Cytokeratin 18 Is Expressed on the Hepatocyte Plasma Membrane Surface and Interacts with Thrombin-Antithrombin Complexes*

During experiments to identify putative hepatic receptors for thrombin-antithrombin (TAT) complexes, a 45-kDa protein was identified by ligand blotting. Following gel purification, amino acid sequencing revealed the 45-kDa TAT-binding polypeptide to be cytokeratin 18 (CK18). The presence of CK18 on the surface of intact rat hepatoma cells was demonstrated by binding of125I-anti-CK18 antibodies. Anti-CK18 antibodies reduced the binding and internalization of 125I-TAT by rat hepatoma cells. Immunocytochemical analysis, to determine the location of CK18 in vivo, revealed a periportal gradient of CK18 staining; with hepatocytes around the portal triads demonstrating striking pericellular staining. In addition, anti-CK18 IgG associated with perfused livers to a significantly greater extent than preimmune IgG. Taken together, these data provide evidence that CK18 is found on the extracellular surface of hepatocytes and could play a role in TAT removal. Finally, these data, in conjunction with recent reports of CK8 (Hembrough, T. A., Li, L., and Gonias, S. L. (1996)J. Biol. Chem. 271, 25684–25691) and CK1 cell membrane surface expression (Schmaier, A. H. (1997) Thromb. Hemostasis 78, 101–107), indicate a novel role for these proteins as putative cellular receptors or cofactors to cellular receptors.

The generation and regulation of thrombin is of central importance in the maintenance of hemostasis (1). Antithrombin, the main physiological regulator of thrombin, is a 60-kDa glycoprotein synthesized by the liver and found in plasma at a concentration of 2-5 M (2). The importance of the regulation of thrombin by antithrombin is demonstrated by the fact that individuals with inherited antithrombin deficiency have an increased tendency to develop thrombosis (3,4) The inhibition of thrombin by antithrombin is catalyzed by heparin and related naturally occurring glycosaminoglycans. This property of heparin accounts for its widespread clinical use as an anticoagulant and antithrombotic agent (5).
Antithrombin is a member of the protein superfamily known as the serine proteinase inhibitors or serpins (6). The serpin family includes both inhibitory proteins, which inhibit their cognate proteases, and proteins with no known inhibitory function. In addition to antithrombin, the inhibitory serpins in-clude: ␣-AC, 1 ␣ 2 -antiplasmin, ␣-PI, HCII, PAI-I, PAI-II, and PN-1 (7). The inhibitory serpins are suicide inhibitors which inhibit serine proteinases through the formation of a 1:1 stoichiometric covalent serpin-enzyme complex (SEC), which renders both serpin and enzyme non-functional. Such complex formation results in a conformational change in the serpin moiety, that exposes residues believed to be responsible for SEC binding to putative hepatic receptors, and which consequently effect their removal from the circulation (8).
Plasma elimination studies in experimental animals have demonstrated that SECs are removed much more rapidly than the corresponding native or cleaved serpin (reviewed in Refs. 8 and 9). Such studies in mice, employing 125 I-labeled SECs in competition with large excesses of non-radiolabeled SECs, suggest the presence of two distinct liver receptors for their removal. These putative receptors have been designated serpin receptors 1 and 2, respectively. Serpin receptor 1 appears to be responsible for the removal of SECs composed of the serpins antithrombin, ␣-PI, ␣-AC, or HCII, while serpin receptor 2 appears to be responsible for the removal of ␣ 2 -antiplasminenzyme complexes (8). A hepatic receptor with binding characteristics similar to that of serpin receptor 1 has been identified on HepG2 cells, monocytes, and neutrophils and has been designated SECR (10 -12). A carboxyl-terminal pentapeptide of ␣-PI, putatively exposed upon complex formation and conserved throughout the serpins, was found to comprise the minimal necessary residues for binding to the SECR (13). Crosslinking of this pentapeptide to a hepatoma plasma cell membrane protein has been reported to result in the formation of an adduct of Ϸ78 kDa, although neither the cloning nor the purification of this membrane protein has yet been reported (14). More recently, the low density lipoprotein receptor-related protein (LRP) has also been demonstrated to bind a number of different SECs, including TAT (15). Furthermore, the urokinase plasminogen activator (uPA) receptor (16,17) and gp330 (18 -20) have also been reported to bind different SECs. In this study, we provide evidence that CK18 may also represent a TAT receptor, or a cofactor to such a receptor.
