Tissue (Type 11) Transglutaminase Covalently Incorporates Itself, Fibrinogen, or Fibronectin into High Molecular Weight Complexes on the Extracellular Surface of Isolated Hepatocytes USE OF 2-[ (2-OXOPROPYL)THIO]IMIDAZOLIUM DERIVATIVES AS CELLULAR TRANSGLUTAMINASE INACTIVATORS*

transglutaminase is shown to act as a binding site for fibrinogen or fibronectin and to covalently incorporate these glycoproteins, in addition to itself, into extracellular high molecular weight complexes. This concept is supported by the observation that a nonpeptidyl, active site-directed transglutaminase inactivator (L683685) elicited concentration-dependent (0.1-10 PM) decreases in the calcium-dependent binding and covalent cross-linking of lZ5I-fibrinogen, lZ5I-fibronectin, or [‘4C]putrescine by hepatocyte suspensions. In corroboration with these findings, an antiserum against rabbit liver transglutaminase, which did not cross-react with rabbit factor XIII, elicited concentration-depend-ent decreases in the calcium-dependent binding and covalent cross-linking of lZ5I-fibrinogen or [14C]pu-trescine by hepatocyte suspensions. Western blots

creases in the calcium-dependent binding and covalent cross-linking of lZ5I-fibrinogen, lZ5I-fibronectin, or ['4C]putrescine by hepatocyte suspensions. In corroboration with these findings, an antiserum against rabbit liver transglutaminase, which did not cross-react with rabbit factor XIII, elicited concentration-dependent decreases in the calcium-dependent binding and covalent cross-linking of lZ5I-fibrinogen or [14C]putrescine by hepatocyte suspensions. Western blots of sodium dodecyl sulfate/Triton-insoluble hepatocyte fractions conducted with this antiserum, with a polyclonal antiserum against human erythrocyte transglutaminase, or with a monoclonal antibody (CUB-7~01) against guinea pig liver transglutaminase detected the 80-kDa tissue transglutaminase, as well as tissue transglutaminase-immunoreactive bands of higher molecular mass (range of 90 to >200 kDa). The higher molecular weight species were preferentially incorporated, in a time-and calcium-dependent manner, into very high molecular weight complexes which did not enter the stacking gel. Incorporation of these tissue transglutaminase-cont~ning bands into the high molecular weight complexes was inhibited by L683685, indicating that cross-linking by the enzyme was responsible for the assembly of the complexes of which tissue transglutaminase was itself a component. Cellular integrins did not mediate ligand binding under the experimental conditions, as evidenced by the failure of the Arg-Gly-Asp-Ser tetrapeptide or anti-integrin antibodies to inhibit binding or cross-linking of "'I-fibrinogen or '251-fibronectin, in the presence or absence of t r a n s g l u t~i n a s e inactivators.
* This work was supported by National Institutes of Health Research Grant HL-20092. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Supported by a grant-in-aid from the Southeastern Pennsylvania Chapter of the American Heart Association and by an institutional research support grant from Thomas Jefferson University. To whom correspondence should be addressed. Tel.: 215-955-5272. Tr~sglutaminases mediate covalent cross-linking between proteins by forming amide bonds between the y-carboxamide groups of appropriate peptide-bound glutamine moieties and the €-amino groups of specific peptide-bound lysine residues (1). These calcium-dependent enzymes are ubiquitously distributed in mammalian tissues and body fluids (2,3). Though many molecular forms of transglutaminase have been purified from a wide variety of tissues (l), cDNA analysis has thus far revealed only three distinct transglutaminase genes. The most recently described cDNA (4,5) encodes for keratinocyte (type I) transglutaminase, a predominantly membrane-bound enzyme which shows homology to the previously sequenced tissue (type 11) transglutaminase (6)(7)(8), as well as to the initially sequenced (9,10) plasma transglutaminase (Factor XIIIa). A fourth transglutaminase, known as epidermal transglutaminase (11,12), is an immunologically distinct enzyme previously referred to as hair follicle transglutaminase (13); however, a cDNA for this transglutaminase has not been reported.
