Purification and Characterization of Novel Heparin-binding Growth Factors in Uterine Secretory Fluids

Uterine growth factors are potential effector molecules in embryo growth signaling pathways. Pig uterine luminal flushings contained a heparin-binding growth factor (HBGF) that required 0.8 m NaCl for elution from heparin columns and was termed HBGF-0.8. This factor, which was heat- and acid-labile and of M r 10,000 as assessed by gel filtration, stimulated DNA synthesis in fibroblasts and smooth muscle cells but not endothelial cells. Two forms of HBGF-0.8, termed HBGF-0.8-P1 and HBGF-0.8-P2, exhibited differential heparin-binding properties. SDS-polyacrylamide gel electrophoresis showed that each form of HBGF-0.8 migrated with an apparent M r of 10,000 under reducing conditions. Amino acid sequencing revealed the N-terminal sequence EENIKKGKKXIRTPKI for HBGF-0.8-P1 and ENIKKGKKXIRT for HBGF-0.8-P2. These sequences corresponded, respectively, to residues 247–262 and 248–259 of the 349-residue predicted primary translation product of porcine connective tissue growth factor (pCTGF). 10-kDa CTGF-mediated fibroblast DNA synthesis was modulated by exogenous heparin, and CTGF-immunoreactive proteins of 10, 16, and 20 kDa were present in unfractionated uterine luminal flushings. These data reveal the identity of a novel growth factor in uterine fluids as a highly truncated form of CTGF and show that the N-terminal two-thirds of the CTGF primary translation product is not required for mitogenic activity or heparin binding.

ine epithelium (1). Polypeptide growth factors and cytokines are emerging as a class of uterine proteins that may be involved in uterine-embryo growth signaling pathways either through their secretion into the uterine lumen or their localization at the implantation site (2)(3)(4)(5). Studies in a variety of species have suggested that epidermal growth factor (EGF), 1 heparin-binding EGF-like growth factor (HB-EGF), insulin-like growth factors-I and -II (IGF-I, IGF-II), acidic and basic fibroblast growth factors (aFGF, bFGF), pleiotrophin (PTN), leukemia inhibitory factor, colony-stimulating factor-1, and transforming growth factor-␣ may be among the uterine growth-regulatory molecules involved in these processes (2, 6 -12). The pig is a useful model for these studies because the embryos of this species produce extensive extraembryonic membranes, exhibit a high incidence (approximately 30%) of mortality around the time of blastocyst expansion and attachment, produce a diffuse noninvasive placenta, and appear highly dependent on uterine secretory components throughout gestation (1,(13)(14)(15). Here we describe the characterization and purification of a principal heparin-binding growth factor (HBGF) in pig uterine secretions that is mitogenic for fibroblasts and smooth muscle cells in vitro. The N termini of two microheterogeneous forms of this factor are 100% conserved with residues 247-262 and 248 -259 of the 349-residue predicted primary translational products of human and porcine connective tissue growth factor (hCTGF, pCTGF). These data reveal the identity of a novel uterine HBGF and provide new insights into structural and functional properties of CTGF.

Animals
Uteri were collected at random from slaughterhouse pigs that were approximately 8 months or less in age. Each uterine horn was flushed with cold (4°C) phosphate-buffered saline (PBS) as described (8) to collect uterine luminal components. These studies involved the use of 323 individual pools of uterine luminal flushings (ULF) (384 liters total volume) obtained from approximately 9,660 pigs; growth factor characterization studies were performed on 1-liter pools of ULF obtained from about 30 individual uteri, and growth factor purification was performed on 4-liter pools of ULF obtained from up to 120 animals. ULF were clarified by centrifugation at 13,500 ϫ g for 30 min at 4°C, and the supernatant was passed through glass wool.
For Western blot analysis of crude ULF samples, pigs (n ϭ 4) of mixed breed (Duroc, Landrace, Yorkshire) were sacrificed on day 14 or 16 of the estrous cycle (day 0 representing the first day of estrus). ULF from both horns of individual animals were pooled and subjected to SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting as described below.

Growth Factor Purification from ULF
Cation Exchange Chromatography-1-or 4-liter samples of clarified ULF supernatant were applied at 4°C to a Bio-Rex 70 cation exchange column (5 ϫ 6 cm; Bio-Rad) that had previously been equilibrated in PBS, 0.2 M NaCl. After sample application, the column was washed with 500 ml of PBS, 0.2 M NaCl, and bound proteins were eluted using a 500-ml gradient of 0.2-2 M NaCl in PBS. The flow rate was 3.5 ml/min throughout, and fractions of 10 ml were collected during treatment of the column with the NaCl gradient. Fractions demonstrating mitogenic activity for Balb/c 3T3 fibroblasts were selected for further use. All subsequent chromatographic steps were performed at room temperature.