Cytokeratins are members of the IF family of proteins that are found primarily in epithelial tissues (21,22). IFs interact to form filamentous cytoplasmic networks that run from the surface of the nucleus to the plasma membrane (23,24). All IFs are composed of a central ␣-helical rod domain which is flanked by non-helical head and tail domains (reviewed in Refs. [25][26][27]. At least 30 different cytokeratin proteins have been identified in epithelial tissues and carcinomas, and these are grouped into two main types: type I, which are acidic; and type II, which are neutral to basic proteins. Unlike other IF proteins, cytokeratins are obligate heteropolymers, requiring the presence of both type I and type II proteins to form filaments (28). Within each type of cytokeratin there is high homology in the ␣-helical rod domains, but with notable diversity in the head and tail domains. Despite the diversity in these domains, all cytokeratins form morphologically similar filaments, implying the primary importance of the ␣-helical domains in filament formation. The sequence heterogeneity in the head and tail domains, along with the diversity of tissue expression, indicates that individual cytokeratins have specialized functions within cells in addition to the ability to form filaments (26,27).
Recently, the view that cytokeratins and other intracellular proteins are confined solely to the cytosol has been challenged. Within the last 5 years, a number of investigators have provided evidence that various intracellular proteins are expressed also on the surface of cells and function as receptors for different plasma ligands. For example, cell surface actin has been reported to bind Factor Va, angiogenin, plasminogen, tissue plasminogen activator, and lipoprotein(a) (29 -31); ␣actinin to thrombospondin (32); annexin II to plasminogen (33); and nucleolin with various lipoproteins (34). Relevant to this present report, CK8 has been reported to be a receptor for plasminogen and tissue plasminogen activator (35,36); and recently CK1 has been demonstrated to be an endothelial cell receptor for high molecular weight kininogen (37). In the present study, data are provided from biochemical, cell culture, immunocytochemical, and liver perfusion studies which support the identification of CK18 as a TAT-binding protein on the hepatocyte surface.
Protein Preparation-Bovine thrombin was purified from Thrombostat (Parke-Davis Co., Detroit, MI) by cation exchange chromatography on SP-Sephadex (Pharmacia, Uppsala, Sweden) as described (38). Purified bovine thrombin was radiolabeled using IODO-GEN-coated glass vials as outlined (39) and gave a specific activity of 2,100 -7,000 dpm/ng thrombin. Free 125 I was removed by extensive dialysis overnight at 4°C. Rabbit antithrombin was purified from rabbit plasma by heparin-Sepharose chromatography. Rabbit plasma (100 ml) was applied to the heparin-Sepharose column and the column was then washed with 0.5 M NaCl buffered with 0.05 M Tris-HCl (pH 7.4) until the OD reading was Ͻ0.03. Rabbit antithrombin was eluted in batch using 0.05 M Tris-HCl (pH 7.4) containing 2.0 M NaCl and dialyzed against 0.025 M Tris-HCl buffer (pH 7.4) containing 0.150 M NaCl and 0.02% sodium azide (TBS) overnight. TAT was formed by reacting bovine thrombin with a 3-fold molar excess of rabbit antithrombin at 37°C for 30 min, before quenching the remaining thrombin with PPACK at a final concentration of 100 M. Proteolytic digestion of blotted protein, peptide purification, and amino acid sequencing was done by Harvard Microchem (Cambridge, MA).
Cell Lines-HepG2 cells, purchased from ATCC (Rockville, MD), were grown in ␣-minimum essential medium, supplemented with 10% fetal calf serum and 100 units/ml penicillin/streptomycin. The rat hepatoma cell line, HTC, was grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 100 units/ml penicillin/streptomycin.
Purification of Rabbit Liver Plasma Membranes-Rabbit liver plasma membranes were purified as described (40) with minor modifications. Briefly, young male New Zealand White rabbits (1.5 kg each) were starved overnight, and anesthetized with sodium pentobarbital prior to euthanization by exsanguination. The livers were exposed and perfused through the portal vein with warm (22°C) STEP buffer at pH 7.4, containing 0.25 M sucrose, 10 mM Tris-HCl, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride. The perfused livers were excised and immersed in cold STEP, then homogenized using a Polytron homogenizer (Brinkmann, Switzerland) at high power for two cycles of 30 s each. Rabbit platelets were prepared as described (41). Rabbit platelet membranes were isolated by lysing the platelets in double distilled water, at ambient temperature, after which the membranes were recovered from the soluble components by centrifugation at 16,000 ϫ g for 10 min. Purified rat liver plasma membranes were a generous gift of Dr. E. Roegeczi (Department of Pathology, McMaster University).