Fibrinogen and fibronectin are adhesive glycoproteins found in plasma and extracellular fluids (14). These molecules share certain structural similarities, the most germane of which are the occurrence of the Arg-Gly-Asp integrin-binding sequence (14)(15)(16) and of tissue transglutaminase-sensitive cross-linking sites (2,(17)(18)(19)(20) within their primary structures. This being the case, it is reasonable to assume that these macromolecules can interact with cell surfaces via their Arg-Gly-Asp sequences and, if transglutaminase is present, undergo covalent cross-linking reactions. In the course of studies on the interaction of these glycoproteins with hepatocytes, however, it was initially demonstrated (21) that when fibrinogen binding to hepatocyte suspensions is analyzed using a 3-h incubation at 4 "C and a ligand concentration of 30 nM (which is below the Kd for fibrinogen binding to the platelet integrin, glycoprotein IIb/IIIa (Ref. 22)), an Arg-Gly-Asp-independent, transglutaminase-mediated interaction could be isolated and studied in detail (23). The same observation proved to be true for fibronectin binding to hepatocyte suspensions (23,24). Under these experimenta~ conditions, binding of fibrinogen and fibronectin to hepatocytes is Arg-Gly-Asp-independent, largely irreversible, and accompanied by cross-linking of the fibrinogen or fibronectin by tissue transglutaminase (23).
In light of the findings described above (21)(22)(23)(24) and the observations that tissue transglutaminase specifically binds to fibrinogen (2), fibrin (25), and fibronectin (26-281, we hypothesized that an hepatocyte surface-expressed transglu-taminase, as opposed to an integrin, may function as a binding and cross-linking site for these glycoproteins. In the present study, we utilized a novel class of nonpeptidyl, active sitedirected' transglutaminase inactivators and specific antisera against tissue transglutaminase (30-32) to provide evidence in support of this hypothesis. Moreover, Western blotting of SDS'/Triton-insoluble hepatocyte extracts with these antitissue transglutaminase antisera revealed the presence of the 80-kDa tissue transglutaminase, as well as tissue transglutaminase-immunoreactive bands of higher molecular mass (range of 90 to >200 kDa). During the 3-h time course of the binding experiments, these latter forms of the enzyme were preferentially assembled into very high molecular weight complexes in a calcium-dependent fashion which was prevented by the nonpeptidyl transglutaminase inactivator. The data are consistent with a mechanism whereby tissue (type 11) transglutaminase serves as an essential component of a binding and cross-linking site for fibrinogen or fibronectin, covalently incorporating these molecules, in addition to itself, into very high molecular weight complexes on the extracellular surface of hepatocytes.
Hepatocyte Preparation-Stock suspensions of hepatocytes (5 x lo6 cells/ml as determined using a Coulter counter) were prepared by a 5-10-min perfusion (60 ml/min) with collagenase (1 mg/ml) as previously described (21). The hepatocytes were suspended in a calcium-free "binding buffer" consisting of 0.15 M NaC1, 0.025 M NaHC03 (pH 7.4) containing hirudin (0.05 units/ml), aprotinin (200 kallikrein inactivating units/ml), leupeptin (50 pg/ml), and pepstatin (0.1 pg/ml), and kept at 4 "C on a horizontal shaker operated at 30 cycles/min for no longer than 15 min prior to the initiation of binding assays. The calcium-free conditions were used to prevent transglutaminase-mediated cross-linking of endogenous (unlabeled) proteins.
Binding Assays-Binding of "'I-labeled fibrinogen or fibronectin to suspended rabbit hepatocytes was measured as previously described (21). Briefly, ligands were incubated a t 4-7 "C with isolated rabbit hepatocytes (2.5 X lo6 cells/ml) suspended in binding buffer containing 5 mM calcium chloride or 0.5 mM EDTA which were added just prior to addition of radiolabeled ligands. Cell-bound ligands were separated from free ligands by centrifugation (12,000 X g for 1 min) of cells through an ice-cold oil mixture (4 parts silicone oil to 1 part light mineral oil) and quantitated by y-scintillation counting of the amputated tube tips containing the cell pellets. Calcium-dependent binding, which averages about 60-70% of total binding (21, 23,24), was determined by subtraction of binding obtained with binding buffer containing 0.5 mM EDTA (calcium-independent binding) from that obtained with binding buffer containing 5 mM calcium chloride. The nonpeptidyl transglutaminase inactivators were dissolved in dimethyl sulfoxide a t a final concentration of 1.0 mM and added to hepatocyte suspensions 5-10 min prior to addition of "'I-Fgn, Fn, or ['4C]putrescine. The final concentration of dimethyl sulfoxide was 1% (v/v) in all incubations including the calcium and EDTA controls. Addition of this amount of dimethyl sulfoxide alone had no effect on binding or cross-linking of '"I-Fgn (not shown). r4C]Putrescine Incorporation into Cells-The incorporation of ['4C]putrescine into hepatocytes was assessed by incubating the cells with 9.1 p~ ["C]putrescine dihydrochloride (1.0 pCi) in the binding buffer containing 5 mM calcium chloride, 0.5 mM EDTA, or 5 mM calcium chloride plus various concentrations of the transglutaminase inactivator L683685. The inactivator was added 5-10 min prior to addition of ["C]putrescine. The incubations were carried out for 3 h at 4 "C, at which time 1-ml aliquots were collected, combined with 100 pl of 0.1 M EDTA, and layered on top of 250 pl of ice-cold oil mixtures in Eppendorf microcentrifuge tubes. The cells were pelleted by centrifugation a t 10,000 X g for 1 min. The binding buffer was then aspirated and the wall of the tube washed 3 times with 1-ml aliquots of distilled water. The oil layer was then aspirated from the cells, which were washed 3 times with 1-ml aliquots of ice-cold binding buffer containing 0.5 mM EDTA. Between washes, the cells were recovered by 1-min centrifugations a t 12,000 X g. The washed cells were resuspended in distilled water to a final volume of 100 pl, and 5O-pl aliquots were added to 450-pl portions of NCS tissue solubilizer. The solubilized cells were kept a t 60 "C for approximately 2 h, and 100-pl aliquots were then added to 10-ml portions of OCS liquid scintillation mixture and the radioactivities were counted in a Beckman liquid scintillation counter. Alternatively, the washed and resuspended pellets (100 p l ) were combined with 100 pl of a mixture consisting of 2% SDS, 8 M urea, and 10% P-mercaptoethanol. After the addition of 10 p1 of bromphenol blue, the samples were boiled for 2 min and then analyzed by SDS-polyacrylamide gel electrophoresis and fluorography as described below.
