EXPERIMENTAL
CELL
RESEARCH
187,&F-89
(1990)
Proteoglycan Synthesis by Cultured Liver Endothelium: The Role of Membrane-Associated Heparan Sulfate in Transferrin Binding EIJIRO OMOTO, The Veterans Administration
JOSE J. MINGUELL,
Medical Center and University
AND MEHDI
of Mississippi
TAVASSOLI’
School of Medicine,
Jackson, Mississippi
39216
noglycans (GAG), are covalently attached. Depending on the type of glycan side chains, three major classes of sulfated PGs are recognized: heparan sulfate (HS), chondroitin sulfate (CS), and dermatan sulfate (DS) (for review see Ref. [l]). PGs are important constituents of the extracellular matrix and also are associated in most cells either with the cell membrane or intracellularly with storage granules. Cell-associated PGs have many different cellular functions including cell adhesion, growth-promoting activities, and regulation of receptor functions [2,3]. In this respect, it has been reported that the core protein of intact heparan sulfate-PG, associated with the cell membrane, displays high affinity for binding the iron transport glycoprotein transferrin (Tf) [4, 51. It therefore has been postulated that the Tf receptors and the proteoglycan core protein share the same polypeptide sequence and, hence, HS core protein is an inactive progenitor molecule of the Tf receptor [6,7]. Liver is a major organ in iron metabolism and it has been demonstrated that liver endothelium possesses Tf receptors. These endothelial cells bind, internalize, and transport Tf which is partially desialylated during the course of transport [8-131. Endothelial cells from vascular tissues have been shown to produce several types of PC, and those from bovine [14] and human [15] aortic endothelia have been well characterized. PG synthesis by transport capillary endothelia has not been studied. To find out if membrane-associated PG, synthesized by liver capillary endothelium, can provide a molecular basis for the Tf binding of this cell, we studied PG synthesis by liver endothelium in culture. We now provide evidence indicating that a fraction of membrane-associated PG, which contains heparan sulfate-PG, accounts for all the specific binding of transferrin.
Liver endothelium has been reported to possess membrane receptors for the iron-binding protein transferrin (Tf). Similarly, the core protein of proteoglycans (PG) associated with cell membrane in many cell systems can bind Tf. To find out if membrane-associated proteoglycans can explain Tf-binding ability of liver endothelium, we investigated the synthesis and distribution of proteoglycans by isolated, cultured liver capillary endothelium. Cells were isolated and cultured for 48 h in sulfate-free medium and pulse-labeled with ?304. The relative distribution of 35S0,-labeled macromolecules, determined in the extracellular (EC), membrane-associated (MA), and intracellular (IC) pools, was respectively 74,15, and 10%. Membrane-associated proteoglycan (MA-PG) was further purified by ion exchange and gel chromatography. Glycosaminoglycan (GAG) chain characterization indicated about 78% chondroitin sulfate, 7% dermatan sulfate, and about 14% heparan sulfate (HS). Similar GAG chain characterization was made for PG in the EC and IC pools. Transferrin-binding ability of MA-PG was studied by affinity column chromatography, using CNBractivated sepharose bound to transferrin. About 15% of the labeled MA-PG was specifically bound to Tf-affinity column and could be eluted by excess soluble Tf. This proportion was similar to the proportion of HS in the total membrane-associated pool. Moreover, the eluted labeled material was susceptible to pretreatment with heparitinase, confirming its HS nature. We conclude that the transport capillary endothelium of the liver can synthesize HS proteoglycans which are membraneassociated and this MA-HS pool can bind transferrin. The finding may provide a molecular basis for transferrin binding to liver endothelium and may explain the subsequent transendothelial transport of iron-trans0 1990hademic press, hc. ferrin complexes into the liver.
MATERIALS INTRODUCTION
AND
METHODS
Under sterile conditions, crude liver cell suspenCell preparation. sions were obtained from 200- to 250-g Sprague-Dawley rats, using the collagenase perfusion method [16]. To purify endothelial cells, crude liver cell suspensions were subjected consecutively to metrizamide gradient centrifugation followed by centrifugal elutriation as described before in detail [9]. The crude liver suspension was prepared in Dulbecco’s phosphate-buffered saline (D-PBS, pH 7.4) and loaded in a Beckman JEG-B standard elutriator rotor (4.2 ml) using a JS-21
Proteoglycans (PG) consist of a central protein core to which many glycan side chains, known as glycosami1 To whom reprint requests and correspondence should he addressed at VA Medical Center (151), 1500 E. Woodrow Wilson Drive, Jackson, MS 39216. 85
0014-4527po
$3.00
Copyright 0 1990 by Academic Press, Inc. All rights of reproduction in any form reserved.
86
OMOTO.
MINGUELL.
