Heparin Binding Induces a Conformational Change in Pigment Epithelium-derived Factor*

Pigment epithelium-derived factor (PEDF) is a noninhibitory serpin found in plasma and in the extracellular space. The protein is involved in different biological processes including cell differentiation and survival. In addition, it is a potent inhibitor of angiogenesis. The function is likely associated with binding to cell surface receptors in a heparin-dependent way (Alberdi, E. M., Weldon, J. E., and Becerra, S. P. (2003) BMC Biochem. 4, 1). We have investigated the structural basis for this observation and show that heparin induces a conformational change in the vicinity of Lys178. This structural change was evident both when binding to intact heparin and specific heparin-derived oligosaccharides at physiological conditions or simply when exposing PEDF to low ionic strength. Binding to other glycosaminoglycans, heparin-derived oligosaccharides smaller than hexadecasaccharides (dp16), or type I collagen did not affect the structure of PEDF. The conformational change is likely to expose the epitope involved in binding to the receptor and thus regulates the interactions with cell surface receptors.

Pigment epithelium-derived factor (PEDF) 3 is a 50-kDa member of the serine proteinase inhibitors (serpin) superfamily (1,2), although no target proteinase has been identified (3). The protein was initially identified as a factor produced by retinal pigmented epithelial cells that promotes neuronal differentiation of retinoblastoma cells (4). PEDF is present in several ocular compartments including the retinal and ciliary epithelia, the cornea, and the vitreous fluid (5). In addition, PEDF is found in a range of other tissues including the central nervous system, skeletal muscle, liver, heart, placenta, and blood (6,7).
The physiological functions of PEDF can be divided into neurotrophic/neuroprotective and antiangiogenic activities. These activities are exemplified by neurotropic activity at concentrations as low as 1 nM (4) and by an in vitro antiangiogenic activity that supersedes angiostatin, thrombospondin-1, and endosta-tin (8). Because PEDF is found in plasma at levels significantly above the physiological relevant concentrations (ϳ100 nM), the activity is likely to be tightly regulated (7).
The three-dimensional structure of PEDF revealed two features that may be involved in the expression and regulation of PEDF activity: (i) the reactive center loop and (ii) an asymmetrical distribution of charged amino acid residues, producing basic and acidic regions at opposite poles of the protein (2). This is unusual in a serpin structure and suggests functional importance. Indeed, it has been reported that PEDF binds to type I collagen (7,9,10) and that the acidic region of PEDF participates in this interaction (10,11). The basic region interacts with glycosaminoglycans (12,13). Interestingly, the binding of PEDF to an unidentified receptor on Y-79 retinoblastoma cells is mediated by heparin (14). It has been speculated that heparin functions as a co-factor or induces a conformational change exposing regions involved in the receptor binding. Heparin binding is known to induce a conformational change in antithrombin III of the coagulation cascade, where binding regulates the proteinase inhibitory activity of this serpin (15).
Here we have analyzed the heparin-induced structural change of PEDF using biochemical and biophysical methods. Collectively, the data suggest that the heparin interaction causes a localized conformational change in the vicinity of Lys 178 , which likely facilitates the interaction with cell surface receptors.
N-terminal Amino Acid Sequence Analysis-PEDF fragments were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore) (18), and analyzed by automated Edman degradation in an Applied Biosystems model 477A/120A protein sequencing system. Analysis of Conformational Changes by Limited Proteolysis-Increasing amounts of heparin, hyaluronic acid, chondroitin sulfate, or collagen were added to PEDF (2 g), and the sam-ples were equilibrated for 1 h at 37°C in 25 l of 20 mM Tris-HCl, pH 7.4, containing 100 mM NaCl. Trypsin was added (enzyme:substrate 1:10 w/w), and the digestion was continued for 15 min at 37°C. Prior to trypsin digestion, heparinase was added (0.5 unit) to some reactions. The trypsin digestions were stopped by adding phenylmethylsulfonyl fluoride (2.5 mM final concentration).
