Alanine Screening Mutagenesis Establishes Tyrosine 60 of Bovine Insulin-like Growth Factor Binding Protein-2 as a Determinant of Insulin-like Growth Factor Binding*

The determinants of insulin-like growth factor (IGF) binding to its binding proteins (IGFBPs) are poorly characterized in terms of important residues in the IGFBP molecule. We have previously used tyrosine iodination to implicate Tyr-60 in the IGF-binding site of bovine IGFBP-2 (Hobba, G. D., Forbes, B. E., Parkinson, E. J., Francis, G. L., and Wallace, J. C. (1996) J. Biol. Chem. 271, 30529–30536). In this report, we show that the mutagenic replacement of Tyr-60 with either Ala or Phe reduced the affinity of bIGFBP-2 for IGF-I (4.0- and 8.4-fold, respectively) and for IGF-II (3.5- and 4.0-fold, respectively). Although adjacent residues Val-59, Thr-61, Pro-62, and Arg-63 are well conserved in IGFBP family members, Ala substitution for these residues did not reduce the IGF affinity of bIGFBP-2. Kinetic analysis of the bIGFBP-2 mutants on IGF biosensor chips in the BIAcore instrument revealed that Tyr-60 → Phe bIGFBP-2 bound to the IGF-I surface 3.0-fold more slowly than bIGFBP-2 and was released 2.6-fold more rapidly than bIGFBP-2. We therefore propose that the hydroxyl group of Tyr-60 participates in a hydrogen bond that is important for the initial complex formation with IGF-I and the stabilization of this complex. In contrast, Tyr-60 → Ala bIGFBP-2 associated with the IGF-I surface 5.0-fold more rapidly than bIGFBP-2 but exhibited an 18.4-fold more rapid release from this surface compared with bIGFBP-2. Thus both the aromatic nature and the hydrogen bonding potential of the tyrosyl side chain of Tyr-60 are important structural determinants of the IGF-binding site of bIGFBP-2.

The insulin-like growth factors (IGF-I 1 and IGF-II) are polypeptides that play a central role in vertebrate growth and development by stimulating cellular proliferation and differentiation (recently reviewed in Ref. 1). The biological activities of the IGFs are mediated mainly through the type 1 IGF receptor which is found on the surface of most cell types (1,2). In turn, the bioavailability of the IGFs is regulated by a family of IGF-specific binding proteins (IGFBPs). The IGFBP family consists of six high affinity IGFBPs (IGFBP-1 to 6) (reviewed in Refs. 3 and 4) and possibly an additional four proteins that can associate with IGFs with lower affinity (5,6). Conserved gene structures and the high degree of sequence identity between the high affinity IGFBPs suggest that these proteins possess three domains and a common IGF binding motif (3,4,7). The sequences of two of these putative domains, the N-and Cterminal cysteine-rich domains, are highly conserved. Where the IGFBP family members differ is in the middle domain and in the possession of phosphorylation and glycosylation sites or sites of association with other biomolecules such as heparin or the integrin receptor (3,4).
At the molecular level, the IGFs have been well characterized. High resolution NMR structures of both IGF-I (8) and IGF-II (9) have been determined, and the overlapping regions that are responsible for IGFBP and receptor interactions have been identified by chemical modification (10,11), epitope mapping (12), and by mutagenesis (13)(14)(15)(16)(17)(18). However, insight into the overall structure of the IGFBPs has been restricted to multiple sequence alignment and secondary structure prediction (19). Furthermore, the growing number of IGFBP mutagenic studies described to date has focused on aspects of IGFBP biology such as heparin binding (20), integrin receptor binding (21), extracellular matrix binding (22), specific proteolysis (23,24) or phosphorylation (25) rather than the systematic identification and characterization of a common IGF binding motif.
In terms of the IGF-binding site, both the N-and C-terminal cysteine-rich domains of IGFBPs are believed to participate. This is suggested by the observations that N-terminal cysteinerich domains of IGFBP-1 (26), IGFBP-3 (27), IGFBP-4 (24), and IGFBP-5 (28) and C-terminal cysteine-rich domains of IG-FBP-2 (29,30) and IGFBP-3 (31) all possess residual IGF binding affinity. Yet, the specific residues of IGFBPs that are directly involved in IGF binding have not been identified.
