Synthesis and Secretion of Thrombospondin by
`Cultured Human Endothelial Cells
`
`DEANE F. MOSHER, MARY JEAN DOYLE, and ERIC A . JAFFE
`Department of Medicine, University of Wisconsin-Madison, Madison, Wisconsin 53706; and
`Department of Medicine, Cornell University Medical College, New York, New York 10021
`
`Thombospondin, the major glycoprotein released from a-granules of thrombin-
`ABSTRACT
`stimulated platelets, is a disulfide-bonded trimer of 160 kilodalton subunits and apparently
`functions as a platelet lectin. Because cultured human umbilical vein endothelial cells synthe-
`size and secrete a glycoprotein (GP-160) which is a disulfide-bonded multimer of 160 Walton
`subunits, the possibility that GP-160 is thrombospondin was investigated . Tritiated GP-160
`could be immunoisolated from [3H]leucine-labeled endothelial cell postculture medium using
`a rabbit antiserum to human platelet thrombospondin . Thrombospondin and GP-160 comi-
`grated in two different two-dimensional electrophoretic systems . Both proteins are disulfide-
`bonded trimers of acidic 160-kdalton subunits. A competitive radioimmunoassay for binding
`of ' 25 1-thrombospondin to the rabbit antibodies indicated that 49 l-rg of thrombospondin
`antigen per 106 confluent endothelial cells accumulated in postculture medium over 24 h .
`Thus, endothelial cells secrete large amounts of a glycoprotein that is identical or very similar
`to platelet thrombospondin .
`
`Thrombospondin, also known as thrombin-sensitive protein (2,
`3) or glycoprotein G (4), is a major platelet a-granule glyco-
`protein that is secreted and then partially bound to platelet
`membranes when human platelets aggregate in response to
`thrombin (5-9). Studies by Lawler and co-workers indicate
`that thrombospondin is a 450-kdalton filamentous protein of
`dimensions 7 X 65 run, has a pl of 4.7, and is composed of
`three large disulfide-linked subunits (10, 11). During the proc-
`ess of aggregation, thrombin-stimulated platelets develop a
`membrane-bound lectin activity (12-14) that originates from
`a-granules and appears to play an important role in mediating
`platelet aggregation by binding to a specific receptor on other
`platelets (15, 16) . We recently found that purified human
`platelet thrombospondin has lectin activity (i.e., agglutinates
`fixed trypsinized sheep erythrocytes) and blocks the aggluti-
`nation of thrombin-treated platelets. Therefore, we suggested
`that thrombospondin is the endogenous lectin of human plate-
`lets (17) . This may explain why platelets from patients with the
`gray platelet syndrome, which lack a-granules and a-granule
`constituents, including thrombospondin, aggregate poorly in
`response to thrombin (18) .
`
`A portion ofthis work was presented to the American Society for Cell
`Biology in Cincinnati, OH, November 14-18, 1980 (1).
`
`THE JOURNAL Of CELL BIOLOGY " VOLUME 93 MAY 1982 343-348
`© The Rockefeller University Press " 0021-9525/82/05/0343/06 $1 .00
`
`Cultured endothelial cells synthesize and secrete a glycopro-
`tein, GP-160, which is a disulfide-bonded multimer of 160-
`kdalton subunits and resists digestion with bacterial collagenase
`(19, 20) . A protein with the same properties is produced by
`cultured HT-1080 human sarcoma cells (21) . Because of the
`similarities between thrombospondin and GP-160, we under-
`took the present experiments which indicate that GP-160 is
`identical or nearly identical to thrombospondin.
`
`MATERIALS AND METHODS
`Materials
`Humanplasma fibronectin was purified andiodinated as described previously
`(22) . Human thrombin (23) and Escherichia coli RNA polymerase were generous
`gifts from Dr. John Fenton, II, New York State Department of Health, Albany,
`NY, and Dr. Richard Burgess, University of Wisconsin-Madison, Madison, WI,
`respectively . The following were purchased: Type I collagenase, Worthington
`Biochemicals, Freehold, NJ ; L-[3,4,5 3H]leucine (110 Ci/mmol), carrier-free
`Na"I, and Enhance, New England Nuclear, Boston, MA; X-Omat XR-2 Film,
`Eastman Kodak, Co., Rochester, NY; protein A-Sepharose CL-4B, Pharmacia
`Fine Chemicals, Piscataway, NJ ; hírudin, molecularsize markers, bovine albumin
`(type V), swine skin type I gelatin, and phenylmethylsulfonyl fluoride (PMSF),
`Sigma Chemical Co ., St . Louis, MO ; Bio-Gel P-300, Bio-Gel A-15M, insolubi-
`lized goat anti-rabbit immunoglobulin (Immunobeads), and reagents for SDS
`PAGE, Bio-RadLaboratories, Richmond, CA, heparin-agarose, Pierce Chemical
`Co., Rockford, IL; and protein A-bearing Staphylococci (IgGsorb), The Enzyme
`Center, Inc., Boston, MA .
