Analysis of Small Latent Transforming Growth Factor-β Complex Formation and Dissociation by Surface Plasmon Resonance

Transforming growth factor-β (TGFβ) is a pluripotent regulator of cell growth and differentiation. The growth factor is expressed as a latent complex that must be converted to an active form before interacting with its ubiquitous high affinity receptors. This conversion involves the release of the mature TGFβ through disruption of the noncovalent interactions with its propeptide or latency associated protein (LAP). Complex formation or dissociation between LAP and TGFβ plays a very important role in TGFβ biological activity at different steps. To further characterize the kinetic parameters of this interaction, we have employed surface plasmon resonance biosensor methodology. Using this technique, we observed real time association of LAP with mature TGFβ1. The complex formation showed an equilibrium K d around 3–7 nm. Furthermore, we observed dissociation of the complex in the presence of extreme pH, chaotropic agents, or plasmin, confirming their effects on TGFβ activation. The same approach was used to examine whether latent TGFβ1 could interact with thrombospondins, previously described as activators of latent TGFβ. Using this method, we could not detect any direct interaction of thrombospondins with either LAP alone, TGFβ1 alone, or the small latent TGFβ1 complex. This suggests that activation of latent TGFβ1 complex by thrombospondins is through an indirect mechanism.

Transforming growth factor-␤ (TGF␤) is a pluripotent regulator of cell growth and differentiation. The growth factor is expressed as a latent complex that must be converted to an active form before interacting with its ubiquitous high affinity receptors. This conversion involves the release of the mature TGF␤ through disruption of the noncovalent interactions with its propeptide or latency associated protein (LAP). Complex formation or dissociation between LAP and TGF␤ plays a very important role in TGF␤ biological activity at different steps. To further characterize the kinetic parameters of this interaction, we have employed surface plasmon resonance biosensor methodology. Using this technique, we observed real time association of LAP with mature TGF␤1. The complex formation showed an equilibrium K d around 3-7 nM. Furthermore, we observed dissociation of the complex in the presence of extreme pH, chaotropic agents, or plasmin, confirming their effects on TGF␤ activation. The same approach was used to examine whether latent TGF␤1 could interact with thrombospondins, previously described as activators of latent TGF␤. Using this method, we could not detect any direct interaction of thrombospondins with either LAP alone, TGF␤1 alone, or the small latent TGF␤1 complex. This suggests that activation of latent TGF␤1 complex by thrombospondins is through an indirect mechanism.
Transforming growth factor-␤1 (TGF␤1) 1 belongs to a family of peptides regulating cell growth and differentiation and extracellular matrix production (for reviews, see Refs. 1 and 2). Its roles in growth inhibition of many epithelial cells, potentiation of wound repair, angiogenesis, regulation of endocrine functions, and immunomodulation are among its many documented actions. TGF␤1 exerts its effects via binding to high affinity receptors that are expressed on virtually all cell types, and as TGF␤1 production is ubiquitous, there must be strict regulation of its activity.
TGF␤1 is synthesized as a 390-amino acid precursor protein, termed prepro-TGF␤1, which undergoes several processing steps. These include proteolytic cleavage of the signal peptide (3), glycosylation and mannose-6-phosphorylation of the prodomain (4,5), and cleavage at a multibasic residue site by a proconvertase, releasing mature TGF␤1 (3). Completion of posttranslational processing results in a pro-domain (amino acid residues 30 -278), called latency-associated peptide (LAP), and a mature TGF␤ (amino acid residues 279 -390). LAP exists as a homodimer, linked by disulfide bonds, and has been shown to noncovalently bind a mature covalent TGF␤1 homodimer with a 1:1 ratio (6). This biologically inactive complex does not bind to TGF␤ receptors. The resulting latent form may then be activated in different ways (for review, see Ref. 7). Dissociation can be accomplished in vitro by heat, extreme pH (above pH 9 or below pH 3), and selected chaotropic agents. In vivo, the mechanism may involve specific proteases such as plasmin (8). Thrombospondins (TSPs), extracellular matrix-associated proteins, have also been described as activators of latent TGF␤1 complex (9,10). It has been proposed that TSP1 activates latent TGF␤ via a protease-and cell-independent mechanism (9). It is now clear that LAP is implicated at different levels in TGF␤ biological activity. Previous studies have indicated that LAP plays important roles in secretion and proper folding of the mature growth factor molecule (5,11,12). Furthermore, specific binding of LAP to TGF␤ may be an important mechanism in regulating the biological functions of secreted TGF␤ (6,13). Indeed, several studies suggest that LAP is a potent inhibitor of bioactive TGF␤ both in vitro and in vivo (6,14). Furthermore, formation of latent complexes increases plasma TGF␤1 half-life about 50-fold (15). Moreover, LAP interacts with the mannose 6-phosphate receptor that has been shown to be required for cellular activation of latent TGF␤1 (16); thus LAP could potentially contribute to enhanced and targeted TGF␤ activity.
