Identification of the sites of phosphorylation in insulin-like growth factor binding protein-1. Regulation of its affinity by phosphorylation of serine 101.

Serine phosphorylation of insulin-like growth factor binding protein-1 (IGFBP-1) has been shown to alter its affinity for the insulin-like growth factors (IGF-I and IGF-II) and to modify its capacity to modulate cellular responses to the IGFs. Because of this, we determined the sites of serine phosphorylation. Purification of 32P-labeled IGFBP-1 was followed by digestion with trypsin and endoproteinase Glu-C and radiosequencing of labeled peptides. Three serines were found to be phosphorylated, with Ser101, Ser119, and Ser169 containing 70%, 5%, and 25% of the incorporated 32P, respectively. A mutated IGFBP-1, substituting alanine for serine at positions 98 and 101, was expressed in CHO cells. On nondenaturing gels, the wild type protein migrated as five isoforms (one non-phosphorylated and four phosphorylated). However, in the mutated protein, the most rapidly migrating band (a phosphorylated form) was not present. The cells containing the mutated cDNA incorporated 60% less 32P into immunoprecipitable IGFBP-1. The mutated protein had a 3-fold reduction in affinity for IGF-I compared to the wild type protein. We conclude that Ser101 represents the major site of phosphorylation containing 63% of the total 32P incorporated and that phosphorylation of Ser101 is important for maintenance of high affinity binding for this growth factor.

(4). Both forms have the same amino acid sequence. In subsequent studies, two groups of investigators have shown that IGFBP-1 is phosphorylated on serine residues (5, 6). Serine phosphorylation results in a 6-fold increase in the affinity of IGFBP-1 for IGF-I. Furthermore, phosphorylation appears to account for the major difference between the two isoforms of IGFBP-1, since the stimulatory form of IGFBP-1 is mostly nonphosphorylated. Likewise, recombinant bacterial derived IGFBP-1, which is nonphosphorylated ( 5 ) , can potentiate IGF-I action in accelerating wound healing (7). In contrast, IGFBP-1 that is purified from HepG2 cells is predominantly phosphorylated (5) and inhibits IGF-I-stimulated growth in chondrocytes (8). Frost and Tseng (6) have shown that in proliferative phase endometrium, IGFBP-1 is secreted as a dephosphorylated form, whereas during the secretory nonproliferative phase, the endometrium secretes a form of IGFBP-1 that is heavily phosphorylated. To date, the only other post-translational modification reported for IGFBP-1 is the capacity of IGFBP-1 purified from amniotic fluid to form disulfide-linked multimers (9). Therefore, phosphorylation may be an important post-translational modification that regulates the capacity of IGFBP-1 to modulate IGF bioactivity. These studies were undertaken to determine the site(s) of phosphorylation in IGFBP-1 and to determine if mutation of a major site of phosphorylation results in an alteration in affinity for the growth factor.

MATERIALS AND METHODS
Chinese hamster ovary cells (CHO-K1) cells were obtained from the American Type Tissue Collection. Rabbit antiserum for human IGFBP-1 was prepared as previously described (10).
[32P]Orthophosphoric acid (8500-9120 Ci/mmol) was purchased from Du Pont-New England Nuclear. Polyvinylidene difluoride (PVDF) transfer membranes were obtained from Millipore. Phenyl-Sepharose CL4B was purchased from Pharmacia LKB Biotechnology Inc. The reverse phase C-4 (0.46 x 3 cm) and C-8 (0.46 X 25 cm) HPLC columns were purchased from Vydac (Hespernia, CAI. The (2-18 (0.46 X 10 cm) and DEAE 5PW columns were purchased from Waters Associates. Acetonitrile, gel-coded silver stain kits, and trifluoroacetic acid were purchased from Pierce Chemical Co. SDS and Vent polymerase were obtained from Bethesda Research Laboratories (Gaithersburg, MD). Tissue culture plates were purchased from Falcon LabWare Division, Becton-Dickinson (Oxnard, CA). a-Modified Eagle's minimal essential medium, phosphate-free Eagle's minimal essential medium, nonessential amino acids, gentamicin, and fetal bovine serum were purchased from Grand Island Biological Co. Recombinant human IGF-I was obtained from Bachem, Inc. lz5I-IGF-I (150-250 pCi/wg) was prepared by a previously described method (23). Agarose (Seakem) was obtained from FMC BioProducts. Reagents used in cDNA sequencing were purchased from U. S. Biochemicals, and reagents and the thermocycler used in the polymerase chain reaction (PCR) were obtained from Perkin-Elmer Cetus Instruments. [cx-~*P]~CTP and the ECL chemiluminescence detection system were purchased from 1125 This is an Open Access article under the CC BY license.
Amersham Corp. The Bluescript plasmid was obtained from Stratagene. Trypsin treated with ~-(tosylamido-2-phenyl)ethyl chloromethyl ketone was from Worthington Biochemical. Calf intestinal alkaline phosphatase and endoproteinase Glu-C (from Staphylococcus aureus V8) were purchased from Boehringer Mannheim. DNA restriction endonucleases were purchased from Promega. All other chemicals except those listed were purchased from Sigma. Sequelon discs were obtained from Milligen.