SDS-Polyacrylamide Gel Electrophoresis and Autoradiography-The molecular integrity of hepatocyte-bound ' T F g n , I2'II-Fn, or ["C] putrescine was examined by subjecting the solubilized cell pellets to SDS-polyacrylamide gel electrophoresis under reducing conditions, as described by Laemmli (38), using a 4% polyacrylamide stalking gel and either a 10% (for "'I-Fgn or ["CCjputrescine) or 7.5% (for I2'I-Fn) polyacrylamide resolving gel. Gels were stained with Coomassie Brilliant Blue and autoradiography (or fluorography, in the case of [I4C]putrescine incorporation) was performed using standard procedures. Scanning of autoradiograms was performed using a CAMAC TLC Scanner I1 on-line with a CAMAC SP4290 Integrator. Transmission was measured using a wavelength of 500 nm and a beam width equal to that of the entire lane. The lanes were scanned a t 0.4 mm/s and the intensity of the material at the top of the stacking gel was corrected using the intensity of the fibrinogen 13 chain (which did not undergo cross-linking) of each lane as an internal standard.
Western Blots-One-ml aliquots of hepatocyte suspension (-5 X 10" cell/ml) were centrifuged through silicone oil, the buffer and oil were aspirated, and the cells solubilized in 250 GI of 0.01 M Tris (pH 7.5) containing 0.1% SDS, 2% Triton X-100, leupeptin (50 pg/ml), pepstatin (1 pg/ml), and phenylmethylsulfonyl fluoride (0.1 mM). The solubilized cells were kept at 4 "C for 15 min and were then centrifuged at 11,000 X g for 10 min. The supernatant was removed, the pellets resolubilized as before, and recentrifuged. The final pellets were suspended in 150 pl of' distilled water and combined with 150 GI of 0.02 M Tris (pH 7.5) containing 4% SDS, 8 M urea, 10% 6mercaptoethanol, 2 mM EDTA, 2 mM N-ethylmaleimide, and 2 mM phenylmethylsulfonyl fluoride. Following the addition of bromphenol blue, the samples were boiled for 5 min and subjected to SDSpolyacrylamide gel electorophoresis using a 7.5% gel. Transfer of proteins to type HA 0.45-pm pore nitrocellulose paper (Millipore Corp., Bedford, MA) was performed with a Hoefer Transphor Model 7E50 Electrophoresis Unit (Hoefer Scientific Instruments, San Francisco, CA). The transfer buffer consisted of 25 mM THAM, 192 mM glycine, 20% (v:v) methanol (39), p H 8.4, and transfer was conducted for 7 2 h (in order to transfer the high molecular weight material on top of the stacking gel) a t constant current of 1 A. The nitrocellulose paper was washed in Tris-buffered saline (0.15 M NaC1, 0.02 M Tris-HC1, pH 7.4) for 3 h at 37 "C in order to block nonspecific binding. The primary antisera were diluted 1:300 (goat anti-rabbit liver transglutaminase, rabbit anti-human erythrocyte transglutaminase, and B.C1 monoclonal antibody (0.6 mg/ml) against human keratinocyte transglutaminase) or 1:600 (ascites fluid containing monoclonal antibody CUB-7401 against guinea pig liver transglutaminase) with Tris-buffered saline containing 1% gelatin and incubated with the nitrocellulose paper overnight a t room temperature, followed by five 15-min washes with 0.05% Tween 20 in Tris-buffered saline. The secondary antibody, peroxidase-conjugated rabbit anti-goat IgG (Cappel Laboratories, Cochranville, PA), was diluted Y l O O O with Trisbuffered saline containing 1% gelatin and incubated with the nitrocellulose paper for 3 h at room temperature. The color development reaction for visualization of transglutaminase-immunoreactive bands was performed using the 4 CN membrane peroxidase substrate system (Kirkegaard & Perry Laboratories, Gaithersburg, MD), following the directions of the manufacturer.

