Influence of interleukin-6 (IL-6) dimerization on formation of the high affinity hexameric IL-6.receptor complex.

The high affinity interleukin-6 (IL-6) signaling complex consists of IL-6 and two membrane-associated receptor components: a low affinity but specific IL-6 receptor and the affinity converter/signal transducing protein gp130. Monomeric (IL-6M) and dimeric (IL-6D) forms of Escherichia coli-derived human IL-6 and the extracellular (“soluble”) portions of the IL-6 receptor (sIL-6R) and gp130 have been purified in order to investigate the effect of IL-6 dimerization on binding to the receptor complex. Although IL-6D has a higher binding affinity for immobilized sIL-6R, as determined by biosensor analysis employing surface plasmon resonance detection, IL-6M is more potent than IL-6D in a STAT3 phosphorylation assay. The difference in potency is significantly less pronounced when measured in the murine 7TD1 hybridoma growth factor assay and the human hepatoma HepG2 bioassay due to time-dependent dissociation at 37°C of IL-6 dimers into active monomers. The increased binding affinity of IL-6D appears to be due to its ability to cross-link two sIL-6R molecules on the biosensor surface. Studies of the IL-6 ternary complex formation demonstrated that the reduced biological potency of IL-6D resulted from a decreased ability of the IL-6D·(sIL-6R)2 complex to couple with the soluble portion of gp130. These data imply that IL-6-induced dimerization of sIL-6R is not the driving force in promoting formation of the hexameric (IL-6·IL-6R·gp130)2 complex. A model is presented whereby the trimeric complex of IL-6R, gp130, and IL-6M forms before the functional hexamer. Due to its increased affinity for the IL-6R but its decreased ability to couple with gp130, we suggest that a stable IL-6 dimer may be an efficient IL-6 antagonist.

The high affinity interleukin-6 (IL-6) signaling complex consists of IL-6 and two membrane-associated receptor components: a low affinity but specific IL-6 receptor and the affinity converter/signal transducing protein gp130. Monomeric (IL-6 M ) and dimeric (IL-6 D ) forms of Escherichia coli-derived human IL-6 and the extracellular ("soluble") portions of the IL-6 receptor (sIL-6R) and gp130 have been purified in order to investigate the effect of IL-6 dimerization on binding to the receptor complex. Although IL-6 D has a higher binding affinity for immobilized sIL-6R, as determined by biosensor analysis employing surface plasmon resonance detection, IL-6 M is more potent than IL-6 D in a STAT3 phosphorylation assay. The difference in potency is significantly less pronounced when measured in the murine 7TD1 hybridoma growth factor assay and the human hepatoma HepG2 bioassay due to time-dependent dissociation at 37°C of IL-6 dimers into active monomers. The increased binding affinity of IL-6 D appears to be due to its ability to cross-link two sIL-6R molecules on the biosensor surface. Studies of the IL-6 ternary complex formation demonstrated that the reduced biological potency of IL-6 D resulted from a decreased ability of the IL-6 D ⅐(sIL-6R) 2 complex to couple with the soluble portion of gp130. These data imply that IL-6induced dimerization of sIL-6R is not the driving force in promoting formation of the hexameric (IL-6⅐IL-6R⅐gp130) 2 complex. A model is presented whereby the trimeric complex of IL-6R, gp130, and IL-6 M forms before the functional hexamer. Due to its increased affinity for the IL-6R but its decreased ability to couple with gp130, we suggest that a stable IL-6 dimer may be an efficient IL-6 antagonist.
Interleukin-6 (IL-6) 1 is a multifunctional cytokine that plays a central role in host defense mechanisms to infection and tissue injury (1)(2)(3). Biological activities ascribed to IL-6 include the ability to induce immunoglobulin secretion in B cells, the maturation of megakaryocytes, acute phase protein synthesis in hepatocytes, support of multipotential colony formation by hemopoietic stem cells, and the growth and differentiation of T cells (1)(2)(3)(4). Dysregulated production of IL-6 is believed to be implicated in the pathogenesis of multiple myeloma, psoriasis, and postmenopausal osteoporosis (3,5,6).
