Characterization of a thrombomodulin cDNA reveals structural similarity to the lowdensity lipoprotein receptor. (protein C/coated-pit receptors/endothelial cell).
Proc. Natl. Acad. Sci. USA Vol. 83, pp. 8834-8838, December 1986 Biochemistry
Characterization of a thrombomodulin cDNA reveals structural similarity to the low density lipoprotein receptor (protein C/coated-pit receptors/endothelial cell)
ROBERT W. JACKMAN*, DAVID L. BEELER*, LIVINGSTON VANDEWATERt, AND ROBERT D. ROSENBERG*: *Department of Biology and Whitaker College, Massachusetts Institute of Technology, Cambridge, MA 02139; and the Departments of tPathology and
*Medicine, Harvard Medical School, and the Beth Israel Hospital, Boston, MA 02215 Communicated by Emilio Bizzi, July 25, 1986
Preparation of Antisera. Two male New Zealand White rabbits were each inoculated by subcutaneous injection with 30 /xg of thrombomodulin in complete Freund's adjuvant and given two booster injections at 3-week intervals with 30 ;kg of thrombomodulin in incomplete Freund's adjuvant. The antisera were screened with an enzyme-linked immunosorbant assay (ELISA) for thrombomodulin and also were shown to inhibit >95% of the biologic activity of the endothelial cell receptor. IgG antibodies were isolated on protein A-Sepharose (7) and further purified on a column of Bio-Rad Affi-Gel 15 to which thrombomodulin had been coupled according to the manufacturer's instructions, using 2 mg of thrombomodulin per 2 ml of matrix. The IgG fraction was filtered through the affinity column, and specific antibodies were eluted with 0.2 M glycine/HCl (pH 3.0). A mouse [(C57B6 x BALB/c)F1, The Jackson Laboratory] was also inoculated with 20 ,fig of thrombomodulin in complete Freund's adjuvant and then boosted on 3 consecutive days with, intravenous injections of 10 ,ug of the protein. Spleen cells were then obtained, and fusions were performed by the method of Kennett et al. (8). Supernatant solutions from growth-positive wells were screened against thrombomodulin by ELISA, antithrombomodulin-secreting hybridomas were cloned in agarose (8), and ascites fluid was produced in mice. Construction of the Xgtll cDNA Library. Primary bovine adrenal capillary endothelial (BACE) cells were isolated from bovine adrenal glands, grown in tissue culture (10), and shown to possess high levels of thrombomodulin activity (data not shown). Polyadenylylated RNA was extracted from these cells (9), and a cDNA library was constructed according to the method of Ginsberg et al. (11). The average length of 10 randomly selected cDNA clones ranged from 1 to 3 kilobases (kb). Computer Analysis. Computer analyses of sequence data were performed using the program IALIGN (National Biomedical Research Foundation, Washington, DC) as modified by W. Gilbert at the Whitaker College Computer Facility of the Massachusetts Institute of Technology.
ABSTRACT We have isolated a partial-length cDNA for bovine thrombomodulin from a Agtll bovine adrenal capillary endothelial cell expression library. This was accomplished by immunoscreening with rabbit anti-thrombomodulin IgG heteroantibody and then rescreening with the initial positive recombinant insert. The cDNA obtained was authenticated by showing that it coded for the primary structure of two separate regions of bovine thrombomodulin. The nucleotide sequence of the largest cDNA allowed us to establish the structure of about 80% of the mature thrombomodulin transcript, which encodes the C-terminal half of the polypeptide. This membrane component is structurally similar to coated-pit receptors and is organized into domains that resemble those of the low density lipoprotein receptor.
Thrombomodulin is a specific endothelial cell receptor that forms a 1:1 stoichiometric complex with thrombin (1). This complex is responsible for the conversion of protein C to activated protein C (protein Ca). Once evolved, protein Ca scissions the activated cofactors of the coagulation mechanism, factor Va and factor VIIIa, and thereby dramatically reduces the amount of thrombin generated (1). Thrombomodulin has been isolated in a homogeneous state from the lungs and placenta of rabbits, cows, and humans (1-3). Immunohistochemical examination of tissue sections reveal that this membrane protein is restricted to the luminal surface of blood vessels, and detailed investigations of cultured cells show that endothelial cells are unique in synthesizing this component (4, 5). However, essentially nothing is known about the molecular structure of this endothelial cell receptor except that it possesses large numbers of proline residues (1). In this communication, we report the cloning and sequencing of a cDNA coding for the C-terminal half of bovine thrombomodulin. The nucleotide sequence allows us to provide a partial picture of the overall structure of this receptor and to surmise how the protein might be oriented across the endothelial cell membrane. Furthermore, the deduced amino acid sequence ofthrombomodulin reveals that this membrane component is structurally similar to coated-pit receptors and is organized into domains that resemble those of the low density lipoprotein (LDL) receptor.