The first purpose of the present work was to analyze the affinity parameters of the complex formation between LAP and TGF␤1. We thought that a method of choice for determining these parameters would be surface plasmon resonance (SPR) using the BIAcore TM apparatus (Pharmacia Biotech Inc.). The principle is that one molecule involved in the interaction to be studied is covalently immobilized to a sensor chip, and the other interactant is then passed over the chip in solution. The detection system measures and records a signal proportional to the mass of the protein bound to the surface. In this way, the association phase can be visualized in real time as the ligandcontaining solution flows over the surface, and the subsequent dissociation is similarly displayed after the flow switches back to buffer containing no ligand (17). Using this technique, we were able to show direct interaction between LAP and mature TGF␤1, that is, formation of the small latent complex with TGF␤1 immobilized and LAP injected as analyte or vice versa.
We measured the affinity parameters of this interaction.
The second goal was to use this technique to visualize the dissociation of the small latent TGF␤1 complex by known activators. We confirmed direct activation of the small latent recombinant TGF␤1 by extreme pH, chaotropic agents, or plasmin. However, we did not detect any direct interaction of thrombospondins with either LAP, TGF␤1, or the small latent TGF␤1 complex.
Immobilization of TGF␤1 or LAP on Sensor Chip-The BIAcore running buffer was 10 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 0.005% P-20 (HBS, Pharmacia), except where indicated. Equal volumes of 0.1 M N-hydroxysuccinimide and 0.1 M N-ethyl-NЈ-(3-diethylaminopropyl)carbodiimide were mixed, and 35 l were injected over the surface of the sensor chip to activate the carboxymethylated dextran. TGF␤1 (6.66 g/ml in 25 l of 10 mM acetate, pH 4) or LAP (13.3 g/ml in 10 l of 10 mM acetate, pH 4) was injected over the activated surface, followed by 35 l of ethanolamine to block remaining active carboxyl groups. The immobilization procedure was carried out at 25°C and at a constant flow rate of 5 l/min. Control immobilization was performed under the same conditions in the absence of TGF␤1 or LAP.
Kinetic Assays on the BIAcore-All experiments were carried out at 25°C with a constant flow rate of 25 l/min in HBS buffer. 50 l of the analyte were injected for 2 min (association phase), followed by a 2-min period when HBS was passed (dissociation phase). Equal volumes of each protein dilution were also injected over a mock blocked surface to serve as blank sensorgrams for subtraction of bulk refractive index background and nonspecific binding of the analyte. All kinetic assays were followed by an injection of 25 l of 6 M guanidine chloride (GnCl) to dissociate remaining bound ligand, leaving only the immobilized interactant (regeneration phase). We worked at low immobilization level, high flow rate (25 l/min), and concentrations of analyte suitable to limit mass transport. Further, the mass transport limitations were checked using the BIAsimulation program. All the steps described were fully automated and carried out by the BIAcore systems robotics. Association, dissociation, and regeneration phases were followed in real time as a change in signal expressed in resonance units (RU). Curves derived from these assays were used to generate kinetic contants.
Data Analysis-Sensorgrams were analyzed by nonlinear least squares curve fitting using the BIAevaluation program (Pharmacia). Kinetic constants were generated from the association and dissociation curves from the BIAcore experiments by fitting to a single-site binding model (A ϩ B ϭ AB). This model gave a single exponential fit with a 2 Ͻ 0.5. Comparison fitting with more complicated models did not give a better interpretation of the data. The equation R t ϭ R 0 exp(Ϫk d (t Ϫ t 0 )) was used for the dissociation phase, where R t was the amount of ligand remaining bound in RU at time t and t 0 was the beginning of dissociation phase. The final dissociation rate constant, k d , was calculated from the mean of the values obtained from a series of injections. To analyze the association phase, the equation R t ϭ R eq (1 Ϫ exp(Ϫk s (t Ϫ t 0 ))) was employed where R eq was the amount of bound ligand (in RU) at equilibrium, t 0 was the starting time of injection, and k s ϭ k a C ϩ k d , where C was the concentration of analyte injected over the sensor chip surface. The association rate constant, k a , was determined from the slope of a plot of k s versus C. The apparent equilibrium dissociation constant K d was determined from the ratio of these two kinetic constants (k d /k a ).