Many of the biological functions of IL-6 overlap with those of other cytokines such as oncostatin M, leukemia inhibitory factor, ciliary neurotrophic factor, cardiotrophin-1, and interleukin-11. These cytokines have been recently found to share a common ␤-chain, gp130, which provides, in part, a rationale for their biological redundancy (18 -22). Whereas homodimerization of gp130 is required for IL-6 signaling (9), leukemia inhibitory factor, oncostatin M, cardiotrophin-1, and ciliary neurotrophic factor have been shown to signal via heterodimerization of the leukemia inhibitory factor receptor and gp130 (19,21,22).
We have demonstrated previously that sIL-6R binds IL-6 monomer (IL-6 M ) with a 1:1 stoichiometry in the low affinity IL-6 M ⅐sIL-6R binary complex (23). We (23) and recently others (24) have shown that the IL-6⅐sIL-6R⅐sgp130 ternary complex is a hexamer comprising two molecules each of IL-6, sIL-6R, and sgp130. The direct interaction of IL-6 with both IL-6R and sgp130 in this complex is suggested by covalent cross-linking studies (25). Contact between the two gp130 molecules in the ternary complex has been inferred from the observation that disulfide linkage between two gp130 molecules occurs upon activation of target cells by IL-6 (9). Both IL-6 oligomerization (25)(26)(27) and IL-6R oligomerization (28) have been implicated in the binding of IL-6 to its receptor, although any role for the direct interaction between the two IL-6 molecules or the two IL-6R molecules in the ternary complex is yet to be confirmed.
Here we investigate the role of IL-6 dimerization, using stable forms of IL-6 M and IL-6 D , in the binding of IL-6 to sIL-6R and in the binding complex of sIL-6R with IL-6 M and IL-6 D to sgp130 and demonstrate that the decreased biological potency of IL-6 D is due to an impaired ability of the IL-6 D ⅐sIL-6R complex to bind sgp130. These data imply that IL-6 D -induced oligomerization of IL-6R does not precede gp130 binding in the formation of the hexameric complex.

EXPERIMENTAL PROCEDURES
Materials-Recombinant human IL-6 was produced in Escherichia coli using the pUC8 vector as a fusion protein with ␤-galactosidase (23,29). IL-6 M was separated from IL-6 D by size-exclusion chromatography using a Sepharose 12 column (300 ϫ 10 mm, inner diameter) equilibrated with phosphate-buffered saline (23). Both IL-6 M and IL-6 D were Ͼ 95% homogeneous and stable for at least two weeks at 4°C. Recombinant sIL-6R and sgp130 were expressed and purified from the conditioned medium of Chinese hamster ovary cells as described previously (23, 29 -31). The concentrations of all recombinant proteins were determined by amino acid composition analysis.
Measurement of Dissociation of IL-6 D into IL-6 M -The time-dependent dissociation of IL-6 D at 37°C was performed at protein concentrations of 50 and 500 g/ml using size-exclusion chromatography as described above.
Measurement of IL-6⅐sIL-6R Interactions-The interactions of IL-6 M and IL-6 D with sIL-6R and of the binary complexes with sgp130 were monitored by surface plasmon resonance (SPR) detection using a BIAcore optical biosensor (Pharmacia Biosensor, Uppsala, Sweden) (32). sIL-6R and sgp130 were immobilized to the sensor surface using procedures described elsewhere (23,29). The sensor surface was regenerated between assays by treatment for 1 min with 4 M MgCl 2 in 10 mM Tris-HCl buffer, pH 7.4 (for immobilized sIL-6R), or by treatment with 1.5 M KSCN in 10 mM Tris-HCl buffer, pH 8.0 (for immobilized sgp130). All reagents were introduced at a flow rate of 5 l/min. Assuming that 1000 response units (RU) corresponds to a surface concentration of 1 ng/mm 2 (33), sIL-6R and sgp130 were immobilized to a surface concentration of 3 and 5.3 ng/mm 2 , respectively. Kinetic analysis was performed utilizing the relationships described by O'Shannessy et al. (34). Equilibrium binding data were fitted assuming either one or two equivalent and independent sites and fitted to the relation: where R P is the SPR plateau response at equilibrium, B 1 and B 2 are the amplitudes of the response for sites 1 and 2, respectively, k AX1 and k AX2 are the association equilibrium constants for sites 1 and 2, respectively, and [S] is the free concentration of analyte (IL-6 M , IL-6 M ⅐sIL-6R complex, IL-6 D , or the IL-6 D ⅐(sIL-6R) 2 complex).