RESULTS
Purification of Bovine Thrombomodulin. The endothelial cell receptor was purified from bovine calf lungs by diisopropylphosphorylthrombin-Sepharose affinity chromatography according to the method of Esmon et al. (1), with minor modifications. The amidolytic activity of purified thrombomodulin was determined as described (6).
Isolation of Bovine Thrombomodulin and Production of Specific Antisera. Bovine thrombomodulin was isolated from calf lungs as outlined in Experimental Procedures. The final product was homogeneous as judged by NaDodSO4/PAGE in the presence of 2-mercaptoethanol, with an apparent molecular weight of 100,000 (Fig. 1, lane 1). Rabbit affinitypurified heterologous anti-thrombomodulin antibodies and mouse monoclonal anti-thrombomodulin antibodies were used for immunoblot analyses of thrombomodulin preparations. The antibodies identified a single band (Fig. 1, lanes 2
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Abbreviations: BACE cell, bovine adrenal capillary endothelial cell; EGF, epidermal growth factor; LDL, low density lipoprotein; bp, base pair(s); kb, kilobase(s).
EXPERIMENTAL PROCEDURES
8834
Biochemistry: Jackman et al. 1
9768-
_-
2
3
4
5
Proc. Natl. Acad. Sci. USA 83 (1986) 6
7
1
2
3
4
56 7
-
+
-
+
+
+
8835
+
-
45 _ 200
16
--
--
1X-11614
97.-.
FIG. 1. NaDodSO4/PAGE and immunoblotting of thrombomodulin and its CNBr fragments. Lane 1: thrombomodulin stained with silver (Bio-Rad). Low M, degradation products are visible at the bottom of the gel. Lanes 2 and 3: immunoblots probed with affinity-purified rabbit antibody (4) and mouse monoclonal antibody 1-4 (5). Lanes 4 and 5: immunoblots of DEAE HPLC-purified CNBr fragment probed with affinity-purified rabbit antibody (4) and mouse monoclonal antibody 1-4 (5). Lanes 6 and 7: autoradiograms of lanes 4 and 5. NaDodSO4/PAGE was conducted according to the method of Laemmli (12), employing 10% gels with added 8 M urea. The amount of thrombomodulin loaded averaged -0.1 /xg per lane. Immunoblots were visualized with the Proto-blot system of Promega Biotec, Madison, WI. Positions and molecular weights (Mr x 10-3) of standards are shown at left.
and 3) that comigrated with the band visualized by silver staining (lane 1). The affinity-purified rabbit heteroantibody precipitated a single component of Mr 100,000 from [35S]methionine-labeled bovine aortic endothelial cells and detected a single band on immunoblots of bovine aortic endothelial cell total protein (data not shown). Further, the affinity-purified rabbit heteroantibody, as well as two anti-thrombomodulin monoclonal antibodies, were bound to Affi-Gel 10, crude preparations of thrombomodulin were filtered through the affinity matrices, and a component with the biologic function of the endothelial cell receptor was eluted with 0.2 M glycine/HCl at pH 3.0 (data not shown). From the above results, we conclude that the rabbit IgG heteroantibody and the mouse monoclonal IgG antibodies are specifically directed against bovine thrombomodulin. Isolation and Identification of a Bovine Thrombomodulin cDNA Clone. Approximately 1.8 x 106 recombinant clones from the BACE cell Xgtll library were screened with the affinity-purified rabbit anti-thrombomodulin heteroantibody by the method of Young and Davis (13). One candidate positive plaque (Xgtll-TM1) was detected and plaque-purified. The fusion protein produced by the Xgtll-TM1 clone was further characterized by preparing lysogens in Escherichia coli host strain Y1089 as outlined by Schwarzbauer et al. (14). Cell lysates were examined by NaDodSO4/PAGE and immunoblot analysis. It is apparent that the synthesis of ,l-galactosidase and the fusion protein are dependent on the presence of isopropyl ,-D-thiogalactoside (Fig. 2, lanes 1-4). The apparent molecular weight of the fusion protein is about 15,000 higher than that of l-galactosidase, which corresponds to an insert of about 380 base pairs (bp). Immunoblotting of cell lysates with affinity-purified anti-thrombomodulin heteroantibody showed that antigenic determinants of the endothelial cell receptor were present within the fusion protein. The existence of these epitopes was confirmed with monoclonal antibodies 1-4 and 25-1 (Fig. 2, lanes 5-7). Of