The rationale for this study is based on our observation that Tyr-60 was protected from iodination in the IGF⅐bIGFBP-2 complex (32). Furthermore, when Tyr-60 was iodinated, it caused a reduction in the binding affinity of bIGFBP-2 for the IGF ligand. However, the residues Val-59, Thr-61, Pro-62, and Arg-63 could also conceivably play a role in IGF binding that is disrupted when Tyr-60 is iodinated. These latter residues are highly conserved across the whole IGFBP family, whereas all of the described IGFBP-1 sequences from various species possess an alanyl rather than tyrosyl residue at the position corresponding to Tyr-60 in bIGFBP-2, as shown in Fig. 1. Therefore, in order to determine which residues in the Tyr-60 region of bIGFBP-2 do influence IGF binding, alanine-scanning mutagenesis has been performed across residues 59 and 63 inclusive. Tyr-60 has also been substituted with Phe to distinguish between the hydrogen bonding properties and the hydrophobic and aromatic properties of the tyrosyl side chain with respect to IGF binding.

EXPERIMENTAL PROCEDURES
Materials-Recombinant bIGFBP-2 was transiently expressed in the COS-1 (ATCC:CRL 1650) monkey kidney cell line and purified from medium conditioned by the transfected cells as described previously (33). Receptor grade IGF-I and IGF-II were the kind gifts of GroPep Pty. Ltd. (Adelaide, Australia). Radiolabeled 125 I-IGF-I and 125 I-IGF-II peptides were prepared to a specific activity of approximately 3 kBq/mol as described previously (34). Carrier-free Na 125 I was purchased from Amersham Pharmacia Biotech (Sydney, Australia). Reverse-phase HPLC columns were purchased from Brownlee Lab (Santa Clara, CA) and Amrad (Sydney, Australia). Pre-siliconized tubes (Sorenson BioScience, Inc., Salt Lake City, UT) were used for the collection of fractions during chromatography. All HPLC was carried out using Waters 510 solvent pumps, a Waters 490 4-channel absorbance detector (Millipore-Waters, Lane Cove, New South Wales), and a Perkin-Elmer LS4 fluorescence spectrometer (Scoresby, Victoria, Australia). The Waters Maxima software package was used to control solvent gradients and for data collection. HPLC-grade acetonitrile was purchased from Merck (Kilsyth, Victoria, Australia) and trifluoroacetic acid from Sigma-Aldrich (Castle Hill, New South Wales, Australia). All other reagents were analytical grade. Nitrocellulose filters were purchased from Schleicher & Schü ll (Dassel, Germany). BIAcore reagents and supplies including CM5 sensor chips, HEPES-buffered saline (HBS), amine coupling reagents, N-ethyl-NЈ-[(dimethylamino)propyl]carbodiimide, N-hydroxysuccinimide, and ethanolamine were kindly provided by Pharmacia & Upjohn AB, Preclinical Research, Stockholm, Sweden, or were purchased from Amrad, Melbourne, Australia.
Mutagenesis-A 344-base pair PstI-SmaI fragment of the bIGFBP-2 pXMT2 based expression vector pGF8 (33) that encompassed the Tyr-60 region was subcloned into the multiple cloning site of the phagemid vector Bluescript pBKS(Ϫ) (Stratagene). Mutagenesis was carried out by the Kunkel method (35) with the MutaGene TM Phagemid In Vitro Mutagenesis Version 2 kit (Bio-Rad, Regents Park, New South Wales, Australia). Six mutagenic oligonucleotides were synthesized by Bresatec Ltd., Thebarton, South Australia, Australia. A summary of the mutagenic oligonucleotide sequences and the resulting amino acid sequence changes in bIGFBP-2 are shown in Fig. 2. For screening purposes, the mismatch oligonucleotides also disrupted the 5Ј-GTAC-3Ј RsaI endonuclease site between the codons GTG and TAC of Val-59 and Tyr-60. Therefore, positive clones were identified by diagnostic RsaI digestion followed by polymerase chain reaction DNA sequencing (36) using the sequencing primers 5Ј-GTTTTCCCAGTCACGAC-3Ј and 5Ј-CACACAGGAAACAGCTATGACCATG-3Ј which were complementary to flanking sequences at the insertion site of pBKS(Ϫ). Finally, pGF8 variants harboring the Tyr-60 region mutants were regenerated by subcloning the mutagenized PstI-SmaI fragments into the parent vector pGF8. The expression vector integrity and correct base changes were confirmed by RsaI restriction analysis and dye terminator DNA sequencing (PRISM, Applied Biosystems, Victoria, Australia) using the primer 5Ј-CTCGCCGTTGTCTGCAACCTGCTCCGGG-3Ј. bIGFBP-2 mutants were expressed in COS-1 cells and purified as described previously (33).