Antibodies-Antibodies against CK18 were generated by injecting laying hens with purified bovine CK18 in complete Freund's adjuvant, followed by later injections with bovine CK18 in incomplete Freund's adjuvant. Chicken IgG was purified as described (42). The reactive IgG was purified by affinity chromatography on CK18-Sepharose by eluting bound IgG using 0.1 M sodium citrate (pH 4.0) into tubes containing 2 M Tris base, to neutralize the acid. The IgG was then dialyzed overnight against TBS and stored at 4°C. Anti-CK18 and preimmune IgG were labeled, using IODO-GEN, to a specific activity of 700 -1400 dpm/ng. Anti-CK18 monoclonal antibodies, CY-90 and KS-B17.2, were purchased from Sigma, while CAM 5.2 monoclonal anti-human cytokeratin was purchased from Becton Dickinson Canada Inc. (Mississauga, Ontario). Horseradish peroxidase-conjugated rabbit anti-chicken IgG was purchased from Jackson Labs (Mississauga, ON). Biotinylated goat anti-mouse IgG, and streptavidin-horseradish peroxidase were purchased from Dako Diagnostics Canada Inc. (Mississauga, ON).
Radioligand Binding Experiments-In a total volume of 125 l of 1% BSA in TBS, 50 g of rabbit liver plasma membranes were incubated with 25 nM 125 I-TAT, alone or in the presence of increasing amounts of non-radiolabeled TAT, or molar excesses of antithrombin, PPACKthrombin, or pentapeptide. The reaction mixtures were agitated in centrifuge tubes coated with 5% BSA in TBS, at 37°C for 40 min, after which 1 ml of cold TBS was added and the tubes centrifuged at 16,000 ϫ g for 2 min. The pellets were washed three times and the tips of the microcentrifuge tubes cut off for counting. For cell binding experiments, cells were grown on 24-well plates at a final density of 4.5 ϫ 10 5 cells/well. Binding experiments were performed in either ␣-minimal essential medium for HepG2 cells or Dulbecco's modified Eagle's medium for HTC cells, in the absence of fetal calf serum and penicillin/ streptomycin, but containing 0.5% BSA and buffered with 10 mM HEPES pH 7.4. The cells were washed in binding buffer and then incubated with the ligand (either 125 I-TAT or 125 I-IgG), with or without competitors, in binding buffer at 4°C for 2 h, after which the reaction solution was aspirated and the cells were washed three times with 10 mM HEPES pH 7.4, containing 0.15 M NaCl, 1 mM CaCl 2 , 2 mM MgCl 2 , and 0.5% BSA (HBSB). The cells and associated radioactivity were solubilized by incubation with 2 M NaOH overnight before counting for radioactivity. Competitive binding curves were obtained by analyzing the data by non-linear regression analysis using the analytical graphics program, Fig Ligand Blot Assay-Rabbit liver plasma membranes, rat liver plasma membranes, or total platelet membranes were separated by SDS-PAGE and electroblotted onto nitrocellulose (transfer conditions: 0.4 A, 4°C, 4 h). The nonspecific binding sites were blocked by incubating the blots overnight in 5% (w/v) skim milk powder in TBS at room temperature. The resulting blots were then incubated with 10 nM 125 I-TAT in 5% skim milk powder in TBS containing 2 mM MgCl 2 and 1 mM CaCl 2 at 37°C for 40 min, with gentle agitation; then washed 4 times with 300 ml of TBS buffer containing 1% skim milk powder and 0.3% Tween 20, air dried, and developed by autoradiography.
Internalization Assay-Internalization assays were done on HTC cells at Ϸ80% confluency. The cells were washed three times with prewarmed binding buffer before being incubated with 50 nM 125 I-TAT, in the absence or presence of competitors, for 5 h at 37°C. The binding media was then removed and the cells incubated with 5 mg/ml proteinase K at 4°C for 30 min, after which the supernatants were aspirated from the culture wells, placed into microcentrifuge tubes, and centrifuged at 16,000 ϫ g for 5 min at room temperature. The supernatants were aspirated off and the tips of the microcentrifuge tubes cut off. The cell pellet-associated radioactivity was then determined.