FIG. 2. Analysis of the fusion polypeptide from Xgtll-TM1. E. coli Y1089 lysogenic for Xgtll or Xgtll-TM1 were either not induced (-) or induced by addition of isopropyl ,8-D-thiogalactoside (+). Cell lysates were prepared, and saturated ammonium sulfate solution was added to 33% (vol/vol). The resulting precipitates were analyzed by NaDodSO4/7% PAGE ('20 ,ug of protein per lane). (Left) Proteins were visualized with Coomassie blue R-250. Lanes 1 and 2: Xgtll (note 8-galactosidase of M, 116,000 in lane 2). Lanes 3 and 4: Xgtll-TM1. (Right) Immunoblot analysis of Xgtll-TM1 samples identical to that in lane 4. Proteins were transferred to nitrocellulose, incubated with anti-thrombomodulin antibodies as indicated below, and visualized by indirect horseradish peroxidase staining using secondary antibodies from Cooper Biochemicals (Malvern, PA). Lane 5: rabbit heteroantibody. Lane 6: mouse monoclonal antibody 1-4. Lane 7: mouse monoclonal antibody 21-5. seven additional monoclonal IgG antibodies directed against thrombomodulin, five recognized the fusion protein, whereas two did not (data not shown). Authentication of Xgtll-TM1. A single plaque of XgtllTM1 was amplified and employed for large-scale preparation of phage. The phage DNA was isolated, inserts were obtained by EcoRI digestion, fragments were electrophoresed in 1.0% low-melting-point agarose gels, and the inserts were recovered by elution. The insert fragments were subcloned into phage vectors M13 mp8 and mp9 and the nucleotide sequence of the 378-bp cDNA was established by the dideoxy sequencing method of Biggin et al. (15). The data showed that the Xgtll-TM1 insert exhibits a single open reading frame encoding 123 amino acids without a stop codon and that the fragment possesses the same reading frame as the 8-galactosidase gene, which is consistent with the expression of a fusion protein. To unequivocally demonstrate that the Xgtll-TM1 insert encodes a region of thrombomodulin, we next obtained amino acid sequence data from the corresponding segment of the protein. Thrombomodulin was reduced with dithiothreitol, alkylated with iodo[14C]acetamide, and cleaved with cyanogen bromide in 70% (vol/vol) formic acid. The final product was lyophilized and then suspended in and dialyzed against 10 mM potassium phosphate at pH 7.1 (buffer A). The radiolabeled thrombomodulin fragment was isolated by HPLC on a TSK 5PW DEAE column (75 mm x 7.5 mm) equilibrated with buffer A with 0.05% Triton X-100 added. Bound components were eluted at a flow rate of 1.0 ml/min with a 30-min linear gradient of 0.2-0.7 M NaCl in buffer A.
8836
Biochemistry: Jackman et al.
Proc. Natl. Acad. Sci. USA 83
