INTRODUCTION

Insulin-like growth factor II (IGF-II)¹ is a peptide hormone related to insulin that is present at high levels during fetal development. Genetic evidence suggests that IGF-II plays an important role in prenatal growth and that its actions are mediated through the IGF-I receptor and another yet uncharacterized receptor (1-3). In addition, IGF-II also binds to the IGF-II/mannose 6-phosphate receptor (IGF-II/MPR). This protein has two distinct functions: first, it mediates the biosynthetic targeting of mannose 6-phosphate containing lysosomal enzymes to the lysosome (for review, see Ref. 4); second, it mediates endocytosis of IGF-II, resulting in its delivery to the lysosome and subsequent degradation (5, 6).

The availability of IGF-II for interaction with its receptors is regulated by its association with binding proteins. To date, six different IGF-binding proteins have been characterized and have molecular masses ranging from 22.8 to 31.3 kDa (for review, see Refs. 7 and 8). In addition, a soluble form of the IGF-II/MPR (sIGF-II/MPR) has been detected in serum and urine (9-13). This protein retains IGF-II and Man-6-P binding activities and is abundant (~5 µg/ml) in fetal bovine serum (FBS) (14). In FBS, 5% of IGF-II is in the free state, with the remainder circulating as high molecular weight (M_r) complexes with the binding proteins and sIGF-II/MPR (14). Complex formation with the binding proteins has been shown to both inhibit and potentiate IGF-II's interactions with its receptors as well as increase the serum half-life of the ligand (for review, see Ref. 8).

IGF-II is synthesized as a preproprotein (15). After cleavage of its signal sequence, the COOH-terminal 88 residues (E-peptide) of the proprotein are removed to yield the 67-residue, mature 7.5-kDa IGF-II. The mature polypeptide can be subdivided into four domains (B, C, A, and D domains) that are homologous to the B and A chains of mature insulin and the B, C, A, and D domains of IGF-I. In addition, the existence of extended forms of IGF-II has been documented (for review, see Refs. 16-18), although their physiological role in the biology of IGF-II is not clear.

To date, high M_r isoforms of IGF-II have been detected in human serum (19-21), cerebrospinal fluid (22), and malignant tissue extracts (23-25). Human IGF-II variants that are substituted at Ser-29 and Ser-33 with Arg-Leu-Pro-Gly and Cys-Gly-Asp, respectively, with or without E-peptide extensions, have also been identified (21, 26). In addition, two IGF-II species with apparent M_r values of 15,000 and 11,500 were purified from human serum Cohn fraction IV₁ and represent 87-88-amino acid polypeptides that are glycosylated at Thr-75 (18).

We recently described the isolation of a high M_r IGF-II fraction from FBS (14). In this study we have separated the high M_r IGF-II isoforms into eight different fractions. Each fraction is composed of a COOH-terminally extended IGF-II that terminates at Gly-87 and contains variable amounts of O-linked sugars. Our studies indicate that all high M_r isoforms are similar to 7.5-kDa IGF-II in terms of both receptor/binding protein interactions and biological activity.

EXPERIMENTAL PROCEDURES

FBS was obtained from Inovar. Yeast phosphomannan was very kindly provided by Dr. M. E. Slodki (U. S. Department of Agriculture, Peoria, IL). Phosphomannan affinity resin was prepared by coupling cyanogen bromide-activated Sepharose CL-6B (Pharmacia) to phosphomannan core as described previously (27). Recombinant human IGF-IIs with and without a 21-amino acid E-peptide extension (rhIGF-IIE₈₈ and rhIGF-II, respectively) were as described (28, 29). Recombinant human IGF-I was obtained from either Pharmacia or PeproTech, Inc. (Rocky Hill, NJ). ¹²⁵I-Labeled IGF-I (specific activity 120.6-127.5 µCi/µg) was prepared as described previously (30). Soluble recombinant human IGF-I receptor (sIGF-IR) and recombinant human IGF-binding protein-1 (IGFBP-1) were kindly provided by M. Jansson and C. Dyring (Pharmacia Upjohn, Stockholm, Sweden) and will be described elsewhere. -[methyl-³H]Aminoisobutyric acid (8.9 Ci/mmol) was purchased from DuPont NEN. Reagents for surface plasmon resonance spectroscopy, including sensor chips CM5 (certified), Surfactant P20, N-hydroxysuccinimide, N-ethyl-N-[(3-dimethylamino)propyl]carbodiimide hydrochloride, and ethanolamine-HCl were from Pharmacia Biosensor AB (Uppsala, Sweden). All other chemicals were reagent grade.

HEPES-buffered saline 1 (HBS-1): 150 mM NaCl, 2 mM EDTA, 20 mM HEPES, pH 7.2; HBS-2: 150 mM NaCl, 3.4 mM EDTA, 0.005% Surfactant P20, 10 mM HEPES, pH 7.2; acid buffer: 60 mM ammonium acetate, 340 mM acetic acid, pH 4.0; propanol/acetate: 20% 1-propanol, 8 mM ammonium acetate, 42 mM acetic acid, pH 4.0; anion exchange (AEX) buffer 1A: 20% 1-propanol, 20 mM NH₄HCO₃, pH 8.0; AEX buffer 1B: 20% 1-propanol, 800 mM NH₄HCO₃, pH 8.0; AEX buffer 2A: 20% 1-propanol, 20 mM bis-Tris, pH 6.0; AEX buffer 2B: 20% 1-propanol, 1 M NaCl, 20 mM bis-Tris, pH 6.0; glycosidase buffer A: 50 mM sodium phosphate, pH 5.0; glycosidase buffer B: 10 mM CaCl₂, 20 mM sodium cacodylate, pH 6.5; phosphate-buffered saline (PBS): 137 mM NaCl, 2.7 mM KCl, 10.1 mM Na₂HPO₄, 1.76 mM KH₂PO₄, pH 7.2; PBST: PBS with 0.05% Tween 20; AIB assay buffer: 5.5 mM glucose, 0.4 mg/ml BSA, Krebs-Ringers phosphate buffer, pH 7.37; TrisT: 1% Triton X-100, 50 mM Tris-MES, pH 7.8; imidazole binding buffer: 0.5 M NaCl, 10% glycerol, 0.075% Triton X-100, 5 mg/ml BSA, 40 mM imidazole buffer, pH 7.4.

