Advertisement

The in Vivo Phosphorylation and Glycosylation of Human Insulin-like Growth Factor-binding Protein-5 *

  • Mark E. Graham
    Footnotes
    Affiliations
    Cell Signalling Unit, Children's Medical Research Institute, Locked Bag 23, Wentworthville, New South Wales 2145, Australia
    Search for articles by this author
  • Dean M. Kilby
    Footnotes
    Affiliations
    Kolling Institute of Medical Research, University of Sydney, Royal North Shore Hospital, Sydney, New South Wales 2065, Australia
    Search for articles by this author
  • Sue M. Firth
    Affiliations
    Kolling Institute of Medical Research, University of Sydney, Royal North Shore Hospital, Sydney, New South Wales 2065, Australia
    Search for articles by this author
  • Phillip J. Robinson
    Affiliations
    Cell Signalling Unit, Children's Medical Research Institute, Locked Bag 23, Wentworthville, New South Wales 2145, Australia
    Search for articles by this author
  • Robert C. Baxter
    Correspondence
    To whom correspondence should be addressed: Kolling Inst. of Medical Research, Royal North Shore Hospital, St. Leonards, New South Wales 2065, Australia. Tel.: 61-2-9926-8486; Fax: 61-2-9926-8484;
    Affiliations
    Kolling Institute of Medical Research, University of Sydney, Royal North Shore Hospital, Sydney, New South Wales 2065, Australia
    Search for articles by this author
  • Author Footnotes
    * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
    The on-line version of this article (available at http://www.mcponline.org) contains supplemental material.
    § Both authors contributed equally to this work.
      Mass spectrometry is often used to determine post-translational modifications by analysis of tryptic digests of proteins. Here we demonstrate that the analysis of tryptic peptides together with analysis of the full-length protein provided optimal characterization of insulin-like growth factor-binding protein-5 (IGFBP-5) phosphorylation and glycosylation. IGFBP-5 binds insulin-like growth factors with high affinity and has important roles in cell survival, differentiation, and apoptosis. Until now, the primary structure of IGFBP-5 has been incompletely defined. We analyzed human IGFBP-5 from T47D cells by mass spectrometry to determine all of the in vivo post-translational modifications. In full-length IGFBP-5, 31% of the protein was unmodified, 37% was monophosphorylated, and 4% was diphosphorylated with no other modification. The remaining 27% was glycosylated, more than half of which was also monophosphorylated. The major phosphorylation site was Ser96 in the central domain, and a minor phosphorylation site was Ser248 near the C terminus. Neither site was phosphorylated in vitro by casein kinase 2, ruling it out as the in vivo kinase. An in vivo phosphorylation site was also found in IGFBP-2 at an analogous position, Ser106. IGFBP-5 was heterogeneously O-glycosylated mainly by sialylated core 1 type glycans. The most abundant structure contained N-acetylhexosamine, hexose, and two N-acetylneuraminic acid carbohydrates. A small amount of sialylated core 2 type glycan was also present. Phosphorylation and O-glycosylation both affected IGFBP-5 binding to heparin but not insulin-like growth factor binding or ternary complex formation with the acid-labile subunit. The results reveal the first description of the in vivo phosphorylation of IGFBP-5 and its glycan composition.
      Insulin-like growth factor (IGF)
      The abbreviations used are: IGF, insulin-like growth factor; IGFBP, IGF-binding protein; ALS, acid-labile subunit; Hex, hexose; HexNAc, N-acetylhexosamine; NeuAc, N-acetylneuraminic acid; CK, casein kinase; RIA, radioimmunoassay; DRB, 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole; S/N, signal to noise.
      1The abbreviations used are: IGF, insulin-like growth factor; IGFBP, IGF-binding protein; ALS, acid-labile subunit; Hex, hexose; HexNAc, N-acetylhexosamine; NeuAc, N-acetylneuraminic acid; CK, casein kinase; RIA, radioimmunoassay; DRB, 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole; S/N, signal to noise.
      -binding protein-5 (IGFBP-5) is one of a family of six structurally related proteins with high binding affinity for IGF-I and IGF-II (
      • Baxter R.C.
      Insulin-like growth factor (IGF) binding proteins: Interactions with IGFs and intrinsic bioactivities.
      ). These proteins are all found in the extracellular environment. Functional roles inside the cell have also been described, including the translocation of IGFBP-3 and IGFBP-5 to the cell nucleus and their interaction with nuclear receptors of the retinoid receptor superfamily (
      • Firth S.M.
      • Baxter R.C.
      Cellular actions of the insulin-like growth factor binding proteins.
      ). All six IGFBPs are released into the bloodstream where they are believed to circulate either free or complexed to IGF-I or IGF-II. In the case of IGFBP-3 and IGFBP-5, binary complexes with IGFs can associate with a third protein, the acid-labile subunit (ALS), to form high molecular weight ternary complexes that act as relatively stable reservoirs of circulating IGF-I and IGF-II.
      In the cellular environment, IGFBPs can modulate IGF access to the IGF-I receptor, either inhibiting or enhancing IGF-I receptor signaling, and can also act in ways that appear to be independent of their IGF binding activity. Paradoxically IGF-dependent actions include both the inhibition and the potentiation of IGF signaling, depending on the cellular context (
      • Firth S.M.
      • Baxter R.C.
      Cellular actions of the insulin-like growth factor binding proteins.
      ). For example, IGFBP-5 has been described as an inhibitor of the cell survival function of IGF in mammary gland (
      • Marshman E.
      • Green K.A.
      • Flint D.J.
      • White A.
      • Streuli C.H.
      • Westwood M.
      Insulin-like growth factor binding protein 5 and apoptosis in mammary epithelial cells.
      ) but as a potentiator of IGF action in stimulating osteoblast proliferation (
      • Andress D.L.
      • Birnbaum R.S.
      Human osteoblast-derived insulin-like growth factor (IGF) binding protein-5 stimulates osteoblast mitogenesis and potentiates IGF action.
      ) in addition to its IGF-independent stimulatory action (
      • Miyakoshi N.
      • Richman C.
      • Kasukawa Y.
      • Linkhart T.A.
      • Baylink D.J.
      • Mohan S.
      Evidence that IGF-binding protein-5 functions as a growth factor.
      ).
      Post-translational modification of members of the IGFBP family is common. Several IGFBPs are secreted as phosphoproteins. In cell culture, IGFBP-3 shows constitutive phosphorylation of serine-containing motifs similar to consensus casein kinase 2 (CK2) phosphorylation sites. Serine phosphorylation of IGFBP-3 by CK2 in vitro affects its cell surface association and susceptibility to proteolysis but not its IGF binding affinity (
      • Coverley J.A.
      • Martin J.L.
      • Baxter R.C.
      The effect of phosphorylation by casein kinase 2 on the activity of insulin-like growth factor binding protein.
      ). In contrast, phosphorylation of IGFBP-3 by DNA-dependent protein kinase abolishes its IGF binding, increases its nuclear retention (
      • Schedlich L.J.
      • Nilsen T.
      • John A.P.
      • Jans D.A.
      • Baxter R.C.
      Phosphorylation of insulin-like growth factor binding protein-3 by deoxyribonucleic acid-dependent protein kinase reduces ligand binding and enhances nuclear accumulation.
      ,
      • Cobb L.J.
      • Liu B.
      • Lee K.W.
      • Cohen P.
      Phosphorylation by DNA-dependent protein kinase is critical for apoptosis induction by insulin-like growth factor binding protein-3.
      ), and has been reported to be essential for the apoptotic and growth-inhibitory effects of IGFBP-3 (
      • Cobb L.J.
      • Liu B.
      • Lee K.W.
      • Cohen P.
      Phosphorylation by DNA-dependent protein kinase is critical for apoptosis induction by insulin-like growth factor binding protein-3.
      ). The ability of IGFBP-1 to potentiate or inhibit IGF action appears to depend on its phosphorylation status. Highly phosphorylated IGFBP-1 has high binding affinity and is inhibitory to IGF action, whereas the hypophosphorylated protein has lower activity and potentiates IGF action (
      • Jones J.I.
      • Clemmons D.R.
      Insulin-like growth factors and their binding proteins: biological actions.
      ). A single preliminary report over a decade ago suggested that IGFBP-5 is secreted as a phosphoprotein, but nothing is known about the phosphorylation sites, the protein kinases involved, or the effect of phosphorylation on IGFBP-5 function.
      IGFBP-3 and IGFBP-4 have established N-glycosylation sites. N-Linked carbohydrate in IGFBP-3 is slightly inhibitory to cell surface association but has no other known function (
      • Firth S.M.
      • Baxter R.C.
      The role of glycosylation in the action of IGFBP-3.
      ). Although the structure of N-linked oligosaccharides on IGFBP-4 has been described, their function is unknown (
      • Chelius D.
      • Wu S.L.
      • Bondarenko P.V.
      Identification of N-linked oligosaccharides of rat insulin-like growth factor binding protein-4.
      ). O-Linked glycosylation has been found in IGFBP-5 (
      • Ständker L.
      • Wobst P.
      • Mark S.
      • Forssmann W.G.
      Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
      ) and IGFBP-6 (
      • Marinaro J.A.
      • Neumann G.M.
      • Russo V.C.
      • Leeding K.S.
      • Bach L.A.
      O-Glycosylation of insulin-like growth factor (IGF) binding protein-6 maintains high IGF-II binding affinity by decreasing binding to glycosaminoglycans and susceptibility to proteolysis.
      ) and also has been noted in an early description of IGFBP-1 (
      • Bohn H.
      • Kraus W.
      Isolation and characterization of a new placenta specific protein (PP12).
      ). Like the N-linked carbohydrate in IGFBP-3 (
      • Firth S.M.
      • Baxter R.C.
      Characterisation of recombinant glycosylation variants of insulin-like growth factor binding protein-3.
      ), the O-glycosylation of IGFBP-6 is inhibitory to cell surface binding (
      • Marinaro J.A.
      • Neumann G.M.
      • Russo V.C.
      • Leeding K.S.
      • Bach L.A.
      O-Glycosylation of insulin-like growth factor (IGF) binding protein-6 maintains high IGF-II binding affinity by decreasing binding to glycosaminoglycans and susceptibility to proteolysis.
      ).
      Part of the tertiary structure of IGFBP-5 has been described. Kalus et al. (
      • Kalus W.
      • Zweckstetter M.
      • Renner C.
      • Sanchez Y.
      • Georgescu J.
      • Grol M.
      • Demuth D.
      • Schumacher R.
      • Dony C.
      • Lang K.
      • Holak T.A.
      Structure of the IGF-binding domain of the insulin-like growth factor-binding protein-5 (IGFBP-5): implications for IGF and IGF-I receptor interactions.
      ) characterized the N-terminal domain by NMR and revealed two disulfide bond pairs at Cys47 to Cys60 and Cys54 to Cys80. Potentially IGFBP-5 has nine disulfide bonds using all of the highly conserved cysteines of the IGFBP family. Ständker et al. (
      • Ständker L.
      • Wobst P.
      • Mark S.
      • Forssmann W.G.
      Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
      ) determined by Edman sequencing that IGFBP-5 was O-glycosylated at Thr152; however, the structure of the glycan is unknown. The central domain that contains the O-glycosylation is also the predicted location of phosphorylation sites because other IGFBP family members are phosphorylated in this domain. This has not been investigated for IGFBP-5.
      We analyzed IGFBP-5 by mass spectrometry to determine the in vivo post-translational modifications. Using 32P-labeling of T47D cells and tryptic digestion of IGFBP-5, we identified two phosphorylation sites. We also determined the identity of the glycans at Thr152. By analyzing the intact protein we obtained quantitative data on the amount of phosphorylation and glycosylation of IGFBP-5, thereby providing information on the amount of protein in different pools of modified IGFBP-5. We also investigated CK2 as a candidate for the protein kinase that might phosphorylate the observed sites and explored the functional relevance of these modifications.