The major radiolabeled species emerged as a sharp peak centered at 0.35 M NaCl. NaDodSO4/PAGE and immunoblot analysis with the affinity-purified IgG heteroantibody as well as the monoclonal antibody I-4 showed that the radiolabeled thrombomodulin fragment exhibits a molecular weight of about 45,000 and retains the Xgtll-TM1 epitope (Fig. 1, lanes 4-7). Prior to and immediately after treatment with trifluoromethanesulfonic acid (16), the fragment was characterized by HPLC on a TSK G3000 column, where it migrated as a sharp peak with an apparent molecular weight of 45,000 and 19,000, respectively. The labeled thrombomodulin fragment was analyzed by vapor-phase amino acid sequencing according to the method of Hewick et al. (17), using an Applied Biosystems 470A sequenator. The derivatized residues were also collected for scintillation counting of radiolabeled cysteine moieties. Structural studies, carried out on three separate occasions, established the presence of a major sequence and a minor sequence that matched almost perfectly with two regions encoded by the Xgtll-TM1 insert (see boxed regions in Fig. 3). The only observed difference was the replacement of an asparagine residue at position 547 in the deduced sequence with an aspartic acid residue in the protein sequence. This alteration may have been caused by treatment of thrombomodulin with formic acid during cleavage with CNBr. Nucleotide Sequence of a Large cDNA That Encodes a Major Segment of Thrombomodulin. The Xgtll-TM1 insert was labeled by nick-translation and then used to screen 106 recombinants by the plaque-hybridization technique (18). Positive phage were detected with a frequency of 0.001%. 1
COCOOCOCACOAOOOOAGACCOAGOOCCOCTOOACCACCOAAOCCCCOOCC OCCTOCOC R
121
241
361 461
OA R G
E
T
E
G
R
W
S
R
E
A AP
O
A
*
A
(1986)
The 10 recombinant clones were plaque-purified and grown in large scale, DNA was isolated and digested with EcoRI to obtain inserts, and the various fragments were examined by electrophoresis in 1.0% agarose. Nine of the phage DNA inserts each contained three classes of fragments with sizes of 1.3-1.6 kb, 0.6-0.75 kb, and 0.2 kb, respectively. The tenth phage possessed an insert with a single fragment of -0.8 kb. Xgtll-TM2, the phage with the greatest total length of insert, was further characterized by subcloning each of its three classes of fragments into M13 mp8 and mp9 and determining 5'- as well as 3'-terminal sequences of each species. The data revealed that only the 5' end of the 1.6-kb fragment and the 3' end of the 0.7-kb fragment possessed synthetic linker sequence. Correspondence of the 1.6-kb and 0.7-kb insert fragments of Xgtll-TM2 to the large and intermediate-size insert fragments of the other eight phage clones was verified by Southern blot analyses with probes constructed from the M13 clones of Xgtll-TM2. Thus, the three fragments isolated from Xgtll-TM2 originate from a single insert, as judged by their occurrence in nine independent recombinant phage and the presence of synthetic linker sequences on the appropriate ends of the fragments. Subsequent examination of the tenth phage, with a solitary insert fragment, demonstrated that this species was contained within the primary structure of Xgtll-TM2. The nucleotide sequence of Xgtll-TM2 is provided in Fig. 3. This sequence was confirmed by determining the sequence of two independent recombinant clones with inserts of similar size. The orientation of the 214-bp internal EcoRI fragment of the cDNA was elucidated by Southern blot analyses of the COCOTCOAGCOC OCOc v
E
R
°
° W
CCAOCACOAZ^pCAAtOOcAOTOCTOGOOCTTCOAAC Q
H
E
ZOCOCGCTOCCOOCCOAOCACCCOOCCAACCAACT LOCOACCACTT gOCTTQZ.TCCOOCCOACOCCOCCCTOCAOOCOOACOOACOCTC E H F L 9P A D A A L q A D O R S 9 0 L P A E H PL,
C
K
a
S A O A S
120
N
OCCACCTCCACOOOCTCOcCAACTAC HN L H O L O N Y
240
OCCOTCAACACOGACOOCOOC
50
AC99CATCICOAOOCAC OCTACCAOCTaCCOCCOACCACCCAOC CaOTOAGOACOTCOACOA .COCCGOCAOTTOCCCAaTCC CCCGCAAC A q L E A Y Q L A A D q H R R L P S T ( E D V D D9 I (
QC)
V
N T
E
G
C
TaC ACC S ACC TTTOACAA'TAA5kCAOTACCA ZCCAOCCOaTOGOCCOOTCCOAACAC TTCCAn5OCCAWCaACACTOOCTAO^aTTaOTOOACOaCOAI Q P V G R S E H F H kg D T C Y I = D P WF D N N QC E Y Q
460
AAp9CATTCOCCOCAOOCTTCOCOCCAaTCCCCOOTOCCCCOCACAAYCCAGA 9 * E F A P V P 0 A P H K Q mI
600
K
T
S
Pr A