High M_r IGF-II was isolated from FBS using a modification of the procedure described in Ref. 14. Briefly, expired lots of FBS that contained 3-5 mg/liter of soluble insulin-like growth factor II/mannose 6-phosphate receptor (sIGF-II/MPR) were used as source material. Serum (50 or 100 liters) was buffered by the addition of 0.04 volumes of a HEPES stock solution (0.5 M, pH 7.2), filtered through 0.2-µm cellulose acetate membranes (Millipore) and loaded onto a phosphomannan-Sepharose CL-6B affinity column (11.3 × 12 cm) at 2-2.5 liters/h. After loading, the column was disassembled, the resin batch was washed with 2 column volumes of HBS-1 (8-10 times), the column was re-assembled, and the column was further washed at 2-2.5 liters/h until A₂₈₀ reached base line. The column was eluted at 200 ml/h with 2.5 mM mannose 6-phosphate in HBS-1. Recovered proteins were concentrated to 20 or 40 ml (for 50- or 100-liter preparations, respectively) by ultrafiltration using a YM-100 membrane (Amicon). The retentate was acidified to pH 4.0 by the addition of 0.1 volumes 10 × acid buffer. After 6 h, 10-ml samples were filtered through 0.2-µm cellulose acetate membranes (Corning) and chromatographed at 1.0 ml/min on a Superdex 75 column (Pharmacia, 2.6 × 60 cm) equilibrated with 1 × acid buffer. The material was resolved into three peaks that represented sIGF-II/MPR, high M_r IGF-II, and mature 7.5-kDa IGF-II, respectively (see Ref. 14 for representative chromatogram). Material eluting in the high M_r IGF-II region from two to four individual column runs was pooled, lyophilized, resuspended in 1 × acid buffer (~5 ml), and re-chromatographed at 0.8 ml/min on the Superdex 75 column equilibrated in 1 × acid buffer. This step achieved near base-line separation of the high M_r IGF-II fraction from the flanking sIGF-II/MPR and mature IGF-II peaks. All steps were performed at 4 °C.

Separation of individual high M_r IGF-II isoforms was performed using anion exchange chromatography on a Pharmacia SMART system at room temperature. Fractions containing high M_r IGF-II from the second acid gel filtration step (eluting between 165 and 250 ml) were consecutively pooled (two to three 6-ml fractions per pool), lyophilized, resuspended in propanol/acetate, and the buffer exchanged to propanol/acetate using a fast desalting PC 3.2/10 column at 100 µl/min. Void volume fractions were pooled, dried, and resuspended with AEX buffer 1A (100 µl). Samples were chromatographed on a Mono Q HR 1.6/5 column equilibrated in AEX buffer 1A at a flow rate of 100 µl/min. Five minutes after sample injection (when A₂₈₀ reached base line), a linear gradient from 0 to 47.5% AEX buffer 1B over 27 min was performed and 50-µl fractions collected. Different IGF-II isoforms eluted between 15 and 39% AEX buffer 1B (120-320 mM NH₄HCO₃) and could be grouped into nine different peaks denoted A through I. Equivalent fractions from different anion exchange runs were pooled, dried, and desalted as described above. Proteins were rechromatographed on the anion exchange column as described above and fractions pooled as indicated under "Results." Some fractions were further purified by anion exchange chromatography using a different buffer system. Here, samples were exchanged into propanol/acetate, dried, and resuspended in AEX buffer 2A (100 µl). Anion exchange chromatography was as described above but used a linear gradient of 8-28% AEX buffer 2B (80-280 mM NaCl) over 19 min, and 20-µl fractions were collected. Appropriate fractions were pooled as indicated under "Results," exchanged into propanol/acetate, and stored at 80 °C.

Peaks A-I representing the different bovine IGF-II isoforms and 7.5-kDa rhIGF-II were iodinated using Na¹²⁵I (NEZ-033A, DuPont NEN) and lactoperoxidase (Boehringer Mannheim) as described previously (31) except the reactions were quenched with tyrosine. Radiolabeled high M_r IGF-II was separated from free iodine using prepacked G-25 M columns (PD10, Pharmacia) with PBS as the mobile phase. BSA was included as a carrier in the column buffer of early experiments, but later omitted due to its uptake of label. As each of the radiolabeled high M_r IGF-II isoforms eluted as a single peak after analytical gel filtration chromatography on a Superose 12 column (1.6 × 85 cm), these were used without further purification (data not shown). Radiolabeled 7.5-kDa IGF-II (peak A and rhIGF-II) was applied to a Sephadex G50 fine column (Pharmacia, 0.7 × 28 cm) using PBS as the mobile phase. Three radioactive peaks were detected, and fractions comprising the second peak (IGF-II monomer) were pooled. The specific activity ranged from 120 to 770 Ci/mmol for high M_r IGF-II and was ~25 Ci/mmol for rhIGF-II.

All digests were carried out at 37 °C. For the experiment shown in Fig. 5, samples of iodinated high M_r IGF-II and rhIGF-II (~30,000 cpm/reaction) were incubated with either 2.5 milliunits of recombinant neuraminidase (Glyko), 1 milliunit of recombinant O-glycosidase (Glyko), or both enzymes in a total volume of 10 µl of glycosidase buffer A. Digests were carried out for 4 h. For double digests, samples were first incubated with neuraminidase for 1 h before addition of O-glycosidase. Digested samples were resolved on a 12.5-17.5% polyacrylamide gel under nonreducing conditions. For the isoelectric focusing experiment shown in Fig. 6, approximately 80,000 cpm of radiolabeled IGFs were used per reaction. Samples were digested with 10 milliunits of neuraminidase (Arthrobacter; Boehringer Mannheim) for 2 h followed by the addition of 2 milliunits of O-glycosidase (Oxford GlycoSystems) and incubation for 18 h. Digests were carried out in a total volume of 48 µl of glycosidase buffer B. Both untreated and digested samples were resolved by isoelectric focusing on precast gels (pH 3-7) according to the manufacturer (Novex). Isoelectric gels were fixed in 3.5% sulfosalicylic acid, 11.5% trichloroacetic acid for 30 min prior to staining.