      EXPERIMENTAL PROCEDURES

       Metabolic 32P Labeling and Purification of IGFBP-5 from T47D Human Breast Carcinoma Cells—

      The method of in vivo radiolabeling of IGFBP-5 was similar to that used by Pattison et al. (
      • Pattison S.T.
      • Fanayan S.
      • Martin J.L.
      Insulin-like growth factor binding protein-3 is secreted as a phosphoprotein by human breast cancer cells.
      ). T47D breast carcinoma cells were cultured to subconfluence in RPMI 1640 medium containing 20 mm HEPES, 10% FCS, 10 mg/liter bovine insulin, and 4 mm glutamine. Cells were then incubated overnight in a similar but serum-free medium containing 0.1% BSA before a 1-h phosphate washout period using phosphate-free Dulbecco's modified Eagle's medium also containing BSA, insulin, and glutamine. Cells were then incubated for 24 h in fresh phosphate-free medium, as described above, with [32P]orthophosphoric acid (100 μCi/ml) and the anti-proteolytics α2-macroglobulin (5 μg/ml), leupeptin (0.5 μg/ml), and aprotinin (Trasylol, Bayer Pharmaceuticals, 1:20 dilution). Radioactive media were collected, and EDTA was added to a final concentration of 0.5 mg/ml. Initial experiments were performed in 12-place multiwells alongside non-radioactive replicate plates, and media were collected at various time points for analysis by radioimmunoassay (RIA) and Western ligand blot after precipitation using IGF-I-conjugated agarose. Kinase inhibitor studies were similarly performed, and resultant comparisons were made between control wells and those containing the CK2 inhibitor 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole (DRB) (Calbiochem) at 0.01 and 0.1 mm. For peptide mapping experiments, larger scale preparations of 32P-labeled IGFBP-5 were made in 420-cm2 flasks, and IGFBPs were semipurified from clarified medium by affinity chromatography using IGF-I-conjugated agarose. Bound IGFBPs were washed with 50 mm sodium phosphate, pH 6.5, and eluted using 1 m acetic acid. Radioactive fractions were pooled, lyophilized, and stored at −80 °C until used. Non-radiolabeled T47D cell-derived IGFBP-5 was purified from conditioned serum-free RPMI 1640 medium as described above. IGF-I affinity chromatography followed by reverse-phase HPLC yielded pure, intact IGFBP-5 free from IGFBP-2 and IGFBP-4 contamination as determined by mass spectrometry. Recombinant human IGFBP-5 (AdIGFBP-5) was expressed in 911 human retinoblastoma cells using an adenoviral vector as described previously (
      • Firth S.M.
      • Clemmons D.R.
      • Baxter R.C.
      Mutagenesis of basic amino acids in the carboxyl-terminal region of insulin-like growth factor binding protein-5 affects acid-labile subunit binding.
      ).

       Tryptic Digestion and HPLC—

      Approximately 15 μg of purified 32P-labeled IGFBP-5 from T47D cells was prepared and purified by HPLC as described previously (
      • Graham M.E.
      • Ruma-Haynes P.
      • Capes-Davis A.G.
      • Dunn J.M.
      • Tan T.C.
      • Valova V.A.
      • Robinson P.J.
      • Jeffrey P.L.
      Multisite phosphorylation of doublecortin by cyclin-dependent kinase 5.
      ). The protein was dissolved in Laemmli sample buffer and heated to near boiling for 5 min. Reduced cysteines were alkylated by addition of acrylamide to a final concentration of 2% (v/v) for 45 min prior to SDS-PAGE. The gel was zinc-stained (
      • Fernandez-Patron C.
      • Castellanos-Serra L.
      • Rodriguez P.
      Reverse staining of sodium dodecyl sulfate polyacrylamide gels by imidazole-zinc salts: sensitive detection of unmodified proteins.
      ), and the IGFBP-5 band was excised and diced before being destained (
      • Fernandez-Patron C.
      • Castellanos-Serra L.
      • Rodriguez P.
      Reverse staining of sodium dodecyl sulfate polyacrylamide gels by imidazole-zinc salts: sensitive detection of unmodified proteins.
      ) and washed three times with 1 ml of water. The gel pieces were digested with 250 ng of trypsin in 75 μl of 50 mm ammonium bicarbonate for 16 h at 37 °C. A volume of acetonitrile equal to half of the total volume was added to the solution, and the tryptic peptides were extracted after vortexing for 15 min. A further extraction was obtained with 40 μl of 5% formic acid solution after 15 min of vortexing. The combined extract was dried down to ∼4 μl. This solution was made up to 50 μl with 0.1% TFA and injected into a HPLC system (SMART system, GE Healthcare). The peptides were separated by reverse-phase chromatography using a 2.1 × 100-mm column (μRPC C2/C18, 3 μm, 120 Å, GE Healthcare) at 100 μl/min. The gradient was from 0 to 40% acetonitrile in 0.1% TFA in 25 min and then to 100% acetonitrile, 0.1% TFA in 10 min. Fractions (50 μl) were collected every 30 s, dried in a rotational vacuum concentrator (ALPHA-RVC IR, Martin Christ GmbH, Osterode, Germany), and dissolved in 5 μl of 10% acetonitrile solution. A 0.5-μl aliquot of each solution was spotted onto nitrocellulose and allowed to dry. The radioactive spots on the nitrocellulose were detected by a PhosphorImager (Storm 860, GE Healthcare) after exposure for 2 weeks.

       Alkaline Phosphatase Treatment and Mass Spectrometry—

      A 0.5-μl aliquot of each radioactive fraction was mixed with a 2-μl solution containing 50 mm ammonium bicarbonate, 2 mm magnesium chloride, and 0.3 units/μl Antarctic phosphatase (New England Biolabs) and incubated at 22 °C for 3 h. The phosphatase-treated sample was desalted using graphite or Poros R2 microcolumns as described previously (
      • Larsen M.R.
      • Graham M.E.
      • Robinson P.J.
      • Roepstorff P.
      Improved detection of hydrophilic phosphopeptides using graphite powder microcolumns and mass spectrometry: evidence for in vivo doubly phosphorylated dynamin I and dynamin III.
      ). Graphite microcolumns were specifically used to trap hydrophilic peptides eluting in the first few fractions. The peptides were eluted from these columns with matrix solution (0.3% TFA, 60% acetonitrile aqueous solution containing 10 mg/ml α-cyano-4-hydroxycinnamic acid and 10 mm ammonium dihydrogen phosphate) directly onto the MALDI plate. The phosphatase-treated sample was compared with an equal amount of untreated sample. The MALDI spots were analyzed using a Voyager DE-Pro MALDI mass spectrometer (Applied Biosystems, Boston, MA) in positive and negative linear and reflectron modes. External calibration was used. Phosphopeptides in the remaining sample were sequenced by tandem mass spectrometry using a QSTAR XL quadrupole-TOF mass spectrometer (Applied Biosystems/MDS Sciex) using the NanoSpray™ electrospray ion source (ESI-MSMS). Samples analyzed using the NanoSpray were concentrated using microcolumns and centrifuged into nanoelectrospray capillaries (Proxeon, Odense, Denmark) in 0.1% formic acid, 50% acetonitrile aqueous solution and were sprayed at 800–1,200 V. Argon was used as the collision gas. Tandem mass spectra of phosphopeptides were acquired in multiple channel acquisition mode for 5–15 min and were manually interpreted with the aid of Analyst 1.1 and BioAnalyst 1.1 (Applied Biosystems). Mass accuracy was between 20 and 60 ppm. No effort was made to distinguish between Gln/Lys and Leu/Ile. Phosphopeptides were distinguished from sulfated peptides (none found) by use of 32P-labeling and phosphatase treatment.
      Intact IGFBP-5 protein, either AdIGFBP-5 or protein purified from T47D cell medium, was reconstituted in formic acid and diluted to 0.1% formic acid. It was then concentrated and desalted using a Poros R2 microcolumn to a final concentration of 20 μm and sprayed into the QSTAR XL mass spectrometer using a nanoelectrospray capillary in 0.5% formic acid, 50% acetonitrile aqueous solution. AdIGFBP-5 was dephosphorylated by incubation with 0.3 units/μl Antarctic phosphatase as above except that 10 mm magnesium chloride was used. Fresh phosphatase was added after 3 h, and the reaction was continued for another 3 h. The IGFBP-5 was purified from the Antarctic phosphatase by precipitation on IGF-I-agarose beads. A residual amount of bovine serum albumin from the precipitation was removed by loading it onto a Poros R3 microcolumn and eluting in steps of 10% acetonitrile. The purest IGFBP-5 fraction, which eluted at 30% acetonitrile, was concentrated to 20 μm and analyzed by ESI-MS. Mass spectra were deconvoluted using the Bayesian Protein Reconstruct tool of BioAnalyst 1.1. Mass spectra were first background subtracted using BioAnalyst 1.1 with a window width setting of m/z 4. A molecular mass spectrum in the range 25,000–35,000 Da was produced from the m/z range of 850–2,000 with a step mass of 0.5 Da, a signal-to-noise threshold of 1.5, and a minimum intensity of 0.01% for 20 iterations. We have reported the apex molecular mass. Each spectrum was acquired immediately after calibration with [Glu1]-fibrinopeptide B (Sigma-Aldrich). All molecular masses were reproducible to within ±1 Da under three repeated analyses.