TTCAC TACCCOACCATTccCCoc D N Y P T I D tC) 9R
720
CCATAACATCOACCCTC TCTACCCAATToccATCaA TcACCCACC A"GCGGCCACCCTGCOA¢CACTATAOCCCTCaGGACCACC LS C QI 0 I D D P T q V N E E R O T P E D Y G S G E P P D S S4 1 GCOCCCACC OCACCCCCC TCOCCCCCACCTocTGGTCCCCTaCACTC~caaCOTCCTCGTGGGCATC TCCATTOCAAGC CTOTCTCTCCTOOTOaCCCTTCT4CGCCC TCCTMCCCACCTO
621
(9
P
0
O OCCC ACTCCO V S P T P
A
T
A
R
P
S
P
A
P
A
G
P
L
H
640 560
S
961
C OCAAGAAGCACGGCOCCTCGCCOGO¢¢^ACTCCAaTACAAI]OCOCaGOTCCCCOCCS"AGGCTCATGCTCCACCAaGTaAA"CTOAC COACACCTCA¢AAACTCTA"ZACCCTCC R K K q G A S R C E L E Y K C V P A K E L M L Q Q V K T E R T P q K L
1060
10OS1
TATCCA"CTTCCTTCCCTCCTCCCGCCCCCOCAGCCCTTt TC TCCCCATCCCCTGCCCCCAACCACCTTAOCTG¢CCTCA"ATAGA¢ACCCCTCGCC CC TCTTC TTTCACTGCTTCTO CACTTGCCTATCGAGMCAGACGAGOCOCC TTaOOOTTTOOCCCC AaCCACTTCCATATaTCATCAGACAA^GCGOcACGOCATOACAOCTTcTtCATTTACAG~cuOTCOAATTCCT aTOCsCTCACCAAGOaOCTacOGTOGGOOGOOTOTOTOtTcTCTTOTToGCAACCTcOOGCAGACATCATTC CAOCaGACTOATCAOTGACACTCATCAAG"ATOAToCsCCCTAT
1200 1820 1440 1560
1201 1121
(W
TTAtTTTTtTTOTTCTICTOTTTTTTCeCTACeCTaOTOT4TOTC CAACACCaC C TTCCTCCC CTCCCCTCTtCCCTCCTGCTCTATTTATGTGTGTGCTCCCATTTGTTATOO"TCC TTA CACTATCTTOOTOAAGTTTTTC CAOCCCTCTTOTOTTTATCA^aCTAAOCCCCATTTCATCC TTTTCATTTCTC CCAACTTTCCCCTTCCA¢GAACYCGOCTOACCTAC C TGGC CACCC 1041 TOcTCTtATTTaGA CCCTATTTTTCToctcTTAOCToTCTocTTGoocACCACToc -TTTGTGOMACAO^^AAAMAAACACTACAAATAAA¢AAC OaTTGAGATCTITTCTTTCACCAO 1 SO1 AGTToCTATfTTATCCToAGATTTCAaaCTTCCAGAOCAATaAcACTTTTAAGAAAooTT^AAATATAAAGoocATTTATTAoACTCATaTTocTcGAC~cTO AtTTCAACeCAA
1441
1s6 1
1921
II.
. . .
.
Il I I I 15 il I^|1L l LI .. IU;.IUI lUwl . . UAAA.lw AI'w IAA. Iwl U^l
II I l IR l'-'-I*I* v I|
IR
OTTATTTOTOAAAATATTOMAATGOTTOCTCTOAATTOAOAOAOATOOAAOCCCATCCCAOOAOACAOTOTAAOAAAOTTTCAOCTOCA
1660 1600 1o20
3040
ATTCACOCCAGTOOCCATTTGOACTCT CATGGTCCTTOTCATTTOTGOOCCCAAATAAAACCAGCTCTACTGGTTTTTTGATTTOCAOCTTOoaQAT4aAA~CTGCACCATCOCCOOTTAGOOCCTCTTCTTATCAeaCeCCTCC 2261 AGOATTTCTC CCACATTOGAGGAAGAOTTTTCT"GCCCAGCCATTTOGAATOCAOTCCA44^AGTOGGAOCTOC CCCGOCCAACTOTTA -AATaC CCTCAGOCOCCATCC CACC TAC 2401 TOCACCAGATCTACATTTTAACAAGTTCAOCAQ4aQOCToTCTOCTCCOTATATTTGAGG^QAGCTGTCCAGACAGCTOcCMCAATTC TaGAACAGATTAAATATQAGTtAATATAG
2400 2620
2721 C26OCCA70TCAOCCCTATATTCAA70ACTOAATTAACTOAO"ATTTTCCTTTTATAOCTTTOCTTTCT7OTA0AGAGCCTAGOTACACACC0TA"CATTOCTACACAO
2640
TCTTOOTTCAAGTAACTAGAvAT"AATOCTTAGTGTATOGMAAAATOTATCATOTAAATOTATCTTTTOTATOTC~AAGTTTTC CTCTATTOTTTTOTAAACTCAOCACATTTG A 2970 2T6 1 TACATAOTTATTTATTTAtTT44"TAAACTAGAACAC4OGCAAAACTC CTGCTTATOATOTCACTTOTC TAA CCTAATACQTCTC TTOGAAAAAZA^AAAA
2t70
2041
2161
2641
2160
2260
FIG. 3. Partial nucleotide sequence of thrombomodulin cDNA. The structure is divided into three EcoRI fragments of Xgtll-TM2 obtained from the first library (designated by arrowheads at the cut sites) and the 5' sequence generated from the second library. The EcoRI fragments were further cleaved into a total of 40 fragments by digestion with Hae III, Sau3AI, and Alu I, cloned into phage vectors M13 mp8 and mp9, and sequenced by the dideoxy method (15). Sequence was verified by sequencing two additional independent clones, using four primers (17-18 nucleotides long) synthesized to match segments of the original sequence of Xgtll-TM2. The primers were synthesized by use of ,8-cyanoethylphosphoramidites on a Biosearch model 8600 DNA synthesizer. The regions corresponding to the two CNBr polypeptide fragments are boxed, with the ambiguous asparagine marked with an asterisk (see text). The cysteine residues are circled, and the regions of repeats are numbered at the first cysteine of each repeat. Two potential N-linked glycosylation sites are underlined with broken lines. The probable transmembrane region is underlined with a solid line. A polyadenylylation signal is marked with a bracket.