A Pharmacia BIAcore system was used to monitor interaction of the high M_r IGF-II isoforms with the sIGF-II/MPR, sIGF-IR, and IGFBP-1. Details of the instrumentation and analytical methods have been described in several recent reports (32, 33). Briefly, standard amine coupling procedures were used for the immobilization of ligand (sIGF-II/MPR, sIGF-IR, or IGFBP-1) to the dextran matrix of the sensor chip surface (34). Surfaces were activated with a 35-µl injection of a 0.05 M N-hydroxysuccinimide, 0.2 M N-ethyl-N-[(3-dimethylamino)propyl]carbodiimide hydrochloride mixture according to the manufacturer using a flow rate of 5 µl/min HBS-2. sIGF-II/MPR and IGFBP-1 were diluted to a final concentration of 5 µg/ml in 20 mM sodium acetate, pH 4.7, before injection over the activated dextran matrix. sIGF-IR was diluted to a final concentration of 30 µg/ml in 0.2 M acetic acid before injection. Typically 2000-4000 resonance units of sIGF-II/MPR or sIGF-IR or 300-600 resonance units of IGFBP-1 were immobilized per experiment. Unreacted carboxyl groups on the dextran matrix were blocked with a 35-µl injection of a 0.2 M solution of ethanolamine. Sensorgrams were generated using a flow rate of 10 µl/min. Typically, four 2-fold dilutions of analyte (rhIGF-I, rhIGF-II, and different high M_r IGF-II isoforms) were used in each experiment. The concentrations ranged from 10 to 80 nM for analyte passed over sIGF-II/MPR and IGFBP-1 surfaces. For sIGF-IR surfaces, 25-200 nM or 12.5-100 nM of the IGF-II isoforms or of rhIGF-I, respectively, were used. Analytes were applied for 3 min to measure association kinetics and then washed out using HBS-2 to measure dissociation kinetics for an additional 3 min. The surface was regenerated after each round with a 1-min injection of either 10 mM HCl for sIGF-II/MPR and IGFBP-1 surfaces or 50 mM NaHCO₃/1 M NaCl for sIGF-IR surfaces.

Kinetic constants were determined using the BIAevaluation software (version 2.0). Both the association and dissociation phases of the binding curves were fit with simple bimolecular interaction assumptions, using the resonance unit signal as a relative measure of the ligand-analyte complex concentration with respect to time. The association phase of different sensorgrams was fit directly with a nonlinear least squares iterative curve fitting function. The association constants obtained with the four different analyte concentrations were averaged. Dissociation constants were calculated from the highest analyte concentrations used to minimize signal contributions due to rebinding of analyte to ligand at the surface.

Binding of radiolabeled bovine 7.5-kDa (peak A) or the different high M_r IGF-II isoforms (peaks B-I) to serum-binding proteins was determined as described previously (14). Briefly, the different radiolabeled IGFs (~30,000 cpm) were each incubated with FBS (Hyclone; 1 ml). Time course experiments indicated that a 24-h incubation at room temperature was sufficient for both mature and high M_r IGF-II radiotracer to reach equilibrium with serum-binding proteins (see Ref. 14 and data not shown), thus reflecting the distribution of endogenous IGF-II. The serum was applied to a Superose 12 column (1.6 × 85 cm) and eluted at 1 ml/min with PBST at room temperature as indicated. Fractions (2 ml) were counted for 1 min in a Packard Cobra auto -counter.

Fibroblasts from a normal 18-year-old man (WUMS1) were isolated at Washington University School of Medicine, St. Louis, MO (35). Uptake of AIB in human fibroblasts was measured as described previously (36) using an assay volume of 500 µl.

The C2I subclone of C2 myoblasts (37) that requires only insulin or IGF-I to differentiate in vitro was kindly provided by Dr. Peter Rotwein at Washington University School of Medicine, St. Louis, MO. Differentiation in C2I myoblasts was induced as described by Rotwein et al. (38) except that assays were performed in 24-well clusters, using an assay volume of 1.0 ml. After ~70 h, cells were washed and lysed by incubation with 100 µl of TrisT for 15-20 min at room temperature. Creatine kinase activity was determined as described (38) and normalized for total protein content, using the BCA protein assay (Pierce). IGF-binding proteins secreted during C2I myoblast differentiation were analyzed by ligand blotting as described previously (35).

To compare binding of IGF-I, IGF-II, and the different high M_r IGF-II isoforms to the IGF-binding proteins secreted by human fibroblasts, conditioned buffer was collected from WUMS1 fibroblasts after incubation in AIB assay buffer for 3 h and clarified by centrifugation. Competition binding studies were performed as described previously (35) with minor modifications. Briefly, fibroblast-conditioned buffer was incubated for 1 h at 37 °C with ¹²⁵I-labeled IGF-I and increasing amounts of unlabeled IGF-I, IGF-II, or high M_r IGF-II in AIB assay buffer (final volume, 250 µl). AIB assay buffer rather than 0.1 M HEPES, pH 6.0, with 44 mM NaH₂PO₄, 0.01% Triton X-100, 1 mg/ml BSA, and 0.02% NaN₃ (35) was used in order to reproduce more closely the conditions under which IGFs and secreted binding proteins interact during the AIB uptake assay. Complexes of binding protein and IGF were precipitated by the addition of 750 µl of 25% polyethylene glycol, 4.0 mg/ml bovine -globulin (2:1). Binding data were analyzed by LIGAND, a computer program developed by Munson and Rodbard (39).