`
`343
`
`Mylan v. Genentech
`IPR2016-00710
`Merck Ex. 1101, Pg. 1
`
`

`

`
`
`(cid:9)(cid:9)
`
`Culture of Endothelial Cells
`
`Human endothelial cells were derived from umbilical cord veins and cultured
`in medium 199 and 2091 human serum using methods and materials previously
`described (19, 24).
`Radioactively labeled proteins synthesized and secreted by confluent endothe-
`lial cells were prepared by incubating washed 75-cm' cell monolayers for 24 h at
`37°C in leucine-free minimal essential medium, 10 ml per flask containing 200
`ACi of L-[3,4,53H)leucine . In experiments in which [3 Iilleucine-labeled fibronec-
`tin and GP-160 were isolated by gelatin-agarose chromatography, the cells were
`labeled in medium containing 20% pooled human serum. In experiments in
`which [3 H)leucine-labeled GP-160 was immunoisolated using rabbit antithrom-
`
`bospondin, the cells were labeled in medium containing either 20% rabbit serum
`or 0.5% tryptose phosphate broth .
`In experiments in which the accumulation of thrombospondin antigen was
`quantitated by radioimmunoassay, endothelial cells were cultured in medium 199
`containing 20% human serum in 2-cm' wells of multiwell plates. When the cells
`were confluent, the cell layer was washed, and medium containing 20% rabbit
`serum was placed over the cells . After a 24-h incubation, the cell layer was
`washed, and the cells were placed in 1 ml of fresh medium 199 containing 20%
`rabbit serum. At various times after the second medium change, postculture
`medium was removed, clarified by centrifugation at 8,000 g for 2 min in a
`microfuge, and frozen for further analysis . Cells were dispersed with 0.02% (wt/
`vol) collagenase and 0.01% (wt/vol) EDTA in 10 mM HEPES, 140 mM sodium
`chloride, pH 7 .4, and enumerated with a Coulter electronic cell counter. To rule
`out the possibility that human thrombospondin might bind to endothelial cell
`layer during subculture and be released slowly into the culture medium during
`the 24-h preincubation and the experimental incubation in rabbit serum, we
`added "5I-thrombospondin (see below) to serum-containing medium of confluent
`cells. After 24 h, <I% was associated with the cell layer . Human serum contained
`-V65 pg/ml thrombospondin (see below), and medium plus 20% serum would
`contain -13 pg/ml. Thus, <130 ng would be in the cell layer at the beginning of
`the experimental incubation .
`
`Purification and lodination of Thrombospondin
`
`8 U of l- to 3-d-old platelet concentrate was obtained from Badger Red Cross,
`Madison, WI . Platelets were washed and reacted with thrombin as described by
`Lawler et al . (10, 11) ; thrombin was inhibited with hirudin after 2 min of reaction.
`The supernatant, 15-30 nil, was frozen in a dry-ice ethanol bath and thawed in
`a 37°C bath . A small fibrin clot formed in the thawing solution; it was gently
`removed with a stir rod . The supernatant was applied to a 5 x 35-cm column of
`Bio-Gel P-300, equilibrated and eluted with TBS-EDTA. The first protein peak,
`' `m) of 0.2, was applied to a 1 .2
`which typically had a peak absorbance (A"° "m .
`x 6-cm column of heparin-agarose equilibrated in TBS-EDTA . The heparin-
`agarose column was washed with 10 mM Tris, 140 mM sodium chloride, and I
`mM EDTA, pH 7 .4 (TBS-EDTA), and thrombospondin was eluted with 0.55 M
`sodium chloride in 10 mM Tris, I mM EDTA, pH 7 .4. Peak fractions from the
`heparin-agarose column were pooled, portions were frozen in a dry-ice ethanol
`bath, and the frozen solutions were stored at -70°C . The amount of thrombo-
`spondin in the initial platelet releasate, determined by radioimmunoassay (see
`below), was typically 7 .5 mg, and the yield of purified thrombospondin, assuming
`
`that a 1 mg/ml solution has an A'8° ""` ' - of 1 .09 (11), was typically 1 .5 mg
`(20% yield) . The purity of thrombospondin, estimated by densitometry after SDS
`PAGE, was 97% (Fig . 1) .
`Thrombospondin was iodinated by the chloramine-T technique (25) . Free
`iodine was removed by chromatography on Bio-Gel P-300 or extensive dialysis.
