Purification and Characterization of Transmembrane Forms of Heparin-binding EGF-like Growth Factor*

Heparin-binding epidermal growth factor-like growth factor (HB-EGF), whose cDNA has a predicted 208-codon open reading frame, is synthesized as a membrane-span-ning precursor that is processed to release mature mi- togenic proteins of -73-87 amino acids in length. Previous work has focused on the structural and biological properties of secreted HB-EGF. In this study, human recombinant transmembrane HB-EGF, produced by ex- pression of HB-EGF,,,, cDNA in a baculovirus system, has been isolated, purified, and characterized structur-ally and biologically. Two isoforms of transmembrane HB-EGF (HB-EGF,) were purified from membrane fractions of infected insect cells by a combination of heparin affinity chromatography and reversed-phase high per- formance liquid chromatography. The isoform designated as HB-EGF,,, a 21.5-kDa protein, yielded no N- terminal sequence, suggesting that it is N-terminally blocked. However, HB-EGF,.,, a 24-kDa protein, was N-terminally sequenced and found to be initiated at Asp" in the 208-amino acid residue primary translation product. This N terminus is the same as that determined for a 18-kDa isoform of secreted

growth factor-like growth factor; HB-EGF,,, transmembrane HB-EGF; The abbreviations used are: HB-EGF, heparin-binding epidermal family that was first identified as a secreted product of macrophage and macrophage-like U-937 cells (1,2). Structurally, U-937 cell-derived HB-EGF has an apparent molecular mass of 20-22 kDa and appears in multiple forms due to N-terminal truncations and possibly to differences in glycosylation (2). The multiple secreted forms contain between 73 and 87 amino acids, suggesting that a substantial amount of the HB-EGF molecular mass is due to glycosylation. The mitogenic activities of HB-EGF for smooth muscle cells (SMC), fibroblasts, and keratinocytes are mediated by interactions with the EGF receptor (EGFR) and with cell-surface heparan sulfate proteoglycans (HSPG) (1)(2)(3). Binding to HSPG facilitates HB-EGF high afflnity binding to EGFR and appears to be responsible for the 40-fold enhanced bioactivity of HB-EGF for SMC compared with EGF and transforming growth factor-a (TGF-a) (4-6). Structure-function analysis has demonstrated that the heparin-binding domain of HB-EGF is associated with a stretch of 21 mostly cationic amino acids upstream but slightly overlapping the EGF-like domain, which in turn is responsible for EGFR binding (4,5).
The open reading frame of the HB-EGF cDNA encodes a 208-amino acid protein that is predicted to be synthesized as a precursor containing a transmembrane region flanked by an ectodomain and a cytoplasmic domain. This precursor structure is common to other members of the EGF family such as EGF, TGF-a, and amphiregulin (7)(8)(9)(10); in each of these cases, the secreted forms of the mitogens are processed from the precursor ectodomains. In the case of TGF-a, it has been demonstrated that the tethered transmembrane form is itself biologically active in signaling neighboring cells in what has been termed juxtacrine stimulation (11)(12)(13). The first suggestion that a transmembrane form of HB-EGF (HB-EGF,,) is also biologically functional was the demonstration that it acts as the diphtheria toxin receptor (DTR) (14). Binding of diphtheria toxin (DT) to HB-EGFm/DTR results in DT internalization and the subsequent inhibition of protein synthesis (15)(16)(17)(18). We have recently found that the toxicity of DT is markedly up-regulated by a transmembrane protein that binds tightly to HB-EGF,$ DTR known as CD9DRAP27 (19). Besides binding DT, cellassociated HB-EGF,, is active in stimulating phosphorylation of EGFR in a juxtacrine manner' and may promote cell-to-cell contact in the binding of the blastocyst to the mouse uterine wall during implantation (20).