RESULTS
To study the potential role of transglutaminase in mediating the interaction of fibrinogen or fibronectin with suspended hepatocytes, we utilized several compounds which are members of a class of transglutaminase inactivators that have recently been shown to produce rapid and irreversible inhibition of human Factor XIIIa and, by analogy, of human erythrocyte transglutaminase.' These compounds are highly specific transglutaminase inactivators as evidenced by the observations' that ( a ) acetonylation of the active site cysteine of Factor XIIIa is the mechanism responsible for inactivation, ( b ) no other cysteine residues in the Factor XIIIa molecule are affected, (c) the inactivators are about lo5 to lo7 times less reactive against glutathione, a low molecular weight sulfhydryl-containing molecule, and ( d ) the compounds do not show detectable reactivity toward papain or several other sulfhydryl reagent-sensitive enzymes tested.
Initial screening experiments conducted with four different transglutaminase inactivators revealed that each of these low molecular weight compounds, when used at a concentration of 10 k~, markedly inhibited the calcium-dependent binding (data not shown) and the covalent cross-linking of ""I-Fgn (Fig. l a ) or "'I-Fn (Fig. l b ) by isolated hepatocytes. Since no marked selectivity of any compound was evident, L683685 was arbitrarily chosen for more extensive analysis in doseresponse experiments. Fig. 2 illustrates the concentration-dependent inhibitory effect of the transglutaminase inactivator L683685 on the calcium-dependent binding of l2 putrescine to the hepatocytes. The ICs0 for inhibition o f binding of either '"I-Fgn or ['4C]putrescine was about 0.2 pM. When the hepatocyte-bound radioligands obtained from one of the experi-ments represented in Fig. 2 were analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography (fluorography in the case of [I4C]putrescine), it was observed that L683685 also produced corresponding concentration-dependent inhibition in the covalent cross-linking of the bound T -F g n (Fig.  3) and [14C]putrescine (Fig. 4). Scanning of the autoradiogram dipicted in Fig. 3 revealed that the cross-linked "'I-Fgn which did not enter the stacking gel, when corrected for the amount of lz5I-Fgn loaded in each lane, was decreased by L683685 with an ICso of approximately 0.8 p~, somewhat higher than the value of 0.2 PM for inhibition of binding portrayed in Fig. 2.
Since fibronectin is a major component of the extracellular matrix of liver (40,41), and since we previously demonstrated that fibronectin undergoes similar transglutaminase-mediated cross-linking upon interaction with rabbit hepatocytes (23,24), we tested the ability of L683685 to inhibit binding and covalent cross-linking of fibronectin by the hepatocyte suspensions. As portrayed in Fig. 5 , L683685 elicited concentration-dependent decreases in the binding of "'I-Fn to the hepatocytes. The IC,, of 0.4 p~ for inhibition of '"I-Fn binding (Fig. 5) was comparable with the value of 0.2 ~L M for the inhibition of l2'1I-Fgn or ['4C]putrescine binding (Fig. 2). Moreover, when the cell pellets from one of the experiments portrayed in Fig. 5 were analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography, the transglutaminase inactivator was observed to produce corresponding decreases in the covalent cross-linking of the cell-bound '"I-Fn (Fig.  6), with total inhibition of cross-linking occurring at concentrations > 5 ~L M as evidenced by scanning of the autoradiogram. These data support the concept that fibronectin and fibrinogen interact with hepatocytes via a very similar, if not identical, mechanism involving a transglutaminase as a crucial component of a binding locus on the extracellular surface of the cell.