       Phosphorylation in Vitro with CK2—

      A reaction volume of 40 μl was used to phosphorylate purified AdIGFBP-5. The reaction contained 3 μg of dephosphorylated IGFBP-5, 10 mm magnesium sulfate, 20 mm Tris, pH 7.4, 25 units of CK2 (Sigma-Aldrich), and 40 μm ATP with 10 μCi of [γ-32P]ATP. This reaction was allowed to progress for 5 min. Another three lots of 3 μg of IGFBP-5 were phosphorylated with CK2 without 32P label for 30 min. These samples were also phosphorylated with extra magnesium (20 mm) and ATP (400 μm). After SDS-PAGE and tryptic digestion, prior to HPLC (see above), the three non-radioactive samples were added to the radioactive sample to boost the concentration of phosphopeptides, thereby improving the signal in the mass spectral analysis.

       Binding Assays—

      Non-phosphorylated and non-glycosylated samples for ligand binding comparisons were prepared by enzymatic treatment of AdIGFBP-5. Twenty micrograms of AdIGFBP-5 were dephosphorylated using Antarctic phosphatase as described above. Deglycosylated IGFBP-5 was achieved by incubation of AdIGFBP-5 (20 μg) with neuraminidase (50 milliunits) and O-glycosidase (10 milliunits) (Calbiochem) together overnight at room temperature in 50 mm sodium phosphate, pH 7.5, including the protease inhibitors α2-macroglobulin (5 μg/ml), leupeptin (0.5 μg/ml), and aprotinin (Trasylol, 1:20 dilution). Purification from the treatment enzymes was achieved using IGF-I affinity chromatography.
      Heparin binding analysis was performed by manually loading IGFBP-5 (1 μg) onto a heparin sulfate Hi-Trap microcolumn (Amersham Biosciences) in 10 mm sodium phosphate, pH 7.5, 0.1% BSA. A linear 30-ml gradient (0–1 m NaCl in binding buffer) was then used to elute the bound IGFBP-5 at a flow rate of 0.5 ml/min. Fractions of 1 ml were collected from which 50 μl were assayed by RIA.
      Binding of IGFBP-5 to IGF-I and IGF-II was performed by an adaptation of a method described previously (
      • Firth S.M.
      • Clemmons D.R.
      • Baxter R.C.
      Mutagenesis of basic amino acids in the carboxyl-terminal region of insulin-like growth factor binding protein-5 affects acid-labile subunit binding.
      ). Briefly IGFBP-5 (0.025–5 ng) was incubated with 125I-IGF-I or 125I-IGF-II (20,000 cpm/100 μl) in a final volume of 250 μl of buffer containing 50 mm sodium phosphate, pH 6.5, 0.25% BSA for 2 h at 22 °C. Cold γ-globulin (0.25%, 250 μl) was added, and binary complexes were precipitated using polyethylene glycol 6000 (25%, 500 μl) followed by centrifugation. Ternary complex formation assays were performed using 125I-ALS (10,000 cpm/100 μl) to which IGFBP-5 (0.01–5 ng) was allowed to bind in the presence of either IGF-I (10 ng) or IGF-II (10 ng) in a final volume of 300 μl of buffer containing 50 mm sodium phosphate, pH 6.5, 0.25% BSA for 2 h at 22 °C. In-house IGFBP-5 antiserum (SFK Y.1, 5 μl) was added and incubated overnight at 4 °C, and complexes were precipitated using sheep anti-chicken γ-globulin (2 μl) followed by 1 ml of 60 g/liter polyethylene glycol 6000 in 0.15 m NaCl and centrifugation.

      RESULTS

       IGFBP-5 Is Secreted as a Phosphoprotein—

      Metabolic labeling with 32Pi indicated that IGFBP-5 is secreted as a phosphoprotein by T47D human breast carcinoma cells. Proteins were precipitated from conditioned medium on IGF-I-agarose beads. Autoradiography demonstrated a single diffuse band corresponding to IGFBP-5 (Fig. 1a); however, the possibility that other IGFBPs secreted by these cells, IGFBP-2 and IGFBP-4 as shown by ligand blot (Fig. 1c, inset), also carry a low level of labeled phosphate cannot be excluded. The relative production of IGFBP-5 and IGFBP-2 by these cells was ∼2:1 as determined by RIA (Fig. 1b). In a larger scale purification, ∼700 ml of culture medium, heavily conditioned by 5 × 108 T47D cells, contained 420 μg of IGFBP-5, determined by RIA, that after IGF-I affinity chromatography and reverse-phase HPLC gave a yield of ∼300 μg of pure IGFBP-5. Fig. 1c shows that after IGF-I affinity chromatography, reverse-phase HPLC clearly separated IGFBP-5 from IGFBP-2 and IGFBP-4. Peaks at 27 and 36 min were immunologically confirmed as IGFBP-5 and IGFBP-2, respectively. The peak at 29 min was confirmed as IGFBP-4 by N-terminal sequencing (data not shown).
      Figure thumbnail gr1
      Fig. 1Metabolic labeling of IGFBP-5 by T47D breast carcinoma cells.a, cells were cultured in the presence of [32P]orthophosphoric acid, and IGFBP-5 in conditioned media was detected by Western immunoblotting (WIB) and autoradiography after SDS-PAGE. b, conditioned media from non-radioactive replicate wells were collected at time points up to 12 h, and levels of secreted IGFBP-2 (open bars) and -5 (closed bars) were quantitated by RIA. c, T47D breast carcinoma cell-derived IGFBPs were purified by IGF-I affinity chromatography and then fractionated by reverse-phase HPLC. UV absorbance is indicated on the left axis, and percent Phase B (60% acetonitrile in 0.1% TFA) is indicated on the right axis. An autoradiograph of a ligand blot of the HPLC fractions 26–37 inclusive using 125I-IGF-I is shown in the inset and overlaid at the appropriate retention times.

