated by the carboxyl terminus for activity (42,59), the KDEL sequence increases cytotoxicity and provides further evidence that these ATF-containing proteins internalize into cells.

Comparison of the Cytotoxic Activities of ATF-Toxins with Respect to Carboxyl Terminus-As shown in
Binding of ATF-Toxins to uPAR-To characterize the binding of ATF-PE38 to uPAR, ATF-PE38 and ATF-PE38KDEL were radioiodinated and displaced by unlabeled ATF or urokinase from U937 cells and A172 cells. For this analysis, the tyrosine rather than the lysine residues of ATF-toxin were radiolabeled with chloramine T, because most (11 of 17) of the tyrosines of ATF-PE38 or ATF-PE38KDEL are located on the toxin rather than on ATF, whereas most of the lysine residues (9 of 12) are present on the ATF ligand. As shown in Fig. 3A for U937 cells and Fig. 3B for A172 cells, both ATF and urokinase displaced the binding of 125 I-ATF-PE38KDEL. The EC 50 , the concentration of unlabeled protein necessary for 50% displacement, was 1.6 nM for ATF compared with 3.2-3.5 nM for urokinase, attributed to the fact that the clinical grade urokinase contains a significant amount of low molecular weight urokinase that is devoid of the ATF domain. SDS-polyacrylamide gel electrophoresis confirmed that about half of the urokinase was in the low molecular weight form (data not shown), accounting for the ϳ2-fold difference in EC 50 .
To confirm that the ATF-toxins were actually binding to uPAR and not to a different cell surface protein capable of binding both ATF and ATF-toxin, the peptide SLNFSQYLWS (57) was used to displace 125 I-ATF-PE38KDEL from both U937 and A172 cells. This decapeptide is the minimal antagonist of the uPA-uPAR interaction, derived from the peptide AEPMPH-SLNFSQYLWYT isolated by phage display (60). By surface plasmon resonance analysis, SLNFSQYLWS was previously reported to bind to immobilized uPAR with about 1% of the affinity of ATF (57). SLNASQYLWS was a less active control peptide identified by alanine scanning mutagenesis of the active decamer (57). As shown in Fig. 3, A and B, SLNFSQYLWS blocked the binding of 125 I-ATF-PE38KDEL to U937 and A172 cells, with EC 50 values of 1900 and 800 nM, respectively. The alanine mutant SLNASQYLWS was much less active in blocking the binding of ATF-toxin. Thus, the ATF-toxin bound directly to uPAR on cells.
To determine the extent to which the toxin impaired the binding of ATF to uPAR by fusion of toxin to the carboxyl terminus of ATF, recombinant ATF, purified from E. coli, was compared with ATF-toxin in displacement of 125 I-ATF-PE38KDEL. As shown in Fig. 3, fusion of ATF to PE38 or PE38KDEL resulted in only a slight impairment of the binding of ATF to uPAR, with the EC 50 values of ATF-toxins (2.1-2.7 nM) and ATF (1.6 nM) varying less than 2-fold. Thus, the ATFtoxins bind to uPAR with affinity similar to that of ATF. Similar results were obtained in displacement assays using 125 I-ATF-PE38 instead of 125 I-ATF-PE38KDEL (data not shown). The lack of difference in binding of ATF-PE38KDEL compared with ATF-PE38 indicates that the higher cytotoxicity of ATF-PE38KDEL is not due to increased cell binding but rather due to improved intracellular trafficking after internalization into the cells.