Immunoblotting-Rabbit liver plasma membranes (3.3 g/lane) and rabbit tissue homogenates (20 g/lane) were electrophoresed on reducing 10% SDS-PAGE gels and electroblotted onto nitrocellulose. The blots were probed with 1 g/ml chicken anti-CK18 IgG ( 125 I-radiolabeled or not) in 5% skim milk powder in TBS at room temperature for 1 h. They were then washed three times in 0.5% Tween 20 in TBS (TBST). The resulting blots were incubated with a 1/5000 dilution of rabbit anti-chicken IgG conjugated to alkaline phosphatase (Jackson Labs, Mississauga, ON) for 1 h at room temperature in 1% skim milk in TBS, washed three times in TBST, and developed with the substrates nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate. After color development the blots probed with 125 I-IgG were also developed by autoradiography.
Solid-phase Binding to CK18 -Purified bovine CK18 (500 ng in 0.1 M phosphate (pH 8.0)), was dried onto the wells of Immulon 4 enzymelinked immunosorbent assay plates (Dynatech Labs, Chantilly, VA) by incubation overnight at 37°C. The wells were washed two times with TBST (0.1% Tween) before blocking the nonspecific sites with 1.5% BSA in TBS at 4°C for 2 h. The wells were then washed with TBST and incubated with 125 I-preimmune or anti-CK18 IgG at 37°C for 40 min, after which the wells were washed three times with TBST. The individual wells were then separated and subjected to ␥-counting.
Immunocytochemistry-Rabbit and human livers were fixed and prepared for histological analysis by conventional methods. Some paraffin sections were also incubated with 0.1% trypsin in TBS containing 0.1% CaCl 2 at 37°C for 25 min. Identification of the primary monoclonal antibodies was done by using a streptavidin-biotin system from Dako Diagnostics Canada Inc. (Mississauga, ON). The binding of chicken IgG to sections was detected using horseradish peroxidase-conjugated rabbit anti-chicken IgG. Color was developed using aminoethylcarbazole for paraffin and cryosections. Sections and cells where then counterstained with hematoxylin.
Liver Perfusion Experiments-Antibodies were differentially radiolabeled, 131 I for anti-CK18 IgG and 125 I for preimmune IgG, using IODO-GEN as described above. The perfusion of rabbit livers and the analysis of the removal of radiolabeled IgG from the perfusate was done as previously reported (43).

Determination of TAT Binding to Hepatic Plasma Membranes-
To determine whether TAT interacted with hepatic receptors, radioligand binding experiments were performed on purified rabbit liver plasma membranes as well as hepatoma cells in culture; including the human hepatoma cell line, HepG2, and the rat hepatoma cell line, HTC. The competition curve and the logit-plot for 125 I-TAT binding to purified rabbit liver plasma membranes is shown in Fig. 1A. The apparent K d for 125 I-TAT binding to rabbit liver plasma membranes was found to be 550 nM, while that for binding to HepG2 cells and HTC cells were found to be 300 and 100 nM, respectively. The Hill coefficient was determined to be 1.01, indicating that the TAT complexes were interacting with a single set of binding sites on the rabbit liver plasma membranes.
To demonstrate specificity of 125 I-TAT binding to rabbit liver plasma membranes, competitive radioligand binding experiments were performed using an excess of unlabeled TAT, antithrombin, or PPACK-thrombin (Fig. 1B). These data support the specificity of TAT binding to the rabbit liver plasma membranes, as neither a 20-fold molar excess of unlabeled PPACKthrombin or antithrombin competed with 125 I-TAT. However, competition by a 20-fold molar excess of unlabeled TAT was clearly evident. To determine if 125 I-TAT bound to the same SECR-binding site on rabbit liver plasma membranes, as that described by Perlmutter and associates (10), the SECR binding pentapeptide (FVYLI) was used in competitive radioligand binding experiments (Fig. 1B). Unlike that seen with excess TAT, a 20-fold molar excess of the pentapeptide did not inhibit TAT binding. A 20-fold molar excess of unlabeled TAT and pentapeptide together, reduced 125 I-TAT binding to the same extent seen with excess TAT alone.