31315
Given the potentially important biological properties of HB-EGF,,, we decided t o analyze its structure and function more precisely. In general, there is little information regarding the structures and bioactivities of isolated transmembrane forms of members of the EGF family. For example, the transmembrane form of TGF-a has been characterized, but not purified (11)(12)(13)21). In this report, we (i) describe the purification of two isoforms of recombinant HB-EGF, using a baculovirus-insect cell system; (ii) demonstrate that the N-terminal amino acid of one of these is identical to that of the longest form of secreted HB-EGF (HB-EGF,,,); and (iii) demonstrate that the HB-EGF, isoforms are bioactive, although with a lower specific activity than the secreted forms of HB-EGF.

EXPERIMENTAL PROCEDURES
Materials EGF was purchased from Collaborative Research (Waltham, MA). Heparin-Sepharose CL-GB was purchased from Pharmacia Biotech Inc. Radioactive materials were purchased from DuPont NEN. All cell media components were purchased from Life Technologies, Inc. Reagents for electrophoresis were purchased from Bio-Rad. Analytical-grade chemical reagents were purchased from Sigma. HPLC-grade water and acetonitrile were purchased from Pierce.
Mitogenic and Motility Assays HB-EGF mitogenic activity was assayed by measuring the incorporation of PHIthymidine into the DNA of confluent monolayers of BALB/c 3T3 cells and bovine aortic SMC as described previously (22)(23)(24). The motility of bovine aortic SMC was measured as described previously (25) with some modifications. Briefly, wells in 48-well plates (Costar Corp., Cambridge, MA) were precoated with 0.5 ml of DMEM containing 10% fetal calf serum, 100 unitdml penicillin, and 100 pg/ml streptomycin sulfate (DMEIWfetal calf serum/PS) for 1 h, followed by washing with DMEWS. A 90-pl polystyrene bead solution (Polysciences, Inc., Warrington, PA) was diluted into 30 ml of DMEMRS and plated (0.5 ml/well). The plates were centrifuged at 800 x g for 20 min at 10 "C to sediment the beads. To measure motility, SMC grown in DMEWfetal calf serum/PS to 60-70% subconfluence were trypsinized (0.25% trypsin; Life Technologies, Inc.) and resuspended in DMEMRS (at 5-7.5 x lo3 SMC/ml), and 150-200 p l of the cell suspension were plated in bead-coated 48-well plates. After the addition of HB-EGF, plates were incubated at 37 "C for 16 h, after which the cells were fixed by the addition of 2.0 ml of phosphate-buffered 10% Formalin. The wells were photographed, and the areas of individual tracks made by migrating cells were measured using an image scanner (MacImage, Macintosh). About 50 tracks were measured per well (three wells for each sample), and a mean value and a standard error were determined.
Expression of HB-EGF in a Baculouirus-Insect Cell System Expression of HB-EGF forms in a baculovirus-insect cell system was carried out essentially by methods described previously (26) with modifications.
Vectors-To express the membrane-anchored fonds) of HB-EGF, the complete open reading frame of the HB-EGF cDNA was amplified by polymerase chain reaction using synthetic DNA oligonucleotides (primers P20 (GCTCTAGAGCATGAAGCTGTGCCGTCG) and P21 (GCT-CTAGATCAGTGGGAATTAGTCAT)). To express a secreted form of HB-EGF, only the sequence corresponding to the first 149 predicted amino acids of the HB-EGF cDNA was amplified by polymerase chain reaction (primers P20 and P31 (GCTCTAGACTATGGGAGGCTCAGCCCATG)). Although an initial study suggested Leu'48 to be the C terminus of secreted HB-EGF (2), Pro149 was included in this construct because recent studies using comigration of synthetic peptides with tryptic fragments suggest that this residue might be the actual C t e r m i n u~.~ The polymerase chain reaction products were digested with XbaI, ligated into the baculovirus transfer vector pVL1392 (Invitrogen), and propagated in Escherichia coli. Ampicillin-resistant clones were picked and analyzed by restriction analysis for the presence of an HB-EGF DNA fragment, which was then sequenced.
Cell Culture-SF21 cells (obtained from David L. Bishop) were grown in SF900 medium supplemented with 10% fetal calf serum (selected serum batches; Intergen Co., Purchase, N Y ) and 50 pg/ml gentamycin. For production and purification of recombinant proteins, SF21 cells K. Lau and J. A. Abraham, unpublished data.
were adapted to serum-free medium (SF900 medium with 50 pg/ml gentamycin) and grown in Spinner flasks (Wheaton Instruments, Millville, NJ).