`The labeled protein was mixed with PMSF-treated albumin, 1 mg/ml, snap
`frozen in a dry-ice ethanol bath, and stored in portions at -70°C. Specific activity
`was -0 .5 mCi/mg . Radiochemical purity was assessed by autoradiography after
`SDS PAGE and found to be >95% .
`
`Production and Characterization of
`Rabbit Antithrombospondin
`
`Thrombospondin was further purified by preparative electrophoresis after
`reduction on SDS PAGE slab gels (26) . The final product, 10-20 Ag in 0.3 ml of
`0.1% SDS, was emulsified with an equal volume of complete Freund's adjuvant
`and injected into a rabbit at multiple subcutaneous sites . The injections in
`complete adjuvant were repeated twice at monthly intervals . 2 wk later, I mg of
`protein from the heparin-affinity column was given without adjuvant . Antiserum
`was collected over the next 2 mo. Immunoglobulin from the antiserum, when
`
`loaded onto protein A-bearing Staphylococci (27), bound 'z5 1-thrombospondin
`and not ... I-fibronectin. In experiments in which immune complexes were
`
`collected with beads coated with goat anti-rabbit immunoglobulin or with protein
`
`344
`
`THE JOURNAL OF CELL BIOLOGY " VOLUME 93, 1982
`
`Purification of human platelet thrombospondin as mon-
`FIGURE 1
`itored by SIDS PAGE on 8% gels . Samples were reduced before
`analysis . The position of thrombospondin (TSP) is indicated . Lanes :
`(a) Molecular size standards : plasma fibronectin (210 kdaltons),
`phosphorylase (93 kdaltons), albumin (68 kdaltons), ovalbumin (43
`kdaltons), and chymotrypsinogen (24.5 kdaitons) . (b) Releasate from
`platelets (9 x 109/ml) stimulated with thrombin (3 U/ml) . The
`release reaction was performed in 0 .1 M HEPES, 0 .15 M sodium
`chloride, pH 7 .6, containing 5 mM glucose and 2 .5 mM EDTA for 2
`min at 37°C . (c) Releasate from thrombin-stimulated platelets after
`freezing and thawing . (d) Clot that was recovered after freezing and
`thawing of platelet releasate . (e) Purified human plasma fibrinogen .
`(f) First protein peak from Bio-Gel P-300 column . (g) Unbound
`fraction from heparin-agarose affinity column . (h) Protein eluted
`from heparin-agarose affinity column by 0.55 M sodium chloride.
`Protein loads: lane a, 5 fug per band ; lane b, 7 Ag; lane c, 9 .51íg ; lane
`e, 5 Ag; lane f, 3 Ag ; lane g, 0 .4 Ag ; and lane h, 6 .5 líg . In other
`experiments, the f3' subunit of E. coli RNA polymerase,150 kdaltons,
`was used to assign a molecular size of 160 kdaltons to thrombo-
`spondin . The a and # chains of platelet fibrin are slightly smaller
`than the Aa and Bf3 chains of fibrinogen because of loss of the A
`and B fibrinopeptides .
`
`A-bearing Staphylococci, maximum binding of 75 ng of `Z'1-thrombospondin was
`obtained with 1 ml of a 1 :1,000 dilution of antiserum .
`
`Radioimmunoassay of Thrombospondin
`Protein A-bearing Staphylococci were armed with antithrombospondin as
`described by O'Keefe and Bennett (27) . Approximately 90% of ... I-thrombo-
`spondin bound when mixed with an equal volume of a 10% (wt/vol) suspension
`of armed Staphylococci (Fig. 2A). Standard or unknown in physiologic saline
`solution, 100 Al, was mixed with ... I-thrombospondin, -2 x 10° cpm in 100 Al,
`
`and then with 100 Al of a suspension that consisted of one part 10% armed
`Staphylococci and nine parts 10% "unarmed" carrier Staphylococci . After 3 h at
`4°C, the mixture was centrifuged for I min at 10,000 g in a microfuge, the
`supernatant was discarded, and the pellet was analyzed in a gamma counter . The
`assay was sensitive to 0 .7 pg/ml thrombospondin (Fig. 2 B). Fibronectin, 200 ttg/
`ml, did not inhibit in the assay . By the assay, the concentration of thrombospondin
`in serum (formed by allowing human blood in glass tubes to clot at 37°C for I
`h) was 65 ± 23 pg/ml (R ± SD, n = 6) .
`
`lmmunoisolation and Analysis of [3 H]Leucine-
`labeled Endothelial Cell Proteins
`Tritiated leucine-labeled endothelial cell medium was made 3% in SDS and
`6 M in urea, heated in a boiling water bath for 5 min, dialyzed against two
`changes of 3% SDS in phosphate-buffered saline (PBS), diluted 10-fold with 0.12
`
`Merck Ex. 1101, Pg. 2
`
`

`

`Brilliant Blue. The gels containing radioactive proteins were sectioned into 2-mm
`slices, and each slice was processed for liquid scintillation counting as previously
`described (l9).