Biological Assays-The biological activities of IL-6 M and IL-6 D were determined by a mitogenic assay using murine hybridoma 7TD1 cells and by fibrinogen induction from human hepatoma HepG2 cells, as described (35).
The STAT3 phosphorylation assay was performed using HepG2 cells grown to confluency in RPMI medium containing 10% fetal calf serum (100-mm plates, Nunc A/S, Roskilde, Denmark). The cells (1 plate/ sample) were serum-starved for 4 h prior to induction with varying concentrations of IL-6 at 37°C. After IL-6 induction, the cells were rinsed with ice-cold phosphate-buffered saline, lysed on ice with 50 mM Tris-HCl, 150 mM NaCl, 1 mM ethylenediaminetetraacetic acid, 1% (v/v) Triton X-100, 0.1 mM Na 3 VO 4 , 1 mM NaF (lysis buffer) containing 200 ϫ 10 3 IU/ml Trasylol, 1 M pepstatin, 1 mM phenylmethylsulfonyl fluoride, 1 mM 1,10-phenanthroline, and centrifuged (Eppendorf centrifuge model 54145) for 15 min at 4°C. The protein concentrations of the supernatants were determined using bovine serum albumin as a standard and adjusted to be equivalent in all samples. The samples were pre-cleared with protein A-Sepharose (Pharmacia) for 30 min on ice prior to immunoprecipitation with phosphotyrosine antibodies (UBI, Lake Placid, NY) for 2 h on ice. The antigen-antibody complexes were recovered with protein A-Sepharose, washed twice with ice-cold lysis buffer containing 20 mg of bovine serum albumin/ml, and washed once with lysis buffer alone.
SDS-PAGE, Western Blotting, and Densitometric Scanning-Precast 4 -20% gels (Novex, San Diego, CA) were utilized for visualization of IL-6 M and IL-6 D by silver staining. For detection of phosphorylated STAT3, the anti-phosphotyrosine immunoprecipitates were analyzed under reducing gel conditions with SeeBlue prestained standards and transferred to polyvinylidene difluoride membrane (Bio-Rad) using a Novaline semidry blotting apparatus (Novex). Tyrosine-phosphorylated STAT3 was visualized by immunoblot analysis using anti-STAT3 antibodies (Transduction Laboratories) followed by peroxidase-conjugated goat anti-mouse immunoglobulin and enhanced chemiluminescence (Amersham Corp.). Densitometric scanning of tyrosine-phosphorylated STAT3 was performed on a model 300 Series Computing Densitometer (Molecular Dynamics, CA).

RESULTS AND DISCUSSION
Stoichiometry of the IL-6 D ⅐sIL-6R Complex-Purified human IL-6 used in this study was a mixture of dimeric and monomeric forms, with relative molecular masses (M r ) of 40,000 and 20,000, respectively, as judged by size-exclusion chromatography (SEC) (Fig. 1A). IL-6 D was not a covalently linked dimer, as judged by SDS-PAGE. Under both nonreducing and reducing SDS-PAGE conditions, IL-6 D migrated at a M r indistinguish-FIG. 1. Size-exclusion chromatography of monomeric and dimeric forms of IL-6 and the corresponding complexes with sIL-6R. The chromatographic conditions are given under "Experimental Procedures." The elution times (in min) are indicated for each protein peak. A, IL-6 (20 g). B, IL-6 M peak (14.34 min) from A (50% reinjected). C, rechromatography of IL-6 D peak (13.01 min) from A (50% reinjected). D, sIL-6R (5 g). E, IL-6 M ⅐sIL-6R complex. F, IL-6 D ⅐sIL-6R complex. The elution times for standard proteins in A (bovine serum albumin dimer, M r ϭ 132,000; bovine serum albumin monomer, M r ϭ 67,000; soybean trypsin inhibitor, M r ϭ 20,000) are indicated by arrows.