RESULTS AND DISCUSSION
LAP Forms a Molecular Complex with Immobilized TGF␤1-The first goal of our study was to determine whether the association between LAP and mature TGF␤1 could be visual-ized by SPR. Therefore, experiments were performed in two ways. Either TGF␤1 was immobilized on the chip and LAP injected as the analyte or vice versa, where LAP was immobilized and TGF␤1 was injected. Immobilization of TGF␤1 yielded approximately 600 RU, and the maximal binding capacity of the surface (R max ) was 1296 RU for LAP binding. Association was started by injection of 50 l of LAP followed by regeneration. Fig. 1A shows the association and dissociation curves from a representative experiment performed with five different concentrations of LAP (50 -200 nM). Corrected sensorgrams are shown after subtraction of a sensorgram performed with the same concentration of LAP on the control flow cell. The association phase (0 -120 s) was analyzed by nonlinear least squares curve fitting as described under "Materials and Methods" to yield k s values at each concentration. A plot of k s versus concentration of LAP produced a straight line (Fig. 1B)   FIG. 1. LAP forms a complex  with a slope equal to the association rate constant (k a ). The value of k a for LAP binding to immobilized TGF␤1 (Table I) was determined to be 1.2 Ϯ 0.4 ϫ 10 4 s Ϫ1 M Ϫ1 . The dissociation phase (120 -240 s) was also analyzed by nonlinear least squares curve fitting. The dissociation rate constant, k d , was calculated from the curve representing 120 s of dissociation. The k d for LAP binding to immobilized TGF␤1 was calculated as 0.9 Ϯ 0.4 10 Ϫ4 s Ϫ1 (Table I). The apparent equilibrium dissociation constant K d determined from the ratio of these two kinetic constants (k d /k a ) was 7.5 nM. This experiment was repeated on three different chips, producing a mean K d of 5.6 Ϯ 1.4 nM (Table I).
TGF␤1 Forms a Complex with Immobilized LAP -Similar experiments were performed with immobilized LAP (approximately 800 RU, yielding an R max of 370 RU for TGF␤1 binding). Association was started by injection of 50 l of TGF␤1 followed by regeneration. Fig. 2 shows the association and dissociation curves from a representative experiment performed with three different concentrations of TGF␤1 (50, 80, and 100 nM). Constants were determined as described above. The k a value for TGF␤1 binding to immobilized LAP (Table I) was determined to be 31.7 Ϯ 26.6 ϫ 10 4 s Ϫ1 M Ϫ1 . The K d value for TGF␤1 dissociation from LAP surface was calculated as 8.5 Ϯ 2.0 10 Ϫ4 s Ϫ1 (Table I). The apparent equilibrium dissociation constant K d was 2.7 nM. This experiment was repeated on three different chips, producing a mean K d of 3.2 Ϯ 1.3 nM (Table I).
Whether LAP or TGF␤1 was immobilized, we obtained a similar equilibrium constant (K d ϭ 3.2 or 5.6 nM, respectively). However, the respective k a and k d values obtained with each protocol differed by a log. LAP binding to immobilized TGF␤1 yielded smaller k a and k d values than TGF␤1 binding to immobilized LAP. This may reflect a different sensitivity of LAP and TGF␤1 to chemical cross-linking to the sensor chip. The immobilization procedure would appear to more faithfully pre-serve the binding sites of LAP than those of TGF␤1. This K d was in agreement with that previously published by Miller et al. (18) (1.1-1.8 nM) using a covalent cross-linking method. The nanomolar range affinity between LAP and TGF␤1 could have major implications. Indeed, cellular receptors for TGF␤1 have been shown to have apparent K d values between 20 -50 pM (19 -21), around 100-fold smaller than the apparent K d values for TGF␤1 binding to LAP. The greater affinity of TGF␤1 for cell receptors than for LAP suggests that TGF␤1 interaction with the cell is favored. This would be in agreement with the latent TGF␤1 complex representing a circulating form of TGF␤1. In support of this idea, it has been shown that the latent complex of TGF␤1 has a serum half-life of 2 h as opposed to a half-life of 3 min observed for the mature active growth factor (15). Formation of latent complexes could thus switch TGF␤ from an autocrine/paracrine mode of action to a more  endocrine mode involving target organs distant from the site of synthesis.