Although the nonpeptidyl transglutaminase inactivators used in the present study are very specific and potent as transglutaminase inactivators,' their potential effects on other cellular proteins have not yet been studied. For this reason, we also utilized another type of inhibitor, namely a goat antiserum against rabbit liver tissue transglutaminase (30). In corroboration with the results obtained with L683685, we observed that the antiserum, which recognized purified rabbit liver transglutaminase but not rabbit plasma Factor XI11 in Western blots (Fig. 7, inset), also elicited concentration-dependent inhibition in the calcium-dependent binding of '"I-Fgn or ['4C]putrescine to the hepatocytes, while nonimmune goat serum did not (Fig. 7). Moreover, like L683685, the antiserum also decreased '"'I-Fgn cross-linking at the hepatocyte surface (Fig. 8a), while nonimmune goat serum actually increased cross-linking of the bound "'I-Fgn (Fig.  8b). When the lanes in Fig. 8a were scanned with a TLC scanner and corrected for the amount of material loaded in each lane using the p chain as an internal standard, the percent inhibition of "'I-Fgn cross-linking elicited by the antiserum (not shown) was found to be nearly superimposable with the inhibition curves for binding of l2'1-Fgn or ["CC] putrescine shown in Fig. 7. For example, at an antiserum concentration of 100 pl/ml, inhibition of cross-linking was 80% as was the inhibition in the calcium-dependent binding of either '?'II-Fgn or [14C]putrescine. The anti-transglutaminase antiserum not only inhibited the binding and cross-linking of "'I-Fgn by hepatocytes, it also inhibited the enzymatic activity of tissue transglutaminase, purified from rabbit liver, as evidenced by a concentration-dependent decrease of ["C] putrescine incorporation into dimethylcasein (Table I). Taken

FIG. 2. Effect of L683685 on binding of '2sII-Fgn or ['"
Cc] putrescine to rabbit hepatocytes. "'I-Fgn (30 nM) or ["C]putrescine (9.1 p~) were incubated a t 4 "C with rabbit hepatocytes (2.5 X 10" cells/ml) suspended in buffer containing 5 mM CaCI2 and increasing concentrations of L683685 which had been added 5-10 min prior t o addition of the radiolabeled ligands. After 3 h of incubation, calcium-dependent binding of ""I-Fgn or [''C]putrescine was quantitated as described under "Experimental Procedures." Each point represents the mean of four determinations obtained from two experiments.
together with the results obtained with L683685, these data support the idea that tissue transglutaminase, expressed on the extracellular surface of the cell, played an essential role in the calcium-dependent interaction of fibrinogen and fibronectin with hepatocytes.
To determine whether the transglutaminase molecule which mediated binding was itself covalently incorporated into the high molecular weight cross-linked complex, immunoblots of the hepatocyte pellets were performed using the polyclonal antiserum against rabbit liver transglutaminase (30), a polyclonal antiserum against human erythrocyte transglutaminase (31), or a monoclonal antibody (CUB-7401) against guinea pig liver transglutaminase (32). As illustrated in Fig. 9, immunoreactive material was minimally detectable at the top of the stacking or resolving gels after a 5-min incubation (lanes I ), but was clearly demonstrable after 3 h of further incubation of hepatocytes in the absence of L683685 F I~. 5. Effect of L683685 on binding of '''I-Fn to rabbit hepatocytes. ""I-Fn (50 nM) was incubated a t 4 "C with rabbit hepatocytes (2.5 X lO"cells/ml) suspended in buffer containing 5 mM CaCI,! and increasing concentrations of L683685 which had been added 5-10 min prior to addition of the ":'I-Fn. After 3 h of incubation, calcium-dependent binding of "'I-Fn was quantitated as described in "Experimental Procedures." Each point represents the mean of four determinations obtained from two experiments. L683685 (lanes 3 ) . These findings indicate that tissue transglutaminase itself was incorporated into these high molecular weight complexes, in a time-dependent fashion, via a mechanism which depended upon expression of transglutaminase cross-linking activity. Cells incubated with EDTA also lacked transglutaminase immunoreactive material at the top of the gels (not shown),

(lanes 2), but not in the presence of