Identification of CK18 as a TAT-binding Protein-A ligandblot assay was used to identify TAT-binding protein(s) in rabbit liver plasma membranes. Probing the electroblotted rabbit liver plasma membranes with 125 I-TAT resulted in identification of a reactive band of 45 kDa (Fig. 2A). The electrophoretic mobility of the 45-kDa polypeptide was identical when electrophoresed under nonreducing or reducing conditions. The specificity of TAT binding was demonstrated by the ability of a 50-fold molar excess of unlabeled TAT to abrogate interaction of 125 I-TAT with the 45-kDa polypeptide. In contrast, a 50-fold molar excess of AT or PPACK-thrombin did not inhibit interaction with the 45-kDa polypeptide ( Fig. 2A). Rabbit liver plasma membrane solubilization experiments with different detergents indicated that the 45-kDa band was insoluble in Triton X-100. Fig. 2B shows the Coomassie Blue-stained protein profiles and the associated 125 I-TAT ligand blots of the Triton X-100 soluble and insoluble fractions of rabbit liver plasma membranes. TAT binding activity is clearly associated with the 45-kDa polypeptide which remained in the post-solubilization pellets. Utilizing this biochemical property, the Triton X-100 insoluble TAT-binding polypeptide was solubilized in SDS and gel-purified by preparative SDS-PAGE. The purity of the isolated polypeptide was verified by SDS-PAGE, and its ability to bind 125 I-TAT verified by ligand blot (Fig. 2B). The purified protein was electroblotted onto polyvinylidene difluoride membranes, then digested with trypsin, and the resultant peptides purified by high performance liquid chromatography. Two peptides of 13 and 12 amino acids each were used to obtain amino acid sequence data. A homology search performed in GenBank, using the NCBI BLASTN program, revealed that both peptide sequences had high homology to human CK18. Peptide number 1, TFQSLEIDLESMK, was 76% identical to human CK18 and peptide number 2, AQIFANSVDNAR, was 91% identical to human CK18. When semi-conservative amino acid substitutions were allowed, the homology was found to be 92 and 100%, respectively.
To exclude the possibility that CK18 was a contaminant in the gel-purified 45-kDa band from the rabbit liver plasma membranes, purified bovine CK18 was probed with 125 I-TAT in the ligand-blot assay (Fig. 3A). 125 I-TAT was found to bind to purified bovine CK18 but not to any polypeptides in rabbit platelet membranes. When purified bovine CK18 was added to the rabbit platelet membranes and probed with 125 I-TAT, the 125 I-TAT bound selectively to the CK18. In addition, 125 I-TAT was found to interact with a 49-kDa polypeptide from purified rat liver plasma membranes (Fig. 3B). This 49-kDa band likely corresponds to the rat homolog of CK18 as it was recognized with anti-CK18 IgG by Western blot analysis. This molecular mass (49 kDa) also corresponds to the reported electrophoretic mobility for the rat equivalent of CK18 (44). The presence of CK18 in both the rabbit and rat plasma membrane preparations is not unexpected as the association of CK18 with hepatocyte plasma membrane preparations has been reported previously (45).
Characterization of Anti-CK18 Antibodies versus Preimmune Antibodies-Polyclonal antibodies to bovine CK18 were generated for use in immunolocalization experiments. Following immunization of laying hens with bovine CK18, the anti-CK18 IgG from egg yolks was purified to homogeneity by affinity chromatography. Affinity-purified anti-CK18 IgG, whether labeled with 125 I or not, specifically recognized purified bovine CK18 and CK18 from rabbit liver plasma membranes and total rabbit liver homogenates in Western blots, as a 45-kDa band (Fig. 4A). Conversely, the preimmune IgG, whether radiolabeled or not, did not react with any polypeptides from rabbit liver plasma membranes in Western blot analysis. In binding experiments to purified bovine CK18 bound to enzyme-linked immunosorbent assay wells, 125 I-anti-CK18 IgG bound to the  CK18 while the 125 I-preimmune IgG did not (Fig. 4B). The specificity of the anti-CK18 IgG was further indicated by the fact that it did not bind to BSA-coated wells.