Heparin Affinity Chromatography-Biologically active fractions containing the 0.3-0.6 M NaCl eluate from the Bio-Rex column were pooled, diluted 3-fold with 20 mM Tris-HCl (pH 7.4) containing 0.1% CHAPS, passed through a 0.45-m membrane filter, placed in a siliconized polypropylene vessel, and applied with a peristaltic pump to an EconoPac heparin column (0.7 ϫ 3.6 cm; Bio-Rad) at 2 ml/min. The heparin column was then washed with 50 ml of 20 mM Tris-HCl buffer, 0.2 M NaCl, 0.1% CHAPS (pH 7.4) and developed at 1 ml/min with a 40-ml gradient of 0.2-2.0 M NaCl in 20 mM Tris-HCl, 0.1% CHAPS (pH 7.4) using a fast protein liquid chromatography (FPLC) system (Pharmacia Biotech Inc.). Fractions (1 ml) were collected into siliconized tubes during NaCl gradient elution and tested for 3T3 cell mitogenic activity.
To perform a second heparin affinity purification step, biologically active fractions containing the 0.8 M NaCl eluate from the EconoPac heparin purification step were pooled, diluted 3-fold with 20 mM Tris-HCl (pH 7.4), and clarified by passage through a 0.2-m filter. The sample was applied at 2 ml/min to a TSK heparin 5PW column (0.8 ϫ 7.5 cm; TosoHaas, Philadelphia, PA), which was washed and eluted as described above except that CHAPS was omitted from the buffers and fractions of 0.5 ml were collected. Fractions containing proteins that were eluted by 0.8 M NaCl and that demonstrated mitogenic activity for 3T3 cells were divided into two pools consisting of fractions 31-34 (peak 1) and fractions 35 and 36 (peak 2).
Reverse-phase HPLC-Reverse-phase HPLC was performed on a Hitachi HPLC system (Hitachi Instruments Inc., Danbury CT) using a C 8 column (0.46 ϫ 25 cm, 5-m particle size; Rainin Instrument Co., Woburn, MA) that was equilibrated with water containing 10% (v/v) acetonitrile and 0.1% (v/v) trifluoroacetic acid. Pooled fractions containing peaks 1 and 2 from the TSK heparin purification step were individually adjusted so that they contained 10% acetonitrile, 0.1% trifluoroacetic acid and were clarified by passage through a 0.2-m filter. Conditions for the elution of bound proteins were 10% acetonitrile from 0 to 10 min after sample injection and 10 -90% from 10 to 146 min. The flow rate was 1 ml/min throughout, and the chromatogram (A 214 ) was archived as described (16). The eluate was collected as 0.5-ml fractions in siliconized tubes containing 50 l of 125 mM NaOH to immediately neutralize the trifluoroacetic acid. 80-l aliquots of selected fractions were evaporated to dryness in a SpeedVac concentrator (Savant Instruments, Farmingdale, NY) and reconstituted in 25 l of 10 mM Tris-HCl (pH 7.4). 10 l of this concentrate were assayed for their stimulation of 3T3 cell DNA synthesis, and 10 l were used for analytical SDS-PAGE. For second step C 8 HPLC purification, two active fractions from the first HPLC step were pooled (1 ml total volume), diluted 5-fold with water, 0.1% trifluoroacetic acid, and subjected to the same chromatographic elution conditions as described above.
N-terminal Microsequencing-Fractions containing HPLC-purified growth factors were pooled, dried, and subjected to preparative SDS-PAGE. Proteins in the gel were transferred for 90 min at 300 mA to a polvinylidene difluoride membrane using 10 mM CAPS buffer (pH 11). The location of the proteins of interest was determined by staining the blots with 0.1% Coomassie R250 in 50% methanol for 2 min, followed by destaining with 50% methanol, 10% acetic acid. Half of each protein band was excised and submitted for N-terminal amino acid sequencing on a model 470A gas phase sequenator (Applied BioSystems, Foster City CA). Phenylthiohydantoin-derivatives were identified by C 18 reverse phase HPLC.

Growth Factor Characterization
DNA Synthesis Assays-Column fractions were tested for their ability to stimulate DNA synthesis as measured by [ 3 H]thymidine incorporation into the DNA of confluent quiescent Balb/c 3T3 cells grown in 200 l of Dulbecco's modified Eagle's medium, 10% bovine calf serum in 96-well culture plates as described (8). Dose-response curves to purified growth factors from ULF were established by assaying each dose in triplicate, with data computed as mean Ϯ S.D. Statistical significance of the effects of 1-100 g/ml porcine heparin (Sigma) on growth factor activity was determined by Student's t test.
Target cell specificity was studied using Balb/c 3T3 cells, bovine capillary endothelial cells (BCECs), and vascular smooth muscle cells. 3T3 cells were utilized as described above. BCECs were obtained from Dr. J. Folkman (Children's Hospital, Boston, MA) and were maintained in gelatinized culture flasks in Dulbecco's modified Eagle's medium containing 3 ng/ml bFGF and 10% heat-inactivated bovine calf serum.