centrifuge (Beckman Instruments, Inc.). Before cell loading, the rotor was eluted with D-PBS at a flow rate of 11 ml/min. The rotor was then loaded, using a total volume of 4 ml while maintaining the same flow rate. Elutriation was done at 650g at 20°C. The flow rate was increased to 22 ml/min and 400 ml of eluant fluid was collected. The yield was 4 X lo7 endothelial cells per liver with 92% purity as judged by indirect immunofluorescent staining for factor VIII as well as by their distinctive features in electron microscopy [9]. Viability was 99%. Cells were cultured under liquid conditions in the presence of 10% fetal calf serum and endothelial growth factor (150 pg/ml, Collaborative Research, Inc., Bedford, MA). Labeling and isolation of proteoglycam. Before labeling with ‘sS, cells were transferred to Fischers medium (which is sulfate free) containing 10% fetal calf serum. Labeling was initiated by adding to cell suspensions (1.0 X lo7 cells/ml) 20 &i/ml of carrier-free Nai5S04 (ICN Biomedicals). The cells were further cultured at 37°C in a 5% COP atmosphere for 48 h. Cell number and viability remained essentially unchanged during this period. At the end of the labeling period, cultures were chilled on ice and the culture medium was removed. Cells were then washed twice with PBS and the washes were combined with the original culture medium. Labeled material in this pool was designated as extracellular pool (EC). Cells were then incubated with 1 ml of 0.05% trypsin in PBS at 37°C for 10 min to release the membrane-associated PG [17]. This procedure does not alter the proteoglycan nature of the material in our experience and in that of others [17]. The cells were then centrifuged and the supernate was saved. The cell pellet was rinsed twice with PBS and the washes were combined with the trypsin supernate to obtain the membrane-associated pool (MA). To obtain the intracellular pool (IC), the cell pellet was then extracted with a dissociative-extraction buffer enriched with protease inhibitors [18]. This extraction procedure gave almost 100% solubilization of ?+labeled proteoglycans from cells. To the extracellular and membrane-associated pools, 0.4 g/ ml solid guanidine HCL, 1.25 mg/ml N-ethylmaleimide, and 0.18 mg/ ml phenylmethylsulfonyl fluoride (protease inhibitor) were added. Unincorporated isotope, guanidine-HCl, and other chemicals were removed from samples of each pool by extensive dialysis against distilled water containing 1 mM NazSO,. Aliquots of each extract were counted for radioactivity to determine distribution of labeled macromolecules in the EC, MA, and IC pools. For purification of labeled PG in the MA pool, aliquots of the dialyzed material were then applied to an ion-exchange column (DESephacel, Pharmacia, bed volume 3.0 ml) equilibrated with 8 M urea, 0.15 M sodium chloride, 0.05 M sodium acetate, and 0.5% (w/v) Triton X-100, pH 6.0 [19]. After sample application, the column was washed with 10 ml of the same buffer and then eluted with a continuous NaCl gradient (from 0.15 to 1.5 M) prepared in the same solvent, by the use of a gradient mixer (Ultrograde, LKB). Fractions of approximately 0.7 ml were collected at a flow rate of 35 ml/h and radioactivity was measured. The NaCl gradient was monitored by measuring conductivity of fractions. Further purification was performed by analytical Sepharose CL-4B (Pharmacia) gel filtration column. Either the intact PGs or the released GAG chains from the protein core, after an overnight 0.2 N NaOH hydrolysis [17], was applied to the column. Columns (1.6 X 70 cm) were prepared and eluted with 4 M guanidine-HCl, 0.05 M TrisHCl, 0.5% (w/v) Triton X-100, pH 7.0, at a flow rate of 5 ml/h [20]. Fractions were collected and radioactivity was measured. The void volume and total column volume were determined with Blue Dextran and free Na$OI, respectively. To compare relative PG sizes, the distribution coefficient (I&,) for each radioactive peak was calculated as described [21]. For GAG chain characterization, samples were adjusted to contain 0.15 M Tris-HCl, pH 8.0, and were digested with pronase (7 mg/ml) for 10 hat 42°C [22]. Samples were then subjected to one of the following treatments:
AND
TAVASSOLI
(a) Chondroitinase ABC or AC (Sigma, 0.25 u/ml) in enriched Tris buffer, pH 8.0, for 3 h at 37°C to identify chondroitin sulfate and/or dermatan sulfate [20]. (b) Heparitinase (ICN, 0.1 u/ml) in enriched Tris-HCl buffer, pH 7.3, for 3 h at 37°C to identify heparan sulfate [23]. The extent of GAG digestion was assessed by precipitation with cetylpyridinium chloride [24]. Binding of transferrin to MA-PG. Human Tf was obtained from Calbiochem-Behring Corp. The purity was subsequently demonstrated by its migration as a single band on SDS-PAGE. Diferric Tf was prepared as previously described [9,12]. To assure saturation, the absorbance of the final solution in D-PBS was read at 470 and 280 nm, giving an absorbance ratio of 0.046 consistent with full saturation [25]. Unless otherwise specified, herein Tf refers to fully saturated diferric Tf. An affinity column was prepared by covalent binding of 4 mg of Tf to 1 ml of wet gel of CNBr-activated Sepharose 4B (Sigma Chemical Co.) in 0.1 M NaHC03, pH 8.8. The slurry was then applied to a column (10 X 30 mm) and washed with 20 bed vol of D-PBS. Aliquots of MA-PG were applied and allowed to enter the column. After an incubation for 1 h at 4”C, the column was washed with 20 bed vol of D-PBS (2 ml/h) to elute labeled PG and/or GAG not bound to Tf. Thereafter, the column was competitively eluted with a solution containing an excess of Tf (10 mg/ml in D-PBS) in order to specifically displace labeled bound PG from Tf-Sepharose complex. Fractions were collected and radioactivity was measured.
RESULTS
Capillary endothelial cells grew and attached to the plastic vessel, forming an adherent layer which contained cellular elements maintaining the immunological and electron microscopic features of endothelial cells [9]. Cell viability was close to 100% during the first week of culture. No cell proliferation occurred as judged by the absence of mitotic figures and only 20% of original inoculum remained attached to the dishes after 1 week of culture. Cellular morphology remained unchanged as judged by light and electron microscopy. Factor VIII positivity was no longer seen after 2 weeks of culture. When radioactive sulfate was added to the culture medium, the label was incorporated into macromolecules after a labeling period of 48 h. The distribution of 35Slabeled macromolecules in the three pools isolated from cultures of liver endothelial cells is shown in Table 1. Approximately 74% of total 35S-labeled material was present in the EC pool. The remaining label was associated with the MA and IC pools and accounted for 15 and 11% of the radioactivity, respectively. To characterize the types of the GAG moieties of the proteoglycans present in each pool, samples were digested with pronase and then treated with chondroitinases ABC and AC and heparitinase. The labeled enzyme-resistant materials were quantitated after precipitation with cetylpyridinium chloride. Results, shown in Table 1, indicated that liver endothelial cells produced PG containing CS, DS, and HS glycan moieties. CS-PG was the most abundant component in all three pools, accounting for 68 to 79% of total GAG. HS-PG was also
PG SYNTHESIS
TABLE
BY LIVER
ENDOTHELIUM
1
Cellular Distribution of Proteoglycans Synthesized by Endothelial Cells and Characterization of Their GAG Moieties”
Pool Extracellular Membraneassociated Intracellular
a%-labeled macromolecules (So)
GAG types” (So) cs
DS
HS
74
68.4 f 1.9
6.0 f 2.0
26.0 f 0.9
15 11
78.7 k 1.0 68.8 z!z4.6
7.0 k 2.1 9.12 0.5
13.7 AZ2.0 21.2 + 5
a Cultured endothelial cells were labeled with %04 and the labeled materials were fractionated into three pools and extracted with a dissociative buffer containing protease inhibitors as indicated in the text. A portion of the labeled materials in each pool was used to measure the distribution of “S-labeled macromolecules. Enzymatic treatment was used to determine the types of PG. The figures represent means f standard deviations of triplicate experiments.
