Culture Shock SELECTIVE UPTAKE AND RAPID RELEASE OF A NOVEL SERUM PROTEIN BY ENDOTHELIAL CELLS IN VITRO*

A novel protein has been purified from fetal calf serum and from serum-free bovine aortic endothelial cell conditioned culture medium. This protein consists of a single polypeptide chain of reduced M, 70,000 (70K protein) and was separated from bovine serum albumin and other proteins by ion-exchange chromatography and immunoabsorption on Sepharose-cou- pled anti-70K protein antiserum. The 70K protein was shown to be structurally and immunologically distinct from bovine serum albumin, a-fetoprotein, and vitronectin by one- and two-dimensional peptide mapping, amino acid analysis, and enzyme-linked immunosorbent assay and/or immunoblotting. The 70K protein was located in endothelial cell cy- toplasmic granules of irregular size and distribution. Metabolic radiolabeling studies showed that the 70K protein was not a biosynthetic product of these cells; its cytoplasmic location was due to a selective uptake from the fetal calf serum in which the cells were initially grown. After subconfluent cultures of endothelial cells were shifted to serum-free medium, nearly 80% of the total 70K protein that was measurable in the medium was released between 0 and 20 min. Moreover, sparse, rapidly proliferating cells released approxi- mately 18-fold more 70K protein within 2 min as compared to dense, nonproliferating cultures. The concentration of 70K protein in fetal calf serum was estimated to be 400-600 Kg/ml. Proliferating bovine aortic endothelial cells, 24 h after plating at an intermediate density, released approximately 250 pg of 70K protein/cell within

A novel protein has been purified from fetal calf serum and from serum-free bovine aortic endothelial cell conditioned culture medium. This protein consists of a single polypeptide chain of reduced M, 70,000 (70K protein) and was separated from bovine serum albumin and other proteins by ion-exchange chromatography and immunoabsorption on Sepharose-coupled anti-70K protein antiserum. The 70K protein was shown to be structurally and immunologically distinct from bovine serum albumin, a-fetoprotein, and vitronectin by one-and two-dimensional peptide mapping, amino acid analysis, and enzyme-linked immunosorbent assay and/or immunoblotting.
The 70K protein was located in endothelial cell cytoplasmic granules of irregular size and distribution. Metabolic radiolabeling studies showed that the 70K protein was not a biosynthetic product of these cells; its cytoplasmic location was due to a selective uptake from the fetal calf serum in which the cells were initially grown. After subconfluent cultures of endothelial cells were shifted to serum-free medium, nearly 80% of the total 70K protein that was measurable in the medium was released between 0 and 20 min. Moreover, sparse, rapidly proliferating cells released approximately 18-fold more 70K protein within 2 min as compared to dense, nonproliferating cultures.
The concentration of 70K protein in fetal calf serum was estimated to be 400-600 Kg/ml. Proliferating bovine aortic endothelial cells, 24 h after plating at an intermediate density, released approximately 250 pg of 70K protein/cell within the first 20 min after exposure to serum-free conditions. The data provide evidence for a novel protein in serum which is selectively internalized by endothelial cells in vitro and which in turn is released rapidly under conditions such as osmotic imbalance due to serum removal, or during periods of cellular proliferation, conditions which we term "culture shock." The establishment of eucaryotic cells in tissue culture has permitted direct examination of selected, differentiated properties as manifested by homogeneous cellular populations. In particular, studies on vascular endothelial cells have been facilitated by the use of somewhat specialized conditions * This work was supported in part by National Institutes of Health Grants GM 29853 and HL 18645, and by a grant from R. J. Reynolds, Inc. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
f Established Investigator of the American Heart Association, with funding contributed in part by the American Heart Association, Washington Affiliate. designed to maintain the several morphologically and biosynthetically unique phenotypes during i n vitro propagation (1, 2). Recent studies have confirmed, however, that certain metabolic properties characteristic of the endothelium become significantly altered as a result of subcultivation (3).
Fetal calf serum contains a number of factors that have been directly implicated in cellular phenotypic modulation. The use of serum-free culture conditions as described by Barnes and Sat0 (4) has demonstrated for some cell types that the presence of certain serum components causes significant qualitative and quantitative changes in protein biosynthesis (5). Cellular proliferation, a requisite property for subcultivation, is ensured by the presence of mitogens and growth factors in serum. Since in vivo the uninjured endothelium generally exhibits an extremely low replication rate (6), excessive proliferation in vitro, in the presence of platelet mitogens, could mimic a form of cellular injury (see Refs. 7 and 8 for discussion on this point).