`
`Partial Purification of [3H]leucine-labeled GP-
`160 and Fibronectin on Gelatin-Agarose
`Purified plasma frbronectin, 12 mg, was added to 25 ml of [3 H]leucine-labeled
`endothelial cell postculture medium, and the mixture was applied to a 1.5 x 12-
`cm column of gelatin-agarose. The column support consisted of gelatin coupled
`to cyanogen bromide-activated Bio-Gel A-15M by the method of March et al.
`(29). Unlabeled fibronectin, tritiated fibronectin, and tritiated GP-160 bound to
`the column . The column was washed with PBS-EDTA, and bound proteins were
`eluted with IM sodium bromide, 20 mM sodium acetate, pH 5.0. The eluted
`proteins were dialyzed against TBS, snap frozen, and kept at -70°C until further
`studies were performed.
`
`Polyacrylamide Slab Gel Electrophoresis
`One-dimensional SDS PAGE was performed using the discontinuous slab gel
`system of Ames (30) . Size markers included reduced fibronectin, 210 kdaltons ;
`ß' aubunit of RNA polymerase, 150 kdaltons; f3-galactosidase, 116 kdaltons ;
`phosphorylase, 93 kdaltons; albumin, 68 kdaltons ; ovalbumin, 43 kdaltons; and
`chymotrypsinogen, 24.5 kdaltons. Two-dimensional gel electrophoresis, in which
`reduced proteins were separated by isoelectric focusing in 8 M urea in the first
`dimension and SDS PAGE in the second dimension, was performed as described
`by Anderson and Anderson (31) . Nonreduced-reduced two-dimensional SDS
`PAGE was performed by the method of Phillips and Poh Agin (32) with slight
`modifications. The first-dimension discontinuous gel was cast in 0.3 (i .d.) x 12-
`cm tubes. After electrophoresis without reduction, the first-dimension gel was
`incubated in sample buffer containing l% ß-mercaptoethanol . The gelwas bound
`to the top of the second-dimension discontinuous slab gel by 1% agarose made
`up in sample buffer . After polymerization of the agarose, a small slot was made
`to the right (bottom) of the cylindrical gel, and size markers were analyzed along
`side the proteins in the cylindrical gel. Nonreduced plasma fibronectin was
`assumed to have a size of 420 kdaltons . Slab gels were stained with Coomassie
`Brilliant Blue, prepared for fluorography by incubation in Enhance, and dried
`onto filter paper. The comers of the paper were marked with ink that contained
`[' °C]glucose before the gel was placed against film . After the fluorogram was
`developed, alignment of the radioactive ink lines allowed exact comparisons of
`radioactive spots and protein-staining spots.
`
`RESULTS
`Cultured human endothelial cells were incubated with [3H]-
`leucine, and the metabolically labeled proteins that were se-
`creted into the medium were analyzed by SDS PAGE with and
`without reduction (Fig . 3A) . Without reduction, 31% of the
`[3H]leucine migrated in a peak that had an apparent size of
`450 kdaltons. With reduction, the peak at 450 kdaltons disap-
`peared and was replaced by peaks with apparent sizes of 220
`and 160 kdaltons. The 220-kdalton peak, which accounted for
`17% of the labeled proteins, has been shown previously to be
`fibronectin (19, 33, 34). The 160-kdalton peak accounted for
`14% of the labeled proteins.
`When rabbit antithrombospondin serum was reacted with
`[3H]leucine-labeled postculture medium, antithrombospondin
`specifically immunoisolated 3H-labeled protein with apparent
`sizes of 450 kdaltons when analyzed without reduction (Fig.
`3 B) and 160 kdaltons when analyzed with reduction (Fig. 3 C).
`Purified human platelet thrombospondin, electrophoresed as a
`marker, migrated in SDS PAGE with the immunoisolated 160-
`Walton [3H]leucine-labeled peak when analyzed after reduc-
`tion and with the immunoisolated 450-kdalton [3H]leucine-
`labeled peak when analyzed without prior reduction (Fig. 3,
`arrows). No labeled proteins were specifically isolated in con-
`trol experiments with antiovalbumin (Figs. 3 B and C) or with
`preimmune serum from the antithrombospondin rabbit . Im-
`munoisolation of the 160-kdalton protein was not blocked by
`
`MOSHER ET AL .