Tissue Culture-CHO cells were maintained in a-modified Eagle's minimal essential medium containing 10% fetal bovine serum and 50 rg/ml gentamicin. An expression vector containing the cDNA encoding human IGFBP-1 was constructed by ligating a human IGFBP-1 cDNA (11) into a PUC18-derived plasmid (pNUT, a gift of Richard Palmiter, University of Washington) as previously described (5). Transcription of the cDNA was driven by a mouse metallothionine promoter utilizing polyadenylation signals and insertion sites derived from the human growth hormone gene. pNUT contains the gene for dihydrofolate reductase driven by an SV-40 promoter allowing selection of successfully transfected cells with methotrexate. CHO cells were transfected with the expression vector by calcium phosphate precipitation followed by glycerol shock (12). Stable methotrexateresistant cell lines were established by passaging in dialyzed serum and by increasing the concentration of methotrexate in the medium at regular intervals. The cell line designated CHOBP1-D6 secreting the greatest amount of human IGFBP-1 was used as a source of phosphorylated IGFBP-1 in all of the experiments. The cell line designated CHOBP1-SA 98-101-B contained the same expression vector with the serine to alanine mutations described below. After selection, both cell lines were stably maintained in a-modified Eagle's minimal essential medium supplemented with 10% dialyzed fetal bovine serum and 50 p~ methotrexate.
Purification of Radiolabeled IGFBP-Radiolabeled conditioned medium was prepared by labeling each of the two transfected cell lines, grown to confluency in 175-cm2 tissue culture dishes. The cultures were rinsed with phosphate-free Eagle's minimal essential medium containing 0.05% bovine serum albumin (BSA) and nonessential amino acids and then incubated for 16 h in the same media to which 50 pCi/ml [32P]orthophosphate was added. Following this interval, conditioned medium was harvested and centrifuged and diluted with an equal volume of 0.05 M NaHP04, 2 mM EDTA, 0.1 M NaCI, pH 6.6. The solution was then loaded onto an IGF-I affinity column which had been prepared by linking IGF-I to Sepharose, as described previously (13). After sample loading, the column was washed with the loading buffer until absorbance (280 nm) returned to baseline. The proteins were eluted with 0.5 M acetic acid. This solution was applied directly to a HPLC C-4 column (0.46 X 3 cm). A single peak containing the radiolabeled protein was eluted using a linear acetonitrile gradient from 0-100% in 0.4% trifluoroacetic acid. Total recovery of 32P-labeled IGFBP-1 after these steps varied between 80 and 90%. Material eluting from the final peak was proven to be IGFBP-1 by immunoblotting. The final product (60 pg) was denatured and reduced in the presence of 5 M guanidine HCl and 0.1 M dithiothreitol, pH 8.6, for 90 min at 37 'C. Alkylation was then performed by adding iodoacetic acid for a final concentration of 0.11 M and incubating the mixture in the dark for 1 h at 22 "C. The alkylated protein was separated from these reagents by rechromatographing it over the reverse phase HPLC (C4) column as described previously except that the linear gradient was extended to 40 min. To obtain tryptic peptides, the radiolabeled intact protein that had been reduced and alkylated was incubated with 2 units/ml bovine trypsin in 0.13 M ammonium bicarbonate, pH 7.8, at 37 "C for 3 h. The pH of the solution containing the cleavage products was immediately adjusted to pH 6.0 with acetic acid, then diluted in an equal volume of 0.04% trifluoroacetic acid and applied to a reverse phase HPLC, C-8 column. Nine peaks were eluted using a linear acetonitrile gradient from 0 to 70% over 60 min, and the amount of 32P in each peak was quantitated. An aliquot of each of the three peaks containing radioactivity (fractions 19,32, and 34) was sequenced by Edman degradation to determine the location of serine residues, using an Applied Biosystems Model 470A Sequencer as described previously (13). In addition, the material that was not retained by the C-8 column was reapplied to a HPLC, C-18 column. Six peaks were eluted, using a Iinear gradient from 0-100% acetonitrile over 40 min, but only one contained a radiolabeled peptide, and this peptide was also sequenced. The fraction from the tryptic digest containing the greatest amount of radioactivity (fraction 34) was concentrated by lyophilization, incubated with 4.0 pg of staphylococcal endoproteinase Glu-C reconstituted in 0.25 M ammonium carbonate, pH 7.8, and incubated for 18 h at room temperature. The pH of the solution containing the cleavage products was adjusted to pH 6.0 with 1 M acetic acid, and the peptides were separated using a reverse phase HPLC (C-18) column as previously described (14) and sequenced. In an additional experiment, endoproteinase Glu-C digestion was performed using the intact reduced and alkylated protein. The resultant peptides were separated as described previously and sequenced by Edman degradation (13). To identify the sites of phosphorylation, the 32P-labeled peptides were subjected to manual radiosequencing by the method of Sullivan and Wong (15). First, the peptides were applied to arylamine-Sequelon discs and covalently attached by incubation of the disc in a solution of 1.0 M 4morpholineethanesulfonic acid, pH 5.0, for 30 min at room temperature. The disc was then extracted five times with neat trifluoroacetic acid at 50 "C. Each peptide was subjected to repetitive cycles of manual Edman degradation, and the amount of 32P bound and released from the disc was determined after each degradation cycle. Three radiolabeled peptides from the endoproteinase Glu-C digests and two from the tryptic digest were analyzed using this method. The endoproteinase Glu-C peaks numbered 8,9,10,11,12,22,24,25, and 29 ( Fig. 1) were sequenced by Edman degradation. Some of these contained no radioactivity, but were sequenced to identify the positions of the serines that were not labeled.