Endothelial cells, both in vivo and in vitro, undergo fluid endocytosis. Quantitative measurement of the endocytotic rate in bovine aortic endothelial (BAEl) cells in vitro showed significant increases in growing as compared to quiescent cultures and in experimentally wounded cells that were migrating and proliferating in response to this injury (9). The effects of other types of endothelial injury, e.g. cellular exposure to free fatty acids (lo), calcium ionophore (11,12), and oxygen radicals (12) have been measured by the net transfer of albumin across endothelial cell monolayers. An increase in cellular permeability of this serum protein was associated with the injurious agents and was in turn accompanied by overt changes in cell shape and actin filaments. The rate of albumin transfer was subsequently reduced when the cells were re-exposed to tissue culture medium containing fetal calf serum (FCS).
Cells which have been isolated from a tissue matrix and propagated in vitro are the victims of "culture shock." Various factors can be considered as contributors to cellular injury in vitro: sparse plating density, cellular migration and proliferation, exposure to serum or other wound-associated factors, incubation with radioisotope, imposition of a plastic or other alien substratum, and removal of nutritional/survival components required by the cells. During a study of stress-related protein secretion (57), we observed a protein of M, 70 from the FCS in which the cells were initially cultured and represented a unique and novel protein that was structurally different from bovine serum albumin (BSA), a-fetoprotein (aFP), and vitronectin. This 70K protein was released rapidly from BAE cells under conditions of serum deprivation or proliferation. We propose that the selective uptake and rapid release of this novel 70K protein from BAE cells in vitro is indicative of culture shock and concomitant cellular injury.

MATERIALS AND METHODS
Cell Culture-Endothelial cells (provided by Dr. S. Schwartz, University of Washington, Seattle, WA) were obtained from adult bovine aorta by scraping the intimal layer gently with a scalpel, as described by Gajdusek and Schwartz (13). BAE cells were alternatively prepared by collagenase treatment of isolated vessels (14) and were provided by Dr. J. Harlan (University of Washington, Seattle, WA). The cells were grown in either Waymouth's or Dulbecco's modified Eagle's medium (Gibco Laboratories, Grand Island, NY) containing antibiotics and 16%, by volume, fetal calf serum (lot 100424, Hyclone Laboratories, Sterile Systems, Inc., Logan, UT) and were subcultured at a 1:3 split ratio in equal volumes of trypsin (0.2% solution, Gibco Laboratories) and EDTA (0.02% solution, pH 7.5, MCB, Cincinnati, OH). Several different strains of cells, ranging in passage number from 6-12, were used during the course of the experiments. All cultures reached confluence within 3 days after plating and, by morphologic criteria, were homogeneous and not senescent.
Chromatographic Procedures-7OK protein was purified from both FCS (lot 100424, Hyclone) and serum-free BAE cell culture medium. Chromatography on Sephadex G-200 (Pharmacia), DEAE-cellulose (DE52, Whatman), and heparin-Sepharose CL-GB (Pharmacia) was performed as previously described (15). DEAE-Affi-Gel@' Blue (Bio-Rad) (8 ml) was equilibrated at 4 "C in 0.02 M KzHPOl buffer, pH 8.0. In a typical experiment, 1 ml of FCS was dialyzed against the equilibration buffer and applied to the column; following a 40-ml wash to remove unbound material, bound protein was eluted with equilibration buffer containing 1.4 M NaCl, at a flow rate of 0.25 ml/ min. Column effluents were monitored by absorbance at 230 nm.
Antiserum 1256, which was raised initially against a BAE cell glycoprotein of Iw, 43,000 (43K protein) (15), but which also contained antibodies reactive with the 70K protein, was dialyzed against 0.1 M NaC03, pH 8.3 buffer at 4 "C. 2 ml of antiserum were subsequently mixed with 1.5 g of CNBr-activated Sepharose CL-4B in 1 mM HC1, according to the manufacturer's instructions (Pharmacia). Culture medium from BAE cells, incubated in the absence of FCS for 30 h, was dialyzed against PBS at 4 "C. 2 ml of this sample was clarified by centrifugation and applied to the antiserum-Sepharose column, equilibrated in PBS at 4 "C. Bound buffer was eluted with a 0.2 M glycine buffer, pH 2.2. Appropriate fractions from all the chromatographic procedures were pooled, dialyzed against 0.1 N acetic acid, and lyophilized. Other Procedures-One-dimensional peptide mapping was performed on 70K protein, BSA (fraction V, Pentex Laboratories, Elkhart, IN), and bovine olFP (a gift of Dr. E. Ruoslahti, La Jolla Cancer Research Foundation, La Jolla, CA), according to methods described by Sage et al. (15). Trypsin (Worthington) and proteinase K (EM Biochemicals, Darmstadt, Federal Republic of Germany) were diluted from stock solutions of 1 mg/ml in Tris-saline and used at enzyme to substrate ratios as detailed in the legend to Fig. 4. Two-dimensional peptide mapping was according to the technique of Elder et al. (16), with modifications as described by Sage et al. (17). SDS-PAGE was performed according to the method of Laemmli (18) in both the presence and absence of 0.5 M urea. Quantitation of protein after SDS-PAGE was performed by scanning densitometry on a Quick-Scan (Helena Laboratories, Beaumont, TX). BSA, 70K protein, and FCS were also analyzed by PAGE under different conditions: (a) standard Laemmli (18) gels in the absence of SDS and added urea, ( b ) acid-urea gels (19), and (c) a native gel system as described by Ornstein (20) and Davis (21).