Analysis of the Mutant bIGFBP-2 Analogs-Lyophilized samples of approximately 10 g of each bIGFBP-2 mutant were submitted for electrospray mass spectrometry. The analysis was carried out on a Perkin-Elmer SCI-EX API-300 triple quadrupole mass spectrometer at the Australian Research Council electrospray mass spectrometry unit, Adelaide. Peptide concentrations were accurately quantified by both reverse-phase HPLC as described (32) and by amino acid analysis on an AminoQuant II/M High Sensitivity Instrument from Hewlett-Packard, Waldbronn, Germany (37), using the orthophthalaldehyde 9-fluorenylmethyl chloroformate two-stage derivatization procedure.
SDS-PAGE and Western Ligand Blot-bIGFBP-2 samples (200 ng/ lane) were electrophoresed on discontinuous 12.5% SDS-polyacrylamide gels under nonreducing conditions (38). The samples were either stained with silver (39) and quantified by densitometry (Molecular Dynamics) or transferred onto nitrocellulose filters for Western ligand blotting (40). The level of 125 I-IGF-II binding to the bIGFBP samples on the filter was visualized and quantified on a PhosphorImager (Molecular Dynamics).
Circular Dichroism-CD spectra were recorded using a Jasco J720 spectropolarimeter equipped with a PTC-348W1 Peltier Type Temperature Controller set to 20°C. bIGFBP-2 samples were adjusted to a final concentration of 0.25 mg/ml with 10 mM sodium phosphate, 60 mM NaCl, pH 7.4, and placed in a quartz cuvette with a path length of 1 mm. Spectra were recorded from 250 to 190 nm with a step resolution of 0.2 nm and a scanning speed of 20 nm/min. The response time was set to 1 s and the bandwidth was 0.5 nm. Each spectrum is the average of five accumulated scans.
Soluble IGFBP Assay-The relative IGF binding affinities of the bIGFBP-2 mutants were determined at equilibrium by charcoal binding assay, essentially as described previously (41) with the following modifications. The assay buffer pH was 7.4 and the IGFBP/IGF-tracer incubation period was 24 h at room temperature. Each bIGFBP-2 concentration was assayed in triplicate, and the total analysis was performed twice. In assays containing 125 I-IGF-I, approximately 7,000 cpm/tube were used, and the nonspecific binding was 6% of the total radioactivity added. Similarly, 10,000 cpm/tube of 125 I-IGF-II were used with a nonspecific binding of 8% of the total radioactivity added. The experimental data were fitted to a sigmoidal dose-response model with variable slope using GraphPad Prism (GraphPad Inc., San Diego).
BIAcore-All BIAcore analyses were carried out with IGF as the immobilized ligand. Covalent attachment of either IGF-I or IGF-II to the CM5 biosensor chip was achieved by the amine coupling method (42). Briefly, IGF (12.5 g/ml in 50 mM sodium acetate, pH 4.7) was injected onto the activated CM5 surface at 5 l/min with HBS (10 mM HEPES, 150 mM NaCl, 3.4 mM EDTA, pH 7.4) as the running buffer. Residual binding sites were quenched with ethanolamine. The multichannel capability of the BIAcore 2000 was used to generate channels with either IGF-I or IGF-II surfaces (prepared to final resonance values of approximately 60 to 120 RU above the resonance value of the activated but underivatized chip) as well as a reference surface to which no IGF was bound. In a kinetic study, bIGFBP-2 and bIGFBP-2 mutants (5-100 nM in HBS, n ϭ 6 per bIGFBP-2 species) were injected for 5 min at a flow rate of 40 l/min with HBS as the running buffer. The dissociation phase, initiated by switching from the stream of the bIG-FBP-2 sample to HBS, was carried out over a period of 10 min. The IGF surfaces were regenerated by a 90-s injection of 0.1 M HCl. Due to the large number of bIGFBP-2 samples and the extended length of the experiment, samples were injected in random order, and the experiment was carried out twice on two different chips rather than with duplicate samples on a single chip. The apparent analyte association and dissociation rates were derived by fitting the experimental data to either a one-site (IGF-I) or two-site (IGF-II) association model and a two-site dissociation model with the BIAevaluation software (version 2.1) supplied with the instrument.