Smooth muscle cells were isolated from a 2-3-cm length of pig thoracic aorta using established procedures (17) and maintained in 10% Dulbecco's modified Eagle's medium, 10% fetal bovine serum. BCEC and smooth muscle cell DNA synthesis assays were performed in 48-or 96-well plates essentially as described (18). BCEC DNA synthesis assays were also performed in the presence of 100 g/ml porcine heparin.
Gel Filtration Chromatography-0.5 ml of a fraction containing the 0.8 M NaCl eluate from EconoPac heparin affinity FPLC of ULF from 30 animals was applied at 0.5 ml/min to a TSK G2000 SW FPLC column (30 cm ϫ 8 mm, 10-m particle size, M r 500 -100,000 fractionation range; TosoHaas) equipped with a SW guard column (4 cm ϫ 8 mm, 10 m; TosoHaas). Proteins were eluted with PBS containing 0.3 M NaCl. Fractions of 200 l were collected and tested for their ability to stimulate DNA synthesis in 3T3 cells. Column calibration was performed using EGF (6,000 Da), lactalbumin (14,200 Da), trypsin inhibitor (20,100 Da), and ovalbumin (45,000 Da).
SDS-PAGE, Silver Staining, and Western Blotting-SDS-PAGE was performed under reducing conditions using 18% polyacrylamide minigels as described (8). Silver staining of proteins was performed as described (19). SDS-PAGE and Western blotting were performed on (i) HPLC-purified growth factors, (ii) 8 l of unfractionated ULF, or (iii) 100 l of ULF after passage through 20-l beds of heparin-Sepharose in the presence of 10 mM Tris-HCl, 0.5 M NaCl (pH 7.4) and subsequent extraction of the heparin beads with SDS-PAGE sample buffer. Gels were blotted and blocked as described (8) and incubated with a 1:1,000 dilution of rabbit preimmune serum or a 1:1,000 dilution of rabbit anti-pCTGF-(247-260) peptide antiserum (rabbit A; see below). Immunoreactive bands were visualized using alkaline phosphatase-conjugated goat anti-rabbit IgG followed by nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate chromogenic substrates.
All peptides contained one or no Cys residues; Cys 292 in CTGF-(285-292) and Cys 325 in CTGF-(318 -328) were replaced with Ser to prevent intrachain disulfide bridging to Cys 287 or Cys 323 within the respective peptides. Heparin-binding properties were determined using an adaptation of the method of Baird et al. (20). Briefly, 37.5 nmol of each peptide were absorbed in duplicate to nitrocellulose using a dot-blot apparatus. The blot was blocked for 30 min with 10 mM Tris-HCl, 0.15 M NaCl, 0.1% bovine serum albumin (pH 7.4) and then incubated for 3 h at room temperature in this solution containing 10 Ci/ml [ 3 H]heparin (NEN Life Science Products). The blot was washed four times with 10 mM Tris-HCl, 0.15 M NaCl, and individual dots were mixed with scintillation fluid for counting of 3 H.
CTGF Antibody Production-A four-branched multiple antigenic CTGF-(247-260) peptide comprising the sequence EENIKKGKKCIRTP (residues 247-260) was produced on a Synergy 432A peptide synthesizer (Applied BioSystems) and purified by reverse-phase HPLC using a C 18 column (0.46 ϫ 25 cm; Rainin Instruments) that was developed with a 90-min 5-95% acetonitrile gradient in water, 0.1% trifluoroacetic acid. Fractions containing the purified peptide were pooled, evaporated to dryness, and reconstituted in sterile water. Two New Zealand White rabbits (rabbits A and B), which had been bled to collect preimmune serum, were injected subcutaneously with 1 mg of peptide in Freund's complete adjuvant, followed 3 weeks later by an intramuscular injection of 250 g of peptide in Freund's incomplete adjuvant. Animals were bled 7 days later for collection of antiserum. Reactivity of the antisera was validated by Western blotting and immunoprecipitation. Preimmune serum and antiserum from rabbit A were used in these experiments.

Cloning of Porcine CTGF cDNA
Library Construction-Total pig endometrial RNA was obtained as described (8). A poly(A)Tract mRNA isolation system (Promega, Madison, WI) was used to isolate poly(A ϩ ) RNA, 5 g of which was subjected to first strand cDNA synthesis using Moloney murine leukemia virus reverse transcriptase and oligo(dT) linker-primer containing XhoI. Second strand synthesis was primed by treating the mRNA-cDNA complex with RNase. Double-stranded cDNA was blunted using Klenow fragment and ligated to EcoRI adaptors that were subsequently phosphorylated with T4 polynucleotide kinase. 100 ng of XhoI-digested cDNA, purified on a Sephacryl S-400 column, were ligated into 1 g of Uni-ZAP XR vector arms at the XhoI-EcoRI multiple cloning site, and the product was packaged using Gigapack II packaging extract (Stratagene, La Jolla, CA). The primary library was amplified in XL1-Blue MRFЈ cells to a titer of 1.4 ϫ 10 10 plaque-forming units/ml.