Purification of IGFBP-1-To assess the electrophoretic properties and binding capacity of wild type and [Ala98~10']IGFBP-1, the proteins secreted by the transfected CHO cells were purified. Approximately 200 ml of serum-free conditioned medium was collected and used to purify each protein. The medium was loaded onto a 2.0 X 4.4-cm phenyl-Sepharose (CL-4B) column that had been previously equilibrated with 0.05 M sodium phosphate, 0.1 M NaCl, 2 mM EDTA, pH 6.6. After sample loading, the column was washed with the equilibration buffer until the absorbance (A280 " , . J returned to baseline. The proteins were eluted using a 0 to 15% acetonitrile gradient in 0.05 M NaHPO,, 2 mM EDTA, pH 6.6, over 30 min (16). The fractions containing IGFBP-1 were concentrated by lyophilization and loaded onto an IGF affinity column. The remaining purification steps were as described for radiolabeled IGFBP-1. Purity was determined by SDS-PAGE with silver staining (17). IGFBP-1 secreted by HepG2 cells was purified as previously described (5). The protein content of the purified forms of IGFBP-1 was determined by radioimmunoassay Preparation of [Ala98J01]IGFBP-l Expression Vector-Two mutations (SerS8 and Ser"' to alanine) were introduced into human IGFBP-1 using the PCR strategy described by Higuchi (18). A 26base DNA oligonucleotide (5"GGAATTCC CTG GAG ATG TCA GAG GTC-3', arbitrarily designated GW19), and a 26-base oligonucleotide (5"GGAATTCC CAT CTG GTT TCA GTT TTG-3', designated GW20) were designed as outside primers. These primers (GW19 and -20) each had 18 bases that were complementary to the IGFBP-1 sequence (14) at bases 118-135 in the 5' to 3' orientation and bases 909-892 in the 3' to 5' orientation, respectively, and each contained an additional 8-base pair sequence that contains an EcoRI restriction site (italicized). Two 29-base primers (5'-CCT GAA GCC CCA GAG GCC ACG GAG ATA AC-3', designated GW21) and (5' AT CTC CGT GGC CTC TGG GGC TTC AGG GCT 3', designated GW22) were designed to be complementary to the IGFBP-1 sequence at bases 481-509 in the 5' to 3' orientation and bases 506-478 in the 3' to 5' orientation, respectively, except that they contained mutations that change amino acids 98 and 101 from serine to alanine (mutations italicized).
By using GW19 with GW22 and GW20 with GW21 in separate PCR reactions with the wild type IGFBP-1 cDNA ligated into pBS as a template, we amplified 392-and 432-base pair fragments, respectively. To obtain those products, 5 ng of template was amplified by PCR in a 50-pl reaction, containing 10 mM KCl, 10 mM (NHI)PSOI, 20 mM Tris-HC1, pH 8.8, 2 mM MgSO,, 0.1% Triton X-100, 0.1 mg/ ml BSA, 0.8 mM concentration of each dNTP, 0.1 pM concentration of each primer. The reaction mixture was covered with mineral oil and heated for 5 min at 97 'C, and 1 unit of Vent DNA polymerase was added. The first 10 cycles were performed with a denaturing temperature of 97 "C. The remaining 21 cycles were done with a denaturing temperature of 95 "C. Annealing and extension temperatures were 55 "C and 72 "C, respectively, in all cycles.
The two DNA fragments that were prepared were analyzed and purified as described previously (19). These fragments each extended from the outside primers to the regions of the mutations, where they each contained the Alag8*'0' substitution and overlapped 26 bases. The products were then mixed together in equal amounts (100 ng each), and PCR was performed under reaction conditions as described above, of IGFBP-1 (10).
using primers GW19 and GW20. Twenty-one cycles were completed with denaturing, annealing, and extension temperatures of 97 "C, 50 "C, and 72 "C, respectively.
The final product extended the length of the IGFBP-1 cDNA (795 base pairs, including the entire protein coding region) and contained Alag8J01 mutations and EcoRI sites at both ends. It was digested with EcoRI and gel-purified as described previously (19). The overhangs were then filled in with the Klenow fragment of DNA polymerase I to create blunt ends in a 4.0-p1 reaction volume containing 50 mM NaC1, 10 mM Tris-C1, pH 7.5, 10 mM MgC12, 1 mM dithiothreitol, 0.05 mM dATP, 0.05 mM dTTP, and 2.5 units of Klenow fragment. After heat inactivation for 10 min at 70 "C, the DNA was ligated directly into pNUT, which had been linearized with S m I and dephosphorylated (19). The blunt end ligation was performed at 16 "C with a vector-to-insert molar ratio of 1:4. Ligation, transfection of Escherichia coli (strain DH-~cI), and analysis of clones was performed as described previously (19).