Real-time Activation of TGF␤ -We then wondered whether SPR could permit us to visualize the dissociation of the small latent TGF␤1 complex during the activation process. As mentioned above, latent TGF␤1 can be activated by different means. Using SPR technology, we could monitor dissociation of the complex by injecting extreme pH buffers (above pH 9 or below pH 3, data not shown). Regeneration protocols confirmed that this complex can also be dissociated by chaotropic agents such as GnCl (Fig. 3A). This demonstrates that the interaction between these two peptides results at least in part from electrostatic interactions. As illustrated in Fig. 3A, a similar amount of LAP binds to TGF␤1 before and after GnCl treatment. This indicates that the binding capacity of TGF␤1 was preserved during the activation of the latent TGF␤1 complex by GnCl and validates its use of GnCl in surface regeneration for SPR analyses.
A more physiological mode of activation of TGF␤1 is through proteolytic cleavage by plasmin, which results in the removal of the amino-terminal portion of the pro-domain (22). We checked whether we could visualize this activation in real time by SPR. To do this, small latent TGF␤1 complex was formed as described previously (TGF␤1 immobilized and LAP injected afterward), and then 50 l of plasmin (dialyzed overnight against HBS, pH 8.6, 0.1 units/ml) was injected at a flow of 2 l/min. As shown in Fig. 3B, plasmin binds temporarily to the small latent TGF␤1 complex, dissociates from the complex, and displaces approximately 40% of the SPR signal (159 RU bound in the control versus 90 RU after injection of plasmin, Table II). This signal loss cannot be explained by the release of the proteolytic fragment of LAP alone, which constitute one-fifth of the molecule (22). It is more probable that the loss of 40% of the SPR signal corresponds to the release of 40% of LAP molecules bound to TGF␤1. This corresponds to the level of dissociation observed in vitro with plasmin (22). As a control, heat-inactivated plasmin (5 min at 100°C) was added; in that case, inactivated plasmin still binds to the small latent TGF␤ complex but does not trigger measurable dissociation of the complex. To confirm that the entire LAP molecule was released, we reinjected LAP. We were able to reconstitute an equivalent amount of complex as was present prior to plasmin treatment, as shown in Table II. This suggests that plasmin disrupted FIG. 4. Absence of association of TSP1 and TSP2 with LAP, TGF␤1, or small latent TGF␤1 forms. A, TSP1 (50 nM) was injected over immobilized LAP, TGF␤1, and a control (CTL) immobilized flow cell. B, TSP2 (50 nM) was injected over immobilized LAP, TGF␤1, and a control immobilized flow cell. C, LAP was immobilized and then TGF␤1 (50 nM) was injected to form small latent TGF␤1 complex, and over this complex, TSP2 (50 nM) or TSP1 (50 nM) was injected. D, vice versa of C. TGF␤1 was immobilized and then LAP (50 nM) was injected to form small latent TGF␤1 complex, and over this complex, TSP1 (50 nM) or TSP2 (50 nM) was injected. LAP binding to TGF␤1 but did not modify the immobilized TGF␤1. In the inverse experiment with the LAP⅐TGF␤1 complex (LAP immobilized and TGF␤1 as analyte), plasmin displaced approximately 70% (125 RU bound in the control versus 41 RU after injection of plasmin, Table II) of the complex. However, as shown in Table II, when TGF␤1 was reinjected, no LAP⅐TGF␤1 complex was reformed. This suggests that a fraction of LAP had been irreversibly proteolyzed by plasmin. Finally, a recombinant commercial source of preformed small latent TGF␤1 complex immobilized to the sensor chip could also be dissociated by addition of plasmin (data not shown). Altogether, our data indicate that plasmin can dissociate preformed recombinant latent TGF␤1 complex as well as small latent TGF␤1 complex formed on the sensor chip through a direct interaction.