Amino acid analyses were carried out by AAA Laboratories (Mercer Island, WA). Protein samples were hydrolyzed under vacuum in 1 ml of constant boiling HC1 in the presence of phenol at 110 "C for 24 h. Cysteine was determined as cysteic acid after performic acid oxidation (22).
Immunochemical Procedures-An antiserum, developed initially against 43K protein as described by Sage et al. (15), exhibited enhanced reactivity toward 70K protein with each successive bleeding. Antibodies were therefore purified from later bleedings by ammonium sulfate precipitation (20% weight to volume ratio) and subsequently by successive affinity absorption chromatography on columns of Sepharose CL-4B to which were coupled 43K protein (purified from BAE cell culture medium), thrombospondin (a gift from Dr. G. Raugi, University of Washington, Seattle, WA), and BSA (Pentex).
Antibodies were tested for specificity by direct ELISA as previously described (15). BSA, purchased from both Pentex and Sigma, and rabbit anti-bovine BSA antisera (RHRP 16202, Cappel Laboratories, Cochranville, PA) (used at a dilution of lo3), and FCS (Hyclone) were tested in the assays, as well as 70K protein purified from both FCS and BAE cell culture medium. Rabbit antibodies to human von Willebrand protein were purchased from Calbiochem-Behring, and rabbit antibodies to bovine thrombospondin (affinity-purified) were a gift from Dr. P. Bornstein (University of Washington, Seattle, WA).
Antisera were also affinity-absorbed in solution with concentrations of protein in PBS ranging from 50-500 pg/ml.
Immunofluorescence studies on the localization of 70K protein in BAE cells were performed as previously described (15,23), except that the first 30-min incubation in Hanks' solution was substituted with a quick rinse.
Qualitative identification of aFP, BSA, and 70K protein in FCS (100424 from Hyclone, and 29101021 from Flow Laboratories, MacLean, VA) was done by immunoblotting according to the method of Towbin et al. (24) with minor modifications. The blots were exposed in parallel to rabbit antisera against BSA, against bovine 0rFP (a gift from Dr. E. Ruoslahti), and against 70K protein (dilutions from 1:lO to 1:lOO were used). Immune complexes were visualized after addition of 1251-staphylococcal protein A (8 Ci/pg, 95 pCi/ml, New England Nuclear) followed by fluorescence autoradiography.
Cellular Release Studies-BAE cells were plated at different densities, typically onto 6-well, 30-mm Costar@' plates (Costar, Cambridge, MA). To examine the amounts of 70K protein released into the culture medium, growth medium containing 16% FCS was removed, the cells were washed once or twice with PBS, EDTA, or serum-free medium, and 0.5-1 ml of fresh DMEM or Waymouth's medium was added to the cultures. All experiments were performed at room temperature, and in some instances the media were supplemented with sodium ascorbate (50 pglml). Specific details have been included in the figure legends. Cultures of BAE cells were photographed prior to inception of the experiments, and cell counts were determined by hemocytometer.
Trypsin "shave" experiments were conducted as follows: confluent cell layers (30-mm dishes) were washed once with EDTA and incubated for 5 min in 0.2 ml of trypsin solution. At this point, the cells were not rounded hut had undergone cytoplasmic retraction and no longer appeared confluent. From this solution were taken two aliquots: (a) 25 pl were mixed with an equal volume of sample buffer and applied directly to an SDS-polyacrylamide gel, and (b) the remaining solution was made 10% with respect to trichloroacetic acid at 4 "C, and precipitated proteins were subsequently analyzed by SDS-PAGE. A control incubation was performed with 5 pl of Waymouth's medium containing 16% FCS (equivalent to 21 pg of BSA) and 5 p1 each of EDTA and trypsin solution (equivalent to 10 pg of trypsin) at room temperature for 5 min and 30 min, and the digests were analyzed by SDS-PAGE.

RESULTS
In earlier studies we had described the biosynthesis of a secreted glycoprotein, of M, 43,000, by BAE and several other mesenchymal cells in vitro (15). This protein exhibited an extremely high affinity for BSA which was dissociable only in the presence of SDS. Nonaffnity-purified antiserum, generated originally to a 43K/BSA/70K complex, was coupled to Sepharose in an attempt to dissociate the 43K protein from its ligand(s). When dialyzed BAE cell-conditioned medium was passed over this column, the major qpecies that bound was neither 43K protein nor BSA, but a protein of apparent reduced M , 70,000 that exhibited a slightly lower mobility than BSA (Mr 67,000-68,000) on SDS-PAGE. This result is shown in Fig. lA (lune 2). In the absence of reduction, the bound protein displayed a molecular weight of approximately 55,000 (lune 2, -DTT) and therefore appeared to be a single polypeptide chain containing intramolecular disulfide bonds.