Comparison of the Cytotoxic Activities of ATF-Toxins and Their Surface Expression of uPAR-To quantitate the surface expression of uPAR on the surface of some of the cell lines tested, binding studies were performed with increasing amounts of 125 I-ATF-PE38. Nonspecific binding was determined by competition with up to a 1000-fold excess of either unlabeled urokinase or recombinant ATF. As shown in Fig. 4, saturation binding is represented on Scatchard plots. Table II lists the numbers of sites/cell and K d values for each of the cell lines tested. uPAR expression varied from less than 1000/cell in T98G cells to 10 6 sites/cell in A172 glioblastoma cells. The binding affinities of 125 I-ATF-PE38 were similar, with K d values usually between 1.1 and 2.3 nM. uPAR expression correlated roughly with sensitivity to ATF-toxins, with A172 cells having the most uPAR and high sensitivity to ATF-PE38 (IC 50 ϭ 0.75 pM) and T98G cells having low uPAR expression (670 sites/cell) and much less sensitivity to ATF-PE38 (IC 50 Ͼ 190 pM). LNCaP prostate carcinoma cells (61) and lymphocytic leukemia cells or lymphoma cells (10) are reported not to express significant levels of uPAR and were not sensitive to ATF-toxins (Table I). However, among glioblastoma cell lines, uPAR expression (A172 Ͼ 897 Ͼ SN19 Ͼ U251 Ͼ T98G) did not closely correlate with sensitivity to ATF-toxins (SN19 Ͼ A172 Ͼ U251 Ͼ T98G Ͼ 897). Thus, cellular characteristics other than uPAR expression, such as speed of uPAR internalization or intracellular processing or trafficking of ATF-toxins, appeared to play a role in sensitivity of the cells to ATF-toxins.
Stability of ATF-Toxins in Serum-The short half-life previously reported for human uPA in plasma (62) suggests that interpretation of any studies involving the incubation of live cells with uPA-like molecules in serum containing medium must take into consideration the stability of the recombinant protein over time. Recombinant toxins containing PE38 are known to be stable at 37°C for over 7 days (63,64), and both PE38 and PE38KDEL were stable with high concentrations of urokinase (Fig. 2), but ATF-toxins may be unstable due to degradation of ATF ligand in plasma. To investigate this possibility, ATF-PE38 was incubated with human plasma for different time periods and frozen at Ϫ80°C, and the samples were tested simultaneously for cytotoxicity toward U937 cells. As shown in Fig. 5, the cytotoxic activity of ATF-PE38 decreases during incubation in human plasma, but even at 48 h the recombinant toxin is still Ͼ10% active. Thus, ATF-PE38 exhibits a time-dependent decrease in cytotoxicity in human plasma probably related to degradation of ATF. The presence of significant cytotoxic activity at 48 h indicates residual recombinant toxin persists and is available for internalization through uPAR.
Internalization of ATF-Toxin-To compare the rates of internalization of ATF and ATF-toxin into cells, U937 cells were incubated with 125 I-ATF or 125 I-ATF-PE38KDEL, and the amount internalized was measured at different time points by removing bound ligand with acidic (pH 3.0) medium. The time points chosen were 1-4 h due to the stability results (Fig. 5) described above. Fig. 6A shows the calculated number of molecules/cell internalized into U937 cells after incubation at 4°C for 1 h or at 37°C for 1, 2, or 4 h. Between 1 and 4 h at 37°C, the rate of internalization of 125 I-ATF-PE38KDEL was 1760 Ϯ 310 molecules/cell/h, while that of 125 I-ATF was 770 Ϯ 170 molecules/cell/h. Fig. 6B shows the total uptake (bound plus internalized) of 125 I-ATF-PE38KDEL and 125 I-ATF in U937 cells, determined by incubating with the radiolabeled proteins and washing with media of pH 7.4. Between 1 and 4 h, the total amount of cell-associated 125 I-ATF-PE38KDEL increased by 1270 Ϯ 530 molecules/cell/h, while that of 125 I-ATF increased by 1500 Ϯ 570 molecules/cell/h. The uptake and internalization of 125 I-ATF-PE38KDEL was blocked by an excess of ATF (data not shown), indicating that it internalized via uPAR. The ϳ2fold difference between the internalization of 125 I-ATF-PE38KDEL and 125 I-ATF was reproducible when assaying from 0 to 4 h (data not shown). Thus, ATF-PE38KDEL and ATF associated with cells at equal rates, but ATF-toxin, which has an endoplasmic reticulum retention sequence, accumulated inside cells at a higher rate.