RESULTS
Protein Characterization-The Tyr-60 region bIGFBP-2 mutant constructs were sequenced in the expression plasmid pGF8, and the expected DNA sequences were obtained. Transient expression of bIGFBP-2 mutants in transfected COS-1 cells yielded approximately 400 g of each purified protein from 0.5 liters of conditioned medium. All of the bIGFBP-2 mutants migrated as single bands of the same size as wild-type bIGFBP-2 when run on nonreducing SDS-PAGE (Fig. 3a). A minor band corresponding to bIGFBP-2 dimer was also evident in the SDS-PAGE analysis (Fig. 3, a and b). When analyzed by electrospray mass spectrometry, the observed mass of each mutant with the exception of Tyr-60 3 Ala corresponded with the sequence-predicted mass to within 3 mass units. Tyr-60 3 Ala bIGFBP-2 possessed a mass that was 18 mass units greater than expected, possibly due to methionine oxidation. Electrospray mass spectrometry also indicated that in each of the bIGFBP-2 and bIGFBP-2 mutant samples there was a small but consistent contamination (approximately 5%) with a species that was 355 mass units greater than the predicted mass. When the N-terminal sequence of the minor contaminant of wild-type bIGFBP-2 was determined, the larger species was identified as a mis-processed form of bIGFBP-2. The N terminus of the contaminant was Gly-Ala-Arg-Ala, corresponding to the last four residues of the leader peptide prior to the normal cleavage site (43) of mature bIGFBP-2.
Circular Dichroism-The secondary structure composition of the bIGFBP-2 mutants were compared with wild-type bIG-FBP-2 using CD spectroscopy. An overlay of the far UV spectra of Val-59 3 Ala, Tyr-60 3 Ala, Tyr-60 3 Phe bIGFBP-2, and wild-type bIGFBP-2 ( Fig. 4) shows that the spectra of these engineered bIGFBP-2 analogs were essentially the same as wild-type bIGFBP-2. Any slight deviation from the CD spectra of bIGFBP-2 could be explained by minor differences in protein concentration. The CD spectra of Thr-61 3 Ala, Pro-62 3 Ala, and Arg-63 3 Ala bIGFBP-2 were also very similar to wild-type bIGFBP-2 (not shown; these spectra were omitted for clarity).
Western Ligand Blots-The 125 I-IGF-II binding by the bIG-FBP-2 mutants was analyzed by Western ligand blot (Fig. 3b). Val-59 3 Ala, Tyr-60 3 Ala, and Tyr-60 3 Phe bIGFBP-2 (Fig.  3b, lanes 2, 3 and 4) showed reduced 125 I-IGF-II binding compared with wild-type bIGFBP-2 (Fig. 3b, lane 1). An estimate of the relative binding of each bIGFBP-2 mutant was provided by direct comparison of the band intensities from the ligand blot and the silver-stained gel. After the bound 125 I-IGF-II radioac-   2. A summary of the mutagenic oligonucleotides used to produce the bIGFBP-2 mutants. Specific bases that have been altered in the synthetic oligonucleotides to yield point mutations or to facilitate clone selection are shown in bold. The RsaI site used for clone selection is shown above the wild-type bIG-FBP-2 sequence. tivity corresponding to each bIGFBP-2 species was quantified by PhosphorImager analysis (Fig. 3b) and corrected for the amount of peptide present on the silver-stained gel (Fig. 3a), it was estimated that Val-59 3 Ala, Tyr-60 3 Ala, and Tyr-60 3 Phe bIGFBP-2 retained approximately 40, 9, and 15% of the 125 I-IGF-II bound by bIGFBP-2 respectively. In contrast, Thr-61 3 Ala, Pro-62 3 Ala, and Arg-63 3 Ala bIGFBP-2 (Fig.  3b, lanes 5-7) all bound 125 I-IGF-II to a similar extent to bIGFBP-2 (Fig. 3b, lane 1).