The enzymes used for determining orientation of the inserts were EcoRI/BamHI, and this was carried out as previously described (19). The clones containing the correctly oriented inserts were sequenced using the dideoxy sequencing method (20). The final product contained the correct mutation while the rest of the sequence was identical with the sequence published for human IGFBP-1 (11). A clone containing the correct sequence in pNUT was transfected into CHO cells, and the IGFBP-1 expressing clones were selected with methotrexate as described above.
Immunoprecipitation of Phosphorylated Products-To determine the degree of phosphorylation in wild type and mutated IGFBP-1, each of the transfected cell lines was incubated in phosphate-free Eagle's minimal essential medium containing 50 pCi of [3ZP]orthophosphate for 6 h. Duplicate wells were treated identically except that 32P was omitted and the amount of IGFBP-1 contained in the conditioned media was determined by radioimmunoassay (10). These results were used to estimate the volume of 32P-labeled conditioned media that contained 100 ng of IGFBP-1, and this volume was transferred to siliconized tubes. The volume in each tube was adjusted to 1.0 ml with a buffer containing 30 mM Na2P04, pH 7.5, 50 mM sodium fluoride, 10 mM EDTA, 0.1% BSA, 0.01% Tween 20, and 1.0 pl of rabbit polyclonal anti-human IGFBP-1 antiserum was added. After overnight incubation at 4 "C, 8 pl of ovine antisera to rabbit IgG was added to each tube, and they were incubated for 1 h at 4 "C, followed by the addition of 2 pl of normal rabbit serum and incubation for a 1 h more. Antibody-bound IGFBP-1 was separated by centrifugation at 8000 X g for 20 min. The immunoprecipitated pellets were washed three times, dissolved in Laemmli sample buffer (containing 1% SDS and 0.1 M dithiothreitol), and boiled for 10 min. The immunoprecipitated proteins were resolved by 12.5% SDS-PAGE, transferred to PVDF filters, and subjected to autoradiography. In order to confirm that equivalent amounts of [Ala98*101]IGFBP-1 and wild type IGFBP-1 were present on the filters, they were immunoblotted using a 1:lOOO dilution of anti-IGFBP-1 antisera. The immune complexes were detected using chemiluminescence with an Amersham ECL substrate according to the manufacturer's directions.
Nondenuturing Gel Electrophoresis-In order to separate nonphosphorylated IGFBP-1 from phosphorylated forms, nondenaturing PAGE was performed as previously described (5). Briefly, a discontinuous buffer system was used at a reduced pH (8.3) compared to the Laemmli system and with an acrylamide concentration of 10% in resolving gel and 4% in the stacking gel. The sample buffer and both the stacking and nonstacking gels contained n-octyl glucoside at 20 mM. Electrophoresis was performed at 25 "C with tap water cooling. The proteins were transferred to PVDF membranes by electroblotting, and the transfer membranes then were subjected to autoradiography (in the case of 32P-containing samples) or immunoblotting as previously described (5).
Anion Exchange Chromatography-IGFBP-1 isoforms were separated using anion exchange chromatography. Wild type or mutant IGFBP-1 (25 fig) were diluted in equilibrating buffer (20 mM (NH&CO$, pH 8.2) and loaded onto the DEAE-anion exchange column. Protein absorbance was monitored at 214 nm. After the absorbance returned to baseline, isoforms of IGFBP-1 were eluted with buffer B (20 mM (NH&CO3, 1 M NaC1, pH 6.8) using a linear gradient to 12% B over 10 min, followed by a linear gradient to 25% B over the next 30 min and a final 10-min linear gradient to 100% B. IGFBP-1 eluted between 19 and 28% B, and these fractions were analyzed for variably phosphorylated isoforms by nondenaturing gel electrophoresis.
Assay of IGFBP-I Binding of IGF-I-To determine the binding affinities of the phosphorylated and nonphosphorylated IGFBP- MgC12, 0.5 mM ZnC12) for 16 h at 37 "C, and the reaction products were purified by reverse phase HPLC (C-4 column). Ten ng of each of the purified forms of IGFBP-1 were added to 0.25 ml of binding assay buffer containing 0.1 M HEPES, 44 mM sodium phosphate, 0.01% Triton X-100, 0.1% BSA, pH 7.0, in the presence of unlabeled IGF-I (0-100 ng/ml). The Iz5I-IGF-I that was bound was separated from unbound by precipitation in 12.5% polyethylene glycol (4). Nonspecific binding was determined in the presence of 1 pg/ml IGF-I and constituted less than 15% of the bound radiolabeled material.