At this point we investigated the origin of the 70K protein. Biosynthetic studies carried out under standard conditions in the presence of [3H]proline did not reveal a radioactive band in the molecular weight range of 70,000 after SDS-PAGE analysis of both culture medium and cell layer components (data not shown). To examine the possibility that the 70K protein was a plasma membrane or cell surface component of low abundance or metabolic turnover, we incubated sparse, freshly plated cells with [3H]leucine for 76 h. During this interval, the cells underwent at least two population doublings. Again, a radioactive band of M , 70,000 was not observed by SDS-PAGE (data not shown). It therefore appeared that the source of 70K protein recovered from postculture medium was the FCS that had been previously used to culture the cells.
Since BAE cell postculture medium contained a heterogeneous mixture of proteins and excessively large amounts of BSA (see Fig. l A , lane I ) , we took advantage of the preferential association of the 43K and 70K proteins to obtain more homogeneous preparations from which the 70K protein could be purified. Previous biosynthetic studies on radiolabeled BAE cells had shown that the 43K protein could be isolated radiochemically pure (but bound to a nonradioactive ligand) from the culture medium after ammonium sulfate precipitation (20-50% concentration range, weight to volume) and DEAE-cellulose chromatography (15). When this purified fraction was applied to the affinity column, as shown in Fig. lB, the 43K protein (lune 1 ) was dissociated from the 70K protein (lune 2) with a recovery of total cpm of 91%. This result confirmed the observation that later bleedings of this antiserum contained a population of antibodies that reacted preferentially with 70K protein, while the titer to 43K protein was markedly diminished. This method of isolation, as illustrated in Fig. lB, provided larger amounts of 70K protein in a single step; typically, 100-200 ml of serum-free culture medium were concentrated by ammonium sulfate precipitation and subsequently enriched for 70K protein on DEAEcellulose, prior to affinity chromatography on the antiserum. The preparation was also considerably depleted of BSA. As shown in Fig. 1B (FCS lune), BSA is the major protein in FCS (26 mg/ml, as reported by Hyclone for lot 100424). At the loading of FCS as illustrated in Fig. 1B (approximately 5 pg of BSA), the 70K protein was not visible by staining with Coomassie Blue. Based on the level of detection afforded by this dye for M , 50,000 proteins (250-500 ng), a rough estimate of the maximum amount of 70K protein in FCS was 1-2% of the BSA content, or approximately 250-500 pg/ml. 70K protein was also purified directly from FCS by affinity chromatography on Affi-Gel Blue. When the unbound fraction (aFP + 70K protein) and bound fraction (BSA) were analyzed on an 8% separating SDS-polyacrylamide gel, the difference in molecular weight between 70K protein and BSA became more apparent. As shown in Fig. 2  column, thereby interfering with the absorbance of the protein at either 230 or 280 nm. In addition, the recovery of BSA from Affi-Gel Blue was 25 f 5%. Removal of the bound BSA by boiling an aliquot of the resin in Laemmli buffer, followed by SDS-PAGE, showed that this fraction did not differ in molecular weight from that eluted by 1.4 M NaCl (not shown). From the data in Fig. 2, 1.6% of the total BSA content, or 420 pg of 70K protein, was present in 1 ml of FCS (based on 26 mg/ml BSA in FCS and a 100% recovery of 70K protein). This value (420 pg) was within the range estimated for the recovery of 70K protein from BAE cell postculture medium (250-500 pglml).
The structure of the 70K protein was shown to be different from that of BSA, aFP, and several other serum proteins. The mobility of purified 70K protein was slightly lower than that of purified BSA (Pentex) or BSA (FCS) in an acid-urea-PAGE system, and a more marked difference was seen in a "native" gel system as described by Ornstein (20) and Davis (21) (data not shown). The most significant difference in mobility between BSA and the 70K protein was observed on Laemmli gels in the absence of SDS and urea. A difference of similar magnitude was seen after molecular sieve chromatography on Sephadex G-200, in which the 70K protein eluted midway between the molecular weight standard proteins BSA (67,000) and ovalbumin (43,000) (data not shown).
Comparative one-dimensional peptide mapping, based on limited cleavage by trypsin and proteinase K, was performed on BSA, aFP, and 70K protein (data not presented). While BSA apparently contained no trypsin-sensitive sites, a single major band (Mr 65,000) was produced from aFP, and limited cleavage, with no peptides evident of M, < 10,000, was observed with the 70K protein. In contrast, a discrete set of unique peptides was produced from each protein with proteinase K. Two-dimensional mapping of 1251-peptides produced by complete proteinase K digestion showed a lack of coincidence of several major peptides (Fig. 31). The overall differences in topology between the maps of BSA (a) and 70K protein (b) did not reveal a convincing homology, even after very long exposures (c and d).