Biochemistry: Jackman et al. products of a Pvu II digestion of Xgtll-TM2, using a labeled probe derived from the 0.75-kb EcoRI fragment. In an attempt to determine the rest of the 5' sequence of thrombomodulin message, a second Xgtll BACE cell cDNA library was constructed, with 1.25 x 107 independent recombinant phage, and screened with a 160-bp probe obtained from the 5' end of Xgtll-TM2. Positive phage were now detected at a frequency of about 0.00015%. Determination of the structure of the appropriate insert fragments contributed an additional 309 bp of 5'-end sequence (Fig. 3). The extent to which the largest cDNA clone spans the length of the thrombomodulin transcript was established by carrying out RNA blot analysis (9) of poly(A)+ RNA from BACE cells, using the 1.6-kb insert fragment of Xgtll-TM2 as probe. The results revealed a single predominant RNA species, of about 3.6 kb. These data suggest that the sequence of about 2.8 kb of the estimated 3.6 kb of the mature message has been delineated. The cDNA sequence has a stop codon at position 1069, which indicates that only about 1 kb of the DNA fragment represents the coding region of thrombomodulin, whereas 1.8 kb of the DNA constitutes a 3' untranslated domain. Thus, about 0.8 kb of sequence, which encodes the N-terminal region as well as any potential leader sequence and 5' untranslated domain, appears to be absent. Examination of our sequence data reveals an extremely hydrophobic domain of 24 residues encoded by nucleotides 889-960 (Fig. 3). Hydropathy plots, obtained by a modification (19) of the method of Kyte and Doolittle (20) substantiate the hydrophobic nature of this region. Therefore, this segment probably represents a transmembrane domain, a feature frequently observed in membrane receptors (21). This notion is reinforced by the occurrence of three positively charged residues (encoded by nucleotides 961-969), which frequently mark the cytoplasmic boundary of the transmembrane region of membrane proteins (21). The orientation of thrombomodulin is suggested by the binding of the monoclonal antibody 1-4 to an epitope present on the external surface of endothelial cells (data not shown). This observation implies that the segment of the endothelial cell receptor encoded by nucleotides 1-888 constitutes a portion of the extracellular domain of thrombomodulin, provided no additional transmembrane regions exist. By exclusion, the C-terminal region encoded at 961-1068 probably represents a short cytoplasmic tail of the endothelial cell receptor. The extracellular domain exhibits a distal segment, encoded at 61-759, of six repeated blocks of 32-36 amino acids each containing 6 cysteine residues (Fig. 3). The IALIGN program, utilizing the mutation data matrix of Dayhoff et al. (22), locates the above blocks and classifies them as similar to epidermal growth factor (EGF)-precursor type B repeats (terminology of Doolittle et al.) (23). The extracellular domain also possesses a proximal segment of 56 amino acids (nucleotides 721-888) with 10 serine or threonine residues as well as a high density (11/56) of proline residues.
DISCUSSION We have isolated homogeneous preparations of bovine thrombomodulin from calf lungs. The endothelial cell receptor consists of a single polypeptide chain, which exhibits an apparent molecular weight of about 100,000 as judged by NaDodSO4/PAGE with reducing agents added. This estimate is slightly higher than that obtained by other investigators for bovine thrombomodulin (2) but is similar to that reported for rabbit and human thrombomodulin (1, 3). The rabbit and human forms of the protein appear to be highly glycosylated, since the molecular weight of the polypeptide chain of the two endothelial cell receptors is about 75,000 as calculated from the amino acid analyses (1, 3). Similar estimates have been obtained for the bovine endothelial cell receptor -(data not shown). We used the purified bovine
Proc. Natl. Acad. Sci. USA 83 (1986)
8837
thrombomodulin to generate affinity-purified rabbit heteroantibody and mouse monoclonal antibodies. The rabbit antibody was used to screen a Xgtll BACE cell cDNA expression library, and we isolated a recombinant clone (Xgtll-TM1) that encoded a fusion protein with epitope-s recognized by both types of immunologic probes. The majority of the monoclonal antibodies directed against thrombomodulin bind to the same domain of the endothelial cell receptor. Thus, it would appear that this segment constitutes an immunodominant region. The overall strategy employed to authenticate Xgtll-TM1 was based upon the observation that the recombinant insert coded for a portion of the endothelial cell receptor that contained large numbers of cysteine residues and interacted with our immunologic probes. Therefore, we isolated a CNBr fragment of thrombomodulin that exhibited both of these characteristics, and we determined a portion of the primary structure of this fragment. Structural studies of the Nterminal region of this polypeptide revealed a major and a minor sequence that corresponded to two separate domains encoded by the recombinant insert. The major sequence lies at some distance from a methionine residue and must have been generated by proteolysis prior to the addition of CNBr, whereas the minor sequence starts immediately after a methionine residue. Given that the major sequence has been observed with several different preparations of thrombomodulin, we suspect that removal of protease inhibitors prior to reduction and alkylation of the protein may have allowed trace amounts of a contaminating, chymotrypsin-like enzyme to cleave the endothelial cell receptor