To compare binding of IGF-I, IGF-II, and the different high M_r IGF-II isoforms to the human IGF-I receptor, IGF-I receptor was purified from human placenta using wheat germ agglutinin-Sepharose chromatography, insulin affinity chromatography, and IGF-I affinity chromatography (30). Competition binding studies were performed as described previously (30). Briefly, IGF-I receptor was incubated overnight at 4 °C with ¹²⁵I-labeled IGF-I and increasing amounts of unlabeled IGF-I, IGF-II, or high M_r IGF-II in imidazole binding buffer. Receptor-bound IGF-I was precipitated by the addition of 900 µl of 33.3% polyethylene glycol, 1.5 mg/ml bovine -globulin (1:1). Binding data were analyzed by LIGAND.

sIGF-II/MPR concentrations were determined by a two-antibody sandwich enzyme-linked immunosorbent assay as described previously (40). Protein concentrations were determined with BSA standards using the Lowry method (41) adapted to microtiter plates. Molar concentrations were calculated using the following protein M_r values: rhIGF-I, 7639; rhIGF-II, 7475; 7.5-kDa bovine IGF-II (peak A), 7532; high M_r IGF-II (peaks B-I), 9578. SDS-PAGE was performed as described (42). Gels were stained with 0.2% Coomassie Brilliant Blue R-250 (Bio-Rad) and/or silver using the Novex SilverXpress silver staining kit as indicated. Dried radioactive gels were exposed to a phosphor storage screen, scanned, and quantitated using a Molecular Dynamics PhosphorImager 400 and ImageQuant 3.15 software. The molecular mass of the different IGF-II isoforms was determined using both a 252Cf plasma desorption time-of-flight mass spectrometer Bio-Ion 20 (Applied Biosystems, Gothenburg, Sweden) and a VG Quattro electrospray mass spectrometer (Fisons Instruments, Altrincham, United Kingdom). For laser desorption mass spectrometry, samples were dissolved in 0.1% trifluoroacetic acid, applied to nitrocellulose-coated foil (Applied Biosystems), dried, and analyzed in a positive-ion mode for 1 h with an acceleration voltage of 18 kV. Samples for electrospray mass spectrometry were applied in 50% methanol, 1% acetic acid at a flow rate of 5 µl/min. Amino acid analyses were performed at the Cornell Biotechnology Analytical/Synthesis Facility (Ithaca, NY). Amino-terminal sequencing was performed using automated Edman degradation on a Hewlett-Packard G1000A protein sequencer or a Milligen Biosearch Prosequencer type 6600. Predicted isoelectric points were calculated from amino acid compositions using the University of Wisconsin GCG program (43). The contribution of glycosylation to pI (isoelectric point) was estimated using a pK_a of 2.6 for sialic acid (44), and ionizable cysteines were excluded from the analysis as all six cysteines present in IGF-II are disulfide bonded.

RESULTS

We previously reported the co-purification of sIGF-II/MPR and bound IGF-II from FBS (14). In addition to mature 7.5-kDa IGF-II, we also isolated ~12 different high M_r IGF-II isoforms that were associated with the soluble receptor in low abundance. In order to investigate the properties of this high M_r IGF-II fraction in more detail, we modified and scaled-up our original purification scheme to isolate milligram quantities of these IGF-II isoforms (see "Experimental Procedures"). Large-scale isolation of sIGF-II/MPR and bound IGF-II from 50- and 100-liter lots of FBS was achieved by affinity adsorption of receptor to phosphomannan-agarose and elution with mannose 6-phosphate. Mildly acidic gel filtration chromatography was employed to dissociate bound IGF-II from receptor and to resolve high M_r IGF-II species from sIGF-II/MPR and 7.5-kDa IGF-II. Finally, pooled fractions from the high M_r IGF-II peak were re-applied to the acidic gel filtration column to remove contaminating sIGF-II/MPR and 7.5-kDa IGF-II. Using this strategy, we have processed ~1000 liters of FBS yielding ~1.5 mg of total high M_r IGF-II.

The different high M_r IGF-II isoforms were subfractionated by anion exchange chromatography at pH 8.0. For each 50- or 100-liter preparation, the high M_r IGF-II peak from the second acid gel filtration step was divided into five to eight fractions, and each was chromatographed on a Mono Q column using an ammonium bicarbonate gradient (Fig. 1). The elution profiles of the different chromatograms were similar, although in general, when comparing early with late gel filtration fractions (larger versus smaller proteins), the earlier eluting gel filtration fractions tended to be biased toward the later (more acidic) eluting ion exchange fractions. The chromatogram depicted in Fig. 1 is of material eluting late in the acidic gel filtration column, and all the peaks of interest are well represented. The anion exchange column fractions were analyzed by SDS-PAGE and Coomassie staining on 14% gels. The regions labeled A-I contained visualizable protein and sequence analysis on selected peaks (A, B, D, and H) revealed that all contained proteins with the amino-terminal residues AYRPS, identical to 7.5-kDa bovine IGF-II (45).

For further purification, peaks A-I from different anion exchange runs were pooled and rechromatographed (Fig. 2) as described above; fractions were pooled as indicated. In addition, Peaks A, B, D, and H were chromatographed on a Mono Q column at pH 6.0 and eluted with a gradient of sodium chloride (Fig. 3); fractions were pooled as indicated.

The purity of peaks A-I was assessed using SDS-PAGE and silver staining (Fig. 4). Peak A consisted of a single protein species that ran identically to the rhIGF-II standard. Subsequent analysis by mass spectrometry and isoelectric focusing (see below) showed that this species represented mature 7.5-kDa bovine IGF-II that was not resolved from the high M_r IGF-II peak by gel filtration chromatography. The remaining peaks contained one to three different protein species with apparent sizes ranging from 11 to 17 kDa.

We previously examined the nature of the increased M_r seen in the unfractionated high M_r IGF-II pool using a combination of protease and glycosidase sensitivity assays (14). We demonstrated that all proteins comprising this pool contain a common peptide backbone that extends beyond Glu-67 and that this extension is modified with various amounts of sialated, O-linked sugars resulting in the pool's heterogeneity. To confirm and extend these earlier studies, the fractionated high M_r isoforms in peaks B-I were iodinated and then digested with neuraminidase and/or O-glycosidase to remove terminal sialic acids and O-linked core disaccharides, respectively (Fig. 5). No shift in the electrophoretic mobility of peak B (Fig. 5) or of rhIGF-II standard (data not shown) was detected. In contrast, digestion of peaks C-I using a combination of neuraminidase and O-glycosidase caused a mobility shift in all the isoforms, resulting in the appearance of a single major species that ran identically to both undigested and digested peak B (arrow). These data suggest that peak B represents an extended, nonglycosylated isoform of high M_r IGF-II and that the species contained in peaks C-I have the same peptide backbone as peak B but are further modified with various amounts of O-linked sugars. Interestingly, double digestion of peak F demonstrated the presence of two bands, one that ran identically to peak B and another that migrated slightly slower. This behavior may represent multiple IGF-II species with slight differences in their peptide backbones (i.e. longer extensions and/or variant internal sequences) or the presence of other post-translational modifications.