Isolation and Sequencing of pCTGF cDNA-A verified 32 P-labeled CTGF probe, corresponding to the 3Ј end of the predicted hCTGF primary translational product, was obtained by reverse transcriptasepolymerase chain reaction of RNA from human foreskin fibroblasts using the forward and reverse primers,5Ј-GCCGTCTAGAGCGGCCG-CATGGAAGAGAACATTAAGAAGGG-3Ј and 3Ј-CCTCTGTACCGTA-CTTAAGCGCCGGCGACC-5Ј, respectively. The probe was used to screen 10 6 plaques, two of which showed reproducible hybridization and were isolated using a Rapid Excision Kit (Stratagene). Two ϳ5.0-kilobase-pair pBluescript SK pig CTGF clones, termed pBSK-pCTGF1 and pBSK-pCTGF2, were obtained and used for initial sequencing reactions. pBSK-pCTGF1 was then fully sequenced by a combination of manual and automated dideoxy terminator sequencing (21). Sequence data were obtained from both strands of DNA.

Characterization and Purification of an Atypical HBGF in
ULF-Ion exchange chromatography of ULF showed the presence of cationic growth factor activity for Balb/c 3T3 cells that was eluted from Bio-Rex 70 columns by 0.3-0.6 M NaCl (data not shown). Although we have previously shown that constituent growth factors in this cationic extract include PDGF, HB-EGF, bFGF, and PTN (2,8,12), heparin affinity chromatography revealed the presence of an additional unidentified HBGF that required 0.8 M NaCl for elution from an EconoPac heparin column and, in terms of the amount of bioactivity recovered from the column, appeared to be a principal cationic HBGF in ULF for 3T3 cells (Fig. 1A). The elution position of this factor (hereafter termed HBGF-0.8) from heparin affinity columns was clearly distinct from PDGF, HB-EGF, PTN, aFGF, bFGF, and amphiregulin (2,8,12,(22)(23)(24). Stability tests performed by exposing HBGF-0.8 to pH 1.5 for 2 min, 56°C for 30 min, or 100°C for 5 min demonstrated that the 3T3 cell mitogenic activity of HBGF-0.8 was highly susceptible to acid or heat inactivation (data not shown). Gel filtration FPLC showed that HBGF-0.8 was of M r ϳ10,000 under nonreducing and nondenaturing conditions (Fig. 2). When compared with other 3T3 cells mitogens, the level of maximum induction of [ 3 H]thymidine incorporation by HBGF-0.8 was comparable with that of calf serum or purified PDGF or bFGF rather than that of weaker mitogens such as IGF or EGF (Table I). Further, the 3T3 cell mitogenic activity of HBGF-0.8 was synergistically potentiated by 10 ng/ml IGF-I, 10 ng/ml PDGF, 3 ng/ml EGF, or 0.3 ng/ml bFGF (data not shown). HBGF-0.8 was also mitogenic for smooth muscle cells and produced a level of stimulation that exceeded that of a maximal amount of EGF but was less than that of bFGF (Table I). HBGF-0.8 lacked mitogenic activity for endothelial cells when tested alone or in the presence of 100 g/ml heparin (Table I).
Fractions that contained HBGF-0.8 activity (i.e. fractions 15-18; see Fig. 1A) were pooled, diluted, and subjected to a second cycle of heparin affinity FPLC using a TSK heparin FIG. 1. Heparin affinity chromatography of ULF. A, 650 ml of ULF from 24 animals were subjected to Bio-Rex cation exchange chromatography, and the biologically active fractions containing the 0.3-0.6 M NaCl eluate were pooled and applied to an EconoPac heparin column (0.7 ϫ 3.6 cm) using a Pharmacia FPLC System. The column was washed with 20 mM Tris-HCl (pH 7.4) containing 0.2 M NaCl, 0.1% CHAPS and developed with a 40-ml gradient of 0.2-2 M NaCl at 1 ml/min. Fractions of 1 ml were collected during gradient elution and assayed at 20 l/ml for stimulation of DNA synthesis in quiescent Balb/c 3T3 cells as described (8). B, fractions containing 3T3 cell activity from cation exchange chromatography of 4 liters of ULF from 118 animals were pooled and subjected to EconoPac heparin affinity chromatography as shown in A. The 0.8 M NaCl eluate from this column was diluted 3-fold with 20 mM Tris-HCl (pH 7.4) and applied to a TSK heparin column, which was washed in 20 mM Tris-HCl, 0.2 M NaCl and then developed with a 0.2-2 M NaCl gradient in 20 mM Tris-HCl (pH 7.4). 0.5-ml fractions were collected during NaCl gradient elution of the bound proteins and were assayed at 10 l/ml for their stimulation of 3T3 cell DNA synthesis.