BIAcore Analysis-Kinetic analyses of the association and dissociation of bIGFBP-2 and the bIGFBP-2 mutants with immobilized IGF-I and IGF-II were carried out in the BIAcore. Fig. 6, a and b, shows a representative subset of the sensorgram data (for qualitative comparison) that was used to generate the kinetic constants summarized in Table II. The interactions between all of the bIGFBP-2 peptides and both IGF-I and IGF-II were difficult to resolve to a single binding site model. In the case of IGF-I interactions, a single apparent association constant and two apparent dissociation constants produced the best fit of the sensorgram data. In contrast, two apparent association and two apparent dissociation constants were necessary for modeling bIGFBP-2 interactions on the IGF-II biosensor surface. The absolute values of the apparent kinetic constants for the IGFBP/IGF interactions (k on , k off , and K D , Table II) varied by up to 25% of the mean value between the two biosensor chips used in this study. However, the ranking of the bIGFBP-2 mutants relative to bIGFBP-2 (i.e. the fold differences in k on , k off , and K D , Table II) were the same on both biosensor chips used in this study.
Mutagenesis in the Tyr-60 region of bIGFBP-2 produced a range of effects on the association rates, dissociation rates, and hence the overall affinity for IGFs that were clearly evident in the BIAcore experiments. On the IGF-I biosensor surface (Fig.  6a, Table II) Val-59 3 Ala, Tyr-60 3 Ala, and Thr-61 3 Ala bIGFBP-2 exhibited association rates (k on ) that were 5.5-, 5.0-, and 2.2-fold more rapid than wild-type bIGFBP-2, respectively. In contrast, the association rates (k on ) of Pro-62 3 Ala and Arg-63 3 Ala bIGFBP-2 were slightly less than bIGFBP-2. The  slowest association rate (k on ) was observed for Tyr-60 3 Phe bIGFBP-2 which bound 3.0-fold more slowly than bIGFBP-2. Although two dissociation components (k off1 and k off2 ) were necessary to model the behavior of bIGFBP-2 and the bIG-FBP-2 mutants on IGF-I biosensor surfaces, the rapid dissociation component (k off1 ), which accounted for approximately 10% of the total interaction, was essentially the same for all of the bIGFBP-2 peptides. In contrast, the slow dissociation component (k off2 ) was affected by mutagenesis in the Tyr-60 region. Therefore, the slow dissociation component (k off2 ), which represented the major proportion of the dissociating population, was considered the more relevant component in this analysis. Three bIGFBP-2 mutants, Thr-61 3 Ala, Pro-62 3 Ala, and Arg-63 3 Ala bIGFBP-2, exhibited apparent dissociation rates (k off2 ) that were similar to bIGFBP-2 (Fig. 6a, Table II). The apparent dissociation rate (k off2 ) of Tyr-60 3 Phe bIGFBP-2 was 2.6-fold more rapid than bIGFBP-2. Large increases in the apparent rate of dissociation (k off2 ) were observed for Val-59 3 Ala and Tyr-60 3 Ala bIGFBP-2 which were released from the IGF-I biosensor surface 9.4 and 18.4 times more rapidly than bIGFBP-2 (Fig. 6a, Table II). The apparent affinities of bIG-FBP-2 and the Tyr-60 region bIGFBP-2 mutants for the IGF-I surface, expressed as the dissociation rate constant (K D ), were derived from the apparent association rate (k on ) and the slow dissociation rate (k off2 ) according to the relationship K D ϭ k off2 /k on . The only bIGFBP-2 mutants that exhibited apparent K D values that were significantly higher than bIGFBP-2 (0.5 nM) were Tyr-60 3 Ala bIGFBP-2 (2.0 nM) and Tyr-60 3 Phe bIGFBP-2 (4.2 nM) corresponding to a 4.0-and 8.4-fold drop in apparent affinity, respectively.
The interactions of bIGFBP-2 and the Tyr-60 region bIG-FBP-2 mutants with the IGF-II biosensor surface are shown in Fig. 6b. Two association components were derived from the sensorgram data, and the relative contribution of each to the total binding profile was estimated using the BIAsimulation program. Between the bIGFBP-2 mutants, the rapid binding component played a varying role in the interaction. Therefore, for each bIGFBP-2 mutant, the two association constants were  a The maximum level of 125 I-IGF binding attained in the assay is expressed as a percentage of total 125 I-IGF added.