       Phosphorylation Sites of IGFBP-5 from T47D Cells—

      Purified IGFBP-5 from 32P-labeled T47D cells was reduced, alkylated with acrylamide, and resolved by SDS-PAGE (Fig. 2a). Three micrograms of the sample was visualized by autoradiography (lane 1) and stained with Coomassie Blue, which also revealed a minor BSA band (lane 2). The remainder of the sample was zinc-stained (
      • Fernandez-Patron C.
      • Castellanos-Serra L.
      • Rodriguez P.
      Reverse staining of sodium dodecyl sulfate polyacrylamide gels by imidazole-zinc salts: sensitive detection of unmodified proteins.
      ), and the entire ∼34-kDa band was excised and digested with trypsin. The tryptic digest was applied to a reverse-phase column, and peptides were collected in fractions. The radiation detected from each fraction was measured (Fig. 2b). Each radioactive fraction was subjected to Antarctic phosphatase treatment and compared with an untreated portion of the fraction using MALDI-TOF MS. The treated and untreated spectra were examined for signals that were a result of reductions in mass equal to multiples of 80 Da, demonstrating enzymatic dephosphorylation. The molecular mass of the signals identified as phosphopeptides by mass spectrometry were matched to theoretical molecular masses of IGFBP-5 tryptic peptides and are listed in Table I and indicated on Fig. 2b. Examples of the MALDI-TOF MS spectra are shown in Supplemental Fig. S1, a–h.
      Figure thumbnail gr2
      Fig. 2SDS-PAGE of IGFBP-5 from 32P-labeled T47D cells and separation of tryptic phosphopeptides separated by HPLC.a, an autoradiograph (lane 1) is shown next to 3 μg of the Coomassie-stained protein (lane 2). b, a total of 15 μg of IGFBP-5 was digested with trypsin and applied to a reverse-phase column. Fractions were collected, and the radiation was measured in each fraction as shown on the left axis. The percent increase in concentration of organic phase (Phase B) is shown on the right axis. Each phosphopeptide detected by MALDI-TOF MS and dephosphorylated in vitro by phosphatase ( and “Experimental Procedures”) was matched to the IGFBP-5 sequence by molecular mass. The matches to amino acids in the IGFBP-5 sequence are shown for each fraction in which they were detected. An IGFBP-2 phosphopeptide was also detected in fractions 29 and 30 as shown. c, the sequence of mature human IGFBP-5 from Swiss-Prot accession number P24593 without the signal sequence, i.e. the first 20 amino acids. The in vivo phosphorylation sites at Ser96 and Ser248, determined in this work, are shown in white on black. The known glycosylation site at Thr152 is underlined. Leu1, which was also shown in this work to be absent in a portion of the mature chain from T47D cells, is shaded. Putative heparin-binding motifs are indicated in boxes.
      Table IIGFBP-5 sequences matching the molecular mass of phosphopeptides detected by MALDI-TOF mass spectrometry
      SequenceFractionsExperimental m/zTheoretical m/z
      Monoisotopic, +veAverage, +veAverage, −ve
      IGFBP-5
       p92–973855.4
      These phosphopeptides were also sequenced by mass spectrometry.
      855.4
       p95–11518–222,533.0
      These phosphopeptides were also sequenced by mass spectrometry.
      ,
      IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      2,533.0
       p95–115222,614.8, diphospho
      IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      2,614.5
       p92–11518–222,931.3
      These phosphopeptides were also sequenced by mass spectrometry.
      ,
      IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      2,931.2
       p92–115223,012.9, diphospho
      IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      3,013.0
       p92–115+O
      Only the oxidized form of this phosphopeptide was detected.
      223,108.5 triphospho
      IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      3,109.0
       p92–120+O
      Only the oxidized form of this phosphopeptide was detected.
      19–213,588.73,588.8
       p88–115223,417.03,417.5
       p227–252
      The cysteine in this phosphopeptide is alkylated with acrylamide (see “Experimental Procedures”).
      213,122.33,122.3
       p231–252
      The cysteine in this phosphopeptide is alkylated with acrylamide (see “Experimental Procedures”).
      30 and 312,643.0
      These phosphopeptides were also sequenced by mass spectrometry.
      2,642.8
      IGFBP-2
       p100–14229 and 304,384.3
      These phosphopeptides were also sequenced by mass spectrometry.
      4,383.4
      a These phosphopeptides were also sequenced by mass spectrometry.
      b IGFBP-5-(95–115) was detected as a mono- and diphosphopeptide; IGFBP-5-(92–115) was detected as a mono-, di-, and triphosphopeptide.
      c Only the oxidized form of this phosphopeptide was detected.
      d The cysteine in this phosphopeptide is alkylated with acrylamide (see “Experimental Procedures”).
      Of the eight sequence assignments in Table I, three are tentative at this point because peptide sequencing data are required to rule out contamination by an unknown protein, however improbable. Note that many of the detected phosphopeptides contained Met and Cys and so were detected as mono- and dioxidized phosphopeptides (+16 or +32 Da), increasing the evidence for their identification. Generally the more oxidized phosphopeptides eluted earlier (fractions 18–20; Fig. 2b) in the HPLC than the non-oxidized phosphopeptides (fractions 21 and 22), resulting in a small broad peak before a sharper peak. The relatively large amount of starting protein enabled the detection of low abundance, incompletely cleaved tryptic phosphopeptides using the more sensitive modes of MALDI-TOF MS detection for phosphopeptides, i.e. linear mode with positive or negative ion detection, rather than the less sensitive reflectron mode with positive ion detection (Table I). Phosphopeptides detected in these modes were at or near the limit of detection. These phosphopeptides should be regarded as either “difficult” to detect (e.g. large phosphopeptides) or so low in abundance that their physiological relevance is questionable.
      Most of the phosphopeptides detected were from the same sequence of the IGFBP-5, 88EQVKIERDSREHEEPTTSEMAEETYSPKIFRPK120, with up to three missed trypsin cleavages (Fig. 2c). In assigning the 32P radioactive peaks, these phosphopeptides were found to contain the majority of radiation. Fractions 3 and 18–22 contained 88.5% of the total measured radiation (Fig. 2b). The remaining phosphopeptides from IGFBP-5 were from the C-terminal sequence 227YGMKLPGMEYVDGDFQCHTFDSSNVE252. The contribution of phospho-IGFBP-5-(227–252) to fraction 21 is probably negligible because it was detected as a very weak signal, whereas the contribution of 231LPGMEYVDGDFQCHTFDSSNVE252 in fractions 30 and 31 is small but significant. The exact contribution, in terms of radiation, was not possible to calculate because this phosphopeptide co-eluted with a contaminating phosphopeptide matching IGFBP-2-(100–142). However, there is sufficient resolution of the two phosphopeptides, in fractions 29, 30, and 31, to conclude that phospho-IGFBP-5-(231–252) contributed between 4.5 and 7.1% of the total radiation.
      The isolation and detection of an IGFBP-2 phosphopeptide was a serendipitous finding. Most of the IGFBP-2 was excluded in the purification of IGFBP-5 (Fig. 1c), and a significant signal (>1%) for IGFBP-2 was undetectable in subsequent ESI-MS analysis of all the intact protein in our preparation (data not shown). Nevertheless a tryptic phosphopeptide was detected (Supplemental Fig. S1, g and h). Due to the abundance of important fragment ions, we localized a phosphorylation site for IGFBP-2. The phosphorylated serine was detected between b6 and b7 as a dehydroalanine residue (resulting from neutral loss of H3PO4 from a phosphoserine residue) and again as the intact phosphoserine residue between y363+ and y373+ (Fig. 3). Therefore, Ser106 is an in vivo phosphorylation site for IGFBP-2.
      Figure thumbnail gr3
      Fig. 3IGFBP-2 was co-purified with IGFBP-5 from 32P-labeled T47D cells, and a phosphorylation site was found. Using tandem MS, the quadruply charged parent ion at m/z 1,104.7 from fraction 29 was selected for fragmentation. The sequence describes IGFBP-2-(100–142) where Ser106 is phosphorylated. Met131 and Met133 were oxidized (shown underlined). The Asn113 predicted by the sequence (Swiss-Prot accession number P83628) was detected as Asp (or iso-Asp)113, indicating that deamidation had occurred. The parent ion has been truncated, and the m/z range from 950 to 1,650 has been multiplied by a factor of 7 to improve the clarity of the fragment ions in this region of the spectrum.
      Only a subset of the phosphopeptides shown in Table I was suitable for sequencing by tandem MS. The small, hydrophilic phosphopeptide detected in fraction 3 (Fig. 2b) at m/z 855.40 was sequenced by MALDI-TOF MSMS (data not shown). The sequence was confirmed as IGFBP-5 92IERDSR97 where Ser96 is phosphorylated. The location of the phosphoserine at Ser96 explains the strong MALDI-TOF MS signal we observed (data not shown) for phospho-IGFBP-5 92IERDSREHEEPTTSEMAEETYSPK115. It is likely that Ser96 phosphorylation contributed to the inhibited trypsin cleavage after Arg94 and/or Arg97. Sequencing of phospho-IGFBP-5-(92–115) revealed that Ser96 was also the phosphorylation site on this peptide (Fig. 4a). The phosphorylated Ser96 was detected by a transition from the b4 fragment ion to the b5 − 98 as a dehydroalanine residue. The C-terminal part of the sequence is mainly described by abundant y type ions that have no additional phosphate. However, there were some fragment ions, detected with low relative abundance, that show there was another phosphoamino acid residue near the C terminus of this phosphopeptide (data not shown). The fragment ions that attest to an alternative sequence were: y14, S/N = 4 (which was phosphorylated) and b92+, S/N = 3 and b102+, S/N = 16 (which were not phosphorylated). Collectively these data suggest that there was an alternative phosphorylation site within IGFBP-5-(103–113). The alternate sequence data was between 9- and 70-fold less intense than the dominant sequence. Therefore, any extra phosphorylation sites would have a very low stoichiometry. This matches the MALDI-TOF MS data in which low abundance di- and triphosphorylated signals were detected for IGFBP-5-(92–115) and -(95–115), and a low intensity monophosphorylated signal was detected for IGFBP-5-(98–115) (Table I). Ser96 was also phosphorylated when IGFBP-5-(95–115) was sequenced (data not shown). There was no evidence for an alternative phosphorylation site. Therefore, Ser96 is undoubtedly the major phosphorylation site in IGFBP-5, but there may be nearby phosphorylation sites of very low stoichiometry.
      Figure thumbnail gr4
      Fig. 4Tryptic phosphopeptides of 32P-labeled IGFBP-5 from T47D cells separated by HPLC and sequenced by tandem MS.a, tandem MS spectrum of the triply charged parent ion at m/z 983.1 from fraction 21. The sequence matched IGFBP-5-(92–115) where Ser96 is phosphorylated and Met107 is oxidized (shown underlined). The parent ion has been truncated, and the m/z range from 990 to 2,000 has been multiplied by a factor of 30 to show the fragment ions more clearly. Neutral loss of methane sulfenic acid (CH3SOH, −64 Da) is shown. b, tandem MS spectrum of the triply charged parent ion at m/z 886.4 from fraction 31. The sequence matched IGFBP-5-(231–252) where Ser248 is phosphorylated. The deduced presence of oxidized Met234 is shown underlined. Also the Cys243 alkylated with acrylamide (see “Experimental Procedures”) is shown underlined. The m/z range from 900 to 1,400 has been multiplied by a factor of 10 to show the fragment ions more clearly.
      Phospho-IGFBP-5-(231–252) was sequenced, and it was found that Ser248 was phosphorylated (Fig. 4b). The phosphorylation site, detected as a dehydroalanine residue, spanned between b172+ and b182+. Therefore, the 32P labeling and MSMS sequencing showed that IGFBP-5 had a major phosphorylation site at Ser96, accounting for 88.5% of the incorporated phosphate, and a minor phosphorylation site at Ser248.

       Glycosylation Sites of IGFBP-5—

      Signals for glycopeptides (showing neutral loss of carbohydrate) were detected in fractions 15 and 16 (Fig. 2b) when analyzed by MALDI-TOF MS. A base peak signal at m/z 1,255.6 was detected in the same spectra, matching the molecular mass of IGFBP-5-(145–156). This peptide was identified previously as containing an O-linked glycosylation site at Thr152 (
      • Ständker L.
      • Wobst P.
      • Mark S.
      • Forssmann W.G.
      Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
      ). However, in that study the constituents of the O-linked glycan were not determined or studied in any detail.
      The glycopeptide fractions were combined and further analyzed using tandem MS (ESI-MSMS). Several related glycopeptides were detected as shown in Fig. 5a. Each glycopeptide was fragmented, and the carbohydrate constituents were determined. An example of a fragmentation product spectrum is shown in Fig. 5b. Because only molecular mass information was revealed, only the size of the carbohydrates, but not the structure of the glycan, could be determined. However, it was clear that the glycans were combinations of hexose (Hex), N-acetylhexosamines, (HexNAc), and N-acetylneuraminic acid (NeuAc, or sialic acid). Each glycopeptide contained some y and b type ions as well (Fig. 5b) that were a result of the fragmentation of the IGFBP-5-(145–156) sequence conjugated to the glycan. The non-modified signal was also fragmented confirming that the peptide was IGFBP-5-(145–156) (data not shown).
      Figure thumbnail gr5
      Fig. 5Tryptic glycopeptides of IGFBP-5 from T47D cells separated by HPLC detected by ESI-MS and sequenced by tandem MS.a, ESI-MS spectrum of combined fractions 15 and 16 containing glycopeptides. The peptide IGFBP-5-(145–156) is heterogeneously modified by O-linked glycans containing HexNAc, Hex, and NeuAc shown as square, circular, and triangular symbols, respectively. Each glycopeptide signal was sequenced (data not shown, except for the most abundant, see below), and the most likely structure was determined from the sequence information and is shown for each glycopeptide signal. b, tandem MS spectrum of the doubly charged parent at m/z 1,102.0 revealed a glycopeptide with HexNAc, Hex, and two NeuAc carbohydrates. This glycopeptide was the most abundant. Examples of the possible glycan structures and their fragmentation are shown above the spectrum in b.
      Structures based on common core types are shown for each glycopeptide in Fig. 5a. Structures with alternative branching are equally possible except for the rule that Thr152 is always O-linked to HexNAc. Also NeuAc is normally a terminating carbohydrate. The most abundant signal was detected as an intense triple charged ion as well as a double charged ion (Fig. 5a). Tandem MS of this abundant signal revealed a glycopeptide with HexNAc, Hex, and two NeuAc carbohydrates (Fig. 5b). An unconfirmed signal for this glycosylation was detected previously by MS (
      • Ständker L.
      • Wobst P.
      • Mark S.
      • Forssmann W.G.
      Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
      ). Mono- and trisialylated forms of this core 1 type structure were also detected. A glycan with only two HexNAc carbohydrates was relatively abundant, whereas the larger glycans built on the di-HexNAc structure (sialylated core 2 type) were in low abundance. Note that these comparisons of signal intensity are only qualitative. In conclusion, IGFBP-5 is heterogeneously glycosylated at Thr152 with a mixture of sialylated and unsialylated core 1 and core 2 type structures. Core 1 structures are much more abundant, and the most abundant structure is one with HexNAc, Hex, and two NeuAc carbohydrates.