present in all fractions and its concentration varied from 14 to 26%. DS-PG was the least abundant component and its concentration was not higher than 9%. The membrane-associated PG was further analyzed by ion-exchange chromatography and gel filtration. As shown in Fig. 1, when MA-labeled material was applied to a DE-Sephacel column, more than 95% of the material was retained and eluted as a single peak at 0.55 M NaCl. A small peak representing less than 5% of the input was not retained and eluted ahead of the salt gra-
m 1.0. :
0
0.5
0.15
IO 20 30 40 50 FRACTION NUMBER
60
FIG. 1. Purification of membrane-associated %-labeled material by DE-Sephacel. The column was eluted with a 4 M urea buffer containing a linear gradient (0.15 to 1.5 M) of NaCl, shown as a continuous line. The small peak, appearing before the application of gradient, may be various labeled glycoproteins. The labeled proteoglycans, eluting as a sharp peak at 0.55 A4 NaCl and indicated by a bar, were collected for further analysis.
I
FIG. 2. Purification of membrane-associated %-labeled materials by Sepharose CL-4B chromatography. The column was eluted with a 4 M guanidine buffer and radioactivity and Kav values of each fraction were measured. Note the shift in the elution pattern of intact PG (A) and after removal of core protein (B).
Figure 2A shows the elution profile of MA-labeled material after Sepharose CL-4B chromatography. A broad main peak appeared, which eluted at an average K,, value of 0.47. Despite the broadness of this peak, this material might have contained other PG, eluting at Kav values of 0.48 and 0.57. The PG nature of the labeled material eluting at Kav of 0.47 was confirmed by the observed shift in the K,, value after treatment of the intact material with 0.2 M NaOH (Fig. 2B). This procedure, which cleaves the GAG moiety from the core protein, reduced the size of the labeled material, with a concomitant change in the elution pattern (K,, 0.47 to Ka. 0.70). To investigate whether transferrin binds to proteoglycans present in the cell membrane, affinity binding studies were performed. The MA-labeled material eluted at 0.55 M NaCl was collected, pooled (Fig. 1, bar), and extensively dialyzed against D-PBS. Portions of this material were first digested with chondroitinase ABC or heparitinase to remove radiolabeled glycan moieties and then applied to a transferrin-Sepharose 4B column. Binding was allowed to occur by incubation at 4°C for 1 h. The column was then washed with D-PBS and eluted with a D-PBS solution containing an excess of free transferrin. As shown in Table 2, in samples containing either intact (nontreated) PG or chondroitinase ABC-treated PG, approximately 15% of the radioactive input was
I.5 1 1.0
0.5 Kav
88
OMOTO,
MINGUELL,
AND
TAVASSOLI
which participates in the uptake of iron by the cells is a membrane-associated glycoprotein [ll, 271. We thereBinding of Membrane-Associated Proteoglycans fore investigated Tf-binding ability of membrane-assoto Transferrin Affinity Column” ciated PG newly synthesized by cultured liver endotheRadioactivity (% of input)” lium to test the hypothesis that membrane-associated eluted as: HS in these cells might serve as a Tf receptor. As occurs with endothelial cells of other origin [14, Unbound Bound 171, liver endothelial cells in culture synthesize PG from PG to Tf PG to Tf labeled 35S precursor and secrete approximately 70% of 85.2 14.8 Intact PG the labeled material to the culture medium. Of the re16.7 Chondroitinase ABC-treated PG 83.3 maining cell-associated material, 60% is membrane-as0 Heparitinase-treated PG 100 sociated and can be released after a short treatment with trypsin. The membrane-associated material contains n Pooled fractions of MA-PG after DE-Sephacel chromatography approximately 85% of CS/DS PG and 15% of HS-PG. (Fig. 1, bar) were dialyzed against D-PBS. Aliquots of these materials were treated with D-PBS (intact PG), chondroitinase ABC, or hepariWhen the MA material is applied to an ion-exchange tinase. Samples were then applied to a Tf-Sepharose column, precolumn, the bulk of the 35S-labeled material is retained pared as indicated under Materials and Methods. After elution with Dand elutes as a single peak at 0.55 M NaCl, indicating PBS (unbound PG to Tf) and with D-PBS containing free transferrin (bound PG to Tf), aliquots were taken from the measurement of 35S that it is either intact PG or its GAG moiety. The PG nature of this material can be further demonstrated after radioactivity. The radioactive input to the column in each case was 1000 cpm MA-PG/lO mg of bound Tf to Sepharose 4B. alkaline treatment which removes the GAG moiety from its core protein: this treatment causes a shift in the elution profile in Sepharose CL-4B chromatography. When MA-PG, thus isolated, is examined for its bound to the transferrin-Sepharose column and eluted binding capacity by affinity chromatograas a single peak after elution with free Tf. However, in transferrin phy, intact 35S-labeled MA-PG specifically binds to samples containing heparitinase-treated PG, no radioactive material was bound to the column and 100% of transferrin and can be competitively eluted by an excess of free transferrin. Pretreatment of intact MA-PG with the radioactive input was recovered in the washes with chondroitinase ABC does not change the elution pattern D-PBS. of the affinity column indicating that the Tf-binding maThe labeled material applied to the column and not terial is not chondroitin sulfate or dermatan sulfate. By bound to the Tf-Sepharose consisted of intact PG molecontrast, pretreatment with heparitinase eliminates all cules other than free GAG moieties. the radioactivity, indicating that Tf-binding material consists entirely of HS. Consistent with this conclusion DISCUSSION is the finding that both in the intact PG and in the PG after ABC-treatment, 15% of the labeled material apThe synthesis of PG by