able from that of IL-6 M (Fig. 2). These forms are not in rapid equilibrium because both purified IL-6 M and IL-6 D elute in identical positions when they are rechromatographed by SEC (Fig. 1, B and C). Although both IL-6 D and IL-6 M were biologically active, IL-6 D exhibited a 2-3-fold decreased potency (based upon molar equivalents) in both the human hepatoma HepG2 bioassay (Fig. 3A) and the murine hybridoma 7TD1 assay (Fig. 3B). It is interesting to compare these results with previous observations that showed that fibroblast-derived natural IL-6 was a mixture of stable multimeric forms consisting of dimers, trimers, and tetramers but contained very little IL-6 M (26). The higher M r forms were reported to be less active in the murine hybridoma B9 assay but were equipotent in the human Hep3B hepatocyte-stimulating factor assay (26).
In a short term assay (1.5 min of stimulation) measuring the induction of STAT3/APRF phosphorylation (14 -16) in HepG2 cells, IL-6 D was approximately 8-fold less potent than IL-6 M (Fig. 4). Densitometric scanning of the bands corresponding to phosphorylated STAT3 (Fig. 4A) showed 5.2-, 20.3-, and 25.8fold increases over background for 0.31, 1.25, and 5 nM IL-6 M , respectively (Fig. 4B). For the corresponding equimolar concentrations of IL-6 D , the increase over background was 0.6-, 2.7-, and 14.6-fold, respectively (Fig. 4B). The lower differential between the two forms at the highest protein concentration (1.8-fold versus ϳ8-fold at lower protein concentrations) suggests that at the higher concentrations the plateau region of the dose response curve is being approached. The difference in potency between IL-6 D and IL-6 M is clearly more pronounced in the short term assay than in the long term assays on HepG2 and 7TD1 cells. This was explained by a time-dependent dissociation of noncovalently linked IL-6 D into IL-6 M at 37°C, as measured by SEC. Although IL-6 D and IL-6 M are stable at 4°C with no significant interconversion over at least 6 days (data not shown), up to 10 and 41% (molar equivalents) of IL-6 D is dissociated into IL-6 M upon exposure of IL-6 D to 37°C for 20 min and 4 days, respectively. There is, therefore, a close correlation between the amount of IL-6 M generated by the dissociation of IL-6 D in the IL-6 D preparation and the measured activity of IL-6 D in both assays. When the STAT3/APRF assay on HepG2 cells was performed under conditions where IL-6 D did not dissociate into IL-6 M (40 min stimulation on ice), 5 nM IL-6 D in contrast with 5 nM IL-6 M did not significantly stimulate the tyrosine phosphorylation of STAT3 above background (data not shown). Taken together these data suggest that the residual biological activity of IL-6 D (Figs. 3 and 4) is due to IL-6 dimer dissociation into active monomers at 37°C. sIL-6R was monomeric under the conditions employed, the calculated M r of ϳ50,000 observed by SEC (Fig. 1D) being in agreement with previous chromatographic and centrifugation studies (23,29). Complex formation between IL-6 D and sIL-6R was studied by isolation of the complex using SEC. The IL-6 D ⅐sIL-6R binary complex eluted with a retention time of 11.17 min, which corresponded to a M r ϳ130,000 (Fig. 1F). This value correlates well with the theoretical M r of 140,000, which assumes that each IL-6 D binds two molecules of sIL-6R.
The stoichiometry of the IL-6 D /sIL-6R interaction was also investigated by subjecting this complex to molecular weight determination by analytical ultracentrifugation. The M r of IL-6 D was not dependent on protein concentration, and there was a linear dependence of ln c versus r 2 (Fig. 5). The calculated M r for IL-6 D was 44,600 (Table I). A M r of ϳ126,000 was determined for the IL-6 D ⅐sIL-6R complex (Fig. 5 and Table I) using v ϭ 0.704 ml/g, which is the weighted average v of IL-6 and sIL-6R (23,29). This value compares favorably with the predicted M r of ϳ145,000 for the IL-6 D ⅐sIL-6R complex, where IL-6 D binds two molecules of sIL-6R using a M r of 50,200 for sIL-6R (23). Using identical conditions, IL-6 M has been previously shown to bind sIL-6R (Fig. 1E) with 1:1 stoichiometry (23).