The high M_r forms of IGF-II were also digested with the two glycosidases separately (Fig. 5). Treatment with neuraminidase alone caused a small but significant increase in the mobility of all bands indicating the presence of sialic acid. In contrast, treatment with O-glycosidase alone had no effect on any of the high M_r IGF-II isoforms. As O-glycosidase only removes the unmodified disaccharide Gal1-3GalNAc from threonine or serine residues, these data suggest that essentially all of the O-linked sugars in each of the high M_r IGF-II isoforms are protected with sialic acid.

The contribution of oligosaccharide to the net charge of high M_r IGF-II was investigated using isoelectric focusing. The measured pI of recombinant human IGF-II was ~6.65 (Fig. 6, upper panel), close to the pI of 6.80 predicted from its composition. (The minor band in the rhIGF-II lane likely represents contaminating BSA introduced from the column buffer as ascertained by comparison of the iodinated preparation with authentic BSA by SDS-PAGE and isoelectric focusing.) Peak A gave a pI of ~6.85 (data not shown), similar to the pI of 6.84 predicted for unmodified mature 7.5-kDa bovine IGF-II. Peak B gave a pI of ~4.85 (Fig. 6, upper panel), which is consistent with the isoelectric point of 4.81 predicted for an unmodified 87-residue extended IGF-II. The remaining bands displayed increasingly acidic pI's, ranging from ~4.6 to 3.5, respectively (Fig. 6, upper panel). Each sialic acid added to the 87-residue peptide is expected to decrease its isoelectric point 0.12 to 0.23 pH units. Thus, the decreasing isoelectric points of peaks C-I are consistent with increasing numbers of sialic acid residues decorating the core O-linked disaccharides. Treatment with both neuraminidase and O-glycosidase reduced all of the high M_r IGF-II isoforms to a major species that had a pI of ~4.85 (Fig. 6, lower panel), identical to both undigested and digested peak B (Fig. 6, upper and lower panels, respectively). In addition, we noted that the deglycosylated samples contained minor species with pI values of approximately 5.4, 4.7, and 4.5 (Fig. 6, lower panel). These minor species may represent experimentally (iodination)-induced modifications in the peptide backbone as similar variability was not detected by SDS-PAGE of native or radiolabeled samples (see above), Coomassie staining of isoelectric focusing gels (data not shown) or mass spectrometry (see below). Regardless of the identity of the minor species, the important point is that all of the high M_r IGF-II isoforms contain an identical peptide backbone and that peak B represents non-glycosylated, extended bovine IGF-II.

We next determined the molecular mass of the high M_r IGF-II isoforms using mass spectrometry (Table I). Recombinant human IGF-II isoforms were used for method validation. Mass determinations of 7465 and 9802 for rhIGF-II and rhIGF-IIE₈₈ standards, respectively, were within ~0.1% of those predicted from the cDNA (28, 29). Peak A had a M_r of 7527, strongly supporting its assignment as the 67-residue mature 7.5-kDa bovine IGF-II. The molecular weights of peaks B and D were 9572 and ~10,850, respectively. This difference is presumably due to the presence of sugar as deglycosylated peak D appeared identical to peak B by both SDS-PAGE and isoelectric focusing (see above). The measured mass of 9572 for peak B is again consistent with this protein terminating at Gly-87, which would give a predicted mass of 9578. Precise mass determinations on the other high M_r IGF-II peaks were not possible due to their large oligosaccharide component and microheterogeneity within the individual peaks (Table I and data not shown). However, all values measured had molecular masses of ~13,000 or less, providing an upper limit toward the size and degree of glycosylation of the IGF-II isoforms.

Table I. Mass spectrometry analysis of high Mr IGF-II Samples were analyzed as described under "Experimental Procedures." Measured values for rhIGF-II, rhIGF-IIE88, and peaks A, B, and D are presented as the mean ± standard deviation for the major peak over n determinations. The range shown for peaks D, G, and I represents the values of multiple peaks observed in a single analysis (see text). At least one determination was performed using electrospray mass spectrometry for rhIGF-II, rhIGF-IIE88, and peaks A and B; all others were performed using laser desorption time-of-flight mass spectrometry. The deduced masses of rhIGF-II and rhIGF-IIE88 are based on the human cDNA; the deduced masses of peaks A and B are based on the bovine cDNA, assuming COOH termini at Glu-67 and Gly-87, respectively. NA, not applicable, only single peak detected. Molecular weight Range n Deduced Measured rhIGF-II 7475 7465  ± 9 NA 2 rhIGF-IIE88 9807 9802  ± 10 NA 2 Peak A 7532 7527  ± 3 NA 4 Peak B 9578 9572  ± 5 NA 5 Peak D 10854  ± 57 10,298-10,901 5 Peak G 10,360-13,160 1 Peak I 10,462-12,999 1

We performed quantitative amino acid analyses as an additional test to determine the peptide composition of peaks B and D. The lysine content of these species is consistent with the COOH terminus being Gly-87: the analyses indicated the presence of a single lysine for both peaks B and D, while two were observed for rhIGF-IIE₈₈ (data not shown). Based on the cDNA, IGF-II with an E-peptide extension to position 88 would contain two lysines (Lys-65 and Lys-88), while IGF-II terminating before position 88 would contain only one as seen for peaks B and D. In addition, values obtained for the remaining amino acids are close to those predicted for an IGF-II polypeptide extended to Gly-87 (data not shown). These results, taken together with that presented above, indicate that nearly all of the high M_r IGF-II isoforms have a common peptide backbone that is extended 20 amino acids beyond that of the 67-residue, mature IGF-II and that the heterogeneity observed in the different peaks is due to differences in O-linked glycosylation.