before admixture of cyanogen bromide. The Xgtll-TM1 insert was used to rescreen our Xgtll BACE cell cDNA library, and recombinant clones with inserts of about 2.6 kb were obtained at a frequency of about 0.001%. Probes were then prepared from the 5' end of these inserts and used to screen a second, larger Xgtll BACE cell cDNA library, and recombinant clones with an additional 0.3 kb of sequence were identified, but at a frequency of only about 0.00015%. The nucleotide sequence of these recombinant clones showed that we had delineated the structure of about 1.0 kb of the coding region of thrombomodulin message, as well as 1.8 kb of the 3' untranslated domain. We estimate that about 0.8 kb of sequence, which must encode the N-terminal region of thrombomodulin as well as any potential leader sequence and 5' untranslated domain, is absent from our data. This estimate is in reasonable accord with the approximate molecular size of the protein and suggests that the length of 5' untranslated domain is small. It appears likely that either the relative scarcity or the secondary structure of the transcript makes it difficult to isolate a cDNA that encodes this region of thrombomodulin. The data provide an overall picture of the C-terminal half of bovine thrombomodulin and reveal that it is extraordinarily similar to the LDL receptor. Both thrombomodulin and the LDL receptor exhibit short cytoplasmic tails (39 or 50 amino acid residues, respectively), with a cysteine residue located in the center of this region. However, the C terminus of the above domain of the LDL receptor is negatively charged, whereas that of thrombomodulin is neutral. Other coated-pit receptors, such as the insulin receptor or the EGF receptor, have very large cytoplasmic regions with a complex tyrosine kinase segment as well as multiple sites for phosphorylation. Both thrombomodulin and the LDL receptor possess a transmembrane region of about 24 amino acid residues and an immediately adjacent, extracellular, serine/ threonine-rich region of about 60 residues. The domain of the LDL receptor has 18 serine/threonine moieties, which bear 0-linked sugar chains; in thrombomodulin, the domain has 10 serine/threonine residues. Cleavage of the thrombomodulin CNBr fragment with trifluoromethanesulfonic acid suggests
8838
Biochemistry: Jackman et al.
that this region or the two potential sites of the N-linked glycosylation possess carbohydrate chains. Both thrombomodulin and the LDL receptor contain a more distal area of cysteine-rich repeats, termed type B, which are similar to those first noted in the EGF precursor (23). However, the cysteine-rich domain of the LDL receptor is comprised of two type B repeats with an associated 270 amino acid spacer coupled to a third single type B repeat, which appears to have been removed en bloc from EGF precursor or a common precursor (24), whereas this region of thrombomodulin is constructed of six type B repeats with no spacer segment. The N-terminal half of the LDL receptor contains the ligand-binding region of the receptor, which is constructed from multiple cysteine-rich repeats similar to those of complement factor C9 (24). The structure of this region of thrombomodulin is not known. However, from the amino acid analysis of bovine thrombomodulin (data not shown), we conclude that the N-terminal half ofthe protein is particularly rich in lysine, methionine, and phenylalanine residues but possesses few, if any, cysteine groups. Thus, the structure of this domain of thrombomodulin is likely to be quite different from the similarly placed region of the LDL receptor. We suspect that protein C and/or thrombin may bind to this domain of the endothelial cell receptor. The structural similarity of thrombomodulin and the LDL receptor may be of great importance with respect to the possible function of the endothelial cell receptor. The LDL receptor is one of the best-studied membrane receptors involved in the endocytosis and internalization of a specific ligand required for cell growth (24). However, there have been conflicting claims about the endocytosis of thrombin by the endothelial cell receptor. Maruyama and Majerus (25) have reported that both thrombin and its receptor, thrombomodulin, are internalized by various cultured cells and that the receptor is then recycled back to the surface. Evidence against the endocytosis of thrombin and thrombomodulin has been provided by Esmon and Owen (5), whose studies utilized both a Langendorff heart-perfusion system and cultured endothelial cells. The strong resemblance between thrombomodulin and the LDL receptor with regard to those regions thought to be crucial for receptor-mediated endocytosis supports the idea that the endothelial cell receptor may be involved in the internalization of thrombin. However, it is possible that a specific set of molecular events, other than the presence of thrombin, regulates the endocytosis of thrombomodulin and its ligands. We also note that the internalization and degradation of thrombin by an endothelial cell receptor-dependent mechanism could play a role in the growth of endothelial cells, since peptides derived from the enzyme are quite mitogenic (26). The eventual availability of a full-length cDNA for thrombomodulin should allow many of these intriguing issues to be examined. R.W.J. and D.L.13. should be considered as equal first authors. We thank Bruce Zetter for the generous donation of BACE cells used to make the cDNA libraries and Marie-Joseph Rabiet for help in gas-phase polypeptide sequencing. Technical assistance was provided by V. Braman in producing the monoclonal antibodies and by G.