To analyze the effect of ionic strength, PEDF (2 g) was digested with trypsin at a ratio of 1:40 (w/w) in 25 l of 20 mM Tris-HCl, pH 7.4, containing increasing amounts of NaCl (0 -100 mM). The incubation was carried out for 15 min at 23°C. The influence of heparin in these experiments was analyzed by preincubating PEDF (2 g) with heparin (0.2 g) for 30 min at 23°C. The samples were then digested with trypsin in a ratio of 1:1 (w/w) in the presence of increasing amounts of NaCl (100 -250 mM), in 20 l for 15 min at 23°C before the addition of phenylmethylsulfonyl fluoride (2.5 mM final concentration). The reaction products from the proteolytic digests described above were analyzed by reduced SDS-PAGE and in some cases by automated Edman degradation.
Fluorescence Equilibrium Denaturation Measurements and CD Spectroscopy-The data were collected in a LS55 luminescence spectrometer (PerkinElmer Life Sciences) using a 310 -400 nm range with excitation at 290 nm and 10 nm excitation and emission slits. All measurements were performed in 20 mM Tris-HCl, pH 7.4, containing 0 mM NaCl, 100 mM NaCl, or 100 mM NaCl and 2 mg/ml heparin and recorded as the average of three scans at 23°C. PEDF (2 g) in 120 l (ϳ0.4 M) was incubated in the presence of urea concentrations ranging from 0 to 6 M at 25°C for 1 h before measurement was performed. The data were analyzed using the following equation for urea denaturation curves (19,20), where F is the signal (the ratio between the emission intensities at 335 and 355 nm), ␣ N and ␣ D denote the signal at 0 M urea for the native and denatured states, respectively, ␤ N and ␤ D are the slopes of the base lines of the native and denatured states, respectively, [urea 50% ] is the urea concentration where 50% of the protein is denatured, and m D-N is the linear dependence of the equilibrium constant of unfolding, K D-N on [urea]. Based on these data, the stability in the absence of urea ⌬G D-N water can be calculated as follows.
Nonlinear least-squares regression analysis of the data were carried out in Kaleidagraph, version 3.5 (Synergy Software). Far-UV CD spectra were recorded in a JASCO J-810 spectropolarimeter (Jasco Spectroscopic Co. Ltd.) from 200 to 250 nm using a 20 nm/min scanning speed and a resolution of 0.1 nm. The bandwidth was 1.0 nm and response time 2.0 s. Cuvettes of 1 mm path length were used. Four accumulations were averaged to yield the final spectra. All measurements were performed in 20 mM Tris-HCl, pH 7.4, containing 0 mM NaCl, 100 mM NaCl, or 100 mM NaCl and 2 mg/ml heparin. Buffer and heparin contributions were subtracted.
Heparin Binding Assays-The interaction between PEDF and heparin or specific heparin-derived oligosaccharides were investigated by ELISA. Microtiter wells (Maxisorb, Nunc, Denmark) were coated with 0.1 g/100 l heparin-BSA conjugate in 50 mM NaHCO 3 , pH 9.6, overnight at 23°C. The wells were blocked using 200 l of BSA (1%) in 20 mM Tris-HCl, 135 mM NaCl, pH 7.6 (TBS), for 1 h at 23°C and washed extensively (TBS containing 0.1% Tween). Serial dilutions of PEDF were added (100 l/well) and left incubating for 1 h at 23°C. The wells were washed as described above, and bound PEDF was detected using a standard ELISA protocol, including PEDF antiserum (generated toward plasma purified PEDF), horseradish peroxidase-conjugated secondary antibody, and o-phenylenediamine dichloride as the substrate. Each data point was collected in duplicates at 450 nm using a plate reader (THERMOmax, Molecular Devices) operated in end point mode. The relative binding capacity of PEDF was estimated in the linear range of the binding curve.