Internalization of Urokinase through an LRP/␣ 2 MR-independent Pathway-Previous data clearly showed that truncated forms of PE devoid of domain Ia, such as PE38, completely lose the capacity to bind LRP/␣ 2 MR (27,36,49). Nevertheless, experiments were performed in the present study to determine directly whether ATF-toxins are independent of LRP/␣ 2 MR for internalization. Because the 40-kDa RAP ligand for LRP/␣ 2 MR has been shown to displace other ligands, including PE (27), ATF-PE38 was incubated with U937 cells in the presence or absence of an excess of RAP. As shown in Fig.  7A, the cytotoxicity of ATF-PE38 was not significantly prevented by coincubation with 1.25 M of RAP but was essentially completely reversed by coincubation with 1.4 M of uPA. As shown in Fig. 7B, 0.7 M ATF but not 1.25 M RAP could prevent the cytotoxicity of ATF-PE38 to U937. In both experiments involving competition with an excess of RAP-GST (Fig.  7, A and B), the cytotoxicity of full-length PE at 2 nM was significantly prevented by coincubation with RAP but not with uPA ( Fig. 7A) or with ATF (Fig. 7B). It was next determined whether rabbit polyclonal antibody to LRP/␣ 2 MR, which has been shown to block the cytotoxicity of ATF-SAP (34), could also block the cytotoxicity of ATF-PE38. As shown in Fig. 7C, the protein synthesis inhibition of 19 pM ATF-PE38 on U937 cells was 56 -58% in the presence or absence of an excess (0.8 M) of antibody but was completely prevented by coincubation of ATF-PE38 19 pM with 0.7 M of ATF. Finally, a displacement assay was performed to determine whether ATF-PE38KDEL could displace 125 I-RAP-GST from LRP/␣ 2 MR, as has been shown for uPA-SAP (28). U937 cells were incubated with 0.2 nM 125 I-RAP-GST in the presence or absence of unlabeled ATF-PE38KDEL, saporin, or RAP-GST. As shown in Fig. 7D, no detectable displacement of 125 I-RAP-GST was observed even with 500 nM ATF-PE38KDEL, but significant displacement was detected with the same concentration of saporin. Taken together, the data indicate that while saporin or fusion proteins containing saporin bind to LRP/␣ 2 MR, ATF fused to truncated PE does not require LRP/␣ 2 MR to internalize.
Cytotoxicity of ATF-Toxins on PMA-treated U937 Cells-To determine the effect of LRP/␣ 2 MR-depletion on the cytotoxicity of ATF-toxins, they were incubated with U937 cells after pretreatment of the cells with PMA, which previously induced resistance to both saporin and uPA-SAP (28). As shown in Table III, PMA incubation for 72 h prior to toxin addition led to a 4-fold or greater resistance to either full-length PE or to saporin. In contrast, PMA incubation had no significant effect on cellular sensitivity to either ATF-PE38 or ATF-PE38KDEL. Taken together, the data indicate that the urokinase binding domain can internalize via uPAR without relying on LRP/␣ 2 MR. DISCUSSION We found, using recombinant toxins ATF-PE38 and ATF-PE38KDEL, which do not bind to LRP/␣ 2 MR, that uPAR is able to internalize without evidence of ligand binding to LRP/␣ 2 MR. Moreover, these agents are extremely cytotoxic selectively toward uPAR-expressing malignant cells, particularly glioblastoma multiforme.