FIG. 2. Gel filtration chromatography of HBGF-0.8. 0.5 ml of a peak HBGF-0.8 fraction from heparin affinity FPLC was applied to a TSK G2000 SW FPLC column as described under "Materials and Methods." Fractions were tested for their ability to stimulate DNA synthesis in 3T3 cells at 40 l/ml. The figure shows the elution position of protein standards that were used to calibrate the column.
HBGF-0.8-P1 and -P2 were adjusted to 10% acetonitrile, 0.1% trifluoroacetic acid and individually subjected to C 8 reverse-phase HPLC. As shown in Fig. 3, the elution positions of HBGF-0.8-P1 and -P2 were determined by bioassay of aliquots of fractions containing the column eluate after they had been evaporated and reconstituted in PBS, demonstrating that there was sufficient activity in the purified HBGF-0.8 samples to permit their detection and further characterization despite prolonged (approximately 30 -40-min) exposure to pH 2 during the HPLC step. Silver-stained SDS-PAGE analysis of the fractions containing either HBGF-0.8-P1 or -P2 revealed the presence of a 10-kDa band under reducing conditions, the levels of which were directly correlated with the levels of mitogenic activity over the peak fractions (Fig. 3). This and other HPLC separations proved conclusively that the bioactivity was attributable to the 10-kDa moiety only. Second step HPLC purification of HBGF-0.8-P2 resulted in the isolation of a single 10-kDa protein that co-purified with Balb/c 3T3 cell mitogenic activity. The levels of mitogenic activity were directly correlated with those of the 10-kDa protein, which was completely pure as shown by silver staining (data not shown) and Western blotting (see below). Collectively, the results from 18 individual HPLC purifications confirmed a direct, causative relationship between the 10-kDa protein(s) and the mitogenic activity of HBGF-0.8-P1 and -P2.
Analysis of the individual purification steps showed that 0.5-1.1 g of HBGF-0.8-P1 or -P2 were each purified from 342 mg of crude ULF protein and that 10 -22 activity units for HBGF-0.8-P1 or P2 were recovered after the first HPLC step as compared with 66,666 units in 1 liter of starting material (Table II). It should be noted that the apparent low recovery of HBGF-0.8 activity was attributable to (i) a major contribution by IGF, EGF, PDGF, bFGF, HB-EGF, and PTN to the overall 3T3 cell mitogenic activity of the crude and partially purified samples (2,8,9,12,(25)(26)(27) and (ii) acid lability of HBGF-0.8 mitogenic activity during the HPLC separation step(s). Al-though alternative strategies were attempted to recover purified growth factors of higher specific activity, it was not possible to avoid the use of either reverse-phase HPLC or trifluoroacetic acid for ion pairing without compromising the purity of the final product. While, in terms of their biological activity, recovery of HBGF-0.8-P1 and -P2 was somewhat compromised, structural characterization of the proteins was readily achieved, since they retained sufficient activity to be unequivocally attributable to a single, homogenous 10-kDa band in SDS-polyacrylamide gels, and adequate quantities of each protein were isolated from several liters of ULF (Fig. 3, Table II).
Fractions containing the HPLC-purified growth factors were pooled, dried, and subjected to preparative SDS-PAGE, after which proteins in the gel were transferred to a polvinylidene difluoride membrane, and the position on the blots of HBGF-0.8-P1 or -P2 proteins was determined by staining with Coomassie R250. Each 10-kDa band was excised and submitted for N-terminal microsequencing with the result that a 16-residue sequence was obtained for HBGF-0.8-P1 with an undetermined residue at position 10, and a 12-residue sequence was obtained for HBGF-0.8-P2 with an undetermined residue at position 9 (Table III). These data showed that HBGF-0.8-P1 and -P2 were N-terminally identical except for the presence of an additional Glu residue at the N terminus of HBGF-0.8-P1. A search of GenBank TM revealed that these sequences aligned perfectly with predicted internal sequences of hCTGF and mouse fisp-12 (also termed ␤IG-M2), the murine homologue of CTGF (28 -30). The unassigned residue in cycle 10 of HBGF-0.8-P1 and cycle 9 of HBGF-0.8-P2 corresponded to Cys 256 of hCTGF and Cys 255 of fisp-12 (Table III).