       Post-translational Modification of Intact IGFBP-5 in Vivo—

      To quantify the extent of post-translational modification, we analyzed intact or full-length, rather than protease-treated, IGFBP-5. Analysis of purified human AdIGFBP-5 by ESI-MS confirmed the presence of the same phosphorylation sites, including phospho-Ser96, and confirmed similar glycopeptides (data not shown). Therefore a similar amount of post-translational modification of the intact protein was expected. The raw ESI-MS data from intact AdIGFBP-5 was converted to molecular masses (see “Experimental Procedures”) and is shown in Fig. 6a. The mature chain of human IGFBP-5 has a molecular mass of 28,573 Da (Fig. 2c, Swiss-Prot accession number P24593 without the signal sequence, i.e. the first 20 amino acids), which would be reduced to 28,555 Da if all of the 18 cysteines were disulfide bonded as is believed to be the case for all IGFBPs. IGFBP-5 was found to be highly modified. We removed the contribution of phosphorylation by incubating IGFBP-5 with phosphatase (Fig. 6b). Signals at +1,027, +738, +444, +283, +160, and +80 Da were absent or greatly decreased after dephosphorylation. The molecular mass of the unmodified form of the protein was 28,556 Da prior to incubation with phosphatase and 28,564 Da after incubation. The difference in mass between these measurements is probably due to decreased preservation of disulfide bonds as the sample handling increased. Assuming the theoretical molecular mass is correct, then eight to nine disulfide bonds were preserved during the analysis of untreated IGFBP-5, and only four to five disulfide bonds were preserved after phosphatase treatment. The small signals at 15–18 and 30–36 Da larger than the five most prominent signals in the spectrum (Fig. 6a) were probably contributions from completely disulfide reduced and/or oxidized IGFBP-5. The phosphatase-treated sample was probably also affected by oxidation to a greater extent as shown by the presence of strong signals at 15 and 31 Da higher than the dominant signal (Fig. 6b).
      Figure thumbnail gr6
      Fig. 6Full-length AdIGFBP-5 and T47D cell-derived IGFBP-5 analyzed by mass spectrometry.a, intact AdIGFBP-5 was analyzed by ESI-MS. Multiply charged ions were converted to the molecular masses (see “Experimental Procedures”). All molecular masses are shown relative to the 28,556 Da signal, assumed to be IGFBP-5 with eight to nine disulfide bonds retained. The identity and amount of the post-translational modifications are shown in . b, AdIGFBP-5 was treated with Antarctic phosphatase (AP) before being analyzed by ESI-MS. Comparison of a with b allowed phosphorylated forms of the protein to be clearly identified. c, intact human IGFBP-5 from T47D cells was analyzed by ESI-MS to reveal a pattern of modification similar to that of AdIGFBP-5 in a. A pool of full-length T47D IGFBP-5, which was 113 Da less, is marked (−113).
      To arrive at a measure of the amount of modified and unmodified protein, we continued with the assumption that the signal at 28,556 Da was intact and unmodified IGFBP-5 (except for disulfide bonds). Also we grouped oxidized/disulfide reduced signals with the non-oxidized/disulfide bonded signals. From the molecular masses of the modified intact IGFBP-5 and the peptide sequencing data (Figs. 4 and 5) we were able to quantify the degree of phosphorylation as well as the degree and type of glycosylation. Table II shows the quantification of phosphorylated and glycosylated forms of IGFBP-5 from Fig. 6a. Approximately 31% of the protein was unmodified, 37% was monophosphorylated without any other modification, and 4% was diphosphorylated. IGFBP-5 containing at least one phosphate represented 56% of the total protein. The full-length IGFBP-5 data supports the findings from 32P labeling (see above and Fig. 2b) that there was one major phosphorylation site and one minor phosphorylation site. Although there is some evidence of as yet unidentified phosphorylation sites in IGFBP-5-(98–115) (see above), the vast majority of the monophosphorylated IGFBP-5 is undoubtedly phosphorylated on Ser96, and the second, low abundance site is mostly Ser248.
      Table IIQuantification of phosphorylated and glycosylated forms of IGFBP-5
      Molecular massΔMass
      Masses are relative to the unmodified IGFBP-5, i.e. 28,556 Da.
      Modifications
      Identity of post-translational modifications were determined by phosphatase treatment (Fig. 5b) and by sequencing of modified peptides (Figs. 3 and 4).
      ,
      Modification by reduction of disulfide bonds or oxidation not listed.
      Percentage of total protein
      DaDa%
      28,556, 28,571, and 28,5860, 15, and 30None31.3
      28,636, 28,651, and 28,66680, 95, and 110HPO337.2
      28,716 and 28,734160 and 178(HPO3)24.1
      28,839283HPO3 + HexNAc0.8
      28,922366HexNAc and Hex1.0
      29,001444HPO3 + HexNAc and Hex1.5
      29,047491HexNAc and NeuAc or HPO3 + (HexNAc)20.9
      29,212 and 29,228656 and 672HexNAc, Hex, and NeuAc2.8
      29,294 and 29,309738, 753, and 771HPO3 + HexNAc, Hex, and NeuAc3.7
      29,505, 29,520, and 29,535949, 964, and 979HexNAc, Hex, and (NeuAc)26.3
      29,583, 29,602, and 29,6171,027, 1,046, and 1,061HPO3 + HexNAc, Hex, and (NeuAc)29.7
      a Masses are relative to the unmodified IGFBP-5, i.e. 28,556 Da.
      b Identity of post-translational modifications were determined by phosphatase treatment (Fig. 5b) and by sequencing of modified peptides (Figs. 3 and 4).
      c Modification by reduction of disulfide bonds or oxidation not listed.
      Overall approximately one-quarter of the protein was glycosylated, and more than half of this pool was also monophosphorylated. The most abundant glycan, 16% of the total, was a combination of HexNAc, Hex, and two NeuAc carbohydrates. A monosialylated form of this structure was 6.5% of the total. The remainder of the glycans were generally in low abundance. These results are in agreement with the glycopeptide data for adenoviral vector-derived (data not shown) and T47D cells (Fig. 5a). Both analyses of peptides and full-length IGFBP-5 indicate that there is a single glycosylation site containing a mixture of mainly core 1 type glycans.
      We also analyzed IGFBP-5 from T47D cells to determine whether endogenously expressed IGFBP-5 has a similar profile of post-translational modification. Fig. 6c shows the molecular mass spectrum of IGFBP-5 from T47D cells. This spectrum differs from the adenoviral vector-derived IGFBP-5 in that there are signals showing a loss of 113 Da in addition to the main signals of normal mass. A likely reason for this is the absence of Leu1 from the mature IGFBP-5 chain (Fig. 2c). Signals matching the molecular mass of 1LGSFVHCEPCDEK13 and 2GSFVHCEPCDEK13 were both detected by MALDI-TOF MS. This suggests some variability in the site of cleavage for the signal peptide from the IGFBP-5 precursor. It is notable that the IGFBP-5 with loss of 113 Da is not phosphorylated to the same extent as the full-length protein. Apart from the loss of 113 Da, the state of post-translational modification of the main signals is very similar to the AdIGFBP-5. In both signal strength and the extent of modification, there is generally similar phosphorylation and glycosylation. One clear exception is that the signal for monosialylated HexNAc + Hex (+656 Da) is slightly more intense in IGFBP-5 from T47D cells (Fig. 6c) than in AdIGFBP-5 (Fig. 6a).

       Phosphorylation of IGFBP-5 in Vitro with Casein Kinase 2—

      Because the phosphorylation sites at Ser96 and Ser248 are generally located near acidic residues, we sought to determine whether IGFBP-5 was a substrate for phosphorylation by the acidophilic CK2. The metabolic incorporation of [32P]phosphate into IGFBP-5 over 16 h in T47D breast cancer cells was significantly inhibited (p = 0.0075) by DRB, a CK2 inhibitor. At 0.1 mm, DRB reduced [32P]phosphate incorporation by 50% when densitometry analysis was normalized to its own immunostained internal control (Fig. 7, a and b). In vitro assays with 32P labeling also showed that IGFBP-5 could act as a substrate for CK2 before or after dephosphorylation with Antarctic phosphatase (Fig. 7c). The decreased 32P incorporation observed in Fig. 7c, lane 4, may be due to some residual phosphatase activity as a result of incomplete Antarctic phosphatase inactivation. Autophosphorylation of the CK2 catalytic subunit (44 kDa) and regulatory subunit (26 kDa), as seen in Fig. 7c, lane 5, is distinct from IGFBP-5 (30–32 kDa). We therefore phosphorylated phosphatase-treated IGFBP-5 with CK2 in vitro with [γ-32P]ATP, resolved the protein by SDS-PAGE, cut out and digested the band, separated the peptides by HPLC, and analyzed them by MALDI-TOF MS as above for the in vivo phosphorylation site analysis. The pattern of radioactive fractions collected from the HPLC was similar (Fig. 8a) with one main group and some low radiation fractions eluting late. However, an important exception was the lack of an early eluting fraction that should have contained phospho-IGFBP-5-(92–97).
      Figure thumbnail gr7
      Fig. 7The role of CK2 in IGFBP-5 phosphorylation.a, metabolic labeling of T47D breast carcinoma cells using [32P]orthophosphoric acid in the absence or presence of either 0.01 or 0.1 mm DRB. IGFBP-5 was extracted from media on IGF-I-agarose and subjected to SDS-PAGE and Western immunoblotting (WIB) followed by autoradiography (AR). b, subsequent densitometry analysis was used to quantitate the effect of including DRB, internally normalized to immunostaining. Data are the mean of three independent experiments, performed in duplicate, expressed as a percentage of the 16-h control. (▪, 0 mm DRB; •, 0.01 mm DRB; ♦, 0.1 mm DRB). c, IGFBP-5 is a substrate for CK2 in vitro. IGFBP-5 (untreated, lanes 1 and 2; Antarctic phosphatase (AP)-treated, lanes 3 and 4) was incubated with [γ-32P]ATP in the absence (lanes 1 and 3) or presence (lanes 2 and 4) of CK2 and analyzed by both Western immunoblotting (WIB) and autoradiography (AR). Lane 5 illustrates the autophosphorylation of CK2 in the absence of IGFBP-5.
      Figure thumbnail gr8
      Fig. 8Tryptic phosphopeptides of AdIGFBP-5, [γ-32P]ATP-labeled in vitro by phosphorylation with CK2, were separated by HPLC and detected and sequenced by mass spectrometry.a, [γ-32P]ATP-labeled phosphopeptides were collected in fractions, and the radiation was measured as shown on the left axis. The percent increase in concentration of organic phase (Phase B) is shown on the right axis. Phosphopeptides that matched to the IGFBP-5 sequence are shown for each fraction in which they were detected. b, tandem MS spectrum of the triply charged parent ion at m/z 850.3 from fraction 18 of the HPLC separation shown in a. The sequence matched IGFBP-5-(95–115) where Thr103 is phosphorylated and Met107 is oxidized (shown underlined). The parent ion has been truncated, and the m/z range from 980 to 1,700 has been multiplied by a factor of 15 to show the fragment ions more clearly. c, tandem MS spectrum of the triply charged parent ion at m/z 881.0 from fraction 33. The sequence matched IGFBP-5-(231–252) where Ser249 is phosphorylated. Cys243 was alkylated with acrylamide (see “Experimental Procedures”) and is shown underlined. The y4 and the b4 ions were equal in molecular mass; however, this did not affect identification of the phosphorylation site.
      Subsequent MALDI-TOF MS and phosphatase analysis showed that phosphopeptides similar to the in vivo phosphopeptides were detected, i.e. IGFBP-5-(92–115), -(95–115), -(98–115), and -(231–252) (Fig. 8a; c.f. Fig. 2b and Table I). The strongest signals were for IGFBP-5-(95–115) and -(98–115). This together with the lack of IGFBP-5-(92–97) suggested that Ser96 was no longer inhibiting cleavage after Arg94 and/or Arg97. Sequencing of phosphoIGFBP-5-(95–115) by ESI-MSMS showed that the in vitro phosphorylation was on Thr103 (Fig. 8b). The phosphothreonine residue was detected between fragment ions b8 and b9 or alternatively between y12 and y13. The C-terminal phosphopeptide IGFBP-5-(231–252) was detected with relatively low stoichiometry of incorporation of 32P label similar to the in vivo phosphopeptide. However, when sequenced, it was found that Ser249 was phosphorylated (Fig. 8c) rather than Ser248. The phosphoserine residue was detected between fragment ions b18 and b19 or alternatively between y3 and y4. Both Thr103 and Ser249 are in (S/T)XXE consensus sequences. Nevertheless the results indicate that although IGFBP-5 is a substrate in vitro for CK2 on Thr103 and Ser249 it is unlikely to be a substrate for this kinase in vivo.