RESULTS
To determine the site of serine phosphorylation, wild type CHO cells were incubated with [32P]orthophosphate, and the radiolabeled, secreted protein was purified to homogeneity. The pure product was reduced and alkylated then exposed to trypsin, and the released peptides were separated by HPLC. Three radiolabeled peaks (fractions 19, 32, and 34) were detected containing 2, 10, and 65% of the total radioactivity, respectively. Amino acid sequence analysis showed that fractions 32 and 34 had the same N-terminal sequence. Fraction 19 was completely sequenced and contained 2 serines at positions 169 and 174 (Table I). This experiment was repeated three times, and each time fraction 19 contained radioactivity, but it was consistently <2% of the total incorporated 32P.
However, approximately 23% of the total radiolabeled protein did not adhere to the HPLC C-8 column. When this fallthrough material was rechromatographed on a C-18 column, a single radiolabeled peak was detected. This peptide was sequenced and shown to contain the same N-terminal sequence as fraction 19 (Table I). Radiosequencing of the fallthrough peptide showed that 80% of the total radioactivity was released from the filter when S e P 9 was cleaved from the peptide. This strongly supports that serine 169 is phosphorylated and is a major site of phosphorylation in IGFBP-1 ( i e . 25% of the total 32P incorporated). Because fraction 34, which contained >70% of the total radioactivity, contained 9 serines, endoproteinase Glu-C digestion was performed, and the endoproteinase Glu-C digestion products were further purified by HPLC. Four radiolabeled peaks were obtained (Fig. 1). All four were N-terminally sequenced and contained serine residues. Three were subjected to radiosequencing. One peak (fraction 12) was shown to contain the Serg5-Pro-Glu-SerSs-Pro-Glu-Ser"'-Thr-Glu sequence at positions 95-103, and radiosequencing showed that the serines at positions 95 and 98 were not labeled. In contrast, the serine at position 101 was shown to have releasable radioactivity after manual Edman degradation. Fraction 11 also contained the Ser''' as its only serine and was intensely labeled. These two peptides containing Ser"' (Fig. 1, fractions 11 and 12) had >85% of the radioactivity in the tryptic fraction 34. This supports the conclusion that Ser"' is a major site of phosphorylation in IGFBP-1. Fraction 24 from the HPLC column of the endoproteinase Glu-C digest contained <8% of the radioactivity. It had only a single serine residue at position 119. This indicates that serine 119 is definitely labeled but represents a minor site of phosphorylation. Fraction 29, which was labeled, contained a mixture of two sequences, one of which contained serine 119. Fraction 25 did not contain radioactivity and contained serines 124,131, and 136, suggesting that they were not labeled. Fraction 8 was minimally (<2%) labeled and contained Sers3 and S e P . The phosphorylation state of serines 83 and 86 was further analyzed in a separate digestion The amino acid sequences for the three 32P-labeled tryptic peptides that eluted from the HPLC C-8 column and for the 32P-labeled peptide isolated from the fall-through material of this column are shown in the upper panel. The first 15 amino acids of peptide 19 and the first 24 amino acids of peptide 32 were determined by sequencing. The lower panel shows the amino acid sequences for eight peptides that were isolated from the digestion of the tryptic peptide in fraction 34 of the upper panel. Peptides containing significant amounts of 32P are marked with an asterisk. The serine residues that were determined to be PhosRhorvlated are underlined.

34* " G Q G A C V Q E S D A S A P H A A E A G S P E S P E E j T E T E E
EndoDroteinase G1u-C DeDtides 8  . 1. HPLC profile of the endoproteinase Glu-C digested peptides. The tryptic peptide that contained >70% of the total radioadvity was subjected to digestion with endoproteinase Glu-C, and the peptides that were generated were separated by HPLC as described under "Materials and Methods." The fractions that were sequenced are numbered, and those containing 32P have an asterisk. The absorbance (Azl1 " , , , ) is shown as a solid line, and the acetonitrile gradient is plotted as a dashed line. No further peptides were eluted after the gradient reached 40% acetonitrile. experiment. Endoproteinase Glu-C digestion of the whole protein after radiolabeling followed by N-terminal sequencing showed that one fraction contained a peptide with serines at positions 83 and 86. Direct radiosequencing of this peptide showed no releasable radioactivity after Edman degradation of these residues. Additionally, the tryptic peptide fraction 34 was radiosequenced and showed no releasable 32P from the serines at positions 83 and 86. Continued release of amino acids by Edman degradation confirmed the release of radioactivity at the position corresponding to Ser' O' .