nontransport vascular endoplied to the column is bound and eluted with free Tf. thelial cells (large vessels) in culture has been reported This figure is similar to the proportion of HS-PG in the [17]. Irrespective of the tissue origin of endothelial cells, more than 70% of these macromolecules are secreted to total MA-PG pool, which is about 14% (see Table 1). The findings strongly suggest that liver endothelial the culture medium and can be detected in the extracellular pool. These extracellular components may play a cells produce a HS-PG, which becomes associated with the cell membrane in such a way that it exposes a Tfrole in the formation of the basement membrane-like binding site. This is consistent with the reports [4, 51 matrix as revealed by immunofluorescence and electron that the core protein of heparan sulfate contains a Tfmicroscopic studies [ 171. binding site and can also explain the reported Tf-bindCapillary endothelium differs from large vessel endothelium in that the capillary endothelium is involved in ing ability of the liver endothelium [8-131. The observamass exchange of a broad spectrum of molecules be- tions that PG in the cell membrane can be released by tween plasma and interstitial fluid [26]. Many of these trypsin and that its core protein still retains Tf-binding properties, suggest that the binding site, which is located macromolecules, including transferrin, have been shown in the ectodomain of the proteoglycan, is resistant to to be transported transendothelially [ 111. This has particularly been studied in liver endothelium and in the trypsin digestion [4, 281. It has been demonstrated that case of Tf which is transported and modified by this en- the Tf receptor molecule after treatment with trypsin, releases a 70kDa fragment which can still bind Tf and dothelium [8-131. which retains antigenic properties [ 111. Recent reports have shown that in certain connective It has been reported that under certain physiological tissues, the core protein of membrane-associated HSPG is closely related to the transferrin receptor [4, 51. circumstances, membrane-associated HS-PGs are enzymatically released from the cell membrane, only to bind Moreover, it is known that the transferrin receptor, TABLE
2
PG SYNTHESIS
BY LIVER
immediately to specific membrane receptors again and to be internalized [3]. Such a mechanism may provide an explanation as to why the Tf-receptor complexes in liver endothelium take an intracellular pathway different from that in such other Tf-binding cells as K562 [12]. In the latter cells, exemplifying the classic Tf-receptor pathway, Tf-receptor complexes enter the endosomes where in the low pH of these organelles, iron dissociates from Tf, and apo Tf-receptor complex is recycled to the cell membrane where, in the neutral pH, apo Tf dissociates in order to be saturated again with iron [ 111. In liver endothelium, by contrast, Tf is transported across the cell and is externalized in the space of Disse [lo]. During this transport, iron remains in association with Tf [12], indicating that the path of transport is different from that of other Tf-binding cells. It is, therefore, possible that the difference in the membrane-associated Tf-binding surface molecules can explain the difference in the intracellular pathway of Tf.
a9
ENDOTHELIUM
6. Caster, L., Carsfeelt,
I., Kendall, S., Malmstrom, A., Schmidtthen, A., and Fransson, L. A. (1986) J. Biol. Chem. 261,12076. 7. Omary, M. B., and Trowbridge, I. S. (1981) J. Biol. Chem. 256, 12888. 8. Tavassoli, M., Kishimoto, T., Soda, R., Kataoka, M., and Harjes, K. (1986) Exp. Cell Res. 166,369. 9. Kishimoto, T., and Tavassoli, M. (1985) Biochim. Biophys. Acta
846,14. 10.
Kishimoto,
T., and Tavassoli,
M. (1987) Amer. J. Anat.
178,
241. 11. 12.
Irie, S., and Tavassoli, M. (1987) Amer. J. Med. Sci. 292,103. Irie, S., Kishimoto, T., and Tavassoli, M. (1988) J. Clin. Invest.
13. 14.
Irie, S., and Tavassoli, M. (1989) Biochem. J. 259,427. Radhakrishnamurthy, B., Ruiz, H. A., Jr., and Berenson, G. S. (1977) J. Biol. Chem. 252,4831-4841. Salisbury, B. G. J., and Wagner, W. D. (1981) J. Biol. Chem. 256,
82,508.
15.
8050. 16. 17.
Seglen, P. 0. (1976) Methods Cell Biol. 13,29. Oohira, A., Wight, T. N., and Borustein, P. (1983) J. Biol. Chem.
18.
Yanagishita,
258,2014. M., and Hascall,
V. C. (1984) J. Biol. C&m.
269,
10260. This work was supported by Grant DK-30142 to Mehdi Tavassoli. Jose J. Minguell is the recipient of a sabbatical award from the Foundation Andes and the University of Chile, Santiago, Chile.
REFERENCES 1.
Hook, M., Kjellen, them. 53,847.
L., and Johansson,
147a.
Received August 22,1989 Revised version received November
7,1989
20. 21. 22.
S. (1984) Annu. Rev. Bio-
2. Gallagher, J., Lyon, M., and Steward, W. D. (1986) Biochem. J. 236,313. 3. Ishihara, M., Fedarko, N. S., and Conrad, H. E. (1987) J. Biol. Chem.262,4708. 4. Fransson, L. A., Caster, L., Carlsted, I., and Malmstrom, A. (1985) Biochem. J. 231,683. 5. Bentley, S. A., Kirby, S. L., and High, K. A. (1987) Blood 7O(Suppl.),
19.
23. 24. 25. 26. 27. 28.
Chang, Y., Yanagishita, M., Hascall, V. C., and Wight, T. N. (1983) J. Biol. Chem. 258,5679. Rapraeger, A., and Bernfield, M. (1985) J. Biol. Chem. 260,4103. Carney, S. L. (1986) in Proteoglycans (Chaplin, M. F., and Kennedy, J. F., Eds.), p. 97, IRL Press, Oxford. Whiteside, T. L., Worral, J. G., Prince, R. K., Buckingham, R. B., and Rodnan, G. P. (1985) Arthritis Rheum. 28,188. Kirby, S. L., and Bentley, S. A. (1987) Blood 70,1777. Elias, J. A., Krol, R. C., Freundlich, B., and Sampson, P. M. (1988) J. Clin. Znuest. 81,325. Schneider, C., Sutherland, R., Newman, R., and Greaves, M. (1982) J. Biol. Chem. 257,8516. Palade, G. E. (1982) Ann. NYAcad. Sci. 401,265. Jalkanen, M., Rapraeger, A., Saunders, S., and Bernfield, M. (1987)J. CellBiol. 105, 3087. Aasa, R., Malmstrom, B. G., Saltman, P., and Vanngard, T. (1963) Biochim. Biophys. Acta 75,203.