indicating that the assembly of the transglutaminase-containing high molecular weight complexes required calcium, consistent with the known calcium-dependency of transglutaminase-mediated cross- linking events (1,2). The L683685-treated ( Fig. 9) and EDTA-treated (not shown) hepatocytes demonstrated immunoreactive bands a t 80 kDa, but also bands with molecular masses of 90 to >200 kDa (TG,) as seen in Fig. 9   (lunes 1 and 3 ) . The temporally coordinated disappearance of "'I-Fgn (30 nM) or ["C]putrescine (9.1 pM) were incubated at 4 "c with rabbit hepatocytes (2.5 X lo6 cells/ml) suspended in buffer containing 5 mM CaCI, and increasing concentrations of goat antiserum against rabbit liver transglutaminase or nonimmune goat serum which had been added 30 min prior to addition of the radiolabeled ligands. After 3 h of incubation, calcium-dependent binding of lY'I-Fgn or ["Clputrescine were quantitated as described under "Experimental Procedures." Each point represents the mean of four determinations obtained from two experiments. Inset, Western blot of 10 pg of either purified rabbit plasma Factor XI11 (lane I ) or rabbit liver transglutaminase (lane 2) using the goat antiserum against rabbit liver transglutaminase (30). these higher molecular weight forms of tissue transglutaminase antigen from the control hepatocytes, along with the concomitant appearance of very high molecular weight complexes which did not enter the gel (compare lanes 1 with lanes 2 ) , indicate that the higher molecular mass forms of tissue transglutaminase, as opposed to the 80-kDa transglutaminase   (42), was also undetectable in hepatocytes by Western blotting (not shown) with monoclonal antibody B.C1 (33). We initially demonstrated (21) that when fibrinogen binding to hepatocyte suspensions is analyzed using a 3-h incubation a t 4 "C and a ligand concentration of 30 nM, which is below the K d for fibrinogen binding to the platelet integrin glycoprotein IIb/IIIa (22), that an Arg-Gly-Asp-independent, transglutaminase-mediated interaction could be isolated and studied in detail (23). The same observation proved to be true for fibronectin binding to hepatocyte suspensions (23,24). However, since integrins play such a prominent role in mediating the interaction of various cytoadhesive glycoproteins with cells (43), and because it is possible that the rapid transglutaminase-mediated cross-linking of the glycoproteins precluded binding to an integrin, we determined whether the binding of l2'1-Fgn or "'I-Fn to hepatocytes treated with transglutaminase inactivators could be mediated by cellular integrins. As seen in Table 11, neither the Arg-Gly-Asp-Ser FIG. 9. Western blots of tissue transglutaminase in covalently cross-linked, high molecular weight protein complexes. ""I-Fgn (30 nM) or l'JI-Fn (50 nM) were incubated at 4 "C with rabbit hepatocytes (2.5 X loG cells/ml) suspended in buffer containing 5 mM CaClp with or without L683685 (10 p~) which had been added 5-10 min prior to addition of the radioligands. After either 5 min or 3 h of incubation, 1-ml aliquots of cell suspension were collected, as described under "Experimental Procedures," and the cell-bound radioligands extracted with 0.1% SDS, 2% Triton X-100. The precipitates were analyzed by SDS-polyacrylamide gel electrophoresis followed by Western blotting with goat antirabbit liver transglutaminase (panel A ) or rabbit antihuman erythrocyte transglutaminase (panel B ) . Lanes I , control cells after 5 min of incubation; lanes 2, control cells after 3 h of incubation; and lanes 3, L683685-treated cells after 3 h of incubation. TG, represents a tissue transglutaminase-immunoreactive major band which migrated with a molecular mass > 200 kDa. TG represents native liver transglutaminase.

TABLE I1
Effect of RGDS and anti-integrin antibodies on binding of lZ5I-F'gn or "'I-Fn to L68368.5-treated rabbit hepatocytes "'1-Fgn (30 nM) or "'I-Fn (50 nM) were incubated a t 4 "c with rabbit hepatocytes (2.5 X 10" cells/ml) suspended in buffer containing 5 mM CaCL (control), 0.5 mM EDTA, or 5 mM CaCI? plus the listed agents which had been added prior to the '"I-Fgn or ""I-Fn. After 3 h of incubation, ligand binding was quantitated as described under "ExDerimental Procedures." FnR. fibronectin receptor.
Ligand binding (76 of control")  (cu,p3), or a fibronectin receptor (ad%) monoclonal antibody (FnR-Ab) were able to inhibit either total binding (control binding in buffer containing 5 mM calcium chloride) or the L683685-resistant component of the calcium-dependent binding of '"'I-Fgn or '"I-Fn. Moreover, binding of '"I-Fgn or '"'I-Fn to hepatocytes in the presence of EDTA was also resistant to inhibition by these antiintegrins and by L683685 (data not shown), indicating that the inhibitory effect of L683685 on binding was on the cal-cium-dependent component, and that integrins did not participate in the binding process as analyzed using the conditions described above. That functional integrins are present in rabbit hepatocytes was evident by the observations that the Arg-Gly-Asp-Ser tetrapeptide, as well as both antibodies, impaired attachment and spreading of hepatocytes to plastic dishes coated with fibronectin or fibrinogen (not shown).