`
`Thrombospondin Production by Endothelial Cells
`
`345
`
`100
`
`tn
`
`80
`
`60
`
`40
`
`20
`
`Z Omó
`
`°
`
`(cid:9)
`(cid:9)
`(cid:9)
`
`0.02
`0.5
`0.2
`0.1
`0.05
`FRACTION ARMED STAPH
`
`1 .0
`
`00
`
`.01
`
`FRACTION MEDIUM
`
`FIGURE 2
`Radioimmunoassay of thrombospondin . Panel A shows
`the binding of . ..1-thrombospondin to various ratios of armed and
`carrier protein A-bearing Staphylococci . Nonspecific binding to
`Staphylococci armed with normal rabbit serum was 15% (dashed
`line) . The arrow points to the ratio of armed and "unarmed"
`Staphylococci that was used in the radioimmunoassay . Panel B
`shows typical inhibition curves in the radioimmunoassay: ", purified
`human platelet thrombospondin standard ; ", supernatant from
`confluent human umbilical vein endothelial cells after 24 h in
`culture medium containing 20% rabbit serum; and A, medium after
`24 h in culture plates that contained no cells. The lower dashed line
`represents nonspecific binding to Staphylococci armed with normal
`rabbit serum and the upper dashed line represents binding to
`antibody-armed and carrier Staphylococci in the absence of unla-
`beled thrombospondin . Total 1251-thrombospondin was 1.9 x 10°
`cpm per assay.
`
`M sodium chloride, 33 mM Tris acetate, pH 8.5, and made 1.7% in Triton X-100
`(28) . The mixed micellar solution, 30 ml, was incubated with 5 lal of rabbit
`antithrombospondin, preimmunization serum from the antithrombospondin rab-
`bit, rabbit antifrbronectin (19), or rabbit antiovalbumin (19) previously coupled
`to 10 lrl of protein A-Sepharose. After overnight incubation with end-over-end
`rotation at 20°C, the beads were centrifuged, washed eight times with 2 M urea,
`0.1 M glycine, 1% Triton X-100 (28), and eluted by boiling in 2% SDS, 6 M urea.
`Immunoisolated labeled proteins were analyzed by electrophoresis in SDS in
`continuous 3% acrylamide-0 .5% agarose gels with and without reduction (19) .
`Unlabeled purified fibronectin and thrombospondin were analyzed in simulta-
`neously run gels . Unlabeled proteins were visualized by staining with Coomassie
`
`Merck Ex. 1101, Pg. 3
`
`

`

`
`
`
`
`
`
`(cid:9)(cid:9)(cid:9)(cid:9)(cid:9)(cid:9)(cid:9)
`(cid:9)
`
`6000
`
`4000
`
`E
`
`2000
`
`i
`TSP
`
`6
`
`TSP
`á
`
`EE
`
`300-
`
`loo
`
`300
`
`B
`
`C
`
`á
`° loo
`
`Analysis
`FIGURE 3
`of
`[3H]leucine-la-
`beled
`postculture
`medium and immu-
`[3H]leu-
`noisolated
`cine-labeled
`GP-
`160 . The following
`samples were ana-
`lyzed by SDS PAGE :
`A,
`unfractionated
`postculture medium
`without (solid
`line)
`or
`with
`(dashed
`line)
`prior
`reduc
`tion;
`8, labeied pro-
`tein
`that was im-
`munoisolated
`with
`antithrombospon-
`din
`(dashed
`line)
`25
`or
`antiovalbumin
`Gel slice #
`(solid line) and an-
`alyzed without reduction ; and C labeled protein isolated with
`antithrombospondin (dashed line) or antiovalbumin (solid line) and
`analyzed with reduction . Arrows point to the positions of migration
`of nonreduced (450 kdaltons) and reduced (160 kdaltons) marker
`thrombospondin . In parallel gels, we found that [3 H]fibronectin
`bound to antifibronectin and migrated at slice 10 without reduction
`and slices 14 and 15 with reduction and that there was only back-
`ground binding to preimmune serum. Size standards are described
`in Fig. 1.
`
`5
`
`15
`
`35
`
`45
`
`preadsorption of the antithrombospondin antiserum (5 ILI) with
`purified fibronectin (100 ug) .