Dansfection and Virus IsolationSF21 cells were transfected by lipofection with recombinant baculovirus expression vectors and linearized wild-type Autographa californica nuclear polyhedrosis virus using a linear transfection module (Invitrogen) according to the manufacturer's recommendations. Individual virus clones were obtained using a soft agar plaque assay (27). These clones were then used to infect SF21 cells, and 5 days later, cell supernatants and extracts were screened for growth factor activity on BALB/c 3T3 cells. Three clones expressed active transmembrane forms of HB-EGF, and five clones expressed secreted forms. One clone of each (AcHB-EGF,,,, and ACHB-EGF,_,,~) was selected and expanded for further experiments.
Expression of Recombinant Proteins-SF21 cells grown serum-free in Spinner flasks were infected at a mid to late logarithmic growth phase with recombinant baculovirus clones at a multiplicity of infection of 1-2. Conditioned medium (CM) or cells were harvested 72-96 h post-infection and used for HB-EGF purification.
Purification of Secreted HB-EGF Secreted HB-EGF produced by infecting insect cells with recombinant baculovirus AcHB-EGF,_,, was purified from insect cell CM by a modification of the method used to purify HB-EGF from U-937 cell CM (1,2). Insect cell CM (1.5-2 liters) was applied to a heparin-Sepharose CL-GB column (2.5 x 8 cm) equilibrated with 0.2 M NaCl, 20 m M Tris-HCl, pH 7.4. After extensive washing with the equilibration buffer, secreted HB-EGF was eluted batchwise with 2 M NaCl, 20 m M Tris-HC1, pH 7.4. The eluant was diluted 1 : l O with 20 m M Tris-HC1, pH 7.4, and applied to a heparin-Sepharose CL-GB column (16 x 200 mm) using a fast protein liquid chromatography system. After washing with the equilibration buffer, secreted HB-EGF was eluted with a 300-ml linear gradient of 0.2-2.0 M NaCl, 20 m M Tris-HC1, pH 7.4, at a flow rate of 2 mumin. Active fractions were collected and applied to a C, reversedphase column (4.6 x 250 mm; Vydac, Hesperia, CA) equilibrated with 5% acetonitrile in 0.1% trifluoroacetic acid using a Hitachi L-6210 HPLC system. The column was washed extensively with the equilibration buffer, and bound proteins were then eluted at a flow rate of 1 ml/min with a 1-ml gradient of 5 2 5 % acetonitrile in 0.1% trifluoroacetic acid, followed by a 60-ml gradient of 25-35% acetonitrile in 0.1% trifluoroacetic acid. In a representative purification starting with 1 liter of SF21 CM, the recovery was -lo%, and the yield of purified HB-EGF,, was 500 pg. Purified HB-EGF,,, was stored in 25 m~ acetic acid, pH 3.0, at -80 "C.