DISCUSSION
In the present study, we demonstrate that rabbit hepatocyte surface-expressed tissue transglutaminase can serve as a crucial component of a binding site for exogenously added fibrinogen or fibronectin, covalently incorporating these glycoproteins, in addition to itself, into high molecular weight complexes on the outside of the cell. That the transglutaminase activity is not Factor XIIIa-related is based on the demonstration that the goat antiserum against rabbit liver transglutaminase (30) used in these studies did not cross-react, in Western blots, with purified rabbit Factor XIII, and by our prior observations (23) that, characteristic of the action of tissue transglutaminase (2,18,44), the Aa chain of fibrinogen is cross-linked much more rapidly and extensively than is the y chain, and no yy-dimers are formed. The participation of keratinocyte (type I) transglutaminase in these phenomena was ruled out based on the failure of the monoclonal antibody B.C1 (33), which recognizes type I transglutaminase in Western blots of epidermal keratinocytes (13, 45), to detect any type I transglutaminase in Western blots of hepatocyte proteins. Moreover, immunohistochemical analysis conducted with this antibody has revealed the absence of type I transglutaminase in rat liver sections (42).
The superimposability of the binding inhibition curves for " putrescine obtained with the low molecular weight active site-directed1 transglutaminase inactivator L683685 (Fig. 2) is suggestive of a mechanism whereby alteration of the active site cysteine residue of the enzyme was responsible for the decreases in the binding of lZ51-Fgn or lZ51-Fn. The important role of the active site in the binding process was corroborated by the demonstration that the anti-transglutaminase antiserum which inhibited lZ51-Fgn binding and cross-linking also produced corresponding decreases in hepatocyte-expressed transglutaminase activity as evidenced by concentration-dependent inhibition of [14C]putrescine incorporation into the hepatocytes. These data suggest that the antiserum probably recognized epitopes located near the active site of the hepatocyte surface-expressed transglutaminase. This conclusion is supported by the previous demonstration that the antiserum inhibited the catalytic activity of purified rabbit liver transglutaminase measured by quantitating putrescine incorporation into casein (30), an observation which was reproduced in the present investigation (Table I).
We have also previously shown (23) that, depending on the divalent cation employed in the assays, the extent of binding and cross-linking of fibrinogen or fibronectin to hepatocytes is directly proportional to the level of enzymatic activity, with calcium supporting maximum levels of enzyme activity which correlate with the degree of ligand binding and cross-linking. These data (23), along with those presented in the present report, indicate that accessibility of the ligands to the active site of the transglutaminase molecule is an important step for the binding of these glycoproteins to the hepatocyte surface.
Although accessibility of the active site of the surfaceexpressed transglutaminase is important for the binding of fibrinogen or fibronectin to hepatocytes, it is unclear whether the active site cysteine residue itself directly participated in the binding process or whether conformational changes in-duced in the enzyme as a result of inactivation precluded binding of IZ5I-Fgn or lZ51-Fn to some other domain of the transglutaminase molecule. In this regard, it is important to note that the interaction of fibrinogen with hepatocyte surface-expressed transglutaminase described in the present investigation differs from the binding of fibrinogen (2), fibrin (25), or fibronectin (27, 28) to tissue transglutaminase in purified systems. In these systems, the binding process has been shown to be calcium-independent and therefore not related to the ability of transglutaminase to express catalytic activity. Moreover, a region of the enzyme distinct from the active site appears to mediate this mode of interaction, since tissue transglutaminase retains catalytic activity following binding to fibronectin, and the transglutaminase and fibronectin molecules can be completely dissociated from each other by SDS-polyacrylamide gel electrophoresis (27,28).
Multiple molecular weight forms of transglutaminase have been detected in liver (46, 47), but the biochemical identification of these higher molecular weight forms has not been resolved. In the present study, the 80-kDa form of tissue (type 11) transglutaminase, along with tissue transglutaminase-related antigens with molecular masses of 90 to >200 kDa, were detected in Western blots of SDS/Triton-insoluble hepatocyte proteins using antisera against rabbit liver transglutaminase (30), human erythrocyte transglutaminase (31), and guinea pig liver transglutaminase (32). Each of these antisera also immunoblotted tissue transglutaminase present in very high molecular weight cross-linked complexes isolated from cells incubated with fibrinogen or fibronectin in the absence, but not in the presence, of the transglutaminase inactivator L683685 or EDTA, indicating that the enzyme itself was also covalently incorporated, in a calcium-dependent manner, into high molecular weight complexes on the extracellular surface of the hepatocytes. These findings suggest the possibility that the 80-kDa tissue transglutaminase molecule may be crosslinked to other hepatocyte proteins, accounting for the 90-to >200-kDa forms, and that these tissue transglutaminasecontaining structures mediated the binding and/or crosslinking of lZ5I-Fgn or "'I-Fn to form very high molecular weight complexes. It is interesting to note that the particulate transglutaminase of rat liver has recently been demonstrated to cross-react, in Western blots, with monoclonal and polyclonal antibodies against the soluble form of tissue transglutaminase (48), suggesting that, under certain conditions, the 80-kDa cytosolic tissue (type 11) transglutaminase of liver can be associated with cellular membranes, thus accounting for the transglutaminase activity associated with the particulate fraction of liver which has been observed by many investigators (46-53).