Binding of IL-6 D and IL-6 M to Immobilized sIL-6R-The influence of IL-6 dimerization on receptor binding was studied using a biosensor employing SPR detection. Typical sensorgrams for the binding of IL-6 D (5 nM) and IL-6 M (100 nM) to immobilized sIL-6R are shown in Fig. 6A. The measured binding response in both cases was specific, because the response could be inhibited by preincubation with anti-IL-6 neutralizing antibodies (data not shown). IL-6 D bound with increased affinity, the dissociation rate constant being drastically reduced relative to IL-6 M (Fig. 6A). IL-6 M binding was characterized by both rapid association and dissociation phases. The association and dissociation rate constants were independent of IL-6 M concentration with values of 3.8 ϫ 10 5 M Ϫ1 s Ϫ1 and 0.018 s Ϫ1 being determined for k a and k d , respectively. From the ratio of the respective rate constants a K D ϭ 4.7 ϫ 10 Ϫ8 M was determined. This value compares favorably with the K D ϭ 4.4 ϫ 10 Ϫ8 M determined by equilibrium binding measurements (Fig.  6B); the latter measurements were obtained from the dependence of the plateau RU values upon concentration (5-200 nM), assuming a single class of binding sites. Both values are in close correspondence with the K D ϭ 4.2 ϫ 10 Ϫ8 M calculated for the binding of sIL-6R to immobilized IL-6 (29). Likewise, the above K D values concur with the value of 2 ϫ 10 Ϫ8 M for the interaction of IL-6 with sIL-6R in solution, as determined by competition experiments (29).
No kinetic parameters describing the binding of IL-6 D to immobilized sIL-6R have been presented because these data could not be readily fitted to first-order kinetics for single or multiple sites relations (36), which were successfully used to describe binding of IL-6 M . Although this deviation could, in part, be attributable to such things as rebinding in the dissociation phase, the major contribution is clearly due to the bivalence of IL-6 D (see Figs. 1F and 5). Ligand multivalency and, therefore, simultaneous attachment of binding sites to immobilized proteins, has not been considered in the derivation of the above kinetic relations. That IL-6 D is binding via a cross-linking mechanism is reinforced by equilibrium binding data (Fig. 6B). A biphasic Scatchard plot for IL-6 D , as opposed to a linear Scatchard plot for IL-6 M , was observed. Although curvilinear plots can also be representative of multiple classes of binding sites or negative cooperativity, the observed data are typical of that expected for a bivalent ligand. At high concentrations of IL-6 D , the slope of the Scatchard plot approaches that for IL-6 M (Fig. 6B). In the cross-linking model, at high concentrations of ligand, the bivalent ligand only attaches via one contact region, and the affinity approaches that of monovalent ligand. The increased value of the abscissa intercept (Fig. 6B) also suggests that IL-6 D is attaching to immobilized sIL-6R via a single linkage at high concentrations of IL-6 D , where the doubling of the molecular weight of IL-6 D results in a corresponding increase in the maximum RU bound. At low ligand concentrations, multiple attachments can occur, and the affinity of IL-6 D is drastically increased relative to that observed for IL-6 M (Fig. 6B). These studies emphasize the need to immobilize the multivalent species when dealing with an interaction between multivalent and monovalent species in order to avoid multiple attachments complicating the binding analyzes when relating kinetic and equilibrium binding data derived from BIAcore analysis to those derived in solution (36,37). For example, if IL-6 D is immobilized and the binding of sIL-6R is monitored, kinetics similar to those observed for the binding of IL-6 M to immobilized sIL-6R (29) were obtained.