CELL
RESEARCH
187,&F-89
(1990)
Proteoglycan Synthesis by Cultured Liver Endothelium: The Role of Membrane-Associated Heparan Sulfate in Transferrin Binding EIJIRO OMOTO, The Veterans Administration
JOSE J. MINGUELL,
Medical Center and University
AND MEHDI
of Mississippi
TAVASSOLI’
School of Medicine,
Jackson, Mississippi
39216
noglycans (GAG), are covalently attached. Depending on the type of glycan side chains, three major classes of sulfated PGs are recognized: heparan sulfate (HS), chondroitin sulfate (CS), and dermatan sulfate (DS) (for review see Ref. [l]). PGs are important constituents of the extracellular matrix and also are associated in most cells either with the cell membrane or intracellularly with storage granules. Cell-associated PGs have many different cellular functions including cell adhesion, growth-promoting activities, and regulation of receptor functions [2,3]. In this respect, it has been reported that the core protein of intact heparan sulfate-PG, associated with the cell membrane, displays high affinity for binding the iron transport glycoprotein transferrin (Tf) [4, 51. It therefore has been postulated that the Tf receptors and the proteoglycan core protein share the same polypeptide sequence and, hence, HS core protein is an inactive progenitor molecule of the Tf receptor [6,7]. Liver is a major organ in iron metabolism and it has been demonstrated that liver endothelium possesses Tf receptors. These endothelial cells bind, internalize, and transport Tf which is partially desialylated during the course of transport [8-131. Endothelial cells from vascular tissues have been shown to produce several types of PC, and those from bovine [14] and human [15] aortic endothelia have been well characterized. PG synthesis by transport capillary endothelia has not been studied. To find out if membrane-associated PG, synthesized by liver capillary endothelium, can provide a molecular basis for the Tf binding of this cell, we studied PG synthesis by liver endothelium in culture. We now provide evidence indicating that a fraction of membrane-associated PG, which contains heparan sulfate-PG, accounts for all the specific binding of transferrin.
Liver endothelium has been reported to possess membrane receptors for the iron-binding protein transferrin (Tf). Similarly, the core protein of proteoglycans (PG) associated with cell membrane in many cell systems can bind Tf. To find out if membrane-associated proteoglycans can explain Tf-binding ability of liver endothelium, we investigated the synthesis and distribution of proteoglycans by isolated, cultured liver capillary endothelium. Cells were isolated and cultured for 48 h in sulfate-free medium and pulse-labeled with ?304. The relative distribution of 35S0,-labeled macromolecules, determined in the extracellular (EC), membrane-associated (MA), and intracellular (IC) pools, was respectively 74,15, and 10%. Membrane-associated proteoglycan (MA-PG) was further purified by ion exchange and gel chromatography. Glycosaminoglycan (GAG) chain characterization indicated about 78% chondroitin sulfate, 7% dermatan sulfate, and about 14% heparan sulfate (HS). Similar GAG chain characterization was made for PG in the EC and IC pools. Transferrin-binding ability of MA-PG was studied by affinity column chromatography, using CNBractivated sepharose bound to transferrin. About 15% of the labeled MA-PG was specifically bound to Tf-affinity column and could be eluted by excess soluble Tf. This proportion was similar to the proportion of HS in the total membrane-associated pool. Moreover, the eluted labeled material was susceptible to pretreatment with heparitinase, confirming its HS nature. We conclude that the transport capillary endothelium of the liver can synthesize HS proteoglycans which are membraneassociated and this MA-HS pool can bind transferrin. The finding may provide a molecular basis for transferrin binding to liver endothelium and may explain the subsequent transendothelial transport of iron-trans0 1990hademic press, hc. ferrin complexes into the liver.
MATERIALS INTRODUCTION
AND
METHODS
Under sterile conditions, crude liver cell suspenCell preparation. sions were obtained from 200- to 250-g Sprague-Dawley rats, using the collagenase perfusion method [16]. To purify endothelial cells, crude liver cell suspensions were subjected consecutively to metrizamide gradient centrifugation followed by centrifugal elutriation as described before in detail [9]. The crude liver suspension was prepared in Dulbecco’s phosphate-buffered saline (D-PBS, pH 7.4) and loaded in a Beckman JEG-B standard elutriator rotor (4.2 ml) using a JS-21
Proteoglycans (PG) consist of a central protein core to which many glycan side chains, known as glycosami1 To whom reprint requests and correspondence should he addressed at VA Medical Center (151), 1500 E. Woodrow Wilson Drive, Jackson, MS 39216. 85
0014-4527po
$3.00
Copyright 0 1990 by Academic Press, Inc. All rights of reproduction in any form reserved.
86
OMOTO.
MINGUELL.