To verify that the partial sequences of HBGF-0.8-P1 and-P2 were actually present in the pCTGF molecule, a full-length pCTGF cDNA was isolated by hybridization screening of a pig endometrial cDNA library using a 32 P-labeled hCTGF probe. The cloned pig CTGF cDNA was determined to be 1.51 kilobase pairs, with an open reading frame of 1,047 base pairs (data not shown). The primary translational product of pCTGF is predicted to comprise 349 amino acids and contains HBGF-0.8 sequence between residues 247 and 262 (Table III). At the amino acid level, pCTGF is ϳ92% identical to fisp-12 and hCTGF. After cleavage of its presumptive 26-residue signal peptide, pCTGF is predicted to comprise 323 amino acids and to contain 38 Cys residues that are fully conserved in hCTGF and fisp-12.
Since HBGF-0.8 proteins represented microheterogeneous forms of truncated CTGF, we wished to address the question of whether their presence in the purified samples was attributable to breakdown of 38-kDa CTGF during the purification process itself. As shown in Fig. 4, the presence of 10-kDa CTGF in the starting material was confirmed by Western blotting of unfractionated ULF samples using a CTGF antibody that reacted with HPLC-purified HBGF-0.8. While two additional mass forms of CTGF (16 and 20 kDa) were also present in ULF, convincing evidence for 38-kDa CTGF was not obtained. All three immunoreactive proteins were shown to bind to heparin-Sepharose in the presence of 0.5 M NaCl (Fig. 4). The Western blot further verified that HPLC-purified HBGF-0.8-P1 comprised a single immunoreactive 10-kDa protein (lanes A and B). Comparison of the staining intensity of 10-kDa CTGF from defined volumes of undiluted uterine fluid (i.e. 0.7-2.3 l) with the staining intensity of mitogenic amounts of HBGF-0.8 (Fig.  4) showed that highly mitogenic concentrations of 10-kDa CTGF exist in uterine fluid in vivo. Taken together, these data suggested that ULF contained nondetectable levels of 38-kDa CTGF and that 10-kDa CTGF was most likely identified in these studies due to its natural occurrence in utero.
Heparin-binding Properties of HBGF-0.8 -Previous studies have shown that heparin modulates receptor binding and biological activity of several HBGFs including bFGF, HB-EGF, and amphiregulin (31)(32)(33)(34)(35)(36)(37). Since HBGF-0.8 exhibited strong affinity for heparin, we examined the effect of this glycosaminoglycan on the mitogenic activity of HBGF-0.8. As shown in Fig. 5, the activity of a high stimulatory dose of HBGF-0.8 was significantly potentiated by 1-3 g/ml heparin but was inhibited by 30 -100 g/ml heparin. The same heparin doses had no effect on basal or calf serum-stimulated DNA synthesis in 3T3 cells (Fig. 5).
The presence of an additional acidic Glu residue at the N terminus of HBGF-0.8-P1 was correlated with the lower heparin affinity of this molecule as compared with HBGF-0.8-P2, suggesting that the N terminus of HBGF-0.8 may be part of a heparin-binding domain. To test the heparin-binding properties of the N-terminal region as well as other portions of the CTGF molecule, the ability of 18 peptides spanning the entire C-terminal 103 residues of hCTGF to bind [ 3 H]heparin was investigated. Table IV shows that the highest level of heparin binding was obtained for peptides containing residues 247-260, 274 -286, and 305-328. Although the functional significance of heparin binding of these domains requires further investigation, none of the peptides had HBGF-0.8 agonist or antagonist activity in the 3T3 cell DNA synthesis assay (data not shown). DISCUSSION Although the presence of CTGF-like activity in uterine flushings was initially demonstrated several years ago (2,8,38), its identification and structural characterization was not possible until the development of the purification protocol described herein, which allowed for its separation from other unrelated HBGFs such as PDGF, bFGF, HB-EGF, and PTN (2,8,12). The data presented in this report show that a fibroblast and smooth muscle cell growth factor in uterine secretory fluids is a CTGF protein that is heparin-binding and substantially N-terminally truncated in comparison with the predicted CTGF primary translation product. CTGF belongs to a recently recognized immediate early gene family, members of which are transcribed within minutes of stimulation by growth factors such as After application of each sample, the column was washed with 10% acetonitrile in water, 0.1% trifluoroacetic acid from 0 to 10 min and then treated with a 10 -90% acetonitrile gradient in water from 10 to 146 min. The flow rate was 1 ml/min, and 0.5-ml fractions were collected 20 min after the samples were injected. 80 l of selected fractions were dried and redissolved in 25 l of 10 mM Tris-HCl (pH 7.4), and 10 l were tested for their ability to stimulate 3T3 cell DNA synthesis. The insets show silver-stained SDS-PAGE analysis of successive active fractions from each HPLC separation. 10 l of the appropriate reconstituted fractions were mixed with 10 l of 2 ϫ SDS-PAGE sample buffer containing 50 mM dithiothreitol and electrophoresed in an 18% polyacrylamide minigel for 1 h at 200 V. Proteins in the gel were visualized with silver staining as described (19). The arrow indicates the 10-kDa protein that co-eluted with mitogenic activity. transforming growth factor-␤ (30,39). Based on the homologies of their encoded proteins, the members of this "CCN" family constitute three distinct paralogues, namely CTGF, cyr61, and nov (40). The 100% homology between the N termini of both HBGF-0.8 isoforms and internal pCTGF sequences shows that the purified proteins are highly N-terminally truncated forms of CTGF. Of importance is that HBGF-0.8-P1 and -P2 would be predicted to be of approximate M r 11,500 -12,000 if they corresponded to the C-terminal 102-103 residues of the primary pCTGF translation product, a size very similar to that determined experimentally by SDS-PAGE and gel filtration. While there are several mechanisms by which a 10-kDa CTGF protein could arise, the most obvious possibilities are that either (i) it is a processed form of the CTGF primary translational product or (ii) it arises through differential splicing of the CTGF gene.