The binding of IL-6 D to the IL-6R on cells has been investigated using covalent cross-linking experiments (25)(26)(27). In addition to a band of M r ϳ120,000 resulting from binding of IL-6 D to the IL-6R, Rose-John et al. (27) also observed a band of M r ϳ200,000 that was postulated to result from a dimeric form of IL-6 cross-linking two receptor molecules. Oligomerization of IL-6 was also observed in the absence of cells (26), which suggests that the interaction of IL-6 D with the IL-6R may result from the binding of pre-existing dimers, such as those characterized in this study, to cellular receptors. D'Alessandro et al. (25) also demonstrated the direct interaction of IL-6 D with cellular receptors, although, in this case, all pre-existing dimers were removed prior to performing cross-linking experiments. A tendency of the IL-6 M ⅐IL-6R complex to oligomerize,   v ) were determined by sedimentation equilibrium (Fig. 5).
b The value of v for IL-6 M and IL-6 D were calculated from the amino acid composition of IL-6 (23). The values of v for IL-6 M ⅐IL-6R were calculated on the basis of zero volume change upon association, using a stoichiometry of 1:1 and 2:1 for the monomer and dimer complexes, respectively. The values used for M and v for the receptor were 0.69 and 50,200, respectively (23). d Predicted molecular weights were calculated using a stoichiometry of 1:1 and 2:1 for the monomer and dimer complexes, respectively. e As reported previously (23).
bringing the two IL-6 molecules in the higher order complex into close spatial proximity, was therefore postulated to explain these results (25). We did not observe any tendency of the sIL-6R to self-associate upon binding IL-6 M (23). This does not rule out the possibility, however, that the presence of the transmembrane domain or other factors such as gp130 facilitates self-association of IL-6R in the cellular environment. Using covalent cross-linking techniques, Stoyan et al. observed the oligomerization of sIL-6R on binding IL-6 (28). However, it was not clear whether there were any pre-existing dimers of IL-6 present in their IL-6 preparation. Consequently, the oligomerization could readily be explained on the basis of the IL-6 D binding two receptor molecules, as we report in this study.
Binding of the IL-6⅐sIL-6R Complex to sgp130 -The ability of gp130 to interact with complexes of IL-6 M ⅐sIL-6R and IL-6 D ⅐sIL-6R was investigated upon immobilizing sgp130 to the sensor surface (Fig. 7, A-C). In all cases, two sequential injections of identical concentrations of the IL-6⅐sIL-6R mixtures were introduced to the sensor surface to ensure that equilibrium was reached. The concentration of IL-6 was kept constant at 200 nM, and the concentration of the IL-6⅐sIL-6R complex was calculated assuming a K D of 20 nM (29). Because both IL-6 D and IL-6 M bind sIL-6R identically in solution, both sites on IL-6 D being equivalent and independent (29), identical concentrations of the respective sIL-6R complexes would be present at equivalent concentrations of IL-6 D and IL-6 M . Increased relative responses were obtained when using IL-6 M ⅐sIL-6R compared with IL-6 D ⅐sIL-6R, suggesting that the IL-6 M ⅐sIL-6R complex couples more effectively with gp130 than the corresponding dimeric species. For example, a plateau response of 900 RU for IL-6 M compared with 350 RU for IL-6 D was obtained when using 144 nM sIL-6R (Fig. 7, compare A and B). The dissociation phases of both were biphasic, with both rapidly and slowly dissociating species being observed. For IL-6 M the more rapidly dissociating phase predominates at higher complex concentrations. Scatchard analysis of the equilibrium binding data reveals that IL-6 M ⅐sIL-6R binding is clearly curvilinear, which is indicative of more than one class of binding site (Fig. 7C). The data for IL-6 M were fitted to two equivalent and independent sites with dissociation constants and amplitude values of k AX1 ϭ 2.7 ϫ 10 7 M Ϫ1 , B 1 ϭ 936 and k AX2 ϭ 8.9 ϫ 10 8 M Ϫ1 , B 2 ϭ 179 for sites 1 and 2, respectively. As for IL-6 M , biphasic dissociation kinetics were observed for the binding of IL-6 D . The Scatchard plot, however, was described by a linear relation, with calculated k AX1 and B 1 values of 3.1 ϫ 10 7 M Ϫ1 and 470, respectively. Failure to observe more than a single site in the equilibrium binding experiment with the immobilized sgp130 probably reflects the weaker binding of the IL-6 D ⅐sIL-6R complex relative to the IL-6 M ⅐sIL-6R complex, the higher affinity site not being significant over the concentration range used in this study (18 -144 nM total IL-6 D ).