       Functional Evaluation of IGFBP-5 Phosphorylation and Glycosylation—

      To evaluate the functional consequences of IGFBP-5 post-translational modifications, the recombinant protein expressed using an adenoviral vector was dephosphorylated using Antarctic phosphatase, O-deglycosylated using a combination of O-deglycosidase and neuraminidase, and then repurified to remove the enzymes. Fig. 9a shows a Western immunoblot of the purified treated preparations. Dephosphorylation consistently resulted in more highly resolved IGFBP-5 variants after SDS-PAGE (Fig. 9a, lane 2), and deglycosylation produced a single IGFBP-5 band (Fig. 9a, lane 3). The Western ligand blot using 125I-IGF-I demonstrated no resultant major loss of ligand binding activity, and enzymatic treatment had no effect on immunological recognition as determined by RIA (data not shown). Because IGFBP-5 is known to associate with a variety of cell surface and matrix proteins, the effect of treatments on the interaction of IGFBP-5 with heparin (a typical sulfated glycosaminoglycan) was first evaluated. Fig. 9b shows that both enzymatic treatments enhanced the association of IGFBP-5 with heparin; that is, deglycosylated and dephosphorylated IGFBP-5 preparations were eluted from a heparin-agarose affinity column at a higher salt concentration than the untreated protein. Despite equivalent loading of the IGFBP-5 preparations onto the heparin column, the treated samples consistently displayed decreased recovery after elution, ∼50% compared with untreated IGFBP-5. A second peak at elution fraction 30 was also observed with dephosphorylated IGFBP-5, suggesting a second structural subspecies.
      Figure thumbnail gr9
      Fig. 9Functional studies of dephosphorylated and deglycosylated adenoviral vector-derived IGFBP-5.a, untreated IGFBP-5 (lane 1), Antarctic phosphatase-treated IGFBP-5 (lane 2), and neuraminidase- and O-glycosidase-treated IGFBP-5 (lane 3) were compared in Western immunoblotting (WIB) and 125I-IGF-I ligand blotting followed by autoradiography (WLB). b–f, functional comparisons of untreated (▪), dephosphorylated (♦), and deglycosylated (•) IGFBP-5. b, heparin-bound IGFBP-5 was eluted with a linear NaCl gradient from 0 to 1 m. Elution fractions were collected (x axis) and assayed for IGFBP-5 by RIA (y axis). c–f, solution binding analysis of increasing concentrations of IGFBP-5 (0.025–5 ng) to 125I-IGF-I tracer (c), 125I-IGF-II tracer (d), and 125I-ALS in the presence of either IGF-I (e) or IGF-II (f). Binding is expressed as a percentage of total tracer bound corrected for nonspecific binding. B/T, bound/total radioactivity.
      Effects of post-translational modifications on ligand interactions with IGFBP-5 were evaluated by measuring the binding of IGF-I and IGF-II to the treated and untreated preparations as well as the formation of ternary complexes containing IGFBP-5, IGF-I or -II, and ALS. Fig. 9, c and d, shows solution binding curves of the IGFBP-5 variants to IGF-I and IGF-II, respectively. Statistical analysis at half-maximal binding (0.5 ng of IGFBP-5) showed no significant difference between the preparations. The ability of dephosphorylated and deglycosylated IGFBP-5 to bind ALS is shown in Fig. 9, e and f. Treatment of IGFBP-5 with Antarctic phosphatase appeared to shift the ALS binding curve to the right in the presence of either IGF-I (Fig. 9e) or IGF-II (Fig. 9f); however, when tested at approximately half-maximal binding (1 ng of IGFBP-5), differences between the two treatment groups (dephosphorylated and deglycosylated IGFBP-5) and untreated IGFBP-5 were not statistically significant.