The structural relationship between the 70K protein and several serum or plasma proteins was also studied with immunochemical probes. Purified aFP, FCS (containing BSA, 70K protein, and aFP as principal components), and an impure preparation of 70K protein were exposed to anti-aFP antibodies, followed by 1251-protein A, in a Western blot assay. In similar immunoblotting experiments, the 70K protein did not react with antisera toward either von Willebrand protein or BSA (data not shown).
In Fig. 4 are summarized data from several ELISAs, all of which were performed with rabbit anti-bovine 70K protein antibodies. This antibody preparation, which had initially been affinity-purified against the 43K protein as bound to BSA and 70K protein, demonstrated no reactivity toward fibronectin, fibrinogen, von Willebrand protein, and thrombospondin (data not presented; 15). In panel A , the antiserum was passed over a BSA-Sepharose column and showed maximal reactivity toward 70K protein, a somewhat reduced response to FCS, and a low but residual activity toward BSA.  When the antiserum was, in addition, absorbed in solution with thrombospondin (which also served as a control for dilution of the antibody with a protein-containing solution), a result similar to that shown in panel A was obtained (panel C). In an effort to eliminate the small fraction of BSA reactive antibodies, the antiserum was absorbed in solution with BSA. As shown in panel B, removal of this reactivity resulted in an approximately 20% decrease in the response of the antiserum to the 70K protein and a slightly decreased reactivity to FCS. We have found by ELISA and immunoblotting that preparations of BSA, as supplied from both Sigma and Pentex, contained small amounts of 70K protein. Similarly, the anti-BSA antiserum fraction, as supplied by Cappel, displayed a low but detectable activity toward the 70K protein. Prior absorption of anti-BSA antiserum with 70K protein abolished this activity, but the initial titer against BSA was retained (data not shown). When the anti-70K antiserum was absorbed in solution with the 70K protein, the antibody population recognized neither 70K protein nor FCS (panel D). Since these ELISAs were not quantitative, we did not determine the amount of the 70K protein in FCS by this procedure. The data provide strong evidence, however, that BSA and the 70K protein are also immunologically distinct. Further confirmation of the difference between BSA and the 70K protein was provided by amino acid analysis (Table   I). There were highly significant differences in at least 4 residues: Ser, Gly, Cys, and Lys. While both BSA and the 70K protein exhibited a high percentage (approximately 20%) of potentially acidic amino acids, BSA contained a higher proportion of residues with bulky, hydrophobic side chains (Val, Leu, Phe, Lys).
Unlike many proteins that have been isolated from serum or plasma, the 70K protein did not bind to heparin-Sepharose (data not shown). It was also distinct from a recently described serum protein of similar molecular weight, vitronectin, by amino acid composition (25). In addition, studies showed that the 70K protein did not promote cell attachment and was unreactive when immunoblotted with anti-bovine vitronectin antisera.* The distribution of the 70K protein in BAE cells was studied by immunofluorescence and by SDS-PAGE of secreted, cell surface, and cytoplasmic components (Fig. 5). Antibodies specific for the 70K protein (as confirmed by ELISA) were localized to perinuclear granules in BAE cells which had been rendered permeable by prior ethanol treatment ( A and B ) . In the absence of this step, there was no apparent staining of either the cell surface or the substrateattached material (as seen between cells at subconfluent den-' E. Ruoslahti and M. Pierschbacher, unpublished data.

FIG. 5. Distribution of 70K protein in BAE cells in vitro.
BAE cells were grown on glass coverslips and were fixed and rendered permeable as described under "Materials and Methods." Panel A shows cells exposed to a sample of anti-70K protein antiserum that had been previously absorbed with BSA, thrombospondin, and 43K

ANTIBODY DILUTION
sities) (not shown). For these immunofluorescence experiments, it was essential to fix the cells after a very brief wash with either DMEM or Hanks' solution, as the 70K protein was released from the cells very rapidly in the absence of FCS (see below). The location of the 70K protein after a 2-h incubation of BAE cells in serum-free DMEM is shown in Fig. 5 (D-F). Among the three compartments, the 70K protein was found essentially exclusively in the culture medium (lane D); the identity of this single band was confirmed by peptide mapping and by ELISA. "Trypsin shave" experiments did not reveal 70K protein or BSA adsorbed to the cell surface (lane E ) . Control experiments on the lability of these proteins to the trypsin-EDTA solution, performed as described under Materials and Methods, showed that both BSA and 70K protein were not appreciably degraded under the conditions used in these experiments. After 2 h, only a small amount (or possibly none) of the 70K protein was apparent in the cells themselves (lune F). These results indicate that the 70K protein was not adsorbed to the BAE cell surface. Within 1 min after transfer of the cells to serum-free medium, the 70K protein could be seen as cytoplasmic granules following fixation ( A ) . 2 h after this transfer, the 70K protein was located principally in the culture medium by SDS-PAGE analysis

(D).