Stronger coupling of the IL-6 M ⅐sIL-6R complex to gp130 was also verified using the alternative strategy of immobilizing sIL-6R and sequentially adding IL-6 and sgp130. Whereas strong coupling was observed when utilizing IL-6 M (Fig. 7D), this binding being characterized by slow dissociation kinetics, no coupling of IL-6 D to sgp130 could be detected up to the maximum concentration of 43 nM sgp130 used in this study. In the assay of ternary complex formation using immobilized sgp130 (Fig. 7, A-C) biphasic dissociation kinetics were observed, with rapid and slow dissociating species being measured. A stable ternary complex and corresponding slow dissociation rate is indicated from cellular binding studies (8), and the demonstration that a hexameric high affinity IL-6 receptor complex can be isolated by SEC (23). The studies of Paonessa et al. (24) may help to rationalize these findings. In these studies (24) it was shown that there are two functionally distinct sites for gp130 binding on IL-6 and that by using IL-6 mutated in either one of these gp130-binding sites, it is possible to form a ternary complex consisting of one molecule each of IL-6, sIL-6R, and sgp130 (trimer complex) but to interfere with hexamer formation. Similarly, we recently showed that IL-6 with mutations in one gp130 binding site was able to induce IL-6R-dependent association with gp130 but could not induce the formation of a stable hexameric complex (38). The biphasic dissociation kinetics in Fig. 7 (A-C) are consistent with slow dissociation kinetics expected for the hexamer as opposed to the trimer. Possible explanations for the failure to observe hexamer formation in all cases shown in Fig. 7 include differing orientation of gp130 on the sensor surface and the immobilization of gp130 at its dimerization site. The latter interpretation is consistent with the predominance of hexamer (slow dissociation kinetics) when the alternative strategy of immobilizing sIL-6R to the sensor surface was utilized.
For reasons that are not clear, IL-6 mitogenic assays such as the mouse 7TD1 (Fig. 3B), mouse B9 and human XG-1 assay are more sensitive (EC 50 ϭ 0.05-1 pM) than assays that meas- FIG. 6. Comparison of the kinetics of binding of IL-6 M and IL-6 D to immobilized sIL-6R using SPR detection. A, the sensorgrams correspond to 100 nM IL-6 M and 5 nM IL-6 D , respectively. The arrows correspond to the beginning of dissociation phases where free IL-6 was replaced with HBS buffer containing 0.005% Tween 20. IL-6 M and IL-6 D , in the same buffer, were introduced at flow rates of 5 and 2 l/min, respectively. B, concentration dependence of the binding of IL-6 M (E) and IL-6 D (q) to immobilized sIL-6R expressed in Scatchard format. At the lower ligand concentrations, it was necessary to multiply inject ligand and use lower flow rates to ensure that equilibrium was reached, the attainment of equilibrium being assessed by constancy of signal with respect to time. For example, with 5 nM IL-6 D a flow rate of 1 l/min and injection volume of 50 l was utilized (contact time ϭ 3000 s). For 1 nM IL-6 D , four such injections were required. ure induction of secreted proteins, such as fibrinogen (Fig. 3A), C1 esterase inhibitor from human HepG2 cells, and IgG1 from human CESS cells (EC 50 ϭ 10 -200 pM) (39). The exact molecular mechanisms responsible for this seeming anomaly are not known but presumably involve intracellular signaling machinery coupling to the activated receptor rather than interactions between IL-6 and the extracellular domains of the receptor. We have shown that human IL-6 D is less potent than human IL-6 M in three different biological assays and that the activity of IL-6 D can be attributed to its dissociation into IL-6 M . Taken together, our data therefore suggest that the in vitro interactions of the IL-6 M /sIL-6R and the IL-6 D /sIL-6R complexes with sgp130 reflect receptor activation on the cell surface.