Determination of CK18 on the Surface of Hepatoma Cells-Since affinity-purified anti-CK18 IgG was highly specific for CK18 and retained the ability to bind even after being radiolabeled, 125 I-anti-CK18 IgG was used in binding experiments in cell culture experiments. To demonstrate that CK18 is located on the cell membrane surface, the binding of 125 I-anti-CK18 to HepG2 cells and HTC cells was compared with that of 125 Ipreimmune IgG (Fig. 5). 125 I-Anti-CK18 IgG bound to HTC cells to a much greater extent than preimmune IgG, and comparable results were obtained with HepG2 cells. Similar results were found when the experiment was performed using paraformaldehyde-fixed, but non-permeabilized, HTC and HepG2 cells; demonstrating the presence of cell membrane CK18 epitopes even after cell fixation. To demonstrate that the anti-CK18 IgG bound specifically to the cells, anti-CK18 IgG binding to intact HTC cells was performed in the presence of an 80-fold molar excess of preimmune IgG. This excess of preimmune IgG only partially inhibited 125 I-anti-CK18 IgG binding to HTC cells, suggesting that the anti-CK18 IgG had both specific and nonspecific binding. To substantiate the antibody specificity to CK18, the cells were fixed and then permeabilized with Triton X-100 to expose the bulk of the CK18 intermediate filaments. As expected, permeabilization increased antibody binding by approximately an order of magnitude at identical antibody concentrations. Similarly, with increasing concentrations of 125 I-TAT there was a dose-dependent increase in TAT binding to permeabilized cells over intact cells such that there was a 3-5-fold increase in TAT binding at 100 nM 125 I-TAT depending on the cell type. The binding of TAT to permeabilized cells was found to be specific such that a 100-fold molar excess of unlabeled TAT reduced binding by Ϸ60%. Such data are consistent with the hypothesis that TAT binds to CK18. Finally, immunofluorescence experiments, done on both HepG2 and HTC cells, showed punctate staining evenly distributed over the cell monolayers (data not shown). This was seen only with the anti-CK18 antibodies and not with the preimmune IgG, or in the absence of specific antibodies.
Competition of TAT Binding to Hepatoma Cells with Anti-CK18 IgG-Competitive radioligand binding studies were done in the presence of anti-CK18 IgG to demonstrate that TAT binds to CK18, or to a CK18-like protein, on the cell surface. Fig. 6 shows the binding of 125 I-TAT to HTC cells in the absence or presence of anti-CK18 IgG or preimmune IgG. With a 25-fold molar excess of anti-CK18 IgG, there was a Ϸ40% reduction in TAT binding, while the binding was not reduced in the presence of 25-fold molar excess of preimmune IgG. The extent of anti-CK18 IgG to inhibit TAT binding to fixed and permeabilized HTC cells was similar (Fig. 6).
There was a similar reduction in 125 I-TAT binding to fixed and permeabilized HTC cells in the presence of a 100-fold molar excess of unradiolabeled TAT. The relative inability of unlabeled TAT to compete with 125 I-TAT binding suggests that there is high nonspecific binding to these cells. This phenomenon was observed in a number of binding experiments with HTC cells. The ability of anti-CK18 IgG to compete to a similar degree as cold TAT suggests that CK18 possibly is the only TAT-binding site on HTC cells. However, anti-CK18 IgG did not inhibit TAT binding to HepG2 cells to the same extent as excess unlabeled TAT, indicating that other proteins may also play a role in TAT binding to these cells.
Anti-CK18 Antibodies Inhibit 125 I-TAT Internalization-To determine whether CK18 has a functional role in 125 I-TAT uptake by hepatic cells, internalization experiments were done using HTC cells with anti-CK18 Fab fragments as the competitor. The results of these experiments are shown in Fig. 7. In the absence of a competitor the amount of internalized 125 I-TAT corresponds to Ϸ17 fmol/10 5 cells. In the presence of competing anti-CK18 Fab fragments this was reduced by 41%. Equivalent amounts of preimmune Fab fragments did not in- hibit 125 I-TAT internalization. A 50-fold molar excess of cold TAT also inhibited 125 I-TAT internalization by Ϸ45%. These data are consistent with the binding data and further indicate that CK18 may be the main TAT-binding protein on HTC cells. To determine whether the internalized 125 I-TAT was degraded in lysosomes, 100 M chloroquine, an inhibitor of lysosomal degradation, was used. The presence of chloroquine resulted in the retention of Ϸ31% more internalized 125 I-TAT than that seen in control cells, indicating that internalized TAT was degraded in lysosomes. In other internalization experiments, the anti-CK monoclonal antibody CAM 5.2 was used to block the internalization of 125 I-TAT (data not shown). In these experiments CAM 5.2 reduced 125 I-TAT internalization to Ϸ40% over control cells.