The interactions of high M_r IGF-II with sIGF-II/MPR, a soluble fragment of recombinant human IGF-I receptor (sIGF-IR), and recombinant human IGF-binding protein 1 (IGFBP-1) were compared with those of mature IGF-II by measuring their association and dissociation rate constants using the BIAcore system (see "Experimental Procedures"). Either the individual receptors or binding protein were covalently attached to the dextran matrix of the sensor chip and different concentrations of the IGF-II isoforms (analyte) passed through the flow cell. The signal is proportional to the amount of IGF-II bound to the immobilized sIGF-II/MPR. The regions of the sensorgrams following introduction or washout of the analyte were used to determine the association and dissociation rate constants, respectively. Kinetic rate constants for all high M_r IGF-II isoforms with the two receptors and binding protein are presented in Tables II and III. In addition, equilibrium dissociation constants (K_d) were determined from the kinetic data and are presented in Table IV.

Table II. Receptor/binding protein association rates Association rate constants were calculated from sensorgrams generated on the BIAcore as described under "Experimental Procedures." All values represent the mean ± standard deviation from three independent determinations. ND, not determined. Ligand Association rate constants sIGF-II/MPR sIGF-IR IGFBP-1 ×105 M1 s1 rhIGF-I ND 5.07  ± 1.77 ND rhIGF-II 10.7  ± 2.0 1.47  ± 0.5 9.4  ± 3.7 Peak A 6.1  ± 3.8 1.03  ± 0.08 5.7  ± 3.5 Peak B 7.7  ± 2.0 0.88  ± 0.17 6.0  ± 2.6 Peak C 8.5  ± 2.0 1.06  ± 0.08 6.6  ± 1.8 Peak D 9.4  ± 4.1 0.95  ± 0.13 7.5  ± 3.5 Peak E 10.7  ± 5.3 1.35  ± 0.55 9.3  ± 2.4 Peak F 13.1  ± 3.5 0.99  ± 0.24 8.7  ± 3.3 Peak G 7.9  ± 2.3 0.88  ± 0.15 7.1  ± 1.2 Peak H 5.9  ± 0.7 1.47  ± 0.81 5.8  ± 1.9 Peak I 9.3  ± 1.7 0.97  ± 0.11 8.0  ± 2.6

Table III. Receptor/binding protein dissociation rates Dissociation rate constants were calculated from sensorgrams generated on the BIAcore as described under "Experimental Procedures." All values represent the mean ± standard deviation from three independent determinations. ND, not determined. Ligand Dissociation rate constants sIGF-II/MPR sIGF-IR IGFBP-1 ×103 s1 rhIGF-I ND 2.74  ± 0.27 ND rhIGF-II 3.25  ± 0.12 4.77  ± 0.73 2.59  ± 1.0 Peak A 2.73  ± 0.38 4.60  ± 0.64 2.44  ± 0.93 Peak B 2.69  ± 0.33 4.90  ± 0.37 2.32  ± 0.82 Peak C 2.70  ± 0.10 4.57  ± 0.13 2.26  ± 0.78 Peak D 2.56  ± 0.22 4.95  ± 0.47 2.26  ± 0.81 Peak E 2.60  ± 0.27 4.37  ± 0.32 2.23  ± 0.68 Peak F 2.75  ± 0.14 5.11  ± 0.64 2.25  ± 0.78 Peak G 2.58  ± 0.08 4.79  ± 0.31 2.29  ± 0.84 Peak H 2.39  ± 0.11 4.71  ± 0.23 2.32  ± 0.8 Peak I 2.55  ± 0.06 4.71  ± 0.29 2.27  ± 0.7

Table IV. Receptor/binding protein affinities calculated from kinetic data Equilibrium dissociation constants (Kd) were calculated from the measured kinetic parameters presented in Tables II and III (Kd = dissociation rate/association rate). All values represent the mean ± standard deviation from three independent determinations. ND, not determined. Ligand Equilibrium dissociation constants sIGF-II/MPR sIGF-IR IGFBP-1 nM rhIGF-I ND 5.9  ± 2.2 ND rhIGF-II 3.10  ± 0.50 35.2  ± 11.5 2.80  ± 0.35 Peak A 5.45  ± 2.27 44.5  ± 5.1 4.83  ± 1.25 Peak B 3.64  ± 0.74 57.7  ± 16.4 4.00  ± 0.52 Peak C 3.30  ± 0.78 43.3  ± 3.8 3.43  ± 0.68 Peak D 3.18  ± 1.64 52.6  ± 9.0 3.20  ± 0.61 Peak E 2.97  ± 1.61 36.4  ± 16.5 2.40  ± 0.36 Peak F 2.20  ± 0.53 54.8  ± 19.9 2.47  ± 0.23 Peak G 3.43  ± 1.10 54.9  ± 6.5 3.15  ± 0.80 Peak H 4.07  ± 0.32 38.2  ± 17.6 4.63  ± 2.87 Peak I 2.80  ± 0.53 49.3  ± 8.0 2.87  ± 0.38

Our data demonstrate that all of the high M_r IGF-II isoforms bind with similar affinities as those measured for mature IGF-II (both bovine and human) and reflect nearly identical association and dissociation rates. Whereas binding to the sIGF-II/MPR and IGFBP-1 was in the low nanomolar range for all species, interaction with the sIGF-IR was of considerably lower affinity (K_d > 35 nM). In contrast, rhIGF-I displayed 6-10-fold higher affinity for the sIGF-IR than any of the IGF-II isoforms. These data suggest that E-peptide extensions and glycosylation do not influence binding of IGF-II to the sIGF-II/MPR, sIGF-IR, or IGFBP-1.

Radiolabeled tracer experiments were performed to: 1) estimate the fraction of high M_r IGF-II carried by the sIGF-II/MPR in FBS and 2) compare the endogenous distribution of high M_r and mature IGF-II among different serum-binding proteins in FBS. Serum was incubated with each of the eight different iodinated high M_r IGF-II isoforms or mature IGF-II under conditions that allow approach to equilibrium (see Ref. 14; similar kinetics were observed for mature and high M_r IGF-II) and fractionated by gel filtration chromatography as described under "Experimental Procedures."