Proc. Natl. Acad. Sci. USA 83
(1986)
Tayrien in HPLC isolation of peptides for sequencing. We thank J. Rosenberg and L. Fritze for assistance in the isolation of phage clones for DNA sequencing and M. Krieger for many helpful discussions. This work was supported by Grant P01-HL33014 from the National Institutes of Health and by funds provided by a private foundation. 1. Esmon, N. L., Owen, W. G. & Esmon, C. T. (1982) J. Biol. Chem. 257, 859-864. 2. Jakubowski, H. V., Kline, M. D. & Owen, W. G. (1986) J. Biol. Chem. 261, 3876-3882. 3. Salem, H. H., Maruyama, I., Ishii, H. & Majerus, P. W. (1984) J. Biol. Chem. 259, 12246-12251. 4. DeBault, L. E., Esmon, N. L., Olson, J. R. & Esmnon, C. T. (1986) Lab. Invest. 54, 172-178. 5. Esmon, C. T. & Owen, W. G. (1981) Proc. Natl. Acad. Sci. USA 78, 2249-2252. 6. Marcum, J. A., McKenney, J. B., Galli, S. J., Jackman, R. W. & Rosenberg, R. D. (1986) Am. J. Physiol. 250, H879-H888. 7. Ey, P. L., Prowse, S. J. & Jenkin C. R. (1978) Immunochemistry 15, 429-436. 8. Kennett, R. H., McKearn, T. J. & Bechtol, K. B. (1980) in Monoclonal Antibodies Hybridomas: A New Dimension in Biological Analyses (Plenum, New York). 9. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1972) in Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). 10. Folkman, J., Haudenschild, C. C. &-Zetter, B. R. (1979) Proc. Natl. Acad. Sci. USA 76, 5217-5221. 11. Ginsburg, D., Handin, R. I., Bonthron, D. T., Donlon, T. A., Bruns, G. A. P., Latt, S. A. & Orkin, S. H. (1985) Science 228, 1401-1406, 12. Laemmli, U. K. (1970) Nature (London) 227, 680-685. 13. Young, R. A. & Davis, R. W. (1983) Proc. Natl. Acad. Sci. USA 80, 1194-1198. 14. Schwarzbauer, J. E., Paul, J. I. & Hynes, R. 0. (1985) Proc. Natl. Acad. Sci. USA 82, 1424-1428. 15. Biggin, M. D., Gibson, T. J. & Hong, G. F. (1983) Proc. Natl. Acad. Sci. USA 80, 3963-3965. 16. Edge, A. S. B., Faltynek, C. R., Hof, L., Reichert, L. E., Jr., & Weber, P. (1981) Anal. Biochem. 118, 131-137. 17. Hewick, R. M., Hunkapiller, M. W., Hood, L. E. & Dreyer, W. J. (1981) J. Biol. Chem. 256, 7990-7997. 18. Benton, W. D. & Davis, R. W. (1977) Science 1%, 180-182. 19. Gribskov, M., Burgess, R. R. & Devereux, J. (1986) Nucleic Acids Res. 14, 327-334. 20. Kyte, J. & Doolittle, R. F. (1982) J. Mol. Biol. 157, 105-132. 21. Sabatini, D. D., Kriebich, G., Morimoto, T. & Adesnik, M. (1982) J. Cell Biol. 92, 1-21. 22. Dayhoff, M. O., Schwartz, R. M. & Orcutt, B. C. (1979) in Atlas of Protein Sequence and Structure (National Biomedical Research Foundation, Washington, DC), Vol. 5, Suppl. 3, pp. 345-362. 23. Doolittle, R. F., Feng, D. F. & Johnson, M. S. (1984) Nature (London) 307, 558-560. 24. Goldstein, J. L., Brown, M. S., Anderson, T. G. W., Russell, D. W. & Schneider, W. J. (1985) Annu. Rev. Cell Biol. 1, 1-39. 25. Maruyama, I. & Majerus, P. W. (1985) J. Biol. Chem. 260, 15432-15438. 26. Bar-Shavit, Z., Bing, D. H., Kahn, A. J. & Wilner, G. D. (1985) in Membrane Receptors and Cellular Recognition (Liss, New York).