We examined the kinetics of the apparent release of the Coomassie Blue. D, 5% by volume of clarified culture medium, diluted 1:l with Laemmli buffer and applied directly to gel; E, after removal of the medium, cells were subjected to a brief trypsin shave, as described under "Materials and Methods," and 12.5% by volume of the total releasate was diluted 1:l with Laemmli buffer prior to SDS-PAGE F, after the trypsin shave, cells were solubilized directly in hot Laemmli buffer, and 25% of the total volume was applied to the gel. Mobilities of protein molecular weight standards are shown on the far right. Bands of M, 20,000-30,000 in lune E were present initially in the trypsin solution. tures of BAE cells. One dish each of cells a t sparse and confluent densities was washed once with PBS and then exposed to 2 ml of 70K Glycoprotein 70K protein by BAE cells in uitro. Additional experiments had indicated that two variables might, in fact, be significant: cell density and early time points. These parameters were studied as shown in Fig. 6. BAE cells were plated simultaneously onto 30-mm dishes a t sparse (2.13 X lo5 cells, 3.0 X lo4 cells/cm*, Fig. 6ZA) or at confluent density (5.24 X 10% cells, 7.42 X lo4 cells/cm2, Fig. 6lB). The cells were allowed to attach in the presence of FCS for 18-24 h prior to the inception of the experiment. When the culture media were analyzed by SDS-PAGE at 0, 5, and 15 min after removal of serum, the major component was in fact the 70K protein, which was present in a 6-fold higher concentration in the sparse cultures as compared to the dense cultures. These data are summarized as densitometric scans in Fig. 6ZZA and B. As will be shown in subsequent experiments, approximately 80% of the total 70K protein that could be measured in the medium after 2 h was released between 0 and 20 min. The identity of the major peak in Fig. 6ZZA and B as the 70K protein was confirmed by peptide mapping and/or ELISA.
The cell layers were solubilized in toto in hot Laemmli buffer after the last aliquot had been removed from the culture medium (15 min after exposing the cells to serum-free medium). As shown in Fig. 6ZZC, there was significantly more 70K protein in the cells plated a t sparse density than was present in the dense cultures (D). Comparisons cannot be made directly between the level of 70K protein in the medium uersus the cell layer (e.g. panel A with C, or B with D), as the total protein analyzed by SDS-PAGE represented 1.2% of the culture medium and 8% of the cell layer (see legend to Fig.  6). The data suggest that more 70K protein was taken up and subsequently released by BAE cells a t sparse density. If the data in Fig. 611 were normalized to represent equal numbers of cells at the two different densities, the increase in the 70K protein in the sparse cultures would be nearly 15 times greater than the level found in the dense cultures (&fold increase by scanning densitometry x 2.5 cell number).
Release of the 70K protein was examined in greater detail as a function of cell growth and density. BAE cells were used 24 h after plating at sparse (1.25 X lo5 cells, 3.0 X lo4 cells/ cm2), subconfluent (3.0 X 10% cells, 4.2 X lo4 cells/cm2), and confluent density (5.2 x lo5 cells, 7.4 x lo4 cells/cm2). To eliminate the cell division associated with attainment of confluence, cells were also plated at confluent density 2 h before inception of the experiment. A protocol similar to that described for Fig. 6 was used, except that fresh medium was added to the cultures after the first 2-min time point. The major protein released into the culture medium within 2 min after shifting the cells to a serum-free environment was the 70K protein; negligible amounts were obtained from 2-30 min (data not shown).  of time (B). A, the amounts of 70K protein released into the culture medium within 2 min after placing BAE cells in serum-free scanning densitometry of the gel lanes corresponding to 2 min for each cell density. The 70K protein was measured relative to the levels of actin ( A I r 40,000) in the cell layers, and this value was plotted as a function of cell density (Fig. 7A). There was a greater than 12-fold increase in 70K protein released from sparse cells than from subconfluent cells and an approximately 18-fold increase from sparse over confluent cells. The amounts of 70K protein released from confluent cells were similar between 24-and 2-h postsubculture (freshly plated, FP) dishes (Fig. 7A). When the 70K protein, in arbitrary absorbance units, was plotted directly as a function of cell number, graphs were similar to those in Fig. 7A (data not  shown).
Release of the 70K protein as a function of time is illustrated in Fig. 7B. The data points were generated from several different experiments in which cells of intermediate density were utilized. If 0 min represented the time at which the cells were exposed to serum-free medium, greater than ?h of the 70K protein recovered from the culture medium (up to a maximum of 2 h) was released within the first 20 min. After 5 min, only negligible amounts of the 70K protein could be recovered from the cells (Fig. 7B; see also Fig. 611).