It is not clear what relation IL-6 D has to previously identified oligomeric forms of fibroblast-derived natural human IL-6 (May et al., 1991). Like IL-6 D , oligomers of natural human IL-6 are less active in hybridoma growth factor assays (May et al., 1991). However, unlike IL-6 D they are equipotent in their ability to activate acute phase assays using Hep3B cells (May et al., 1991). Dissociation of IL-6 D to IL-6 M at elevated temperatures is largely responsible for observed biological activity of IL-6 D . It is unclear whether such a dissociation occurs for natural IL-6 oligomers and indeed whether it is concentration-dependent in nature. A concentration-dependent dissociation of natural IL-6 oligomers could be used to rationalize increased activity in the hybridoma growth factor assay (pg/ml range) as opposed to acute phase assays (ng/ml range). For IL-6 D , however, this is not the case where elevation of the temperature seems to drive the dissociation; decreasing the protein concentration below 5 mg/ml has no measurable effect. 2 A greater insight into the relationship between E. coli-derived IL-6 D and fibroblast-derived natural IL-6 oligomers must await an increased understanding of the protein-protein interactions involved in the respective dimeric forms.
IL-6 D binds to immobilized sIL-6R with a greatly increased apparent affinity relative to IL-6 M due to its ability to interact simultaneously with two immobilized receptors on the sensor surface. Such an increased affinity of IL-6 D relative to IL-6 M for the IL-6R would also be expected in cellular binding assays because the lateral mobility of receptors in the plasma membrane should allow their juxtapositioning enabling them to simultaneously bind both sites on IL-6 D . The presence of these dimers in IL-6 preparations explains some but not all reports of IL-6 D binding to sIL-6R in cellular cross-linking studies. Our data argue against a possible mechanism suggested by crosslinking studies whereby IL-6 D induces dimerization of IL-6R, which in turn couples with gp130 to form a hexameric complex. By contrast, the present study supports an alternative model whereby monomeric IL-6 cross-links IL-6R and gp130 to form a trimeric complex, two such complexes further dimerizing to The arrows a and b correspond to the commencement and the end of the first injection of sample, arrow c corresponds to the corresponding points relating to injection of a second sample of the same solution, and arrow d corresponds to the start of the dissociation phase. All solutions were introduced at a flow rate of 1 l/min. C, Scatchard plot describing the equilibrium binding of the binary complexes to immobilized sgp130 (IL-6 M ⅐sIL-6R (E) and IL-6 D ⅐sIL-6R (q)). As in A, the total concentration of IL-6 was kept constant at 200 nM, and the concentration of the IL-6⅐sIL-6R complex was calculated assuming a K D of 20 nM (27). Lines through data points correspond to theoretical relations, calculated assuming, for IL-6 M , k AX1 ϭ 2.7 ϫ 10 7 M Ϫ1 , B 1 ϭ 936; k AX2 ϭ 8.9 ϫ 10 8 M-1; B 2 ϭ 180, which was calculated via nonlinear regression analysis, and, for IL-6 D , k AX1 ϭ 3.1 ϫ 10 7 M-1 and B 2 ϭ 470, which was calculated via a linear relation. D, sIL-6R was immobilized to the sensor surface. IL-6 M (100 nM) and sgp130 (43 nM) were introduced as indicated by the arrows.
form the hexameric species (Fig. 8). It should be stressed that this does not rule out an IL-6-IL-6 contact in the hexamer and may reflect that the IL-6 in the stable dimers are orientated differently to those purported to occur in the hexameric complex (24).
Various groups have demonstrated that IL-6 antagonists can be generated by mutating IL-6 at either of the two gp130 sites. Such antagonists have therapeutic potential in the treatment of IL-6-related disease states such as multiple myeloma. One likely limitation of such antagonists is their reduced affinity for IL-6R relative to native IL-6. Although this can be overcome by introducing further amino acid substitutions in IL-6 to increase its affinity for IL-6R, the studies presented here suggest that by generating stable covalently linked dimeric forms of IL-6, it may be possible to both increase the affinity for the IL-6R and decrease any biological activity due to reduced interactions with gp130. Studies are underway to further develop this finding.