centrifuge (Beckman Instruments, Inc.). Before cell loading, the rotor was eluted with D-PBS at a flow rate of 11 ml/min. The rotor was then loaded, using a total volume of 4 ml while maintaining the same flow rate. Elutriation was done at 650g at 20°C. The flow rate was increased to 22 ml/min and 400 ml of eluant fluid was collected. The yield was 4 X lo7 endothelial cells per liver with 92% purity as judged by indirect immunofluorescent staining for factor VIII as well as by their distinctive features in electron microscopy [9]. Viability was 99%. Cells were cultured under liquid conditions in the presence of 10% fetal calf serum and endothelial growth factor (150 pg/ml, Collaborative Research, Inc., Bedford, MA). Labeling and isolation of proteoglycam. Before labeling with ‘sS, cells were transferred to Fischers medium (which is sulfate free) containing 10% fetal calf serum. Labeling was initiated by adding to cell suspensions (1.0 X lo7 cells/ml) 20 &i/ml of carrier-free Nai5S04 (ICN Biomedicals). The cells were further cultured at 37°C in a 5% COP atmosphere for 48 h. Cell number and viability remained essentially unchanged during this period. At the end of the labeling period, cultures were chilled on ice and the culture medium was removed. Cells were then washed twice with PBS and the washes were combined with the original culture medium. Labeled material in this pool was designated as extracellular pool (EC). Cells were then incubated with 1 ml of 0.05% trypsin in PBS at 37°C for 10 min to release the membrane-associated PG [17]. This procedure does not alter the proteoglycan nature of the material in our experience and in that of others [17]. The cells were then centrifuged and the supernate was saved. The cell pellet was rinsed twice with PBS and the washes were combined with the trypsin supernate to obtain the membrane-associated pool (MA). To obtain the intracellular pool (IC), the cell pellet was then extracted with a dissociative-extraction buffer enriched with protease inhibitors [18]. This extraction procedure gave almost 100% solubilization of ?+labeled proteoglycans from cells. To the extracellular and membrane-associated pools, 0.4 g/ ml solid guanidine HCL, 1.25 mg/ml N-ethylmaleimide, and 0.18 mg/ ml phenylmethylsulfonyl fluoride (protease inhibitor) were added. Unincorporated isotope, guanidine-HCl, and other chemicals were removed from samples of each pool by extensive dialysis against distilled water containing 1 mM NazSO,. Aliquots of each extract were counted for radioactivity to determine distribution of labeled macromolecules in the EC, MA, and IC pools. For purification of labeled PG in the MA pool, aliquots of the dialyzed material were then applied to an ion-exchange column (DESephacel, Pharmacia, bed volume 3.0 ml) equilibrated with 8 M urea, 0.15 M sodium chloride, 0.05 M sodium acetate, and 0.5% (w/v) Triton X-100, pH 6.0 [19]. After sample application, the column was washed with 10 ml of the same buffer and then eluted with a continuous NaCl gradient (from 0.15 to 1.5 M) prepared in the same solvent, by the use of a gradient mixer (Ultrograde, LKB). Fractions of approximately 0.7 ml were collected at a flow rate of 35 ml/h and radioactivity was measured. The NaCl gradient was monitored by measuring conductivity of fractions. Further purification was performed by analytical Sepharose CL-4B (Pharmacia) gel filtration column. Either the intact PGs or the released GAG chains from the protein core, after an overnight 0.2 N NaOH hydrolysis [17], was applied to the column. Columns (1.6 X 70 cm) were prepared and eluted with 4 M guanidine-HCl, 0.05 M TrisHCl, 0.5% (w/v) Triton X-100, pH 7.0, at a flow rate of 5 ml/h [20]. Fractions were collected and radioactivity was measured. The void volume and total column volume were determined with Blue Dextran and free Na$OI, respectively. To compare relative PG sizes, the distribution coefficient (I&,) for each radioactive peak was calculated as described [21]. For GAG chain characterization, samples were adjusted to contain 0.15 M Tris-HCl, pH 8.0, and were digested with pronase (7 mg/ml) for 10 hat 42°C [22]. Samples were then subjected to one of the following treatments:
AND
TAVASSOLI
(a) Chondroitinase ABC or AC (Sigma, 0.25 u/ml) in enriched Tris buffer, pH 8.0, for 3 h at 37°C to identify chondroitin sulfate and/or dermatan sulfate [20]. (b) Heparitinase (ICN, 0.1 u/ml) in enriched Tris-HCl buffer, pH 7.3, for 3 h at 37°C to identify heparan sulfate [23]. The extent of GAG digestion was assessed by precipitation with cetylpyridinium chloride [24]. Binding of transferrin to MA-PG. Human Tf was obtained from Calbiochem-Behring Corp. The purity was subsequently demonstrated by its migration as a single band on SDS-PAGE. Diferric Tf was prepared as previously described [9,12]. To assure saturation, the absorbance of the final solution in D-PBS was read at 470 and 280 nm, giving an absorbance ratio of 0.046 consistent with full saturation [25]. Unless otherwise specified, herein Tf refers to fully saturated diferric Tf. An affinity column was prepared by covalent binding of 4 mg of Tf to 1 ml of wet gel of CNBr-activated Sepharose 4B (Sigma Chemical Co.) in 0.1 M NaHC03, pH 8.8. The slurry was then applied to a column (10 X 30 mm) and washed with 20 bed vol of D-PBS. Aliquots of MA-PG were applied and allowed to enter the column. After an incubation for 1 h at 4”C, the column was washed with 20 bed vol of D-PBS (2 ml/h) to elute labeled PG and/or GAG not bound to Tf. Thereafter, the column was competitively eluted with a solution containing an excess of Tf (10 mg/ml in D-PBS) in order to specifically displace labeled bound PG from Tf-Sepharose complex. Fractions were collected and radioactivity was measured.
RESULTS
Capillary endothelial cells grew and attached to the plastic vessel, forming an adherent layer which contained cellular elements maintaining the immunological and electron microscopic features of endothelial cells [9]. Cell viability was close to 100% during the first week of culture. No cell proliferation occurred as judged by the absence of mitotic figures and only 20% of original inoculum remained attached to the dishes after 1 week of culture. Cellular morphology remained unchanged as judged by light and electron microscopy. Factor VIII positivity was no longer seen after 2 weeks of culture. When radioactive sulfate was added to the culture medium, the label was incorporated into macromolecules after a labeling period of 48 h. The distribution of 35Slabeled macromolecules in the three pools isolated from cultures of liver endothelial cells is shown in Table 1. Approximately 74% of total 35S-labeled material was present in the EC pool. The remaining label was associated with the MA and IC pools and accounted for 15 and 11% of the radioactivity, respectively. To characterize the types of the GAG moieties of the proteoglycans present in each pool, samples were digested with pronase and then treated with chondroitinases ABC and AC and heparitinase. The labeled enzyme-resistant materials were quantitated after precipitation with cetylpyridinium chloride. Results, shown in Table 1, indicated that liver endothelial cells produced PG containing CS, DS, and HS glycan moieties. CS-PG was the most abundant component in all three pools, accounting for 68 to 79% of total GAG. HS-PG was also
PG SYNTHESIS
TABLE
BY LIVER
ENDOTHELIUM
1
Cellular Distribution of Proteoglycans Synthesized by Endothelial Cells and Characterization of Their GAG Moieties”
Pool Extracellular Membraneassociated Intracellular
a%-labeled macromolecules (So)
GAG types” (So) cs
DS
HS
74
68.4 f 1.9
6.0 f 2.0
26.0 f 0.9
15 11
78.7 k 1.0 68.8 z!z4.6
7.0 k 2.1 9.12 0.5
13.7 AZ2.0 21.2 + 5
a Cultured endothelial cells were labeled with %04 and the labeled materials were fractionated into three pools and extracted with a dissociative buffer containing protease inhibitors as indicated in the text. A portion of the labeled materials in each pool was used to measure the distribution of “S-labeled macromolecules. Enzymatic treatment was used to determine the types of PG. The figures represent means f standard deviations of triplicate experiments.