Collectively, these data indicate that the release of the novel serum protein, of apparent M, 70,000, from BAE cells in vitro was (a) rapid (>80% released within 30 min), and ( b ) significantly enhanced in sparse, proliferating cultures. In order to measure accurately the levels of 70K protein in the culture medium as a function of time, we found that extremely rapid (<30 s) washes were necessary, prior to addition of serumfree medium, to minimize losses of 70K protein from the cells. Such release occurred regardless of the wash solution used (EDTA, PBS, DMEM, or Waymouth's medium). There appeared to be no reproducible differences in the levels of 70K protein released in either the presence or absence of ascorbate, or with DMEM as compared to Waymouth's medium. The release also appeared to be unaffected by the addition of [3H]    Fig. 7B): a minimum value.
amount of the 70K protein based on (a) its recovery from Affi-Gel Blue chromatography, relative to aFP, and (b) its release from BAE cells in culture, relative to BSA. In both cases, the published values for aFP and BSA, as shown in Table 11, were used, and a standard curve for the Coomassie Blue-staining reaction was generated with BSA over a concentration range of 0.1-100 pg. Studies by Lai and co-workers (26,27) have shown in fetal bovine plasma a progressive increase in the concentration of BSA with fetal maturation and a concomitant decrease in the levels of aFP. While the concentration of fetuin did not exhibit progressive changes, at partuition the concentration of BSA was 22 mg/ml, and that of aFP, 0.2 mg/ml (Table 11).
The FCS used in the present study (Hyclone lot 100424) was pooled from approximately 4000 individuals and contained 26 mg/ml of BSA (Table 11). This level is considerably higher than those reported for several other mammalian species, in which the gene for serum albumin is activated immediately prior to birth (28). For this lot of FCS, we calculated a maximum concentration for the 70K protein of 600 pg/ml, based on the recoveries of 70K protein and aFP from Affi-Gel Blue chromatography. A minimum value of 400 pg/ml was derived from the cellular release studies. In addition, 2.55 X IO5 cells/30-mm dish released into the culture medium 66 pg of 70K protein (based on standard curve with BSA), within the first 20 min of exposure to serum-free conditions, or 250 pg 70Kprotein/cell. These calculations define the limits for the concentration of the 70K protein in FCS, from approximately 400-600 pg/ ml. The value, 250 pg of 70K protein released/cell, was reproducible within 10% for cultures of BAE cells at the same density and on the same size dish (containing the same amount of culture medium). When cells grown on 100-or 150mm dishes (in medium containing 42 or 83 mg of BSA, respectively) were compared to those on 30-mm dishes, values somewhat higher than 400 pg/ml for the concentration of the 70K protein were obtained. This difference was essentially due to the concomitant uptake and release of BSA (up to 20% of the total 70K protein) that was characteristic of cultures containing higher numbers of cells, especially after exposure to serum-free conditions for more than 2 h.

DISCUSSION
This study provides structural characterization of a novel protein that was purified both from fetal bovine serum and from culture medium conditioned by bovine aortic endothelial cells. The protein is a single, disulfide-bonded polypeptide chain with an apparent molecular weight of 70,000 by SDS-PAGE in the presence of DTT. By peptide maps, amino acid analysis, and immunologic criteria, the 70K protein exhibited a unique primary structure. Its levels in FCS were estimated to be 400-600 pg/ml, and it was not a biosynthetic product of the BAE cells in uitro. An interesting observation was the apparently selective uptake and subsequent release of the 70K protein by BAE cells in culture. Within 20 min after exposure to serum-free medium, the cells released nearly 80% of their total internalized 70K protein, which by immunofluorescence had been irregularly distributed within cytoplasmic granules. Sparse, proliferating cells released, within 2 min, approximately 18-fold more 70K protein per cell than that released by confluent, nondividing cells. The data indicate that a novel protein in FCS is taken up by BAE cells during routine cell culture, and that its rapid release is triggered by some form of culture shock, such as exposure to serum-free conditions or during proliferation from subconfluent densities.
Since BSA (Mr 68,000) is the major protein component of FCS, experiments were performed to examine whether there was a structural relationship between BSA and the 70K protein. An effective separation of these two proteins was achieved by chromatography on Affi-Gel Blue. Albumins bind selectively to several dyes, including Cibacron Blue (29), while the 70K protein did not demonstrate this affinity. By several criteria including amino acid composition, one-and twodimensional peptide mapping, ELISA and Western blotting, and apparent molecular weight by PAGE under both native and denaturing conditions, the 70K protein was structurally distinct from BSA. These experimental data were sufficient to establish that the 70K protein was not an alternate and less prevalent form of serum albumin, such as prealbumin (30), proalbumin (31), glycosylated albumin (32,33), and variant albumins as have recently been shown in alloalbuminemia and by restriction enzyme fragment length polymorphism (34).