b The concentration of bIGFBP-2 peptide required to bind 50% of the maximum 125 I-IGF binding level attained in the assay. R 2 values for the dose/response curve fits were equal to or greater than 0.95. weighted to yield a single apparent association constant k on(app) . In the case of Val-59 3 Ala and Tyr-60 3 Ala bIG-FBP-2, the rapid association component accounted for approximately 30% of the total association interaction with the IGF-II surface. It was estimated that for bIGFBP-2, Pro-62 3 Ala, and Arg-63 3 Ala bIGFBP-2 only 5% of the total IGF-II association interaction was due to the rapid association phase. In the case of Tyr-60 3 Phe and Thr-61 3 Ala bIGFBP-2, a single association component adequately modeled the association interaction to yield a single association constant k on(app) . The trends of the apparent association rates, dissociation rates, and overall affinities of bIGFBP-2 mutants on the IGF-II surface were similar to those observed for the mutants on the IGF-I surface, yet the magnitudes of the changes were generally less marked. Thus Val-59 3 Ala and Tyr-60 3 Ala bIGFBP-2 bound 1.7-and 1.9-fold more rapidly (k on(app) ) to the IGF-II surface than did bIGFBP-2 (Fig. 6b, Table II). Thr-61 3 Ala and Arg-63 3 Ala bIGFBP-2 bound with similar apparent association rates k on(app) to bIGFBP-2. The bIGFBP-2 mutants that exhibited reduced apparent association rates k on(app) were Pro-62 3 Ala and Tyr-60 3 Phe bIGFBP-2 which bound 2-and 5.5-fold more slowly than bIGFBP-2 to the IGF-II surface, respectively. As was observed for the IGF-I surface, the dissociation of bIG-FBP-2 and the bIGFBP-2 mutants from the IGF-II surface occurred in two phases to generate two dissociation constants (k off1 and k off2 ). Again, the rapid dissociation component k off1 was essentially the same for all of the bIGFBP-2 peptides and accounted for approximately 10% of the total dissociation phase of the IGF/IGFBP interaction. Tyr-60 3 Phe, Thr-61 3 Ala, and Arg-63 3 Ala bIGFBP-2 all exhibited dissociation rates (k off2 ) that were similar to the dissociation rate of bIGFBP-2. Pro-62 3 Ala bIGFBP-2 was released 5.0-fold more slowly from the IGF-II surface than bIGFBP-2. On the other hand, Val-59 3 Ala and Tyr-60 3 Ala bIGFBP-2 were released 2.3-and 6.2-fold more rapidly than bIGFBP-2, respectively. In terms of their overall apparent dissociation constants (K D ), calculated by the relationship K D ϭ k off2 /k on(app) , Val-59 3 Ala (0.3 nM), Thr-61 3 Ala (0.3 nM), and Arg-63 3 Ala bIGFBP-2 (0.2 nM) were similar to bIGFBP-2 (0.2 nM) on the IGF-II surface. With an apparent K D of 0.04 nM, Pro-62 3 Ala bIGFBP-2 bound to the IGF-II surface with a greater affinity than bIGFBP-2. In contrast, Tyr-60 3 Ala and Tyr-60 3 Phe bIGFBP-2 bound to the IGF-II surface with K D values of 0.7 and 0.8 nM, respectively, which corresponded to 3.5-and 4.0-fold lower affinities for IGF-II than bIGFBP-2, respectively. DISCUSSION In this study, the role of the Tyr-60 region of bIGFBP-2 in IGF binding has been investigated by alanine screening mutagenesis. In our chemical modification study (32), Tyr-60, which lies in the N-terminal cysteine-rich domain of bIGFBP-2, was shown to be protected from iodination when either IGF-I or IGF-II was bound. The high degree of sequence homology in the Tyr-60 region of the whole IGFBP family (Fig. 1) suggested that other residues in this vicinity might also play a role in IGF binding. To investigate this possibility, Val-59 3 Ala, Tyr-60 3 Ala, Tyr-60 3 Phe, Thr-61 3 Ala, Pro-62 3 Ala and Arg-63 3 Ala bIGFBP-2 were recombinantly expressed, purified, and characterized for changes in IGF affinity. The only bIGFBP-2 mutants that were observed to exhibit significantly reduced affinity for IGF-I and IGF-II using a variety of complementary analyses were Tyr-60 3 Ala and Tyr-60 3 Phe bIGFBP-2. All of the Tyr-60 region bIGFBP-2 mutants produced essentially the same CD spectra (Fig. 4) and all ran as single bands at the same molecular weight as bIGFBP-2 in SDS-PAGE analysis (Fig. 3). Therefore, any changes in the IGF binding characteristics of the bIGFBP-2 mutants were considered to be due to the loss of side chain interactions at the mutagenic site in question and not to gross changes in protein structure.