present in all fractions and its concentration varied from 14 to 26%. DS-PG was the least abundant component and its concentration was not higher than 9%. The membrane-associated PG was further analyzed by ion-exchange chromatography and gel filtration. As shown in Fig. 1, when MA-labeled material was applied to a DE-Sephacel column, more than 95% of the material was retained and eluted as a single peak at 0.55 M NaCl. A small peak representing less than 5% of the input was not retained and eluted ahead of the salt gra-
m 1.0. :
0
0.5
0.15
IO 20 30 40 50 FRACTION NUMBER
60
FIG. 1. Purification of membrane-associated %-labeled material by DE-Sephacel. The column was eluted with a 4 M urea buffer containing a linear gradient (0.15 to 1.5 M) of NaCl, shown as a continuous line. The small peak, appearing before the application of gradient, may be various labeled glycoproteins. The labeled proteoglycans, eluting as a sharp peak at 0.55 A4 NaCl and indicated by a bar, were collected for further analysis.
I
FIG. 2. Purification of membrane-associated %-labeled materials by Sepharose CL-4B chromatography. The column was eluted with a 4 M guanidine buffer and radioactivity and Kav values of each fraction were measured. Note the shift in the elution pattern of intact PG (A) and after removal of core protein (B).
Figure 2A shows the elution profile of MA-labeled material after Sepharose CL-4B chromatography. A broad main peak appeared, which eluted at an average K,, value of 0.47. Despite the broadness of this peak, this material might have contained other PG, eluting at Kav values of 0.48 and 0.57. The PG nature of the labeled material eluting at Kav of 0.47 was confirmed by the observed shift in the K,, value after treatment of the intact material with 0.2 M NaOH (Fig. 2B). This procedure, which cleaves the GAG moiety from the core protein, reduced the size of the labeled material, with a concomitant change in the elution pattern (K,, 0.47 to Ka. 0.70). To investigate whether transferrin binds to proteoglycans present in the cell membrane, affinity binding studies were performed. The MA-labeled material eluted at 0.55 M NaCl was collected, pooled (Fig. 1, bar), and extensively dialyzed against D-PBS. Portions of this material were first digested with chondroitinase ABC or heparitinase to remove radiolabeled glycan moieties and then applied to a transferrin-Sepharose 4B column. Binding was allowed to occur by incubation at 4°C for 1 h. The column was then washed with D-PBS and eluted with a D-PBS solution containing an excess of free transferrin. As shown in Table 2, in samples containing either intact (nontreated) PG or chondroitinase ABC-treated PG, approximately 15% of the radioactive input was
I.5 1 1.0
0.5 Kav
88
OMOTO,
MINGUELL,
AND
TAVASSOLI
which participates in the uptake of iron by the cells is a membrane-associated glycoprotein [ll, 271. We thereBinding of Membrane-Associated Proteoglycans fore investigated Tf-binding ability of membrane-assoto Transferrin Affinity Column” ciated PG newly synthesized by cultured liver endotheRadioactivity (% of input)” lium to test the hypothesis that membrane-associated eluted as: HS in these cells might serve as a Tf receptor. As occurs with endothelial cells of other origin [14, Unbound Bound 171, liver endothelial cells in culture synthesize PG from PG to Tf PG to Tf labeled 35S precursor and secrete approximately 70% of 85.2 14.8 Intact PG the labeled material to the culture medium. Of the re16.7 Chondroitinase ABC-treated PG 83.3 maining cell-associated material, 60% is membrane-as0 Heparitinase-treated PG 100 sociated and can be released after a short treatment with trypsin. The membrane-associated material contains n Pooled fractions of MA-PG after DE-Sephacel chromatography approximately 85% of CS/DS PG and 15% of HS-PG. (Fig. 1, bar) were dialyzed against D-PBS. Aliquots of these materials were treated with D-PBS (intact PG), chondroitinase ABC, or hepariWhen the MA material is applied to an ion-exchange tinase. Samples were then applied to a Tf-Sepharose column, precolumn, the bulk of the 35S-labeled material is retained pared as indicated under Materials and Methods. After elution with Dand elutes as a single peak at 0.55 M NaCl, indicating PBS (unbound PG to Tf) and with D-PBS containing free transferrin (bound PG to Tf), aliquots were taken from the measurement of 35S that it is either intact PG or its GAG moiety. The PG nature of this material can be further demonstrated after radioactivity. The radioactive input to the column in each case was 1000 cpm MA-PG/lO mg of bound Tf to Sepharose 4B. alkaline treatment which removes the GAG moiety from its core protein: this treatment causes a shift in the elution profile in Sepharose CL-4B chromatography. When MA-PG, thus isolated, is examined for its bound to the transferrin-Sepharose column and eluted binding capacity by affinity chromatograas a single peak after elution with free Tf. However, in transferrin phy, intact 35S-labeled MA-PG specifically binds to samples containing heparitinase-treated PG, no radioactive material was bound to the column and 100% of transferrin and can be competitively eluted by an excess of free transferrin. Pretreatment of intact MA-PG with the radioactive input was recovered in the washes with chondroitinase ABC does not change the elution pattern D-PBS. of the affinity column indicating that the Tf-binding maThe labeled material applied to the column and not terial is not chondroitin sulfate or dermatan sulfate. By bound to the Tf-Sepharose consisted of intact PG molecontrast, pretreatment with heparitinase eliminates all cules other than free GAG moieties. the radioactivity, indicating that Tf-binding material consists entirely of HS. Consistent with this conclusion DISCUSSION is the finding that both in the intact PG and in the PG after ABC-treatment, 15% of the labeled material apThe synthesis of PG by