`In the presence of purified fibronectin (0.5 mg/ml in PBS),
`[3H]leucine-labeled endothelial cell fibronectin and GP-160
`bound to gelatin-agarose and could be eluted with acetate-
`buffered 1 M NaBr, pH 5 (Fig . 4). SDS PAGE of unbound
`and bound fractions (Fig . 4) and calculations of yields of
`labeled proteins indicated that affmity chromatography on
`gelatin-agarose in the presence of unlabeled fibronectin re-
`sulted in nearly quantitative recovery of [3H]leucine-labeled
`fibronectin and [3H]leucine-labeled GP-160 . To make further
`comparisons between thrombospondin and [3H]Ieucine-labeled
`GP-160, thrombospondin was added to the mixture of unla-
`beled fibronectin, [3H]leucine-labeled endothelial cell fibronec-
`tin, and [3H]leucine-labeled endothelial cell GP-160 . The four
`proteins were then analyzed by two different two-dimensional
`electrophoretic systems : isoelectric focusing of reduced proteins
`in 8 M urea followed by SDS PAGE (Fig. 5A) and SDS PAGE
`without reduction followed by SDS PAGE after reduction
`(Fig. 5B). GP-160 was acidic (Fig. 5A, left), had an apparent
`size of 450 kdaltons in its unreduced form (Fig. 5 B, left), and
`was apparently a disulfide-bonded trimer of 160-kdalton sub-
`units (Fig. 5 B, left). GP-160 migrated at exactly the same
`positions as thrombospondin in both two-dimensional systems.
`GP-160 and thrombospondin exhibited the same limited isoe-
`lectric heterogeneity (Fig. 5 A). In contrast, labeled endothelial
`cell fibronectin was slightly larger than plasma fibronectin in
`its reduced (220 vs. 210 kdaltons) and unreduced forms and
`thus migrated slightly higher and more to the left in the
`nonreduced-reduced system (Fig. 5 B). There were also slight
`differences in isoelectric heterogeneity between endothelial cell
`and plasma fibronectins (Fig. 5A) .
`Thrombospondin antigen was first detectable by radioim-
`munoassay in the postculture medium of confluent endothelial
`cells after 8 h of culture ; after 24 h the postculture medium
`
`346
`
`THE JOURNAL OF CELL BIOLOGY " VOLUME 93, 1982
`
`Chroma-
`FIGURE 4
`tography of a mix-
`ture
`of
`purified
`plasma
`fibronectin
`and [3H]leucine-la-
`beled
`postculture
`medium on gelatin-
`agarose
`as
`moni-
`tored by SDS PAGE
`on 6% gels. Samples
`were reduced
`be-
`fore
`analysis, and
`radioactive bands
`were detected
`by
`fluorography . Lanes
`are (a) unfraction-
`ated proteins,
`(b)
`proteins not bound
`to
`gelatin-agarose,
`and
`(c)
`proteins
`to
`bound
`gelatin
`agarose and eluted
`with 1 M sodium
`bromide, pH 5.0. Of
`the many proteins in
`the serum-contain-
`ing growth medium
`that could be de-
`tected
`with
`Coo-
`massie
`Brilliant
`Blue, only fibronec-
`tin bound to gela-
`tin-agarose .
`Unla-
`beled
`fibronectin
`disturbed the band
`ing pattern of [3H]leucine-labeled fibronectin and caused the rare-
`faction on the fluorogram in the fibronectin (FN) region, The
`positions of thrombospondin ( TSP, electrophoresed as a marker)
`and of human albumin (68 kdaltons, detected by protein staining in
`lanes a and b) are indicated . Labeled proteins < 60 kdaltons in size
`migrated at the dye front.
`
`contained 49 Itg/106 cells (Table I).' The curves for inhibition
`of 'z.I-thrombospondin binding to rabbit antibodies were sim-
`ilar for purified platelet thrombospondin and endothelial cell-
`conditioned medium, although we could not construct a full
`inhibition curve for conditioned medium because ofits limited
`antigen concentration (Fig. 2B) . The value of 4.4 wg throm-
`bospondin/ml of 24-h conditioned medium is comparable to
`the value of 5.2 jLg fibronectin/ml of72-h conditioned medium
`that we determined earlier (19) .
`
`DISCUSSION
`Several different comparisons of GP-160 synthesized and se-
`creted by cultured endothelial cells and thrombospondin se-
`creted by thrombin-stimulated platelets indicate that these two
`proteins are the same . GP-160 and thrombospondin migrated
`identically in two different two-dimensional gel systems. There-
`fore, it seems likely that the differences between GP-160 and
`thrombospondin, if present, are minor.
`Since submission of this manuscript, McPherson et al. (35)
`have described the isolation and characterization of a glyco-
`
`' Recently, we developed an enzyme-linked immunoabsorbent assay
`(ELISA) for human thrombospondin based on a mouse monoclonal
`IgG antibody to thrombospondin isolated from platelets. Analyses of
`human thrombospondin in the 24-h tissue culture medium by ELISA
`agree within 40% of the values obtained by radioimmunoassay.