Comparison of mature 7.5-kDa bovine IGF-II (peak A) and two high M_r IGF-II isoforms (peaks B and E, representing extended, nonglycosylated and extended, glycosylated isoforms, respectively) demonstrated similar binding profiles with essentially no tracer eluting as the free polypeptide (Fig. 7, upper three panels). In addition, ~35% of the tracer in each case was associated with a peak that elutes identically to sIGF-II/MPR standard (Fig. 7, upper three panels, region 1). Likewise, values ranging from 31-39% for radiotracer association with the sIGF-II/MPR peak were determined for the other high M_r IGF-II isoforms (data not shown). Most of this IGF-II binding activity represents authentic sIGF-II/MPR as receptor-depleted serum has greatly diminished activity (Fig. 7, lower panel, region 1). Interestingly, while enzyme-linked immunosorbent assay analysis demonstrated complete removal of receptor (data not shown), 1.5-5% of the IGF-II tracer still migrated in this region when the different high M_r IGF-II isoforms were tested (Fig. 7, lower panel and data not shown). Thus, while another protein makes a minor contribution to IGF-II binding in this region, these results clearly demonstrate that the endogenous sIGF-II/MPR carries ~1/3 of the total IGF-II (both mature and high M_r isoforms) in FBS.

The different radiolabeled IGF-II isoforms also eluted in two other regions. Region 2 represents tracer associated with a binding protein(s) that elutes similarly to the 150-kDa standard and may represent ternary complex formation between tracer, IGFBP-3, and the acid labile subunit (8). Region 3 represents tracer association with any of the five remaining IGF-binding proteins that have apparent sizes ranging from 25 to 40 kDa. The similarity in the binding profiles for the different high M_r IGF-II isoforms and mature bovine IGF-II suggests that these molecules distribute similarly among different IGF-binding proteins in FBS. It is worth noting that peak E appears to have diminished binding activity in region 2 compared with the other IGF-II isoforms (Fig. 7). While this could be interpreted as reflecting differences in affinity for binding proteins, the apparent decreased magnitude of region 2 may simply represent incomplete resolution of the different peak E-binding protein complexes in regions 2 and 3 on the gel filtration column.

It has been demonstrated that uptake of the amino acid analog, -aminoisobutyric acid (AIB), is stimulated by IGF-I in human fibroblasts (36). Fig. 8A, shows a representative experiment comparing stimulation of AIB uptake by rhIGF-I, mature bovine IGF-II, and high M_r IGF-II isoforms C-H. In six experiments, the ED₅₀ of rhIGF-I-stimulated AIB uptake in human fibroblasts was 0.29 ± 0.03 nM, as reported previously (36). Stimulation of AIB uptake by mature and high M_r IGF-II is summarized in Fig. 8B. The ED₅₀ of IGF-II-stimulated AIB uptake was 1.41 ± 0.05 nM, 4.8-fold higher than that of IGF-I. Surprisingly, the ED₅₀ values for stimulation of AIB uptake by high M_r IGF-II ranged from 0.59 ± 0.04 nM to 0.96 ± 0.09 nM, suggesting that they were slightly more effective than mature IGF-II in this system. Maximal stimulation of AIB uptake (mean ± S.E.) by rhIGF-I (2.42 ± 0.12-fold, n = 6), mature IGF-II (2.36 ± 0.12-fold, n = 5), and high M_r IGF-II (2.35 ± 0.16-fold to 2.59 ± 0.10-fold, n = 5) did not differ significantly in these experiments.

The small, but reproducible, differences between mature and high M_r IGF-II in stimulating AIB uptake in fibroblasts might be explained by reduced affinity of high M_r IGF-II for binding proteins secreted by the fibroblasts, resulting in greater availability for IGF receptor binding. To investigate this possibility, binding of rhIGF-I, mature bovine IGF-II, and high M_r IGF-II isoforms C-H to the IGF-binding proteins secreted by fibroblasts was investigated. The major IGF-binding protein present in fibroblast-conditioned buffer is IGFBP-3 (35). Fig. 9 shows displacement curves from a representative study. The mean dissociation constant derived from LIGAND analysis of IGF-I binding to secreted binding proteins in two independently-performed experiments was 0.21 nM, slightly higher than reported previously (35). IGF-II bound with a very similar affinity, whereas high M_r IGF-II bound with a similar or slightly higher affinity. These results indicate that the enhanced effectiveness of high M_r IGF-II in stimulating AIB uptake cannot be explained by reduced affinity for fibroblast IGF-binding proteins.

The biological activity of high M_r IGF-II was also investigated using the C2I subclone of C2 myoblasts selected by Pinset et al. (37). This cell line requires only insulin or IGF-I to differentiate in vitro. In preliminary experiments, we found that human or bovine IGF-II also induced terminal differentiation, as evidenced by equivalent maximal induction of creatine kinase activity and myotube formation. Fig. 10 shows a representative experiment comparing the dose-dependent induction of creatine kinase activity in C2I myoblasts by rhIGF-I, mature bovine IGF-II, and high M_r IGF-II isoforms D through G. In three experiments, the ED₅₀ of IGF-II-induced differentiation was 2.2-2.8-fold higher than that of rhIGF-I. The high M_r IGF-II isoforms were at least as or slightly more effective than mature IGF-II in this system.

Rotwein et al. (38) have shown that the expression of IGFBP-5 is stimulated during C2I myoblast differentiation. To examine IGF-binding protein expression during differentiation, media were collected from C2I myoblasts after incubation for ~70 h with rhIGF-I, mature bovine IGF-II, or high M_r IGF-II isoforms C-H and analyzed by ligand blotting. As shown in Fig. 11, the major IGF-binding protein stimulated during differentation migrated with an apparent molecular mass of 33.3 kDa. We confirmed by immunoblot analysis that this binding protein was IGFBP-5 (data not shown). Similar amounts of IGFBP-5 were present in conditioned media from myoblasts incubated with rhIGF-I, IGF-II, and high M_r IGF-II. An IGF-binding protein with an apparent molecular mass of 27.6 kDa was present in conditioned media from myoblasts incubated without IGF or insulin and did not change in amount during differentiation.