      DISCUSSION

      The strategy used in this study has enabled the first complete structure of IGFBP-5 modifications to be determined. Vital to the strategy was the analysis of both tryptic peptides and the full-length protein, which provided intersecting and complementary data. We demonstrated that IGFBP-5 is secreted as a phosphoprotein and O-linked glycoprotein and that there are two main in vivo phosphorylation sites in human IGFBP-5 at Ser96 and Ser248. Unexpectedly we also demonstrated one in vivo phosphorylation site, Ser106, in IGFBP-2. IGFBP-5 glycosylation on Thr152 was confirmed, and the carbohydrate composition was shown to be a mixture of sialylated and unsialylated glycans.
      We determined that the majority of secreted IGFBP-5 was phosphorylated (56%). Using 32P labeling it was evident that most of the phosphorylation was detected at Ser96, and it had the highest turnover of phosphate. Also diphosphorylation was uncommon (4%).
      Tryptic peptides provided information on the type of glycans at a specific site. Analysis of the full-length protein confirmed the distribution of the types of glycans and also ruled out any significant amount of alternative post-translational modifications (e.g. acetylation or myristoylation). However, we clearly observed a variability in the site of cleavage of the signal peptide with IGFBP-5 from T47D cells with Leu1 (Leu21 of the IGFBP-5 precursor) being cleaved more often than not. Apart from this variability, there was generally little difference between adenoviral vector-derived IGFBP-5 and IGFBP-5 purified from T47D cells. The similar result for adenoviral vector-derived IGFBP-5 corroborates that we have correctly determined the in vivo phosphorylation and glycosylation state of IGFBP-5.
      It should be noted that comparisons of mass spectral signal intensity are only valid when the molecules are alike. Small phosphopeptides are detected with different efficiency compared with the same peptides that lack a phosphate group (
      • Craig A.G.
      • Hoeger C.A.
      • Miller C.L.
      • Goedken T.
      • Rivier J.E.
      • Fischer W.H.
      Monitoring protein kinase and phosphatase reactions with matrix-assisted laser desorption/ionization mass spectrometry and capillary zone electrophoresis: comparison of the detection efficiency of peptide-phosphopeptide mixtures.
      ). However, in the case of large phosphopeptides (>8,000 Da), the phosphate group makes no detectable difference (
      • Graham M.E.
      • Dickson P.W.
      • Dunkley P.R.
      • von Nagy-Felsobuki E.I.
      Determination of phosphorylation levels of tyrosine hydroxylase by electrospray mass spectrometry.
      ). This is in agreement with observations that protein phosphorylation often has a much smaller than predicted effect on the pI of the protein (
      • Zhu K.
      • Zhao J.
      • Lubman D.M.
      • Miller F.R.
      • Barder T.J.
      Protein pI shifts due to posttranslational modifications in the separation and characterization of proteins.
      ) and thus its ability to be protonated (or the effect is smaller than the error in the measurement). No comparable studies have been made on the effect of glycosylation on detection efficiency, and therefore it must be noted that the mass spectral analysis of the intact IGFBP-5 may not be strictly quantitative. However, the amount of phosphorylation and glycosylation determined by ESI-MS was clearly in agreement with observed gel staining and 32P distribution (c.f. Fig. 6 and Table II with Fig. 2, a and b).
      The major phosphorylation site at Ser96 of IGFBP-5 is at the beginning of the non-conserved central domain. Ser96 on IGFBP-5 shares the same proximity to the N-terminal domain as Ser101 on IGFBP-1 (
      • Jones J.I.
      • Busby W.J.
      • Wright G.
      • Smith C.E.
      • Kimack N.M.
      • Clemmons D.R.
      Identification of the sites of phosphorylation in insulin-like growth factor binding protein-1. Regulation of its affinity by phosphorylation of serine 101.
      ), Ser111 and Ser113 on IGFBP-3 (
      • Hoeck W.G.
      • Mukku V.R.
      Identification of the major sites of phosphorylation in IGF binding protein-3.
      ), and Ser106 on IGFBP-2 (determined in this study). This proximity indicates a possible role in regulating the folding between the N-terminal domain and the central domain or even regulating IGF binding affinity as shown for IGFBP-1 (
      • Jones J.I.
      • Busby W.J.
      • Wright G.
      • Smith C.E.
      • Kimack N.M.
      • Clemmons D.R.
      Identification of the sites of phosphorylation in insulin-like growth factor binding protein-1. Regulation of its affinity by phosphorylation of serine 101.
      ). However, this has been shown not to be the case for IGF-I binding to IGFBP-3 (
      • Hoeck W.G.
      • Mukku V.R.
      Identification of the major sites of phosphorylation in IGF binding protein-3.
      ), and we similarly saw no effect of IGFBP-5 dephosphorylation on IGF-I or IGF-II binding. Alternatively the phosphorylation may regulate binding to the short heparin-binding motif at residues 133–136 (Fig. 2c), although the glycosylation at Thr152 is much nearer to this motif. We found that both deglycosylation and dephosphorylation appeared to enhance the interaction between IGFBP-5 and heparin.
      The minor phosphorylation site at Ser248 is five residues from the C terminus (Fig. 2c). It is therefore in a flexible position to simultaneously interact with any protein folds or binding partners near the C terminus. IGFBP-5 has many potential binding partners outside and within the cell, including fibronectin (
      • Xu Q.
      • Yan B.
      • Li S.
      • Duan C.
      Fibronectin binds insulin-like growth factor-binding protein 5 and abolishes its ligand-dependent action on cell migration.
      ), retinoid X receptor-α (
      • Schedlich L.J.
      • O’Han M.K.
      • Leong G.M.
      • Baxter R.C.
      Insulin-like growth factor binding protein-3 prevents retinoid receptor heterodimerization: implications for retinoic acid-sensitivity in human breast cancer cells.
      ), Ras-association domain family 1 protein (
      • Amaar Y.G.
      • Baylink D.J.
      • Mohan S.
      Ras-association domain family 1 protein, RASSF1C, is an IGFBP-5 binding partner and a potential regulator of osteoblast cell proliferation.
      ), four and a half LIM protein 2 (
      • Amaar Y.G.
      • Thompson G.R.
      • Linkhart T.A.
      • Chen S.T.
      • Baylink D.J.
      • Mohan S.
      Insulin-like growth factor-binding protein 5 (IGFBP-5) interacts with a four and a half LIM protein 2 (FHL2).
      ), and others (
      • Firth S.M.
      • Baxter R.C.
      Cellular actions of the insulin-like growth factor binding proteins.
      ). A flexible C-terminal domain that can be phosphorylated may provide a regulatable docking site for these or other proteins. A C-terminal crystal structure of IGFBP-1 (
      • Sala A.
      • Capaldi S.
      • Campagnoli M.
      • Faggion B.
      • Labò S.
      • Perduca M.
      • Romano A.
      • Carrizo M.E.
      • Valli M.
      • Visai L.
      • Minchiotti L.
      • Galliano M.
      • Monaco H.L.
      Structure and properties of the C-terminal domain of insulin-like growth factor-binding protein-1 isolated from human amniotic fluid.
      ) is available for comparison with IGFBP-5, but the final residues after the last conserved cysteine are not clearly defined. In addition to the ligand binding comparisons performed in this study, mutagenesis and other phosphosite-specific studies will be required to advance the understanding of the roles of the two phosphorylation sites determined in this work.
      Because both IGFBP-1 and IGFBP-3 are phosphorylated at central domain residues close to the cysteine-rich N-terminal domain, and the kinases involved were considered likely to be CK1, CK2, or a related enzyme (
      • Hoeck W.G.
      • Mukku V.R.
      Identification of the major sites of phosphorylation in IGF binding protein-3.
      ,
      • Ankrapp D.P.
      • Jones J.I.
      • Clemmons D.R.
      Characterization of insulin-like growth factor binding protein-1 kinases from human hepatoma cells.
      ), we also considered whether CK2 might phosphorylate IGFBP-5. Ser96 was predicted as a potential CK2 phosphorylation site by the kinase specificity prediction program NetPhosK 1.0, and we observed that DRB, a specific inhibitor of CK2, reduced 32P incorporation into IGFBP-5 in cellular metabolic labeling experiments. Despite this, CK2 phosphorylated IGFBP-5 at Thr103 and Ser249 but failed to phosphorylate at Ser96 and Ser248 and is therefore not likely to be an in vivo protein kinase for IGFBP-5. However, there was some evidence that there may be more in vivo phosphorylation sites of very low abundance within the region 103–115. It remains to be determined whether these very low stoichiometry sites can have a significant effect on IGFBP-5 function. Ser96 and Ser248 share a slightly acidic context, DSXXXE, and may be phosphorylated by the same in vivo kinase, which also remains to be determined.
      Variable glycosylation of IGFBP-5 may differentially affect its susceptibility to different proteases. IGFBP-5 has been reported to be cleaved C-terminally to residues 138 (
      • Imai Y.
      • Busby Jr., W.H.
      • Smith C.E.
      • Clarke J.B.
      • Garmong A.J.
      • Horwitz G.D.
      • Rees C.
      • Clemmons D.R.
      Protease-resistant form of insulin-like growth factor-binding protein 5 is an inhibitor of insulin-like growth factor-I actions on porcine smooth muscle cells in culture.
      ), 143 (
      • Overgaard M.T.
      • Boldt H.B.
      • Laursen L.S.
      • Sottrup-Jensen L.
      • Conover C.A.
      • Oxvig C.
      Pregnancy-associated plasma protein-A2 (PAPP-A2), a novel insulin-like growth factor-binding protein-5 proteinase.
      ), 188 (
      • Ständker L.
      • Wobst P.
      • Mark S.
      • Forssmann W.G.
      Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
      ), and 169 (
      • Andress D.L.
      • Loop S.M.
      • Zapf J.
      • Kiefer M.C.
      Carboxy-truncated insulin-like growth factor binding protein-5 stimulates mitogenesis in osteoblast-like cells.
      ,
      • Campbell P.G.
      • Andress D.L.
      Plasmin degradation of insulin-like growth factor-binding protein-5 (IGFBP-5): regulation by IGFBP-5-(201–218).
      ). This implies that, in various cell types, either N-terminal or C-terminal IGFBP-5 fragments could contain O-glycosylation at Thr152. The detection of both core 1 and core 2 type glycans at Thr152 indicates that there may be different pools of IGFBP-5 that have different functional glycosylation. This could occur if IGFBP-5 was exposed to different amounts of specific glycosyltransferases. For example, in normal breast, mucin O-glycans are largely core 2 type structures, whereas in breast carcinomas they are often smaller core 1 type structures (
      • Burchell J.M.
      • Mungul A.
      • Taylor-Papadimitriou J.
      O-Linked glycosylation in the mammary gland: changes that occur during malignancy.
      ). Differential glycosylation may affect cell association as reported for IGFBP-3 (
      • Firth S.M.
      • Baxter R.C.
      Characterisation of recombinant glycosylation variants of insulin-like growth factor binding protein-3.
      ) and IGFBP-6 (
      • Marinaro J.A.
      • Neumann G.M.
      • Russo V.C.
      • Leeding K.S.
      • Bach L.A.
      O-Glycosylation of insulin-like growth factor (IGF) binding protein-6 maintains high IGF-II binding affinity by decreasing binding to glycosaminoglycans and susceptibility to proteolysis.
      ).
      Variants of IGFBP-5 lacking either covalently linked phosphates or carbohydrates were prepared by treatment with Antarctic phosphatase or neuraminidase and O-glycosidase, respectively. Consistent with original data published by Conover and Kiefer (
      • Conover C.A.
      • Kiefer M.C.
      Regulation and biological effect of endogenous insulin-like growth factor binding protein-5 in human osteoblastic cells.
      ), deglycosylation of adenoviral vector-derived IGFBP-5 resulted in a single band resolved after SDS-PAGE. Dephosphorylation of IGFBP-5 was determined to be complete by ESI-MS and produced more tightly resolved isoforms after SDS-PAGE and immunoblotting. Enzymatic treatments had no effect on immunorecognition and therefore allowed accurate quantitation of purified preparations by RIA. The ability of IGFBP-5 to interact with glycosaminoglycans, including heparin sulfate, has been documented previously (
      • Arai T.
      • Parker A.
      • Busby Jr., W.
      • Clemmons D.R.
      Heparin, heparan sulfate, and dermatan sulfate regulate formation of the insulin-like growth factor-I and insulin-like growth factor-binding protein complexes.
      ). The enhanced retention of both deglycosylated and dephosphorylated IGFBP-5 preparations by heparin-agarose is consistent with an inhibitory effect of the negative charge carried by phosphates and carbohydrates, especially sialic acids, on the central or C-terminal domain heparin-binding motifs. This modulation of heparin binding affinity may similarly affect IGFBP-5 interactions with other glycosaminoglycan and other extracellular matrix components (
      • Parker A.
      • Rees C.
      • Clarke J.
      • Busby Jr., W.H.
      • Clemmons D.R.
      Binding of insulin-like growth factor (IGF)-binding protein-5 to smooth-muscle cell extracellular matrix is a major determinant of the cellular response to IGF-I.
      ). Functional studies provided no evidence that IGFBP-5 phosphorylation and O-glycosylation can modulate IGF bioavailability at the levels of the binary complex with IGF-I or IGF-II or the ternary complex with ALS in the presence of either IGF-I or IGF-II. However, the dephosphorylated preparation did show a trend to lower ALS binding in the presence of either IGF-I or IGF-II; this is in contrast to the observed effect of phosphorylation on the binding of IGFBP-3 to ALS (
      • Coverley J.A.
      • Martin J.L.
      • Baxter R.C.
      The effect of phosphorylation by casein kinase 2 on the activity of insulin-like growth factor binding protein.
      ). Whether phosphorylation affects the molecular distribution of IGFBP-5 between binary and ternary complexes in the circulation, and perhaps influences the size redistribution observed in pregnancy serum (
      • Baxter R.C.
      • Meka S.
      • Firth S.M.
      Molecular distribution of IGF binding protein-5 in human serum.
      ), remains to be determined.
      In conclusion, we demonstrated a major phosphorylation site in human IGFBP-5 at Ser96 and a minor site at Ser248. Although CK2 inhibition in T47D breast carcinoma cell culture inhibited IGFBP-5 phosphorylation, exposure of IGFBP-5 to CK2 in vitro led to phosphorylation at Thr103 (major site) and Ser249 (minor site), suggesting that CK2 is unlikely to be an important protein kinase for IGFBP-5 at least in these breast cancer cells. Whereas the full functional significance of the complex pattern of post-translational modification of IGFBP-5 and other IGFBPs remains to be established, this study has provided the first detailed demonstration of IGFBP-5 post-translational modifications and a firm biochemical basis on which to further investigate the biological consequences of phosphorylation and glycosylation.

      Acknowledgments

      The assistance of Xiao-Lang Yan in the isolation and purification of IGFBP-5 preparations is gratefully acknowledged.