`
`Merck Ex. 1101, Pg. 4
`
`

`

`(cid:9)
`(cid:9)(cid:9)
`
`
`Two-dimensional analyses of fibronectin (FN), GP-160, and thrombospondin ( TSP) . The following mixture of proteins
`FIGURE 5
`was analyzed : plasma fibronectin, platelet thrombospondin, and [3HJIeucine-labeled endothelial cell fibronectin and GP-160. Two
`in 8 M urea followed by electrophoresis through 6%
`systems were used : A, isoelectric focusing (pH, acid side to the left)
`polyacrylamide in SDS after reduction (R) ; and 8, electrophoresis through 4% polyacrylamide in SIDS without reduction (NR)
`followed by electrophoresis through 6% polyacrylamide in SDS after reduction (R) . The slabs were stained for protein and dried
`onto filter paper (right) and analyzed by fluorography (left) . Apparent size in kilodaltons is indicated on the right. Standards are
`1 . The figure has been made to allow one-for-one comparisons of fluorographic and protein-staining patterns .
`described in Fig.
`We could make a more critical evaluation by overlaying the dried stained gel with the developed fluorography film.
`
`TABLE I
`Accumulation of Thrombospondin Antigen in Culture
`Medium of Human Umbilical Vein Endothelial Cells as
`Ascertained by Radioimmunoassay
`
`Culture
`h
`0
`4
`8
`14
`24
`*z±SD, n=3.
`
`Thrombospondin
`Aglml
`<0 .7
`<0 .7
`0.8 ± 0.1
`3.4±0 .8
`4.4±0.2
`
`antigen*
`jug/708
`
`cells
`
`<7
`<7
`10 ± 1
`36±6
`49±4
`
`protein from serum-free conditioned media of bovine aortic
`endothelial cells which is apparently homologous to human
`endothelial cell GP-160 . Following our lead (1), McPherson et
`al. (35) compared the glycoprotein to thrombospondin . The
`amino acid composition of the glycoprotein was similar to that
`published for human platelet thrombospondin, antibodies to
`the glycoprotein reacted with bovine thrombospondin after
`electrophoretic transfer from SDS polyacrylamide gels onto
`nitrocellulose paper, and two-dimensional peptide maps ofthe
`iodinated glycoprotein and iodinated thrombospondin were
`
`similar. Thus, thrombospondin synthesis by endothelial cells
`in culture must be a general phenomenon.
`Although thrombospondin and GP-160 are disulfide-bonded
`trimers of 160-kdalton subunits, there is no evidence that
`thrombospondin and GP-160 are collagenous . We have been
`able to degrade human endothelial cell GP-160 using purified
`bacterial collagenase (unpublished experiments carried out in
`both of our laboratories) . Similarly, Sage et al. (20) were unable
`to degrade bovine endothelial GP-160 by purified bacterial
`collagenase under conditions in which type III procollagen was
`readily degraded . By amino acid analysis, thrombospondin is
`not rich in proline or glycine (3, 10) .
`We are unsure of the determinants ofbinding of [3H]leucine-
`labeled GP-160 to gelatin-agarose in the presence of unlabeled
`fibronectin. In preliminary studies, we found that >60% of "sI-
`thrombospondin bound to gelatin-agarose, underived agarose,
`or Bio-Gel P-300. The presence of fibronectin had little influ-
`ence on this apparently nonspecific binding . Thus, while chro-
`matography on gelatin-agarose in the presence of fibronectin
`allowed us to partially purify GP-160, a better experimental
`system is needed to allow investigation of specific binding.
`We can only speculate on the function(s) of endothelial cell
`thrombospondin . As described above, thrombospondin is se-
`creted and bound by thrombin-activated platelets (2-9) and
`347
`
`Thrombospondin Production by Endothelial Cells
`
`MOSHER ET AE.
`
`Merck Ex. 1101, Pg. 5
`
`

`

`(cid:9)
`
`apparently mediates thrombin-induced aggregation (12-17).
`Two proteins, von Willebrand factor and fibronectin, are also
`found in platelet a-granules (36-39), bind to the surface of
`thrombin-stimulated platelets (40-42), and are synthesized by
`endothelial cells (19, 33, 34, 43,44). Von Willebrand factor and
`fibronectin are found in plasma and in the subendothelium
`(45, 46). Von Willebrand factor mediates platelet adhesion to
`subendothelium (47, 48) whereas the role of fibronectin in
`platelet function is unknown. Although plasma thrombospon-
`din levels as measured by radioimmunoassay are quite low (- 1
`lug/ml, our unpublished data), endothelial cells may, when
`appropriately stimulated, release thrombospondin into local
`microenvironments and thus support platelet interactions with
`endothelial cells and the subendothelium. Thrombospondin,
`because of its lectin activity, may also play important roles in
`the interaction of endothelial cells with each other and with
`the underlying extracellular matrix. Experiments to test these
`hypotheses are currently underway.