" ' L L D N F H L M A P S E E 24* ' 0 7 E E L L D N F H L M A P E j E 25 ' * ' D H S I L W D A I S T Y D G S K 29* 1 0 s L L D N F H L M A P E j E E D H S I L W D A I S T Y D G
CHO cells containing an expression vector for human IGFBP-1 had been shown previously to secrete high levels of IGFBP-1 (5). When these cells were labeled with [32P]orthophosphate, immunoprecipitation of an aliquot of conditioned medium containing 100 ng of IGFBP-1 demonstrated a 32Plabeled band corresponding in molecular weight to IGFBP-1 (Fig. 2, lune 1 ). If this immunoprecipitation was conducted in the presence of excess unlabeled IGFBP-1, the band intensity was markedly reduced, indicating that the immunoprecipitation was specific. In order to block phosphorylation of Ser'O', this residue was changed to alanine by site-directed mutagen- (lanes 1 and 3) or mutant (lanes 2 and 4 ) cDNAs were incubated with [32P]orthophosphate as described under "Materials and Methods." An equivalent amount (100 ng) of each radiolabeled protein was immunoprecipitated, and the immunoprecipitates were electrophoresed as described under "Materials and Methods" and then transferred to PVDF filters. The figure shows an autoradiograph of the 32P-labeled proteins (lanes 1 and 2) and an immunoblot of the identical filter (lanes 3 and 4 ) . The arrow denotes the immunoprecipitated band corresponding to IGFBP-1. Multiple other immunoreactive unlabeled bands are present in lanes 3 and 4, representing rabbit immunoglobulin and other rabbit serum components in the immunoprecipitate that react with the goat anti-rabbit IgG conjugated to horseradish peroxidase that was used for detection. esis of the IGFBP-1 expression vector. Because of an ambiguity in the initial phosphopeptide analysis experiments suggesting that Serg8 was also phosphorylated (later disproven as described above), we also changed this serine to alanine during the mutagenesis procedure. CHO cells were transfected with this construct containing base substitutions to encode alanines at positions 98 and 101. These cells also secreted phosphorylated IGFBP-1 into their conditioned media, and immunoprecipitation of 100 ng of [Alag8*'O']IGFBP-1 demonstrated a "P-labeled band of identical molecular weight (Fig.  2, lune 2). However, the 32P-labeled band immunoprecipitated from the media containing [Ala98*'o']IGFBP-1 was reduced in intensity 58% as determined by scanning densitometry, compared to the band immunoprecipitated from media of CHO cells transfected with the wild type cDNA. Immunoblotting followed by scanning densitometry of the chemiluminescenceexposed film showed that 13% more total (labeled plus unlabeled IGFBP-1) protein was present in the [AlagR*'o']IGFBP-1 lane. Therefore, there was a net decrease of 63% in 32P band intensity in the lane containing [Ala98~'o']IGFBP-1 compared to wild type IGFBP-1. These results strongly suggest that mutagenesis had successfully blocked phosphorylation and that serine 101 was a major site of phosphorylation in IGFBP-1, since a 63% reduction in phosphate radiolabeling was detected in spite of the fact that similar amounts of protein were immunoprecipitated.

FIG. 2. Immunoprecipitation of S2P-labeled IGFBP-1 from transfected CHO cell conditioned media. CHO cells that had been transfected with wild type
To further confirm that the cells containing the mutated cDNA secreted a form of IGFBP-1 whose phosphorylation state had been altered, the technique of nondenaturing gel electrophoresis was used. Conditioned medium from cells expressing the wild type protein and the [Ala98*'o']protein were electrophoresed using n-octyl glucoside instead of SDS to permit resolution of phosphorylated and nonphosphorylated IGFBP-1. As we have previously shown, IGFBP-1 from CHO cells secreting the wild type protein resolved into five bands (four phosphorylated and one nonphosphorylated) which could be visualized by immunoblotting (Fig. 3). The most rapidly migrating form, which is believed to be the most intensely phosphorylated, was easily detectable. However, when the media from the cells secreting [Alag8.'o']IGFBP-1 were analyzed, only four bands were detected. The most rapidly migrating 5th band was not visualized by immunoblotting (Fig. 3), and the intensity of the 4th band was significantly reduced. In contrast, the intensity of the upper nonphosphorylated band was increased. This suggests that the most heavily phosphorylated band migrates most rapidly through the gel and that it has been eliminated by mutagenesis. HepG2 cell-conditioned media contained an increased amount of the most rapidly migrating bands compared to the media from CHO cells, suggesting that in this cell type phosphorylation is more extensive.
The results were confirmed using anion exchange chromatography. Conditioned media from cells secreting the wild type and mutant proteins were subjected to anion exchange HPLC using a DEAE-anion exchange column as described under "Materials and Methods." The phosphorylated isoforms eluted in four peaks of activity with the first peak containing the dephosphorylated protein and lesser amounts of the three phosphorylated bands. The next three peaks contained a mixture of the three middle bands and lesser amounts of the remaining two bands. The last peak to elute contained the two most rapidly migrating bands. The fractions containing the [Alag8*'O']1GFBP-1 had a major reduction in the amount of protein in the last peak, indicating that loss of the Ser'O' phosphorylation site resulted in an alteration in charge (data not shown).