Although not yet isolated or structurally characterized, the 20and 16-kDa CTGF proteins seen by Western blotting of ULF may be intermediates in the proteolytic cleavage of 38-kDa CTGF to 10-kDa CTGF. Also, by analogy with the genes for hCTGF, fisp-12 and nov, there is no intron/exon boundary that could directly give rise to the N terminus of the truncated pCTGF proteins (29,41,42), suggesting that alternative splicing alone is unlikely to be responsible for their production. Although the mechanism is as yet undefined, 10 -12-kDa forms of immunoreactive CTGF are also present in conditioned me-dium from cultured mouse or human fibroblasts, 2 suggesting that production of low mass CTGF is common to several species and biological systems. The relationship between the predicted pCTGF primary translational product and the truncated uterine CTGF proteins is shown schematically in Fig. 6. We have yet to determine whether our lack of detection of 38-kDa CTGF reflects its actual absence from ULF or its immediate processing after secretion or is a result of differences between the 38and 10-kDa forms of CTGF in their immunoreactivity, chromatographic behavior, and/or mitogenic activity. Although CTGF mitogenic and chemotactic activities were reported to correlate with the presence of 36 -39-kDa proteins in human endothelial cell conditioned medium (28), these proteins were not individually isolated, tested for biological activity, or structurally characterized. Also, while the mouse CTGF gene product has been characterized as a nonglycosylated 35-kDa protein that is rapidly produced in and secreted by NIH 3T3 cells after serum stimulation (29), its intrinsic growth factor activity has not been reported. The susceptibility of CTGF to acid inactivation suggests that 10-kDa HBGF-0.8 proteins are likely to be more potent in their native state than after purification (i.e. ED 50 of 25 ng/ml for twice HPLC-purified HBGF-0.8-P1; see Table II). However, even in comparison with the purified proteins, Western blotting showed that 10-kDa CTGF is present in uterine secretions at levels that are predicted to be highly mitogenic, which is strongly suggestive of a physiological role for this protein in utero. Native hCTGF, partially purified on a PDGF immunoaffinity column, was reported to stimulate DNA synthesis in NIH 3T3 cells with an ED 50 of ϳ10 ng/ml (28). A similar ED 50 was shown for the 2  Glu-Glu-Asn-Ile-Lys-Lys-Gly-Lys-Lys-Cys-Ile-Arg-Thr-Pro-Lys-Ile a Repetitive yield ϭ 88%; initial yield ϭ 7 pmol. b Repetitive yield ϭ 90%; initial yield ϭ 3 pmol. c See Bradham et al. (28). d See Ryseck et al. (29). e From cDNA analysis in this study. stimulation of mitosis in normal rat kidney 49F fibroblasts by 38-kDa hCTGF produced in baculovirus, although contaminating proteins were present in the heparin-purified recombinant protein sample (43). Therefore, with the caveats associated with the use by other investigators of different target cells and incomplete purification methods, the CTGF fragments isolated in this study appear to have activity comparable with that of the full-length protein. However, the same isolation strategy, cellular source, and bioassay will need to be utilized in the future for accurate determination of the relative specific activities of 38-and 10-kDa CTGF. Regarding previous studies of the bioactivity of 38-kDa CTGF (28,43), it will be of interest to determine whether these growth factor preparations actually contained hitherto unrecognized 10-kDa CTGF and whether 38-kDa CTGF is processed to low mass forms prior to activating target cells in vitro. Whatever the relative mitogenic properties of each form of CTGF, the two proteins may exhibit additional differences (e.g. protease sensitivity, half-life, matrix interactions) that could impact their efficacies in vivo. For example, expression in chicken embryo fibroblasts of a 63-amino acidtruncated version of the CTGF-related nov gene resulted in cell transformation, whereas expression of full-length nov did not (41).