The interaction between PEDF and the heparin-derived oligosaccharides was examined using competition assays based on PEDF binding to heparin-BSA-coated microtiter wells. A fixed amount of PEDF (1 g) was preincubated with increasing amounts of intact heparin or heparin-derived oligosaccharides, including tetrasaccharides (dp4), octasaccharides (dp8), and hexadecasaccharides (dp16) in 100 l of TBS for 1 h before the addition to the wells. Wells were developed as described above and all data points were collected in duplicates.

The Susceptibility of PEDF to Limited Trypsin Proteolysis
Was Increased by Heparin-To investigate putative structural changes induced by ligand binding, PEDF was titrated with heparin, hyaluronic acid, chondroitin sulfate, or collagen and studied by means of limited proteolysis. In the presence of hyaluronic acid, chondroitin sulfate, and collagen, PEDF remained resistant to proteolysis (data not shown). However, heparin significantly modified the proteolytic susceptibility, and PEDF was readily cleaved into smaller fragments (Fig. 1). One fragment of 29 kDa migrated as a fuzzy band, and two smaller fragments of ϳ23 kDa migrated as a closely spaced doublet. At increasing amounts of heparin, the rate of proteolysis was accelerated, as observed by a decrease in staining intensity of the band representing intact PEDF. The specificity of the interaction was ver-ified by including heparinase in the reaction mixture prior to trypsin digestion, which reversed the effect of heparin (Fig. 1). Collectively, these data are consistent with a change in the PEDF structure following binding of heparin.
Identification of the Initial Trypsin Cleavage Site-The PEDF fragments produced by limited digestion with trypsin were characterized by automated Edman degradation. The 29-kDa fragment produced the sequence Glu-Ile-Pro-Asp-Glu, suggesting that this band represents the C-terminal part of PEDF spanning Glu 179 -Pro 399 (theoretical molecular mass 25 kDa) ( Figs. 1 and 2). The fragment includes the N-linked glycan at Asn 266 (7), and the fuzzy appearance in SDS-PAGE is likely caused by heterogeneity of the attached glycan (ϳ2 kDa) and/or C-terminal truncation at Arg 380 , Lys 391 , or Arg 397 . The lower 23-kDa closely spaced doublet did not sequence during Edman degradation, suggesting that these fragments represents the pGlu-blocked N terminus of PEDF (7). The theoretical molecular mass of the pGlu 1 -Lys 178 fragment (20 kDa) corresponds well with the observed mass of 23 kDa (Fig. 1). The appearance of a doublet is most likely caused by C-terminal truncation.   Taken together these data suggest that the initial trypsin cleavage occurs at Lys 178 -Glu 179 (Fig. 2).
The Proteolytic Susceptibility of PEDF Is Affected by Ionic Strength-Because the interaction between PEDF and heparin is likely to be electrostatic in nature (12), we tested whether the proteolytic susceptibility of PEDF was affected by the ionic strength of the buffer. In the presence of 100 mM NaCl the protein was resistant to cleavage by trypsin. However, in the absence of NaCl, PEDF was completely digested (Fig.  3). The digestion of PEDF at low ionic strength generated fragments similar to those observed during the digestion at physiological conditions in the presence of heparin (Fig. 1). The increase in proteolytic susceptibility at low ionic strength could be reversed by the addition of NaCl (data not shown). These findings show that the structure of PEDF in the vicinity of Lys 178 is altered by low ionic strength and that the structural change is reversible.
Increase in Ionic Strength Can Reverse Heparin-induced Conformational Change in PEDF-As described above, the presence of heparin or low ionic strength induces a conformational change, rendering PEDF receptive to limited proteolysis. To investigate whether the two circumstances generated similar structural alterations, PEDF was titrated with NaCl in the presence of heparin and analyzed by reduced SDS-PAGE following tryptic digestion (Fig. 4). It was apparent that an increase in the ionic strength reversed the effect of heparin and caused PEDF to become resistant to proteolysis at higher NaCl concentrations. Because the structural change induced by heparin is reverted by an increase in the salt concentration, it is likely that the structural change induced by heparin and low ionic strength is similar in nature. This is further supported by the fact that appearance of identical reaction products are obtained following limited proteolysis of PEDF at low ionic strength and in the presence of heparin (Figs. 1 and 3).