Purification of Dansmembrane HB-EGF Transmembrane HB-EGF produced by infection of insect cells with baculovirus AcHB-EGF,,,, was purified from insect cell lysates. Insect cell suspensions ( 2 4 liters) were collected and centrifuged at 5000 x g for 30 min at 4 "C. The cell pellets were resuspended in a 10 x pellet volume of 0.15 M NaC1, 20 m M Tris-HC1, pH 7.4, and then lysed by Materials Inc., Danbury, CO). After centrifugation of the lysates at continuous sonication for 1 min on ice using a cell sonicator (Sonics & 12,000 x g for 60 min, a microsomal fraction was prepared by centrifuging the supernatants at 105,000 x g for 60 min. HB-EGF, was then purified by three cycles of RP-HPLC. Active fractions from the second heparin-Sepharose column were collected and applied to a semipreparative reversed-phase phenyl column (COSMOSIL 5 Ph-AR-300, 10 x 250 mm; Nacalai Tesque, Kyoto, Japan) equilibrated with 5% acetonitrile in 0.1% trifluoroacetic acid. The column was washed extensively, and bound proteins were eluted at a flow rate of 3 mumin with a 15-ml gradient of 5 1 5 % acetonitrile in 0.1% trifluoroacetic acid, followed by a 180-ml gradient of 1540% acetonitrile in 0.1% trifluoroacetic acid. For the second RP-HPLC step, active fractions were collected from the semipreparative phenyl RP-HPLC column, diluted 1:2 with 5% acetonitrile in 0.1% trifluoroacetic acid solvent, and applied to an analytical reversed-phase phenyl column (COSMOSIL 5 Ph-AR-300, 4.6 x 250 mm). Bound protein was eluted with a 5-ml gradient of 5 1 5 % acetonitrile and then with a 60-ml linear gradient of 1540% acetonitrile at a flow rate of 1 mumin. For the third RP-HPLC step, active fractions were pooled and applied to the analytical reversed-phase phenyl column using the same gradient program. In a representative purification, starting with 2 liters of SF21 cell suspension and using this protocol, the recovery was -l%, and the yield of purified HB-EGF,, was 5 pg. Purified HB-EGFT, was stored in 25 m M acetic acid, pH 3.0, containing 0.5% CHAPS at -80 "C. The 0.5% CHAPS had no adverse effects on the mitogenic and motility activities of HB-EGF,,.

SDS-Polyacrylamide Gel Electrophoresis and Western Blot Analysis
Samples were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) using a 15% polyacrylamide gel, and protein bands were visualized by silver staining as described previously (28). To prepare antibodies for Western blotting, two synthetic peptides were prepared (see Fig. 1): P-2911 (ICHPGYHGERCHGLSL), which corresponds to HB-EGFIaslM, at the C terminus of mature secreted human HB-EGF; and P-3100 (DVENEEKVKLGMTNSH), which corresponds to HB-EGF,,,,,, at the C terminus of the cytoplasmic tail of transmembrane HB-EGF. The synthetic peptides were injected into rabbits by Lampire Biological Laboratories (Piperville, PA), and two polyclonal antibody preparations were obtained (Ab 2911 and Ab 3100). For Western blotting, purified samples of HB-EGF were subjected to electrophoresis by SDS-PAGE and transferred using a Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad) to a nitrocellulose membrane (0.45 pm; Schleicher & Schuell) as described previously (3). The membrane proteins were incubated with Ab 2911 or Ab 3100 and then with alkaline phosphatase-conjugated goat anti-rabbit IgG as the second antibody. Protein bands were visualized using nitro blue tetrazolium and 5-bromo-~-chloro-3-indolyl phosphate (Promega) as described previously (3).

N-terminal Sequencing of Dansmembrane and Secreted HB-EGF Forms
Purified species of HB-EGF, (7 pg each) and of HB-EGF,,, (9 pg each) were subjected to SDS-PAGE, and protein bands were transferred at 500 mA for 90 min at 4 "C onto polyvinylidene difluoride membranes (0.45 pm; Millipore Corp.) equilibrated with 100 m M CAPS (Sigma), pH 11.0, containing 20% methanol. The proteins were stained with 0.1% Coomassie Brilliant Blue R-250 (Sigma), and after destaining with 10% acetic acid and 50% methanol, each protein band was cut out and subjected to N-terminal microsequencing using an Applied Biosystems 477A microsequenator with an on-line Applied Biosystems Model 120A phenylthiohydantoin-derivative analyzer.

RESULTS
Purification of Dansmembrane and Secreted Forms of HB-EGF-The 208-amino acid primary translation product encoded by human HB-EGF cDNA is shown in Fig. 1. The predicted protein includes signal peptide, propeptide, mature secreted HB-EGF, juxtamembrane, transmembrane, and cytoplasmic domains. Mature HB-EGF, for which the largest isoform identified so far contains the 86 amino acids corresponding to residues 63-148 (boxed in Fig. 1) and possibly Pro149 as well, is processed from a larger transmembrane precursor molecule and secreted. However, the N and C termini of HB-EGF, have not been identified. To do so, it was necessary to purify transmembrane HB-EGF on a large scale. Accordingly, a baculovirus vector expressing the entire 208-amino acid protein encoded by human HB-EGF cDNA (AcHB-EGF,_,,,) was used to produce HB-EGF, in insect cells. For comparison, a second baculovirus vector (AcHB-EGF,-,,,) in which the transmembrane and cytoplasmic domains were deleted was constructed in order to produce secreted forms of HB-EGF.