The detection of 90->200-kDa tissue transglutaminaseimmunoreactive bands in Western blots of the SDS/Tritoninsoluble fraction of hepatocytes, and the covalent incorporation of these materials into very high molecular weight complexes which did not enter the stacking gel, is a novel finding of the present investigation. However, such a result should not have been entirely unexpected, in light of the earlier report demonstrating that purified guinea pig liver transglutaminase can both incorporate [I4C]putrescine into itself and autocatalytically cross-link itself into high molecular weight polymers (54). Furthermore, evidence has been presented (49) which suggests that the transglutaminase present in liver cytosolic and particulate fractions may serve as both the enzyme and acceptor molecule. Further evidence that transglutaminase may be capable of cross-linking substrates to itself was provided by the observation that purified guinea pig liver transglutaminase and purified &-microglob-ulin, a 12-kDa protein found on the surface of a variety of mammalian cells, comigrate to the top of the gel when incubated together in the presence of calcium (55). Similarly, murine leukemic macrophage cell lines, following differentiation, express transglutaminase activity present in a high molecular weight complex which elutes in the void volume following Sepharose 4B chromatography (56). Whether this high molecular weight complex contained covalently incorporated transglutaminase, as shown in the present study using hepatocytes, or whether the transglutaminase was noncovalently associated with cellular proteins, such as fibronectin (27, 28), was not determined (56). Apart from or in addition to the cross-linking of tissue transglutaminase to hepatocyte proteins, it is possible that some of the higher molecular weight forms of tissue transglutaminase observed in the present investigation may have been polymers of the enzyme formed by autocatalytic cross-linking, as previously reported to occur with the purified enzyme (49, 54, 55). Additional work will be required to address this issue and to determine the functional significance of the assembly of surface-expressed tissue transglutaminase into high molecular weight complexes on the extracellular surface of the hepatocyte.
Finally, the involvement of integrins in the binding process is highly unlikely, since neither the calcium-dependent nor the calcium-independent components of fibrinogen and fibronectin binding could be inhibited by the Arg-Gly-Asp-Ser tetrapeptide or by monoclonal antibodies reactive with the vitronectin or fibronectin receptors. These observations were true even when binding was analyzed under conditions where cross-linking did not occur ( i e . in the presence of L683685). Thus, by using appropriate experimental conditions (21), we have isolated and characterized (21, 23, 24) an Arg-Gly-Aspindependent transglutaminase-mediated pathway for the interaction of fibrinogen or fibronectin with hepatocytes. We have also shown this system to be operative for fibrinogen binding by human umbilical vein endothelial cells (57), while others have demonstrated a strikingly similar, Arg-Gly-Aspindependent interaction of fibrinogen with a murine melanoma cell line (58). Both with the endothelial cells (57) and the melanoma cells (58), the transglutaminase-mediated covalent cross-linking of the cell-bound fibrinogen suggest the possibility that surface-expressed transglutaminase may also serve as a binding site for the glycoproteins in these cells, as presently shown for hepatocytes. The low level of the calciumdependent components of fibrinogen (15%) and fibronectin (25%) binding which was not inhibited by the transglutaminase inactivators or the anti-integrins may have involved cellular proteins distinct from tissue transglutaminase or integrins, and this possibility is presently under investigation.
The functional significance of tissue transglutaminase-mediated cross-linking of fibrinogen, fibronectin, or itself at the hepatocyte surface is presently unknown. However, based on the accumulated findings by many investigators, a role for the extracellular expression of the enzyme in mediating hepatocyte-hepatocyte interactions (59-63) or the covalent restructuring of extracellular matrix proteins (19, 20, 26, 64) are possibilities. The extracellular expression of tissue transglutaminase activity by hepatocytes (21,23,24,65) or endothelial cells (57) may possibly play a role in the formation of the highly insoluble matrix found in cirrhotic liver (41). Moreover, fibronectin (40, 66) and tissue transglutaminase (66) colocalize to the space of Disse, and cirrhosis increases fibronectin deposition at this location (41). Hypothetically, transglutaminase may mediate the covalent cross-linking of fibronectin, forming a scaffold upon which other extracellular components may deposit leading to the collagenization of the space of Disse. The recently developed' transglutaminase inactivators used in the present investigation have previously been shown to inhibit fibrin cross-linking by Factor XIIIa (29, 67)'; however, the potential for modulation of tissue tranglutaminasemediated physiologic or pathologic processes awaits further experimentation.