There was good agreement between the three functional analyses that were used to investigate the IGF binding abilities of the bIGFBP-2 mutants. Western ligand blot (Fig. 3b), solution binding assays (Fig. 5, Table I), and the BIAcore analyses (Fig. 6, Table II) all indicated that both Ala and Phe substitution for Tyr-60 resulted in a bIGFBP-2 molecule with reduced affinity for IGFs. However, some differences were noted. For example, Val-59 3 Ala bIGFBP-2 bound 125 I-IGF-II at a visibly reduced level compared with bIGFBP-2 in the Western ligand blot (Fig. 2), yet bound 125 I-IGFs as well as bIGFBP-2 in the solution binding assay (Fig. 4, Table I). Similarly, the decrease in 125 I-IGF-II binding of Tyr-60 3 Ala bIGFBP-2 was far more apparent in the Western ligand blot (Fig. 3b) than in the Chi 2 values for sensogram curve fits were equal to 1.0 or were less. R 2 values generated during the linear regression derivation of binding constants were equal to 0.95, or were greater. Interchip variation was 25% of the value or less for the apparent kinetic constant values k on , k off , and K D .
c The ranking of the bIGFBP-2 mutants relative to bIGFBP-2 with respect to k on , k off , and K D was the same on both biosensor chips. The interchip variation in the fold differences of kinetic constants was 12% of the mean value or less. d IGF-II biosensor data: K D ϭ k off /k on(app) .
Direct measurement of the association and dissociation kinetics of the IGFBP-2/IGF interaction in the BIAcore experiments could explain the observed differences between the Western ligand blot and the solution binding assay results. Whereas Val-59 3 Ala and Tyr-60 3 Ala bIGFBP-2 dissociated 2.3-and 6.2-fold more rapidly from the IGF-II biosensor surface than bIGFBP-2, respectively, both also associated with this surface approximately 2-fold more rapidly than bIGFBP-2. Under the equilibrium conditions of the solution binding assay, the increased dissociation rates of Val-59 3 Ala and Tyr-60 3 Ala bIGFBP-2 were offset by the increases in the association rate. However, in the Western ligand blot, nonspecifically bound 125 I-IGF-II was washed from the filter by buffer replacement, and so the possibility of IGF and IGFBP-2 reaching a binding equilibrium was prevented. Therefore, we propose that the Western ligand blot ranked the bIGFBP-2 mutants with respect to their relative dissociation rates, whereas the solution binding assay ranked the bIGFBP-2 mutants according to their overall affinities.
Insight into the reduced IGF binding affinity of Tyr-60 3 Ala and Tyr-60 3 Phe bIGFBP-2 was provided by BIAcore analysis. Substitution of Tyr-60 of bIGFBP-2 with Ala and Phe produced very different changes in the kinetics of IGF interactions. Tyr-60 3 Ala bIGFBP-2 associated with and dissociated from IGF biosensor surfaces more rapidly than bIGFBP-2. In contrast, Tyr-60 3 Phe bIGFBP-2 exhibited a reduced rate of association with, and an increased rate of dissociation from, IGF biosensor surfaces compared with bIGFBP-2. Interestingly, Tyr-60 3 Ala and Val-59 3 Ala bIGFBP-2 exhibited very similar kinetics on the IGF-I and IGF-II biosensor surfaces. The enhanced association and dissociation kinetics of Val-59 3 Ala bIGFBP-2 with immobilized IGF suggests that this mutation produced subtle changes to the structure of the IGF-binding site of bIGFBP-2, without detriment to the net energy of IGF binding. In contrast, Tyr-60 3 Ala bIGFBP-2 exhibited a net reduction in the energy of IGF binding due to large increases in the dissociation rate from IGF surfaces. Therefore, the shape and volume of the side chains of Val-59 and Tyr-60 may help to define the IGF-binding site of bIGFBP-2. However, the aromatic function and the hydrogen bonding potential of Tyr-60 are clearly the most significant contributors to the stability of the bIGFBP-2⅐IGF complex. The replacement of Tyr-60 with Phe was anticipated to be the most subtle mutation, with the aromatic side chain packing maintained and the loss of the tyrosyl hydroxyl group the only change. Surprisingly, Tyr-60 3 Phe bIGFBP-2 exhibited the lowest affinities for IGFs, thus providing further evidence to suggest that Tyr-60 participates in hydrogen bond(s) that stabilize IGF interactions.