Immunocytochemical Analysis of Liver Tissue-Immunocytochemical techniques were used to determine the CK18 distribution in rabbit liver sections. Both paraffin embedded and frozen sections of rabbit liver revealed an acinar gradient of staining (Fig. 8). Hepatocytes around the portal triads demonstrated striking pericellular staining (Fig. 8, A, C, and E) with a progressive reduction of staining seen moving distally into the lobule (Fig. 8B). Interestingly, there was also increased pericellular staining of hepatocytes adjacent to the central veins, although not to the same extent as that seen around the portal triads (Fig. 8D). This pattern of staining was seen consistently when using the chicken anti-CK18 IgG or two different anti-CK18 monoclonal antibodies, KS-B17.2 (bovine antigen) or CY-90 (human antigen), with both paraffin-embedded sections and frozen cryosections (Fig. 8, A and C). This staining pattern was also visualized on paraffin-embedded human liver sections using the monoclonal anti-CK antibody, CAM 5.2 (human antigen). Mild trypsinization of paraffin-embedded sections resulted in increased staining.
Binding of Anti-CK18 IgG to Hepatocytes-Experiments were also done to determine whether anti-CK18 IgG would bind to CK18 on the surface of hepatocytes in whole organ perfusion studies. A liver perfusion system was used to compare the clearance and binding of 131 I-anti-CK18 IgG and 125 Ipreimmune IgG in isolated rabbit livers. Fig. 9 shows the clearance data from six such liver perfusion experiments, expressed as a ratio of the protein-bound radioactivities in the perfusate of anti-CK18 IgG to preimmune IgG, as a function of perfusion time. In all perfusion experiments the anti-CK18 IgG cleared at a faster rate than that observed with the preimmune IgG. This trend was reflected also in the ratio of bound anti-CK18 IgG to preimmune IgG bound to the liver. In all experiments, there was 1.6-fold more anti-CK18 IgG bound compared with preimmune IgG after 3.5 h of perfusion (p ϭ 0.015). It is important to note that over the course of these experiments the livers continued to synthesize and release proteins and glucose, indicating the viability of the organs.

DISCUSSION
The present studies were undertaken to identify the putative receptor(s) responsible for removal of TAT complexes from the circulation. During the course of these studies CK18 was identified as a TAT-binding protein associated with the plasma membranes of hepatocytes. This conclusion was based on the following observations: 1) TAT-binding protein from rabbit liver plasma membranes has high sequence homology to human CK18; 2) TAT-binding protein has a similar electrophoretic mobility to CK18; and 3) the insolubility of the TATbinding protein in various detergents, a physical property of most IF proteins. The potential physiological relevance of TAT interaction with CK18 is suggested by: 1) the inhibition of 125 I-TAT binding to cultured hepatoma cells by anti-CK18 antibodies; 2) the ability of anti-CK18 antibodies to inhibit 125 I-TAT internalization; 3) immunolocalization of CK18 on the surface of hepatoma cells in culture; and 4) the specific association, in perfusion experiments, of anti-CK18 antibodies with livers.
Our findings are difficult to reconcile with an exclusive intracellular localization of CK18. Moreover, CK8 has been detected on the cell surface of primary hepatocytes, HepG2 cells, and human breast carcinoma cells (35,36,46). Additionally, CK1 also has been demonstrated recently to be present on the cell surface of endothelial cells (37). The cell surface localization for cytokeratins is also supported by surface labeling studies of cultured carcinoma cells (47). Moreover, the cell surface localization of predominantly intracellular proteins is not unprecedented, as shown by the detection of ␣-actin on the surface of calf and human endothelial cells by immunofluorescence microscopy (30,31).
The interaction of serpin-enzyme complexes with four other cell surface membrane proteins has previously been reported; these include LRP, gp330, uPA receptor, and SECR. Of these TAT specifically has been reported to interact with LRP and SECR (10,12,15). In the present study, the affinity of TAT for hepatic-binding sites appears to be low (ranging from 100 to Ͼ500 nM). However, this is not dissimilar to other studies which indicate low affinity TAT binding to SECR (10,12), as well as purified LRP (15). Specific TAT binding to monocytoid cells had a K d Ϸ 80 nM (48), while TAT binding to HepG2 cells was found to have a K d Ϸ 247 nM (49). In the present studies, the use of the SECR-binding pentapeptide did not inhibit 125 I-TAT binding to rabbit liver plasma membranes while a 20-fold molar excess of cold TAT inhibited binding by Ϸ50%. In other studies, similar molar excesses of TAT were found to reduce the SECR-binding pentapeptide, ␣-PI-elastase, and TAT binding to HepG2 cells to a similar degree (10,12).