TO purify HB-EGF,,, solubilized insect cell microsomal fractions were prepared from cells infected with AcHB-EGF,-,,, and applied to a heparin-Sepharose column. Growth factor ac- tivity was eluted with -1 M NaCl ( Fig. 2A 1, approximately the same NaCl concentration required to elute secreted HB-EGF isolated from insect cell CM after infection with ACHB-EGF,-,,~ ( Fig. 2 B ) . The CM of cells infected with AcHB-EGF,-,,, (prepared from a 10-fold greater cell number then cells expressing HB-EGF,_,,,) had no detectable growth factor activity, indicating that this form remains cell-associated and is not processed to the mature secreted form (data not shown). Differences in elution profiles were detected on RP-HPLC columns. After heparin-Sepharose chromatography, pooled active HB-EGF,, samples were applied to semipreparative reversed-phase phenyl columns. Two major peaks of transmembrane HB-EGF activity, designated as HB-EGF,,., and HB-EGF,,.,,, were eluted from reversed-phase phenyl columns with -30 and 35% acetonitrile, respectively (Fig. 2C). These two peaks were purified individually by two more cycles of RP-HPLC that yielded highly purified preparations in which HB-EGF,., (Fig. 3 A ) and HB-EGFTM.II (Fig. 3B) corresponded to single peaks of absorbance at 214 nm. RP-HPLC analysis of secreted HB-EGF using C, columns showed multiple peaks of mitogenic activity (Fig. 2 0 ) . Of these, two growth factor peaks that had detectable absorbance at 214 nm were eluted with 25-27% acetonitrile. The growth factors corresponding to these two peaks were designated as HB-EGF,,-, and HB-EGF,,.,,, respectively.
The purity of RP-HPLC-purified transmembrane and secreted HB-EGF peaks of growth factor activity was ascertained by SDS-PAGE under reducing conditions (Fig. 4). HB-EGF,,,., as shown in Fig. 11, an antibody that is capable of detecting both secreted and transmembrane forms of bioactive HB-EGF (data not shown).

N-terminal Sequence Analysis and Western Blotting of Mature and Precursor
Forms of HB-EGF-To confirm further that the various forms of transmembrane and secreted species were actually HB-EGF, N-terminal microsequencing was carried out (Fig. 51, and the results were compared with the amino acid sequence predicted from the HB-EGF cDNA (see Fig. 1). The N terminus of HB-EGF,,.,, was found to be Asp63. This amino acid is also the N terminus of the longest form of secreted HB-EGF Heparin Affinity Reversed Phase HPLC

FIG. 2. Heparin-Sepharose chromatography and RP-HPLC of secreted and transmembrane forms of HB-EGF.A, heparin affinity chromatography. Solubilized microsomal fractions of insect cells in-
fected with AcHB-EGF,-,,, were prepurified batchwise on an open heparin-Sepharose column, followed by heparin affinity chromatography on a heparin-Sepharose CL-GB column using fast protein liquid chromatography. Fractions were eluted with a linear gradient of 0.2-2.0 M NaCl, and mitogenic activity for BALB/c 3T3 cells was measured. B, heparin affinity chromatography; the same as described for A, except that the insect cells were infected with AcHB-EGF,-,,,. C, RP-HPLC. The active fractions of transmembrane HB-EGF purified by heparin affinity, as shown in A, were applied to a semipreparative reversedphase phenyl column, and HB-EGF was eluted with a linear gradient of 1540% acetonitrile and assayed for mitogenic activity for BALB/c 3T3 cells. D, RP-HPLC; the same as described for C, except that the material applied to the RP-HPLC column was the heparin affinity-purified secreted HB-EGF shown in B, and a C, RP-HPLC column with a 25-35% acetonitrile gradient was used.