In circulation, IGFBPs bind IGFs with a 1:1 stoichiometry (reviewed in Ref. 7), and therefore a single site kinetic model should provide a valid approximation of the IGFBP/IGF interaction. Indeed, the apparent affinity constants that were calculated with the derived association and dissociation constants of bIGFBP-2 and the IGF-I and IGF-II surfaces (0.5 nM for IGF-I and 0.2 nM for IGF-II) correspond very well with published constants generated by competition solution binding assays (44,45). Yet, the interactions of bIGFBP-2 and the bIGFBP-2 mutants with immobilized IGFs deviated from pseudo-first order kinetics in the BIAcore (Fig. 6). Multiple phase kinetics for single site interactions can be due to artifacts of the BIAcore assay conditions (46,47). In this study, steps have been taken to address some of the common artifactual causes for multiple phase kinetics. Thus, very low density IGF biosensor surfaces (60 -120 RU) and moderate flow rates (40 l/min) have been deliberately used to minimize mass transfer limita-tions and the steric masking of binding sites. We have used amine coupling to immobilize IGFs in this study, and this strategy can lead to a mixed population of ligand on the biosensor surface with a range of different affinities for the analyte (47). However, the same non-pseudo-first order kinetic behavior was observed when immobilized hIGFBP-3 was analyzed with free IGF (48). It is therefore possible that the multiple association and dissociation phases present in IGF/IGFBP sensorgrams reflect a physical characteristic of the protein interaction such as multiple step binding. It is interesting to note that Scatchard analysis of competitive solution binding assays have also shown that IGFBPs may exhibit both low and high affinity binding sites for IGFs under some conditions (49,50).
Overall, the effects of mutagenesis in the Tyr-60 region of bIGFBP-2 were more severe for interactions with IGF-I than IGF-II. For example, Tyr-60 3 Phe bIGFBP-2 bound to the IGF-I biosensor surface with a 8.4-fold lower affinity than bIGFBP-2, whereas it bound to the IGF-II biosensor surface with a 4.0-fold lower affinity than bIGFBP-2. This corresponds well with our earlier findings that iodo-bIGFBP-2 exhibited an 8-fold reduction in apparent affinity for IGF-I compared with a 4-fold reduction in apparent affinity to IGF-II (32). The high degree of structural similarity between IGF-I and IGF-II (9) suggests that both IGF molecules interact with IGFBPs through similar side chain contacts. Mutagenic support for this idea can be seen in the similar reduction in IGFBP affinity of [Arg 3 ]IGF-I and its structural homolog [Arg 6 ]IGF-II (16). The disproportionate sensitivity of the Tyr-60 bIGFBP-2 mutants toward IGF-I binding, in addition to the natural preference of bIGFBP-2 for IGF-II (4), suggests that the bIGFBP-2⅐IGF-II complex contains additional points of molecular interaction that are absent in the bIGFBP-2⅐IGF-I complex. Indeed, in a recent truncation study (33), mutagenic removal of 62 amino acid residues from the C terminus of bIGFBP-2 resulted in a dramatic loss of IGF binding. More importantly, this bIGFBP-2 mutant had lost all binding preference for IGF-II over IGF-I. It was thus concluded that the C-terminal cysteine rich domain of bIGFBP-2 contained determinants of IGF-II binding specificity.
The ability of Tyr-60 to form a hydrogen bond that is important for IGF binding is evident by the reduced affinity of both Tyr-60 3 Ala and Tyr-60 3 Phe bIGFBP-2 for IGF-I and IGF-II. Moreover, the presence of an aromatic function at position 60 of bIGFBP-2 reduces the rate of formation of bIGFBP-2⅐IGF complex but also enhances the stabilization of the complex. This study raises the question as to whether the hydrogen bond acceptor of Tyr-60 is within the bIGFBP-2 molecule or within the IGF molecule. We propose that Tyr-60 is located in the IGF binding interface of bIGFBP-2 and that modification or mutagenic replacement of Tyr-60 directly disrupts contacts between IGF and bIGFBP-2. However, the alternative possibility that Tyr-60 modification and mutagenic replacement indirectly affects IGF binding for example, by preventing a change in conformation that is necessary for high affinity IGF binding, has not yet been eliminated. The simplest interpretation of our earlier observation that iodination of Tyr-60 can be blocked by the formation of an bIGFBP-2⅐IGF complex supports a direct interaction between Tyr-60 of bIGFBP-2 and IGF. Clearly, distinction between these two models must await biophysical characterization of the IGFBP-2⅐IGF complex.