Although SECR and LRP have both been reported to interact with TAT, there is evidence to suggest that TAT also interacts with other proteins. Thus, Takeya et al. (48) demonstrated that TAT bound to a receptor distinct from the SECR. Similar to the findings of this present study, the SECR-binding pentapeptide did not inhibit TAT, or ␣-AC-cathepsin G complex binding to cells (48,50). Indeed, alterations in the SECR-binding pentapeptide consensus sequence for HCII have been shown to have no effect on HCII-thrombin complex binding to HepG2 cells (51).
Despite its low affinity for TAT, LRP has been shown to be the vehicle for TAT internalization in vivo in rats (15). However, although LRP has been shown to be involved in the internalization of TAT and other SEC, other proteins are very likely the initial binding sites for such complexes. In addition to TAT having low apparent affinity for LRP, its binding to purified LRP appears very slow (an 18-h incubation was used in binding experiments) (15). This slow binding is not in keeping with the known very rapid removal rate for TAT in vivo. Moreover, some investigators have found no evidence for TAT binding to purified LRP (48). Recently a PN-1 peptide was found to be a potent inhibitor of PN-1-thrombin internalization by LRP, but this peptide was found not to inhibit PN-1-thrombin binding to the cell surface; supporting the possibility that TAT could bind to a receptor protein other than LRP (52). This scenario is not unprecedented as uPA-PAI and uPA-PN-1 complexes have been shown to bind to the uPA receptor before being internalized through interaction with LRP (16,17).
Is there evidence to link CK18 and LRP? CK8 and CK18 have been reported to be found in a complex with the uPA receptor and protein kinase C in the plasma membranes of WISH cells (53). Additionally, an unidentified plasma membrane glycoprotein of 85 kDa was co-immunoprecipitated from a variety of epithelial cell lines with CK18 (54). This unidentified protein could represent the ␤ subunit of LRP, as it had similar biochemical traits. Finally, although the histochemical experiments do not provide direct evidence of CK18 cell surface expression in vivo, they provide an intriguing coincidental pattern of expression with that seen for LRP, in that LRP is also preferentially expressed in hepatocytes proximal to the portal triads (55). Taken together, these observations support the hypothesis that CK18 may be the initial TAT-binding protein on hepatocytes, leading to subsequent interaction with LRP.
High sequence homology in the rod domains of intermediate filaments indicate the necessity of these regions for the formation of filaments, yet the high sequence divergence of the head and tail regions would seem to indicate the possibility of functional differences between the different cytokeratins unrelated to their cytoskeletal role. Recently, the COOH-terminal domain of CK8 has been demonstrated to be expressed on the surface of breast mammary carcinoma cells and to be necessary for plasminogen binding (36). Furthermore, CK8 has been demonstrated to augment the tissue-plasminogen activator activation of plasminogen on the cell surface and in the media of cultured breast mammary carcinoma cells (36,56). In addition, the transfection of mouse L fibroblasts with CK8 and CK18 resulted in their increased migration and invasion of matrigel (57). This novel functional role for cytokeratins is probably mediated by their acting as cell surface receptors (35,36).
At this time the true role for CK18 in TAT removal from the FIG. 9. Preferential binding to perfused livers of 131 I-anti-CK18 IgG over 125 I-preimmune IgG. Livers were isolated from anesthetized rabbits and perfused with 25 ϫ 10 6 cpm of 125 I-preimmune IgG and equivalent amounts of 131 I-anti-CK18 IgG. The IgG clearance over time from the perfusate is represented as the means Ϯ S.E. of the ratios of anti-CK18 IgG to preimmune IgG for six different livers. circulation is unclear. However, in this study we provide evidence for the biochemical interaction between TAT and CK18 as well as the presence of CK18 on the surface of cultured hepatic cells and hepatocytes in liver perfusion experiments. These data thus support the hypothesis that cytokeratins may represent cellular receptors at the hepatocyte cell surface. CK18 may thus be involved in the interaction of TAT, as well as other SECs, with hepatocytes. Such interactions are likely mediated by the variable head and tail domains of cytokeratins as has recently been demonstrated for CK8 for plasminogen binding (36) and CK1 for high molecular weight kininogen (37). Along with the growing body of evidence for surface expression of intracellular proteins, our data indicate potentially new and physiologically relevant roles for cytokeratins as putative cellular receptors, and/or as cofactors of cellular receptors.