FIG. 3. Purification of transmembrane HB-EGF by multiple cycles of RP-HPLC. HB-EGF,,
and HB-EGF,,, purified by RP-HPLC on a semipreparative reversed-phase phenyl column, as shown in Fig. 2C, were collected individually and applied separately to analytic reversed-phase phenyl columns. After elution with a gradient of 1540% acetonitrile, HB-EGF,., (A) and HB-EGF,,, ( B ) were each reapplied to a second analytic reversed-phase phenyl column (third RP-HPLC step overall), and each HB-EGF, was eluted with same linear gradient of acetonitrile.
originally purified from U-937 cells (2). The N terminus of HB-EGF,,, was found to be S e P , corresponding to the N terminus of a shorter form of secreted HB-EGF purified from U-937 cells (2). Thus, HB-EGFw, is an active truncated species lacking the 14 N-terminal amino acids in HB-EGF,,,,. The N terminus of HB-EGF,,, was found to be Aspfi3, the same as for HB-EGF,,,.
N-terminal sequencing of HB-EGF,, was not successful since it appeared to be N-terminally blocked and insufficient material was available for tryptic digestion analysis. These results indicate that at least one species of HB-EGF, has the same N terminus as the longest form of HB-EGF,,, and suggest that Arg'j2-Asp63 might be a natural cleavage site for N-terminal processing of both transmembrane and secreted forms of HB-EGF.
To analyze the C-terminal region of HB-EGF,-,,, Western blotting was carried out with Ab 3100 (see Fig. 1) directed against the 16 C-terminal amino acids of the cytoplasmic tail (HB-EGF,,,), an antibody that should cross-react with transmembrane but not secreted forms of HB-EGF. As shown in Fig. 6, Ab 3100 cross-reacted with HB-EGF,,, (lane 3), but not with the two secreted HB-EGF forms even when overloaded (lunes 1 and 2). Thus, it could be concluded definitively by using a combination of N-terminal sequencing and Western blotting that HB-EGF,,, is a transmembrane protein beginning at Asp63 in the ectodomain and terminating somewhere in the last 16 amino acids of the cytoplasmic domain. Western blotting of 21.5-kDa HB-EGF,, with Ab 3100 was unsuccessful. Because it was the best characterized transmembrane HB-EGF, HB-EGF,,, was used for biological studies. Purified HB-EGF, Stimulates Cell Proliferation and Cell Migrution-Having purified a bona fide transmembrane HB-EGF molecule, the next question was whether it was biologically active. Accordingly, the ability of HB-EGF,,, to stimulate r3H]thymidine incorporation in BALB/c 3T3 cells and SMC was measured (Fig. 7). HB-EGF,,, stimulated half-maximal 3T3 cell proliferation at 1100 PM (24 ng/ml) (Fig. 7A) and halfmaximal SMC proliferation at 1300 PM (28 ng/ml) (Fig. 7B). Although definitely bioactive, the specific activity of HB-EGF,,, for 3T3 cells and SMC was -10-25% of that of the secreted HB-EGF forms.
Secreted HB-EGF purified from U-937 cells stimulates SMC migration (5). Using a motility assay, which measures tracks formed by migrating cells, it could be shown that both HB-EGF,,, (Fig. 8B) and HB-EGF,,, (Fig. 8C) stimulated SMC migration markedly compared with a mock control (Fig. 8A). HB-EGF,,, and HB-EGF,,,, stimulated SMC motility in a dose-dependent manner (Fig. 8). However, HB-EGF,,, was only -25% as active as HB-EGFw,, in the linear range. HB-EGF,.,, also enhanced the motility of BALB/c 3T3 cells (data not shown). Taken together, these results indicate that purified HB-EGF, is biologically active in the same bioassays as the  Fig. 4 were subjected to SDS-PAGE, transferred to polyvinylidene difluoride membranes, and stained with Coomassie Brilliant Blue R-250. After destaining, each protein band was cut out, and its N-terminal sequence was determined using an Applied Biosystems 477A protein sequenator. The top line shows the predicted amino acids encoded by HB-EGF cDNA for positional reference (see Fig. 1).
processed secreted forms of HB-EGF, but a t a lower specific activity.