To determine the effect of phosphorylation of Ser'O' on the affinity of IGFBP-1 for IGF-I, the [Alag8*'o']IGFBP-1 was purified to homogeneity as described under "Materials and Methods.'' Homogeneity was proven by SDS-PAGE with silver staining. The capacity of increasing concentrations of IGF-I to compete for binding of '251-IGF-I to [Alag8*'O']IGFBP-1 and to native IGFBP-1 was compared. Scatchard plots obtained using each protein were linear. The slopes of the two lines were significantly different, however. The association constant (KO) of the wild type protein for IGF-I was 1.9 x lo9 M-'. In contrast, the K, of the [Ala98*'0'] form was 7.2 X 10' M" or approximately 3-fold lower (Fig. 4). Furthermore, the affinity of IGFBP-1 purified from HepG2 conditioned media, which contains more of the most rapidly migrating isoforms, was 2.5-fold greater (KO = 4.8 x lo9 M-') than the wild type CHO protein. To further confirm that serine phosphorylation resulted in changes in affinity, the wild type IGFBP-1 and [AlagR.'o']IGFBP-l purified from CHO media were each dephosphorylated with alkaline phosphatase, then repurified. IGF-I binding was assessed using both proteins. Scatchard analysis showed that dephosphorylation of wild type IGFBP-1 resulted in a 6-fold reduction in affinity, whereas dephos-  (lanes 3 and 4 ) and [Ala98~10']IGFBP-1 (lanes I and 2 ) before (lanes I and 3) and after (lanes 2 and 4 ) treatment with alkaline phosphatase. The arrow indicates the position of nonphosphorylated IGFBP-1. Lunes 2 and 4 demonstrate that dephosphorylation of each form of IGFBP-1 was complete and that the serine-to-alanine mutations did not alter mobility of the dephosphorylated protein in the nondenaturing gel. phorylation of [Alag8~"']IGFBP-1 lowered its affinity by only 2-fold (Fig. 5). After dephosphorylation, both the wild type and [Ala98.'o']IGFBP-1 proteins had identical affinities for IGF-I (K, = 3.4 X lo8 "*), indicating that the changes in the affinity for IGF-I as a result were due to differences in phosphorylation and were not due to conformational changes as a result of the replacement of the serines with alanines. The fact that the alanine substitutions did not alter protein conformation was confirmed by the observation that after dephosphorylation both the wild type and the mutated proteins had identical mobilities during nondenaturing gel electrophoresis (inset, Fig. 5, lunes 2 and 4 ) . DISCUSSION These studies demonstrate that IGFBP-1 is phosphorylated on serine residues at positions 101, 119, and 169. Phosphorylation of Ser"' accounts for greater than 60% of the 32P that is incorporated into the protein. Direct radiosequencing also showed that Ser16' was labeled and accounted for 25% of the 3zP incorporation. Ser"' was not directly radiosequenced due to the small amount (8%) of 32P incorporated, but complete sequencing of the labeled peptide in fraction 24 after endoproteinase Glu-C digestion showed that it was the only serine present. Further indirect evidence that Ser"' is a phosphorylation site is that it is followed by a sequence that is similar to those following serines 101 and 169 (i.e. Ser-X-Glu). Therefore, we conclude that Ser"' is phosphorylated, but it is a minor site of phosphorylation. Importantly, the serines at positions 83, 86, 95, and 98 were definitively proven to be nonphosphorylated by direct radiosequencing. Likewise, sequencing of several unlabeled peptides showed that several other serines were not labeled. We conclude that these 3 serines are the sites of IGFBP-1 phosphorylation and that serines 101 and 169 account for >85% of the total 32P that is incorporated.
Direct radiosequencing showed that the serine at position 101 was a major site of phosphorylation, and the extent of phosphorylation of this residue was confirmed by mutagenesis. Specifically, changing this residue to alanine to prevent its phosphorylation resulted in a 63% reduction in radiolabeling of the immunoprecipitated protein. Western blotting of the immunoprecipitated proteins showed that equivalent amounts of protein were loaded in each gel lane, and therefore the comparison of radiolabeled band intensity is valid. These data taken together confirm that serine 101 is phosphorylated and is a major site of phosphorylation in this protein.
Further definitive evidence that the [Ala98~'o']IGFBP-1 contained an alteration in the degree of phosphorylation of IGFBP-1 was derived from two observations. First, nondenaturing gel electrophoresis showed a significant decrease in the two most rapidly migrating bands containing phosphorylated IGFBP-1 and eliminated the most rapidly migrating band. Reduction in the abundance of this most rapidly migrating band indicates that the rate of migration through the electric field is dependent upon the number of phosphate groups per IGFBP-1 molecule as well as site-specific changes in charge density. This observation supports the hypothesis that different bands detected using this technique are due to differences in the degree of IGFBP-1 phosphorylation and that removal of the most heavily phosphorylated site results in a mobility shift. The results do not definitely determine whether this shift is due to a change in number of serines that are phosphorylated or to the location of specific phosphoserine residues. Second, anion exchange chromatography confirmed these observations. Frost and Tseng (6) have previously shown that this technique will separate the most rapidly migrating bands from the slowest migrating bands. Separation of the isoforms of wild type IGFBP-1 showed that the most rapidly migrating forms eluted from the column at higher salt concentration. The amount of this latter peak of protein was reduced when [Alag8~'"]IGFBP-1 was analyzed, further confirming that mutation had resulted in a reduction in the amount of this phosphorylated isoform.