Structural Analysis of PEDF Using Intrinsic Tryptophan Fluorescence and CD Spectroscopy-To investigate the influence of heparin and ionic strength on the tertiary structure, we subjected PEDF to urea denaturation and monitored the structural impact by Trp fluorescence spectroscopy. The degree of unfolding was evaluated by the ratio of fluorescence at 335 nm (hydrophobic environment) and 355 nm (hydrophilic environment), determined to be the wavelengths of maximum intensity for native and fully unfolded PEDF, respectively (Fig. 5A). The fluorescence spectra of PEDF in the presence of (i) 100 mM NaCl, (ii) heparin and 100 mM NaCl, and (iii) 0 mM NaCl in the absence of denaturant were similar (data not shown). This observation shows that the overall structure of PEDF is maintained under all conditions. In the presence of 100 mM NaCl the [urea] 50% value is 3.30 Ϯ 0.02 M and the m D-N value is 2.11 Ϯ 0.14 M Ϫ1 , corresponding to a calculated stability (⌬G D-N ) of 6.97 Ϯ 0.63 kcal/mol (Fig. 5B). This value is within the range of typical globular proteins (5-15 kcal/mol) (21). When heparin was included in the analysis, the [urea] 50% value was slightly reduced to 3.03 Ϯ 0.04 M with an m D-N value of 1.97 Ϯ 0.21 M Ϫ1 , corresponding to a stability of 5.97 Ϯ 0.64 kcal/mol (Fig. 5B). In the absence of NaCl, the [urea] 50% value was further reduced to 2.47 Ϯ 0.03 M (m D-N value of 2.00 Ϯ 0.17 M Ϫ1 ), corresponding to 4.94 Ϯ 0.42 kcal/ mol (Fig. 5B). These findings suggest that the increase in proteolytic susceptibility observed at low ionic strength or in the presence of heparin at physiological conditions is the result of a  ). B, titration of PEDF in Tris-HCl, pH 7.4, by urea in the presence of 0 mM NaCl (E), 100 mM NaCl (Ⅺ), or 100 mM NaCl, 2 mg/ml heparin (OE). The level of denaturation is depicted as the ratio of the fluorescence signal obtained for native (335 nm) and fully unfolded (355 nm) PEDF. These data show that PEDF is destabilized by heparin and low ionic strength. C, far-UV CD spectra of PEDF (4 M) in the presence and absence of NaCl (100 mM) and heparin (2 mg/ml) as indicated. Data represent the average of three spectra (0.2 nm steps). The spectra were corrected for the appropriate background contributions. No significant change in the secondary structure of PEDF is observed. localized structural change and not a global unfolding of the PEDF structure. This conclusion was further supported by CD spectroscopy, which revealed no significant change in the secondary structure of PEDF in the absence of NaCl or presence of heparin (Fig. 5C).
Heparin-derived Oligosaccharides Smaller than 8 Disaccharide Units (dp16) Do Not Bind to PEDF nor Induce a Conformational Change-The ability to induce a conformational change is likely to depend on the size of the heparin. To investigate this, PEDF was subjected to limited proteolysis in the presence of intact heparin or heparin-derived oligosaccharides, including tetrasaccharides (dp4), octasaccharides (dp8), and hexadecasaccharides (dp16). Only intact heparin and the hexadecasaccharides (dp16) accelerated proteolysis, consistent with a conformational change (Fig. 6). The tetrasaccharides (dp4) and octasaccharides (dp8) were apparently unable to coordinate the intramolecular interactions required for the conformational change to occur.