DISCUSSION
We have isolated and purified a membrane-associated form of HB-EGF, analyzed its primary structure, and shown that it is biologically active in the same bioassays as the secreted forms of HB-EGF. From a baculovirus-insect cell system, two isoforms of HB-EGF,, were purified by heparin affinity chromatography and several cycles of RP-HPLC. One of these isoforms, HB-EGF,,,,, yielded an N-terminal sequence in which the first amino acid was Asp63 in the 208-amino acid primary translation product of HB-EGF. Further evidence that HB-EGF,,,, is actually a transmembrane protein was obtained using a specific anti-HB-EGF antibody directed against the 16 most C-terminal amino acids of HB-EGF, found in the cytoplasmic tail. This antibody cross-reacted with HB-EGF,,,,, but, as expected, not with any of the secreted isoforms of HB-EGF. Thus, we concluded that HB-EGF,,, most likely corresponds to the 146-amino acid stretch between amino acids 63 and 208 of the primary translation product and constitutes one form of authentic transmembrane HB-EGF precursor. However, without knowing the exact epitope of the anti-HB-EGF antibody in the C-terminal region, it cannot be ascertained for certain whether HB-EGF,,, extends all the way to amino acid 208. There have been previous attempts to isolate the HB-EGF precursor. In these experiments, the objective was to purify DTR (not realized at that time to be equivalent to HB-EGF,) from Vero cells, which are highly sensitive to DT and produce high levels of DTR (18). A 106-fold purification was reported of a 14.5-kDa protein that aggregated to form a 60-90-kDa complex. However, insufficient material was obtained from Vero cells to obtain any amino acid sequence and proof of homogeneity. As we have found ourselves using cell lines, purification of HB-EGF,, is difficult without using high expression systems such as the baculovirus-insect cell system.
The transmembrane form of TGF-a (proTGF-a) has also been characterized (21). In those studies, it was found that retrovirus-transformed rat cells release into CM a 17-19-kDa form of TGF-a possessing a potential transmembrane domain. This form of TGF-a was identified using specific antibodies and displayed full biological activity. However, no purification or N-terminal sequencing of proTGF-a was included in those studies.
HB-EGF,,,, produced in the baculovirus system has a molecular mass of -24 kDa compared with two isoforms of secreted recombinant HB-EGF that are -15 and 18 kDa, respec- tively. These numbers are entirely consistent with the processing of HB-EGF from a 146-amino acid precursor into 72-87-amino acid secreted isoforms, a loss of -60-75 amino acids. In our initial structural analysis of natural secreted HB-EGF, we found that U-937 cells released multiple forms of HB-EGF that had molecular masses of 20-22 kDa (1,2), somewhat larger than the 15-and 18-kDa forms produced in insect cells. This discrepancy appears to be due to a lower level of secreted HB-EGF glycosylation in insect cells. U-937 cell-derived HB-EGF, when treated with 0-glycanase, is reduced in molecular mass to -14-16 kDa (2). However, insect cell secreted HB-EGF molecular mass is unaffected by glycanases and remains a t -15-18 kDa, suggesting that it might not be highly glycosylated to begin with and may be more similar to recombinant mature 14-kDa HB-EGF produced in E. coli, which is not at all glycosylated (3, 4).  The shorter form of secreted HB-EGF, 15-kDa HB-EGF,,, was found to start with S e P . Assuming the C terminus to be which is the C-terminal amino acid encoded by the baculovirus vector AcHB-EGF,-,,,, HB-EGF,,, possesses at most 73 amino acids and thus is the smallest HB-EGF form shown to date to be bioactive. One would predict that a 73-amino acid protein, particularly if non-glycosylated, would have a molecular mass e15 kDa. The reason for this discrepancy is unclear, but has also been noted for amphiregulin (10). It is possible that cationic heparin-binding growth factors such as HB-EGF and amphiregulin display retarded mobility on SDS-PAGE, yielding apparently higher molecular masses on SDS-PAGE than expected (2). Purified HB-EGF,,,, is biologically active and stimulates the migration and proliferation of BALB/c 3T3 cells and SMC. Transmembrane HB-EGF, however, has a specific activity that is 10-25% lower than that of secreted HB-EGF. The reasons for lowered bioactivity are not known, but possibilities are that (i) HB-EGF,, loses activity during purification that utilizes three cycles of RP-HPLC, (ii) HB-EGF,, is not optimally folded due to the exposure in solution of its hydrophobic transmembrane domain, (iii) HB-EGF,, might be inserted into the target cell membrane via its hydrophobic tail and be constrained in activity when compared with secreted forms of HB-EGF that are soluble, (iv) HB-EGF,, might have intrinsically less activity than the secreted forms due to its larger and different structure, (v) HB-EGF,, might be inefficiently processed to a more soluble and active form lacking the hydrophobic anchor, and (vi) HB-EGF,, might require a cofactor such as transmembrane CD9/DRAP27 (19) for optimal bioactivity.