Known Potential Pathways for the Internalization of uPAR-In some human cells, the internalization of uPA via uPAR has been previously shown to require binding through other ligands to LRP/␣ 2 MR, since 1) ATF alone does not internalize (31), 2) uPA does not internalize without PAI, which in turn binds to LRP/␣ 2 MR (20), and 3) the cytotoxicity of chimeric toxins containing uPA or ATF and saporin is mediated by binding of the saporin toxin to LRP/␣ 2 MR (28,(32)(33)(34). More recently, Nykjaer et al. (65) reported the presence of uPAR in the lysosomes of human HT1080 fibroblasts even in the presence of excess RAP. These studies identified a new pathway for the internalization of uPAR, namely the cation-independent, mannose 6-phosphate/insulin-like growth factor-II receptor (CIMPR) (65). In cell-free experiments, it was found that uPAR binds to a region on CIMPR that is different from the binding site for the ligands mannose-6-phosphate, insulin-like growth factor-II, and ␤-glucuronidase (65). While studies of the internalization of ligands such as ATF or uPAR by the uPAR-CIMPR pathway have not yet been reported, cell-free experiments showed that uPA did not affect the binding of uPAR to CIMPR (65). Thus, it would be expected that uPAR, upon binding ATF or ATF-toxin, would interact with CIMPR. It is unknown, however, whether this pathway would facilitate the cytotoxicity of ATF-toxin, since delivery of toxins to lysosomes is a process that is thought to prevent toxin molecules from reaching the cytosol. Alternatively, CIMPR might mediate internalization of an amount of ATF (quantitated by 125 I-ATF in Fig. 6A) that was too small to detect in prior studies of U937 cells (31). The few molecules of ATF-PE38 or ATF-PE38KDEL internalized could be rescued by the KDEL receptor for cytosolic trafficking instead of destruction of the toxin in lysosomes. To investigate this possibility, we attempted to block cytoxicity or internalization of ATF-toxin in U937 cells using 5 M insulin-like growth factor-II or 10 M ␤-glucuronidase but found no competition of either internalization or cytotoxicity (data not shown). Although these concentrations of insulin-like growth factor-II or ␤-glucuronidase were sufficient to significantly block the cell-free association of uPAR with CIMPR (65), it is unknown whether such competition would block this association if it were present in cells internalizing ATF-toxin. Murine cells have been isolated that are known not to contain CIMPR, but the murine uPAR on such cells would not bind toxins containing ATF. Further investigation of the role of the CIMPR pathway will require the isolation of human CIMPR-negative/ uPAR-positive cells and their transfection with CIMPR, and such studies are beyond the scope of the present work.
Cytotoxicity as a Sensitive Assay of Ligand Internalization-Several other cell surface proteins have been thought not to internalize based on standard internalization assays but have proven effective targets for internalizing recombinant toxins. Examples include CD25, the ␣-subunit of the interleukin-2 receptor, and carcinoembryonic antigen (47,66). It is known that protein toxins can kill cells after injection of only one or a few molecules into the cytoplasm (67). Because the steps required for intoxication are less than 100% efficient, hundreds or thousands of molecules must bind and internalize in order to kill cells (68). Thus, the number of receptors internalizing may be too few for standard assays of internalization but more than sufficient to allow cytotoxicity by targeted toxins.  Potential Clinical Utility of ATF-Toxins-Since uPAR is present on human liver, ATF-toxins may be too toxic to deliver systemically for the treatment of uPAR-expressing malignancies. However, several agents have been developed for intratumoral therapy of unresectable malignant tumors, particularly recurrent glioblastoma multiforme. Glioblastoma has been targeted in this manner due to its tendency to spread locally but not distantly and lack of acceptable therapy. Human transferrin chemically conjugated to a mutated form of diphtheria toxin has resulted in many responses, with the dose limited by toxicity to normal brain (69). IL4(38 -37)-PE38, a recombinant toxin targeting interleukin-4 receptors on glioblastoma, is currently undergoing clinical testing as an intratumoral agent (70,71). The rationale for the latter agent is that normal brain does not express the receptor, and glioblastoma lines often express several thousand sites/cell. uPAR may be a useful molecule to target in this way, since normal brain does not express uPAR (72), and as shown in Table II glioblastoma lines display up to 10 6 uPAR sites/cell. Moreover, because uPAR appears to play a major role in the local invasiveness of glioblastoma cells through normal brain tissue, targeting such tumors with agents such as ATF-PE38 or ATF-PE38KDEL might result in preferential cytotoxicity to the most invasive tumor cells (73). This strategy may also be applicable for the intratumoral therapy of metastatic solid tumors, such as colon cancer, where uPAR is also overexpressed at the leading edge of invasive tumor cells (74).