Several other serum proteins, with apparent molecular mass in the range of 70,000 daltons, were found to be different from the 70K protein by amino acid composition, immunologic criteria, affinity for heparin, and/or cell attachment and spreading activity. These proteins include a-fetoprotein (35), vitronectin (25,36), an M, 70,000 human serum protein as described by Vuento and co-workers (37), heparin cofactor I1 (38), thrombomodulin (39), hemopexin (40), and leucine-rich a2-glycoprotein (41). Reference tables containing the molecular parameters of purified plasma proteins (e.g. 42) did not provide further possibilities for the identity of the 70K protein. It appears therefore that this protein is both a unique and novel component of fetal calf serum (the lot analyzed in this study was pooled from 4,000 individuals).
At this time the pathway by which albumin (and/or the 70K protein) enters the endothelial cell is not clear. An endocytotic mechanism of transendothelial transport, dependent both on the physiological state of the cell and on the biochemical properties of the macromolecule, is most probable (43, see Ref. 9 for a review). That molecular signals might favor selective internalization of certain serum proteins by the endothelium is suggested by the enhanced transport of glycosylated albumin (32) and uptake of acetylated low-density lipoprotein (44), as compared to the rates observed with the underivatized counterparts of these molecules.
The contribution of endothelial injury to disease processes such as atherosclerosis has provided a rationale for understanding the relationship between cellular behavior and overt cellular injury (7,45). Hansson and co-workers (46) have shown that both IgG and C3 accumulated in injured endothelial cells from normal and atherosclerotic arteries, and that IgG was bound to vimentin-type intermediate filaments within the cells (47). Studies from several laboratories have also demonstrated selective adsorption of serum proteins to the uninjured endothelial surface in vivo, with a subsequent diminution in cellular permeability to other macromolecules (48). Despite the apparent protective effect of albumin and other circulating proteins on endothelial cells, increased rates of macromolecular (especially albumin) transfer across the endothelium have been correlated with cellular injury. Exposure of endothelial cells to oxidants (12,49), calcium ionophore (11,12), and free fatty acids (10) resulted in intracellular accumulation of albumin that was associated with changes in cell shape and cytoskeletal configuration. Injury as a result of cell culture can also have an effect on endocytotic rate and macromolecular accumulation. Davies (9) found that the rate of fluid endocytosis in wounded, endothelial monolayers was higher in proliferating and migrating cells. In this regard, the higher levels of the 70K protein associated with sparsely plated, proliferating BAE cells, in contrast to those in confluent and quiescent cultures, might reflect the close relationship between a polarized endocytic cycle and cell locomotion, as proposed by Bretscher (50).
In proliferating cultures of BAE cells plated at intermediate density, approximately 250 pg of 70K protein was released per cell within 20 min after exposure to serum-free medium. This figure is expected to be, in part, a function of in vitro conditions, as serial subcultivation of endothelial cells is accompanied by changes both in cell cycle kinetics and in secretory properties (3). Although we observed a gradual and sustained release of BSA from BAE cells in the absence of serum, the rapid release of the 70K protein (>€io% of the total internalized 70K protein within the first 30 min) was quite specific. This result is suggestive of a separate pathway and compartmentalization for the uptake, storage, and release of BSA and the 70K protein. One possibility is that the release of BSA reflects the general osmotic imbalance that occurs when cells are placed in serum-free medium, as has been shown for ovalbumin in capillary endothelium (51). In contrast, release of the 70K protein might be diagnostic for the initial events associated with acute cellular injury: e.g. proliferation, migration, or attachment to plastic mediated by the secretion of extracellular matrix macromolecules. In this regard, acute radiation injury has been shown to potentiate the release of von Willebrand protein, but not of fibronectin, from endothelial cells in vitro (52). The differences observed in the release of the 70K protein between sparse and confluent BAE cells could be a direct consequence of the actin, myosin, and vinculin reorganization that occurs upon migration of previously confluent endothelial cells into an experimental wound (53).
The presence of the 70K protein in serum raises the possibility that it might be selectively internalized by endothelial cells that have sustained a local injury, such as viral infection (54,55) or after exposure to certain mitogens (56). Endothelial cells in vitro are the victims of culture shock; some of the metabolic changes that occur in,response to serum factors, disruption of osmotic balance, plating at sparse density, or in conjunction with migration, proliferation, and cytoskeletal rearrangement could mimic those observed in vivo. Since the function of the 70K protein is presently not known, the significance of its uptake, apparent storage, and rapid release under situations of cellular stress cannot be fully appreciated. An intriguing possibility is that the 70K protein is posttranslationally modified to facilitate rapid transcytosis, and in turn is able to function as a carrier protein for selected biosynthetic products of endothelial cells that are up-regulated in response to stress, such as the 43K protein (57). The response of endothelial cells in particular to environmental vicissitudes is in all probability linked to the regulation of genetic transcription for such stress-induced proteins.