The physiologic consequences of the mutation of Serl'' were significant. The [Ala98J01]IGFBP-1 mutant had a $fold reduction in affinity for IGF-I, suggesting that phosphorylation of this site is an important means of enhancing IGFBP affinity. The mutation of Serg8 to alanine had no demonstrable effect on the protein, since after dephosphorylation, both the wild type and the mutant proteins had identical affinities for IGF-I and identical mobilities under nondenaturing gel electrophoresis, indicating that the only effect of the two serineto-alanine mutations was to prevent the phosphorylation of Ser'''. Nondenaturing gel electrophoresis is very sensitive in detecting differences in charge or native conformation of proteins, and the identical mobilities of the wild type and mutant IGFBP-1 after dephosphorylation make it extremely unlikely that the amino acid alterations in the mutant protein had any direct effect on conformation. We confirmed our earlier observation that dephosphorylation of IGFBP-1 reduces affinity for IGF-I 6-fold (5). However, preventing phosphorylation of Ser''' by mutagenesis did not completely reduce the affinity of IGFBP-1 to that of the dephosphorylated protein (Fig. 5 ) . This suggests that IGFBP-1 affinity for IGF-I is increased by phosphorylation not only of Ser"' but also of serines 119 and/or 169.
Nondenaturing gel electrophoresis of IGFBP-1 purified from various sources demonstrates a correlation between the relative abundance of the most rapidly migrating (and most heavily phosphorylated) IGFBP-1 band and relative affinity for IGF-I. When IGFBP-1 purified from the conditioned media of CHO cells is compared to that of HepG2 cells, the HepG2 protein contains much more of the most rapidly migrating isoform (Fig. 3). Likewise when the affinity of the HepG2 IGFBP-1 is compared to the wild type CHO protein, it is consistently 2-3-fold greater (Fig. 4). Thus, the abundance of this most rapidly migrating isoform correlates well with changes in the net affinity of IGFBP-1 for IGF-I. Phosphorylation of Ser''' is necessary for this isoform to be present, since it is absent in the [Alag8~''']IGFBP-1 mutant (even though at least 2 other serines are phosphorylated), and this is consistent with our observation that the mutant protein has a lower affinity for IGF-I. Taken together with the finding that IGFBP-1 secreted by HepG2 cells has the highest relative affinity, we conclude that both the degree of phosphorylation of Ser''' and the total number of phosphates per molecule contribute to an increase in IGFBP-1 affinity for IGF-I.
Our finding that the principal site of phosphorylation is contained in a Ser-X-Glu-X-X-Glu-Glu-Glu sequence and that two other sites contain the Ser-X-Glu sequence with other acidic residues located on their C-terminal side of the phosphorylation site suggests that a member of the casein kinase family could be the kinase responsible for phosphorylation of IGFBP-1. Phosphorylation by casein kinase I1 depends upon acidic residues on the C-terminal side of the phosphorylated serines, and this enzyme has been reported to phosphorylate protein phosphatase inhibitor-2 which contains a Ser-X-Glu sequence (21). Other known serine-threonine kinases do not recognize this motif and neither Ser"', Ser"', nor Ser"j' represents consensus phosphorylation sites for other known protein kinases. We have confirmed the finding reported by Frost and Tseng (6) that casein kinase I1 can phosphorylate IGFBP-1 i n vitro,2 but we have not proven that this is the kinase in CHO cells that phosphorylates IGFBP-1. An important objective of future studies will be to identify the intracellular kinase since it appears to be able to modulate the affinity of IGFBP-1 for IGF-I.
In previous studies, we have demonstrated that nonphosphorylated IGFBP-1 purified from amniotic fluid is a potent potentiator of IGF's effects on porcine aortic smooth muscle cells. Specifically, IGFBP-I added under these conditions potentiated smooth muscle cell response 5.5-fold, and this response exceeded that of 10% human serum (3). In a recent in vivo assay, we demonstrated that nonphosphorylated IGFBP-1 potentiates wound healing in rats, leading to a 37% increase in wound breaking strength. There also is a 67% increase in hydroxyproline content reflecting increased collagen synthesis, and histologic analysis shows that the wounds that have received nonphosphorylated IGFBP-1 plus the IGF-I have increased cellularity (7). All of these responses are consistent with increased trophic action of this growth factor being potentiated by nonphosphorylated IGFBP-1. In contrast, direct addition of phosphorylated IGFBP-1 resulted in no increase in breaking strength or in hydroxyproline content (7). Therefore, phosphorylated IGFBP-1 not only appears to inhibit cell growth in culture, but also appears to inhibit the wound healing response to IGF-I in an in vivo model test system. Potential physiological regulation of the kinase that phosphorylates IGFBP-1 in situations of diminished growth is further substantiated by our finding that when newborn pigs are starved for 48 h, IGFBP-1 becomes heavily phosphorylated (22). Therefore, it is possible that phosphorylated IGFBP-1 can act as a growth inhibitor and that metabolic conditions which are associated with growth inhibition result in increased phosphorylation of this protein. Likewise, our data obtained in pigs suggest that phosphorylation of IGFBP-1 may be an important means for inhibiting the insulin-like actions of IGF-I in response to fasting.
In summary, our findings show that Ser''' is an important phosphorylation site in IGFBP-1. Phosphorylation at this J. I. Jones, unpublished observations. residue markedly alters affinity of the binding protein for the growth factor. Future studies should enable us to determine the effect of substitutions at positions 119 and 169 on the functional activity of IGFBP-I, and these mutants may be useful in confirming the identity of the kinase that phosphorylates IGFBP-1.