      Supplementary Material

      REFERENCES

        • Baxter R.C.
        Insulin-like growth factor (IGF) binding proteins: Interactions with IGFs and intrinsic bioactivities.
        Am. J. Physiol. 2000; 278: E967-E976
        • Firth S.M.
        • Baxter R.C.
        Cellular actions of the insulin-like growth factor binding proteins.
        Endocr. Rev. 2002; 23: 824-854
        • Marshman E.
        • Green K.A.
        • Flint D.J.
        • White A.
        • Streuli C.H.
        • Westwood M.
        Insulin-like growth factor binding protein 5 and apoptosis in mammary epithelial cells.
        J. Cell Sci. 2003; 116: 675-682
        • Andress D.L.
        • Birnbaum R.S.
        Human osteoblast-derived insulin-like growth factor (IGF) binding protein-5 stimulates osteoblast mitogenesis and potentiates IGF action.
        J. Biol. Chem. 1992; 267: 22467-22472
        • Miyakoshi N.
        • Richman C.
        • Kasukawa Y.
        • Linkhart T.A.
        • Baylink D.J.
        • Mohan S.
        Evidence that IGF-binding protein-5 functions as a growth factor.
        J. Clin. Investig. 2001; 107: 73-81
        • Coverley J.A.
        • Martin J.L.
        • Baxter R.C.
        The effect of phosphorylation by casein kinase 2 on the activity of insulin-like growth factor binding protein.
        Endocrinology. 2000; 141: 564-570
        • Schedlich L.J.
        • Nilsen T.
        • John A.P.
        • Jans D.A.
        • Baxter R.C.
        Phosphorylation of insulin-like growth factor binding protein-3 by deoxyribonucleic acid-dependent protein kinase reduces ligand binding and enhances nuclear accumulation.
        Endocrinology. 2003; 144: 1984-1993
        • Cobb L.J.
        • Liu B.
        • Lee K.W.
        • Cohen P.
        Phosphorylation by DNA-dependent protein kinase is critical for apoptosis induction by insulin-like growth factor binding protein-3.
        Cancer Res. 2006; 66: 10878-10884
        • Jones J.I.
        • Clemmons D.R.
        Insulin-like growth factors and their binding proteins: biological actions.
        Endocr. Rev. 1995; 16: 3-34
        • Firth S.M.
        • Baxter R.C.
        The role of glycosylation in the action of IGFBP-3.
        Prog. Growth Factor Res. 1995; 6: 223-229
        • Chelius D.
        • Wu S.L.
        • Bondarenko P.V.
        Identification of N-linked oligosaccharides of rat insulin-like growth factor binding protein-4.
        Growth Horm. IGF Res. 2002; 12: 169-177
        • Ständker L.
        • Wobst P.
        • Mark S.
        • Forssmann W.G.
        Isolation and characterization of circulating 13-kDa C-terminal fragments of human insulin-like growth factor binding protein-5.
        FEBS Lett. 1998; 441: 281-286
        • Marinaro J.A.
        • Neumann G.M.
        • Russo V.C.
        • Leeding K.S.
        • Bach L.A.
        O-Glycosylation of insulin-like growth factor (IGF) binding protein-6 maintains high IGF-II binding affinity by decreasing binding to glycosaminoglycans and susceptibility to proteolysis.
        Eur. J. Biochem. 2000; 267: 5378-5386
        • Bohn H.
        • Kraus W.
        Isolation and characterization of a new placenta specific protein (PP12).
        Arch. Gynecol. 1980; 229: 279-291
        • Firth S.M.
        • Baxter R.C.
        Characterisation of recombinant glycosylation variants of insulin-like growth factor binding protein-3.
        J. Endocrinol. 1999; 160: 379-387
        • Kalus W.
        • Zweckstetter M.
        • Renner C.
        • Sanchez Y.
        • Georgescu J.
        • Grol M.
        • Demuth D.
        • Schumacher R.
        • Dony C.
        • Lang K.
        • Holak T.A.
        Structure of the IGF-binding domain of the insulin-like growth factor-binding protein-5 (IGFBP-5): implications for IGF and IGF-I receptor interactions.
        EMBO J. 1998; 17: 6558-6572
        • Pattison S.T.
        • Fanayan S.
        • Martin J.L.
        Insulin-like growth factor binding protein-3 is secreted as a phosphoprotein by human breast cancer cells.
        Mol. Cell. Endocrinol. 1999; 156: 131-139
        • Firth S.M.
        • Clemmons D.R.
        • Baxter R.C.
        Mutagenesis of basic amino acids in the carboxyl-terminal region of insulin-like growth factor binding protein-5 affects acid-labile subunit binding.
        Endocrinology. 2001; 142: 2147-2150
        • Graham M.E.
        • Ruma-Haynes P.
        • Capes-Davis A.G.
        • Dunn J.M.
        • Tan T.C.
        • Valova V.A.
        • Robinson P.J.
        • Jeffrey P.L.
        Multisite phosphorylation of doublecortin by cyclin-dependent kinase 5.
        Biochem. J. 2004; 381: 471-481
        • Fernandez-Patron C.
        • Castellanos-Serra L.
        • Rodriguez P.
        Reverse staining of sodium dodecyl sulfate polyacrylamide gels by imidazole-zinc salts: sensitive detection of unmodified proteins.
        BioTechniques. 1992; 12: 564-573
        • Larsen M.R.
        • Graham M.E.
        • Robinson P.J.
        • Roepstorff P.
        Improved detection of hydrophilic phosphopeptides using graphite powder microcolumns and mass spectrometry: evidence for in vivo doubly phosphorylated dynamin I and dynamin III.
        Mol. Cell. Proteomics. 2004; 3: 456-465
        • Craig A.G.
        • Hoeger C.A.
        • Miller C.L.
        • Goedken T.
        • Rivier J.E.
        • Fischer W.H.
        Monitoring protein kinase and phosphatase reactions with matrix-assisted laser desorption/ionization mass spectrometry and capillary zone electrophoresis: comparison of the detection efficiency of peptide-phosphopeptide mixtures.
        Biol. Mass Spectrom. 1994; 23: 519-528
        • Graham M.E.
        • Dickson P.W.
        • Dunkley P.R.
        • von Nagy-Felsobuki E.I.
        Determination of phosphorylation levels of tyrosine hydroxylase by electrospray mass spectrometry.
        Anal. Biochem. 2000; 281: 98-104
        • Zhu K.
        • Zhao J.
        • Lubman D.M.
        • Miller F.R.
        • Barder T.J.
        Protein pI shifts due to posttranslational modifications in the separation and characterization of proteins.
        Anal. Chem. 2005; 77: 2745-2755
        • Jones J.I.
        • Busby W.J.
        • Wright G.
        • Smith C.E.
        • Kimack N.M.
        • Clemmons D.R.
        Identification of the sites of phosphorylation in insulin-like growth factor binding protein-1. Regulation of its affinity by phosphorylation of serine 101.
        J. Biol. Chem. 1993; 268: 1125-1131
        • Hoeck W.G.
        • Mukku V.R.
        Identification of the major sites of phosphorylation in IGF binding protein-3.
        J. Cell. Biochem. 1994; 56: 262-273
        • Xu Q.
        • Yan B.
        • Li S.
        • Duan C.
        Fibronectin binds insulin-like growth factor-binding protein 5 and abolishes its ligand-dependent action on cell migration.
        J. Biol. Chem. 2004; 279: 4269-4277
        • Schedlich L.J.
        • O’Han M.K.
        • Leong G.M.
        • Baxter R.C.
        Insulin-like growth factor binding protein-3 prevents retinoid receptor heterodimerization: implications for retinoic acid-sensitivity in human breast cancer cells.
        Biochem. Biophys. Res. Commun. 2004; 314: 83-88
        • Amaar Y.G.
        • Baylink D.J.
        • Mohan S.
        Ras-association domain family 1 protein, RASSF1C, is an IGFBP-5 binding partner and a potential regulator of osteoblast cell proliferation.
        J. Bone Miner. Res. 2005; 20: 1430-1439
        • Amaar Y.G.
        • Thompson G.R.
        • Linkhart T.A.
        • Chen S.T.
        • Baylink D.J.
        • Mohan S.
        Insulin-like growth factor-binding protein 5 (IGFBP-5) interacts with a four and a half LIM protein 2 (FHL2).
        J. Biol. Chem. 2002; 277: 12053-12060
        • Ankrapp D.P.
        • Jones J.I.
        • Clemmons D.R.
        Characterization of insulin-like growth factor binding protein-1 kinases from human hepatoma cells.
        J. Cell. Biochem. 1996; 60: 387-399
        • Imai Y.
        • Busby Jr., W.H.
        • Smith C.E.
        • Clarke J.B.
        • Garmong A.J.
        • Horwitz G.D.
        • Rees C.
        • Clemmons D.R.
        Protease-resistant form of insulin-like growth factor-binding protein 5 is an inhibitor of insulin-like growth factor-I actions on porcine smooth muscle cells in culture.
        J. Clin. Investig. 1997; 100: 2596-2605
        • Overgaard M.T.
        • Boldt H.B.
        • Laursen L.S.
        • Sottrup-Jensen L.
        • Conover C.A.
        • Oxvig C.
        Pregnancy-associated plasma protein-A2 (PAPP-A2), a novel insulin-like growth factor-binding protein-5 proteinase.
        J. Biol. Chem. 2001; 276: 21849-21853
        • Andress D.L.
        • Loop S.M.
        • Zapf J.
        • Kiefer M.C.
        Carboxy-truncated insulin-like growth factor binding protein-5 stimulates mitogenesis in osteoblast-like cells.
        Biochem. Biophys. Res. Commun. 1993; 195: 25-30
        • Campbell P.G.
        • Andress D.L.
        Plasmin degradation of insulin-like growth factor-binding protein-5 (IGFBP-5): regulation by IGFBP-5-(201–218).
        Am. J. Physiol. 1997; 273: E996-E1004
        • Burchell J.M.
        • Mungul A.
        • Taylor-Papadimitriou J.
        O-Linked glycosylation in the mammary gland: changes that occur during malignancy.
        J. Mammary Gland Biol. Neoplasia. 2001; 6: 355-364
        • Conover C.A.
        • Kiefer M.C.
        Regulation and biological effect of endogenous insulin-like growth factor binding protein-5 in human osteoblastic cells.
        J. Clin. Endocrinol. Metab. 1993; 76: 1153-1159
        • Arai T.
        • Parker A.
        • Busby Jr., W.
        • Clemmons D.R.
        Heparin, heparan sulfate, and dermatan sulfate regulate formation of the insulin-like growth factor-I and insulin-like growth factor-binding protein complexes.
        J. Biol. Chem. 1994; 269: 20388-20393
        • Parker A.
        • Rees C.
        • Clarke J.
        • Busby Jr., W.H.
        • Clemmons D.R.
        Binding of insulin-like growth factor (IGF)-binding protein-5 to smooth-muscle cell extracellular matrix is a major determinant of the cellular response to IGF-I.
        Mol. Biol. Cell. 1998; 9: 2383-2392
        • Baxter R.C.
        • Meka S.
        • Firth S.M.
        Molecular distribution of IGF binding protein-5 in human serum.
        J. Clin. Endocrinol. Metab. 2002; 87: 271-276
        • Sala A.
        • Capaldi S.
        • Campagnoli M.
        • Faggion B.
        • Labò S.
        • Perduca M.
        • Romano A.
        • Carrizo M.E.
        • Valli M.
        • Visai L.
        • Minchiotti L.
        • Galliano M.
        • Monaco H.L.
        Structure and properties of the C-terminal domain of insulin-like growth factor-binding protein-1 isolated from human amniotic fluid.
        J. Biol. Chem. 2005; 280: 29812-29819