To investigate whether the tetra-and octasaccharides were able to bind PEDF without inducing a conformational change as observed for intact heparin and the hexadecasaccharides, we used a heparin-based ELISA. As expected, PEDF bound intact heparin in a dose-dependent manner (Fig. 7A). The binding of the oligosaccharides was evaluated in a competition format. Thus, the binding capacity of PEDF was reduced in the presence of intact heparin or hexadecasaccharides (dp16) (Fig. 7B). Conversely, the tetrasaccharides (dp4) and octasaccharides (dp8) did not prevent the binding of the PEDF to the heparin-coated microtiter wells (Fig. 7B). Collectively, these data suggest that only intact heparin and the hexadecasaccharide (dp16) were able to bind to PEDF and induce a conformational change.

DISCUSSION
The binding of PEDF to a receptor on the cell surface of retinoblasoma Y-79 cells is significantly improved by the presence of heparin (14). It has been suggested that the mechanism of enhanced receptor binding involves a structural change of PEDF (14). In this study, we have employed biochemical and biophysical methods to investigate this hypothesis. Limited proteolysis has previously been used as a tool to investigate the inhibitory capacity of bovine PEDF (3,22). These studies concluded that bovine PEDF was resistant to proteolysis and only a single cleavage of Arg 360 -Leu 361 within the reactive center loop was detected when the protein was treated with trypsin in the presence of 150 mM NaCl (3). 4 As shown in this study, human PEDF is not cleaved, using similar conditions, most likely because the homologous peptide bond, His 362 -Leu 363 , is not a substrate for trypsin. However, we observed that PEDF became a substrate for trypsin in the presence of heparin, suggesting that the protein undergoes a conformational change. The addition of heparinase reversed this effect, supporting the fact that the altered proteolytic susceptibility was caused by the heparin-PEDF interaction. Interestingly, the interaction between PEDF and other identified binding partners, including type I collagen (7,9,10) and glycosaminoglycans such as hyaluronic acid and chondroitin sulfate (12), did not increase the proteolytic susceptibility. The fragments generated by limited trypsin digestion were analyzed by automated Edman degradation. This revealed that the initial cleavage event occurred at Lys 178 -Glu 179 , a tryptic cleavage site which has also been reported using recombinant PEDF (23). This result is consistent with an increased segmental mobility or a local unfolding in that region upon heparin binding.
In addition to the presence of heparin, we also observed that low ionic strength increased the proteolytic susceptibility. However, the stability was recovered by the addition of NaCl, suggesting that the impact on the structure was reversible and not the result of a structural collapse. To test whether a similar effect of ionic strength was a property of other noninhibitory serpins in general, we subjected ovalbumin to trypsin digestion in the absence or presence of 100 mM NaCl. Unlike PEDF, ovalbumin was stable at both conditions (data not shown). Although the absence of NaCl represents a nonphysiological condition, we hypothesize that the observed increase in proteolytic sensitivity at low salt concentrations or in the presence of heparin reflects similar PEDF conformations. This was substantiated by (i) the appearance of similar reaction products after limited trypsin digestion and (ii) the observation that the heparin-induced conformational change was reversed when ionic strength was increased. 4 Numbering of bovine PEDF is done by homology to the mature N terminus of the human protein: Gln 1 -Asn-Ala-. . . . . .-Gly-Thr 397 . FIGURE 6. Effect of heparin-derived oligosaccharides on the proteolytic susceptibility of PEDF. PEDF (2 g) was incubated with trypsin 1:1 (w/w) in the presence of 2 g of intact heparin or heparin-derived oligosaccharides (tetrasaccharides (dp4), octasaccharides (dp8), hexadecasaccharides (dp16)) at physiological conditions. Samples were analyzed by SDS-PAGE and Coomassie Brilliant Blue staining. Controls representing PEDF (2 g) and trypsin (2 g) in the presence and absence of heparin were included as indicated.
Only intact heparin and hexadecasaccharides (dp16) render PEDF more susceptible to proteolysis by trypsin. Note that intact heparin interferes with the staining and prevents the detection of low molecular weight bands of trypsin.