We next compared binding of rhIGF-I, mature bovine IGF-II, and high M_r IGF-II isoforms F and G to intact IGF-I receptor purified from human placenta. Fig. 12 shows displacement curves from this study. The mean dissociation constant derived from LIGAND analysis of rhIGF-I binding to the human IGF-I receptor in two independently performed experiments was 0.37 nM, very similar to that reported previously (30). The difference between this and the equilibrium dissociation constant obtained by surface plasmon resonance spectroscopy probably reflects the different receptor preparations employed and suggests that the extracellular portion of the receptor binds IGFs with a lower affinity than the intact receptor. The mean concentration of mature IGF-II which inhibited 50% of ¹²⁵I-labeled IGF-I binding to the receptor was 1.78 nM. Interestingly, the high M_r IGF-II isoforms bound to the purified IGF-I receptor with slightly higher affinity than mature IGF-II, a result which may explain their slightly enhanced biological activity.

DISCUSSION

These studies were initiated to characterize individual high M_r IGF-II isoforms. Preliminary analysis of the unfractionated high M_r IGF-II pool isolated from fetal bovine serum revealed that it contained at least 12 different IGF-II isoforms and that all species contained an identical peptide backbone extended beyond the COOH terminus of mature 7.5-kDa IGF-II (14). In addition, this pool contained a significant amount of sialated, O-linked sugars. In this study, the high M_r IGF-II pool was fractionated into nine distinct fractions to facilitate analysis. Taken together, results from amino-terminal sequence analysis, amino acid analysis, mass spectrometry, glycosidase digestions, isoelectric focusing, and SDS-PAGE experiments demonstrate that all fractions predominantly contain IGF-II species with a 20-amino acid COOH-terminal extension to Gly-87 that is modified with different amounts of sialated, O-linked oligosaccharides.

The presence of O-linked sugars on IGF-II precursors may represent an important modification to promote correct processing to the mature form (46). Mapping of the glycosylation site on an extended IGF-II isoform isolated from human serum indicated that Thr-75 carries this post-translational modification (18). While we did not directly determine the location of the modified residues for the different high M_r IGF-II isoforms, our mass spectrometry and isoelectric focusing results strongly suggest that most isoforms are likely to be glycosylated at multiple positions. The E-peptide extension (residues 68-87) contains five potential O-linked glycosylation sites (two serines and three threonines; Ref. 47). If three sites were modified with a tetrasaccharide consisting of a Gal1-3GalNAc core modified with two sialic acids, this would give a predicted size of ~12,630 and an isoelectric point of 3.9. In comparison, the major species of peak H had an isoelectric point of 3.7 and the M_r range of 10,500-12,500. Similarly, an 87-residue IGF-II with one site modified with the tetrasaccharide would have a predicted isoelectric point of 4.4 and a M_r of 10,600. In comparison, the major species of peak D had an isoelectric point of 4.35 and an M_r of 10,854. Thus, the most likely explanation for the heterogeneity of the different high M_r isoforms is that they contain 1-3 O-linked oligosaccharides, each of which contains one to two sialic acids.

Binding studies with mature 7.5-kDa IGF-II and the high M_r IGF-II isoforms indicate that no substantial differences exist in the association or dissociation rate constants (Tables II and III, respectively) or in the calculated equilibrium dissociation constants (Table IV) for interaction with the sIGF-II/MPR, sIGF-I receptor, or IGFBP-1. In addition, similar distribution profiles among serum-binding proteins (Fig. 7), and similar displacement curves from IGFBP-3 (Fig. 9) and intact human IGF-I receptor (Fig. 12) were observed for both mature and high M_r IGF-II. This suggests that the COOH-terminal extensions do not have adverse effects on the rates or affinities with which these ligands interact with their receptors and binding proteins. This is consistent with other reports. Site-directed mutagenesis studies interpreted in light of the three-dimensional structure of mature IGF-II have implicated three regions of the folded molecule that appear to play important roles in determining protein-protein interactions (48-50). Whereas residues 27 and 43 of the B and A domains, respectively, play key roles in IGF-I receptor interactions, residues 48-50 and 54-55 of the A chain are important in mediating IGF-II/MPR interactions. Likewise, residues 48-50, together with 6-7 of the B domain, are important in IGFBP interactions. More importantly, mutations within, or deletion of, the D domain (comprising the carboxyl-terminal six amino acids of mature IGF-II) has no effect on IGFBP interactions (48). Finally, it has been shown that extended forms of human IGF-II display similar binding affinities for the IGF-II/MPR (20) and the IGF-I receptor (51) as compared with mature IGF-II. Taken together, these data suggest that E-peptide extensions, with or without glycosylation, do not supplement or occlude the important contact regions between IGF-II and the sIGF-II/MPR, IGF-I receptor, or the IGFBPs.

An intriguing question still remaining, therefore, is what is the biological significance of the different IGF-II isoforms? One possible answer is that glycosylation is important for proper folding or export of IGF-II. Evidence suggests that glycosylation of proIGF-II is necessary for proper proteolytic processing to the mature 7.5-kDa form (46). If so, the mechanism may involve binding of glycosylated IGF-II to intracellular lectins that retain and present the glycoprotein to the appropriate processing enzymes (52). In this case, the sIGF-II/MPR-bound high M_r IGF-IIs may simply reflect inefficiencies in this system resulting in the secretion of processing intermediates. Alternatively, the glycosylated IGF-II species may have unique biological functions. This is not unprecedented as N-linked glycosylation of several glycoprotein hormones is necessary for receptor activation but not binding (for review, see Ref. 53). Although no differences in the binding rates or affinities for mature and high M_r IGF-II were observed for the receptors and binding proteins analyzed in this study, this does not preclude the possibility that the high M_r isoforms have specialized roles in signal transduction. For instance, the extensions and/or saccharides could: 1) serve as a targeting signal to concentrate IGF-II in specific tissues or cell types; 2) interact with other cellular components, either when free or complexed with receptors or binding proteins, and/or; 3) alter the efficacy/potency of receptor activation after ligand binding. Such possibilities are supported from previous studies that demonstrated that a high M_r form of IGF-II with an apparent M_r of 15,000 was more potent than mature IGF-II in stimulating the replication of fetal dermal fibroblasts (20) and in the clonal expansion and differentiation of peripheral blood granulocyte colony-forming cells (54). Future studies will be required to determine if the high M_r IGF-II glycoproteins have other specialized biological activities or pharmacological properties.