Recently, 38-kDa hCTGF was reported to bind to heparin-Sepharose, although the salt concentration required for its elution was not stated (43). However, the 41-kDa cyr61 gene product showed peak elution from heparin-agarose with 0.8 M NaCl (44), a salt concentration identical to that required for elution of 10-kDa CTGF from immobilized heparin. Based on sequence homology with thrombospondin, properdin, TRAP, f-spondin, UNC5, and several complement and cicumsporozoite proteins, Bork (40) proposed a binding motif (WSXCSXXCG) for sulfated glycoconjugates between residues 206 and 214 of hCTGF. However, the absence of this motif in the truncated forms of CTGF identified in this study suggests that additional heparin-binding domains exist in the C-terminal 102 residues of CTGF. The N terminus of HBGF-0.8-P1 and -P2 may be involved in heparin-binding, since the two proteins displayed differential binding to heparin yet differ only by a single Nterminal Glu; the lack of this acidic residue may facilitate the binding of heparin to basic residues located in the N-terminal region. There is a very high proportion (50%) of basic amino acids (K, R) in the 17-residue region CTGF-(251-267), a composition that is well conserved in the corresponding regions of all CCN family members. Residues 250 -255 (IKKGKK) of pCTGF conform to a proposed heparin recognition consensus sequence (XBBXBX, where a basic residue is strongly favored at the B position and a nonbasic residue is usually, but not absolutely, found at the X position) that occurs in other heparin-binding proteins (45). While this sequence may account, at least partly, for the binding of [ 3 H]heparin to CTGF-(247-260), the lack of binding of [ 3 H]heparin to CTGF-(247-255) suggests that residues 256 -260 may act cooperatively with the proposed consensus sequence in binding heparin. Our data further highlights a potential contribution from other regions (residues 274 -286 and 305-328) to the overall net heparin-binding characteristics of 10-kDa CTGF. While these domains require further study, the heparin dependence of HBGF-0.8 bioactivity suggests that CTGF-heparin interactions are functionally significant. Although 10 g/ml heparin was recently shown to stimulate the mitogenic activity of partially purified recombinant 38-kDa hCTGF (43), our finding of a biphasic effect of heparin on CTGF activity has not previously been described and suggests that the heparin:CTGF ratio is an important determinant of net CTGF mitogenic activity. Additionally, cyr61 is present in the extracellular matrix, where it may be stabilized by association with heparan sulfate proteoglycans (44).
The predicted sequences of all CCN members conform to four distinct structural modules that span the entirety of each molecule, suggesting that all of the family members are mosaic proteins (40). Based on our cDNA analysis, these modules are fully conserved in the pCTGF molecule (Fig. 6). The fourth module, which spans residues 256 -330 of hCTGF and pCTGF and residues 255-329 of mouse CTGF, has been termed the C-terminal (CT) module and is also present in a variety of unrelated extracellular mosaic proteins including slit, Ndp, Von Willebrand factor, and several mucins (40). Based on predictions for slit and the CCN family (40), the 10 cysteine residues in the CT module conform to a model of five intrachain disulfide bridges, three of which may adopt the "cystine knot" motif that occurs in a structural superfamily of dimeric growth factors including nerve growth factor, TGF-␤, and PDGF-B (40,46,47). Further, it has been suggested that the CT module may also be involved in receptor binding (40), a possibility that is supported by our data, since the CT module is entirely con- FIG. 6. Relationship between HBGF-0.8 proteins and pCTGF primary translational product. HBGF-0.8-P1 and -P2 are shown aligned relative to one another and to the pCTGF primary translational product. The question mark indicates that the C termini of HBGF-0.8-P1 and -P2 have yet to be established, although the apparent molecular masses of the HBGF-0.8 proteins are consistent with the same or similar C terminus as in the primary translational product. Domains within the primary translational product include the signal peptide (SP) and four structural modules that show structural similarity to an IGF binding region (module I), a von Willebrand factor type C repeat (module II), a sulfated glycoconjugate binding motif (module III), and a C-terminal (CT) dimerization/receptor binding domain (module IV) (40).  tained within the truncated uterine CTGF protein.
In conclusion, we have shown that (i) heparin-binding forms of CTGF of M r 10,000, 16,000, and 20,000 are present in uterine flushings; (ii) 10-kDa CTGF corresponds to two microheterogeneous N-terminally truncated forms of the primary translational product, both of which are biologically active; (iii) mitogenic activity of 10-kDa CTGF is heparin-dependent; and (iv) receptor-binding and heparin-binding domains of CTGF lie in the C-terminal 103 residues of the molecule. Since CTGF is a uterine secretory growth factor, it may play a role in growth and remodeling of the endometrium and, during pregnancy, growth and development of the placental membranes. Future work will involve investigations of the physiological role of CTGF in female reproductive tract function and embryogenesis, as well as detailed structure-function studies of CTGF and determination of the mechanism of production of 10-kDa CTGF.