Trans-
These studies of purified recombinant HB-EGF,, are significant in that they allow insights into transmembrane HB-EGF structure-function relationships involved in processing, juxtacrine activity, and DT sensitivity. Purification of HB-EGF,, and identification of its N terminus by microsequencing along with use of specific antibodies to detect the C-terminal region constitute the first biochemical determination of the structure of the transmembrane forms of HB-EGF. Interestingly, the Nterminal amino acid of one transmembrane form of HB-EGF and of the longest form of secreted HB-EGF produced either in the baculovirus system or by U-937 cells (2) is Asp63. These results suggest that both transmembrane and secreted forms of HB-EGF are generated by N-terminal processing of the primary translation product a t A r g % k~~~. On the other hand, the N-terminal cleavage site for TGF-a has been shown to be Ala-Val (7). The proteolytic enzyme responsible for the potential cleavage of HB-EGF at its N terminus is not known, but one possibility could be furin, a membrane-bound protease that recognizes Arg residues and cleaves at a consensus sequence of (29), consistent with the RDRKVR sequence of HB-EGF at amino acids 57-62 shown in Fig. 1.
The bioactivity of purified HB-EGF,, in stimulating cell proliferation and SMC migration is consistent with its being a jwtacrine growth factor mediating cell-to-cell interactions. Recent evidence indicates that cells transfected with HB-EGF,, stimulate A431 EGFR phosphorylation in co-culture? HB-EGF,, has the same heparin binding properties as HB-EGF,. Both forms contain the HB-EGF heparin-binding domain (2,4, 51, which begins -30 amino acids downstream of Asp63, and both forms bind equally well to heparin-Sepharose. The heparin binding properties of HB-EGF,, suggest that juxtacrine interactions might occur not only between HB-EGF and EGFR, but between HB-EGF and cell-surface HSPG as well. This could be significant given the importance of HSPG binding for optimal HB-EGF bioactivity for SMC (1,2,5). Alternatively, cell-associated HB-EGF,, could bind to the extracellular matrix via interactions with matrix HSPG. By comparison, TGF-a does not bind to HSPG.
Finally, knowledge of HB-EGF,, helps to delineate the possible binding sites for DT. DT binds to HB-EGF,,/DTR, and the whole complex is internalized (18,19). If HB-EGF,, begins at Aspm, then it appears that the propeptide region of HB-EGF (most probably HB-EGFzw,,) is not involved in DT binding. Consistent with this, recent data indicate that the DT-binding site on HB-EGF/DTR lies within the EGF-like domain." It has become quite apparent in the last few years that there are a number of bona fide insoluble forms of growth factors capable of bioactivity including membrane-anchored TGF-a (7lo), kit ligand (30), colony-stimulating factor (31), and tumor necrosis factor (32,33). Purification and characterization of HB-EGF,, and other transmembrane growth factors might be important initial steps in understanding their function.