nontransport vascular endoplied to the column is bound and eluted with free Tf. thelial cells (large vessels) in culture has been reported This figure is similar to the proportion of HS-PG in the [17]. Irrespective of the tissue origin of endothelial cells, more than 70% of these macromolecules are secreted to total MA-PG pool, which is about 14% (see Table 1). The findings strongly suggest that liver endothelial the culture medium and can be detected in the extracellular pool. These extracellular components may play a cells produce a HS-PG, which becomes associated with the cell membrane in such a way that it exposes a Tfrole in the formation of the basement membrane-like binding site. This is consistent with the reports [4, 51 matrix as revealed by immunofluorescence and electron that the core protein of heparan sulfate contains a Tfmicroscopic studies [ 171. binding site and can also explain the reported Tf-bindCapillary endothelium differs from large vessel endothelium in that the capillary endothelium is involved in ing ability of the liver endothelium [8-131. The observamass exchange of a broad spectrum of molecules be- tions that PG in the cell membrane can be released by tween plasma and interstitial fluid [26]. Many of these trypsin and that its core protein still retains Tf-binding properties, suggest that the binding site, which is located macromolecules, including transferrin, have been shown in the ectodomain of the proteoglycan, is resistant to to be transported transendothelially [ 111. This has particularly been studied in liver endothelium and in the trypsin digestion [4, 281. It has been demonstrated that case of Tf which is transported and modified by this en- the Tf receptor molecule after treatment with trypsin, releases a 70kDa fragment which can still bind Tf and dothelium [8-131. which retains antigenic properties [ 111. Recent reports have shown that in certain connective It has been reported that under certain physiological tissues, the core protein of membrane-associated HSPG is closely related to the transferrin receptor [4, 51. circumstances, membrane-associated HS-PGs are enzymatically released from the cell membrane, only to bind Moreover, it is known that the transferrin receptor, TABLE
2
PG SYNTHESIS
BY LIVER
immediately to specific membrane receptors again and to be internalized [3]. Such a mechanism may provide an explanation as to why the Tf-receptor complexes in liver endothelium take an intracellular pathway different from that in such other Tf-binding cells as K562 [12]. In the latter cells, exemplifying the classic Tf-receptor pathway, Tf-receptor complexes enter the endosomes where in the low pH of these organelles, iron dissociates from Tf, and apo Tf-receptor complex is recycled to the cell membrane where, in the neutral pH, apo Tf dissociates in order to be saturated again with iron [ 111. In liver endothelium, by contrast, Tf is transported across the cell and is externalized in the space of Disse [lo]. During this transport, iron remains in association with Tf [12], indicating that the path of transport is different from that of other Tf-binding cells. It is, therefore, possible that the difference in the membrane-associated Tf-binding surface molecules can explain the difference in the intracellular pathway of Tf.
a9
ENDOTHELIUM
6. Caster, L., Carsfeelt,
I., Kendall, S., Malmstrom, A., Schmidtthen, A., and Fransson, L. A. (1986) J. Biol. Chem. 261,12076. 7. Omary, M. B., and Trowbridge, I. S. (1981) J. Biol. Chem. 256, 12888. 8. Tavassoli, M., Kishimoto, T., Soda, R., Kataoka, M., and Harjes, K. (1986) Exp. Cell Res. 166,369. 9. Kishimoto, T., and Tavassoli, M. (1985) Biochim. Biophys. Acta
846,14. 10.
Kishimoto,
T., and Tavassoli,
M. (1987) Amer. J. Anat.
178,
241. 11. 12.
Irie, S., and Tavassoli, M. (1987) Amer. J. Med. Sci. 292,103. Irie, S., Kishimoto, T., and Tavassoli, M. (1988) J. Clin. Invest.
13. 14.
Irie, S., and Tavassoli, M. (1989) Biochem. J. 259,427. Radhakrishnamurthy, B., Ruiz, H. A., Jr., and Berenson, G. S. (1977) J. Biol. Chem. 252,4831-4841. Salisbury, B. G. J., and Wagner, W. D. (1981) J. Biol. Chem. 256,
82,508.
15.
8050. 16. 17.
Seglen, P. 0. (1976) Methods Cell Biol. 13,29. Oohira, A., Wight, T. N., and Borustein, P. (1983) J. Biol. Chem.
18.
Yanagishita,
258,2014. M., and Hascall,
V. C. (1984) J. Biol. C&m.
269,
10260. This work was supported by Grant DK-30142 to Mehdi Tavassoli. Jose J. Minguell is the recipient of a sabbatical award from the Foundation Andes and the University of Chile, Santiago, Chile.
REFERENCES 1.
Hook, M., Kjellen, them. 53,847.
L., and Johansson,
147a.
Received August 22,1989 Revised version received November
7,1989
20. 21. 22.
S. (1984) Annu. Rev. Bio-
2. Gallagher, J., Lyon, M., and Steward, W. D. (1986) Biochem. J. 236,313. 3. Ishihara, M., Fedarko, N. S., and Conrad, H. E. (1987) J. Biol. Chem.262,4708. 4. Fransson, L. A., Caster, L., Carlsted, I., and Malmstrom, A. (1985) Biochem. J. 231,683. 5. Bentley, S. A., Kirby, S. L., and High, K. A. (1987) Blood 7O(Suppl.),
19.
23. 24. 25. 26. 27. 28.
Chang, Y., Yanagishita, M., Hascall, V. C., and Wight, T. N. (1983) J. Biol. Chem. 258,5679. Rapraeger, A., and Bernfield, M. (1985) J. Biol. Chem. 260,4103. Carney, S. L. (1986) in Proteoglycans (Chaplin, M. F., and Kennedy, J. F., Eds.), p. 97, IRL Press, Oxford. Whiteside, T. L., Worral, J. G., Prince, R. K., Buckingham, R. B., and Rodnan, G. P. (1985) Arthritis Rheum. 28,188. Kirby, S. L., and Bentley, S. A. (1987) Blood 70,1777. Elias, J. A., Krol, R. C., Freundlich, B., and Sampson, P. M. (1988) J. Clin. Znuest. 81,325. Schneider, C., Sutherland, R., Newman, R., and Greaves, M. (1982) J. Biol. Chem. 257,8516. Palade, G. E. (1982) Ann. NYAcad. Sci. 401,265. Jalkanen, M., Rapraeger, A., Saunders, S., and Bernfield, M. (1987)J. CellBiol. 105, 3087. Aasa, R., Malmstrom, B. G., Saltman, P., and Vanngard, T. (1963) Biochim. Biophys. Acta 75,203.