Having validated this method through the determination of the kinetic parameters of LAP⅐TGF␤1 complex formation and through the observation of the dissociation of this complex by known activators, we used it to investigate the interaction of TSPs with the small latent TGF␤1 complex. TSPs are extracellular matrix-associated proteins that have also been described as activators of the latent TGF␤1 complex (9,10). Murphy-Ullrich and co-workers (23) have found that TSP1 activates latent TGF␤ via a protease-and cell-independent mechanism, whereas TSP2 does not. Their work indicates that TSP1 activation of TGF␤ could result from the binding of the tripeptide RFK, not present in TSP2, to the amino-terminal portion of LAP, inducing a conformational change of the com-plex and the liberation of mature TGF␤1 (23). In contrast, our laboratory has reported previously that both TSP1 and TSP2 activate the large latent TGF␤1 complex (containing LTBP (latent TGF␤ binding protein)) (10). This prompted us to investigate the possible association of TSP1 and TSP2 with LAP or the small latent TGF␤1 complex using SPR. TSP1 (Fig. 4A) or TSP 2 (Fig. 4B) was injected over immobilized LAP or TGF␤1 or a control flow cell in the presence of 2 mM Ca 2ϩ , which is essential to TSP structure and function (24,25). No specific association of TSP1 or TSP2 (50 nM) with LAP or TGF␤1 could be detected as the SPR signal returned to the same level as before the injection. Similar results were obtained in the absence of Ca 2ϩ (data not shown). The significant change in the signal after injection of TSP2 or TSP1 was due to the bulk refractive index of the buffer in which the TSPs are purified, which can also be observed on the control immobilization flow cell. We then tested if TSP1 or TSP2 could interact with the small latent TGF␤1 complex bound to the sensor chip through either LAP (Fig. 4C) or TGF␤1 (Fig. 4D). Again, there was no association between TSPs and the small latent TGF␤1 complexes. These experiments were performed with a wide range of TSP concentrations and gave similar results (data not shown). Nor did we detect any interaction in the inverse experiment (TSP immobilized and TGF␤1 or LAP as analytes, data not shown). To confirm that TSPs could bind to a known ligand in this system, we performed binding studies with a TSP-interacting molecule, namely heparin. When injected over biotinylated heparin immobilized onto a streptavidin chip, TSP1 and TSP2 (50 nM) bound to the surface and increased the SPR signal by 1000 RU (Fig. 5A) and 2000 RU (Fig. 5B), respectively.
The absence of interaction between TSPs and either LAP or latent TGF␤1 complex suggests that TSPs cannot activate the small latent TGF␤1 complex through direct protein-protein interaction as has been suggested (23). Activation by TSPs may be indirect and require other proteins, e.g. LTBP or the mannose 6-phosphate receptor. We have been unsuccessful so far in purifying the LTBP⅐LAP⅐TGF␤1 complex (large latent TGF␤1 complex) to homogeneity and, therefore, could not examine possible large latent TGF␤1 complex interactions by SPR. We have tried to reconstitute this large complex by passing LTBP-1 over immobilized small latent TGF␤1 complex. No association was seen, probably because the formation of the disulfide bond between cysteine 33 of LAP1 and the third 8-cysteine repeat of LTBP-1 (26) requires an oxidative environment and chaperone proteins.
Taken together, our data show that SPR technology can be used successfully to determine the kinetic parameters of the formation of the latent complex between TGF␤1 and LAP. It also demonstrates that SPR can be used to monitor in real-time the dissociation of preformed latent TGF␤1 complex in the presence of activators. Furthermore, it allows to discriminate between direct activators (like plasmin) and indirect activators (like thrombospondins). This technique should be useful in studying the possible association of latent TGF␤ with proteins such as cation-independent mannose 6-phosphate/insulin-like growth factor type II receptor or transglutaminase, which have been shown to be required for activation of latent TGF␤ at the cell surface (16,27). SPR should also be a method of choice to characterize the interaction between mature TGF␤ and some of its neutralizing molecules such as decorin, biglycan, ␣2-macroglobulin, or soluble beta-glycan.
Acknowledgments-We thank Dr. Anna Chinn for careful review of this manuscript and Dr. Odile Cochet for helpful discussion of the results.