`
`We thank Douglas Armellino, Linda Griese, and Peter Schad for their
`excellent technical assistance.
`This work was supported by grants from the National Institutes of
`Health (HL-18828, HL-21644, and HL-24885). It was done during the
`tenure of an Established Investigatorship from the American Heart
`Association and its Wisconsin Affiliate to Deane F. Mosher and a
`National Institutes ofHealth Career DevelopmentAward and a Career
`Scientist Award from the Irma T. Hirschl Trust to Eric A. JafPe.
`
`Received for publication 26 October 1981, and in revised form 15
`December 1981 .
`
`REFERENCES
`
`1 . Doyle, M . J ., D . F. Mosher, and E . A. Jaffe. 1980 . Endothelia l cells synthesize and secrete
`thrombospondin . J. Cell Biol. 87 (2, Pt. 2) : 306 a (Abstr .).
`2 . Baenziger, N . L., G.N. Brodie, and P. W. Majerus. 1971 . A thrombin-sensitive protein of
`human platelet membranes . Proc. Nail. Acad. Sci. U. S. A . 68 :240-243 .
`3 . Baenziger, N. L., G. N . Brodie, and P . W . Majerus. 1972 . Isolation and properties of a
`thrombin-sensitive protein of human platelets . J. Biol. Chem. 247 :2723-2731 .
`4. George, l . N . 1978. Studies on platelet plasma membranes. Quantitative analysis of
`platelet membrane glycoproteins by ('"I)-diazotized diiodosulfanilic acid labeling and
`SDS-polyacrylamide gel electrophoresis. J. Lab. Clin. Med. 92:430-446 .
`5. Hagen, I. 1975 . Effects of thrombin on washed, human platelets: changes in subcellular
`fractions. Biochim . Biophys. Asia. 392 :242-254 .
`6 . Hagen, I ., T. Olsen, and N . O. Solum. 1976 . Studies on subcellular fractions of human
`platelets by the lactoperoxidase-iodination technique . Biochim. Biophys. Acta . 455 :214-
`225.
`7 . Lawler, J . W., F . C. Chao, and P .-H. Fang . 1977 . Observation of a high molecular weight
`platelet protein released by thrombin . Thromb. Haemostasis 37 :355-357 .
`8 . George, J. N., R . M. Lyons, and R. K. Morgan . 1980. Membran e changes associated with
`platelet activation. Exposure of actin on the platelet surface after thrombin-induced
`secretion . J. Clin. Invest. 66 :1-9.
`9 . Phillips, D. R ., L. K . Jennings, and H . R . Prasanna . 1980. Ca'*-mediated association of
`glycoprotein G (thrombin-sensitive protein, thrombospondin) with human platelets . J.
`Biol Chem. 255 :11629-11632 .
`10 . Lawler, J . W., H. S. Slayter, and J. E . Coligan. 1978. Isolatio n and characterization of a
`high molecular weight glycoprotein from human blood platelets. J. Biol. Chem. 253:8609-
`8616 .
`11 . Margossian, S . S., J . W . Lawler, and H . S. Slayter. 1981 . Physica l characterization of
`platelet thrombospondin. J. Biol. Chem. 256 :7495-7500.
`l2 . Gartner, T. K., D . C. Williams, and D. R . Phillips. 1977 . Platelet plasma membrane lestin
`activity. Biochem. Biophys Res. Commun . 79:592-599.
`13 . Gartner, T. K ., D. C . Williams, F. C . Minion, and D. R . Phillips . 1978. Thrombin-induced
`platelet aggregation is mediated by a platelet plasma membrane-bound lestin. Science
`(Wash . D . C.) . 200:1281-1283 .
`14 . Gartner, T. K ., D. R. Phillips, and D . C . Williams. 1980. Expression of thrombin-enhanced
`platelet lestin activity is controlled by secretion . FEBS (Fed. Eur. Biochem . Soc.) Letr.
`113 :196-200.
`
`15. Gartner, T. K ., J . M . Gerrard, J . G . White, and D. C. Williams. 1981 . Fibrinogen is the
`receptor for the endogenous lestin of human platelets . Nature (Lond) . 289 :688-700.
`16. Gartner, T. K ., J . M. Gerrard, J. G. White, and D. C. Williams . 1981 . The endogenous
`lestin of human platelets in an a-granule component. Blood. 58 :153-157 .
`17. Jaffe, E. A ., L. L. K ., Leung, R. L . Nachman, R . I. Levin, and D. F. Mosher . 1981 .
`Thrombospondin is the endogenous lestin of human platelets . Nature (Lond). In press .
`18. Gerrard, J . M ., D . R. Phillips, G . H . R. Rao, E. F. Plow, D. A . Wak, R.