Fluorescence resonance energy transfer reveals interleukin (IL)-1-dependent aggregation of IL-1 type I receptors that correlates with receptor activation.

Fluorescence resonance energy transfer (FRET) was used to investigate whether interleukin-1 (IL-1) causes the aggregation of IL-1 type I receptors (IL-1 RI) at the cell surface. For these experiments, a noncompetitive anti-IL1 RI monoclonal antibody, M5, was labeled separately with a donor probe, fluorescein isothiocyanate, or with an acceptor carbocyanine probe, Cy3. Donor-labeled M5 and acceptor-labeled M5 were simultaneously bound to transfected mouse IL-1 RI on either C-127 mouse mammary carcinoma cells or on Chinese hamster ovary (CHO)-K1 cells, and the ratio of acceptor emission at 590 nm to donor emission at 525 nm (excitation at 488 and 514 nm) was monitored with flow cytometry as an indicator of FRET. Addition of a saturating concentration of human IL-1 alpha at 22 degrees C causes a time-dependent increase in FRET for both cell lines that indicates IL-1-dependent self-association of IL-1 RI. Binding of the IL-1 receptor antagonist at 22 degrees C causes little or no FRET for both cell lines, indicating a correlation between receptor aggregation and the ability of the ligand to stimulate a functional response. When donor-labeled and acceptor-labeled Fab fragments of M5 are used to monitor FRET, IL-1 alpha causes efficient energy transfer in the CHO-K1 cells at 22 degrees C, but not at 4 degrees C. In contrast, IL-1 alpha causes much less FRET at 22 degrees C in C-127 cells when the M5 Fab fragments are used instead of the intact bivalent M5. In a striking parallel, IL-1 alpha-dependent activation of prostaglandin E2 production depends on the bivalent M5 antibody in the C-127 cells, but is independent of this monoclonal antibody in the CHO-K1 cells. These results provide a strong correlation between the ability of IL-1 to cause the aggregation of IL-1 RI and the stimulation of a functional response.

Fluorescence resonance energy transfer (FRET) was used to investigate whether interleukin-1 (IL-1) causes the aggregation of IL-1 type I receptors (IL-1 RI) at the cell surface. For these experiments, a noncompetitive anti-IL1 RI monoclonal antibody, M5, was labeled separately with a donor probe, fluorescein isothiocyanate, or with an acceptor carbocyanine probe, Cy3. Donor-labeled M5 and acceptor-labeled M5 were simultaneously bound to transfected mouse IL-1 RI on either C-127 mouse mammary carcinoma cells or on Chinese hamster ovary (CHO)-K1 cells, and the ratio of acceptor emission at 590 nm to donor emission at 525 nm (excitation at 488 and 514 nm) was monitored with flow cytometry as an indicator of FRET. Addition of a saturating concentration of human IL-1␣ at 22°C causes a time-dependent increase in FRET for both cell lines that indicates IL-1dependent self-association of IL-1 RI. Binding of the IL-1 receptor antagonist at 22°C causes little or no FRET for both cell lines, indicating a correlation between receptor aggregation and the ability of the ligand to stimulate a functional response. When donor-labeled and acceptor-labeled Fab fragments of M5 are used to monitor FRET, IL-1␣ causes efficient energy transfer in the CHO-K1 cells at 22°C, but not at 4°C. In contrast, IL-1␣ causes much less FRET at 22°C in C-127 cells when the M5 Fab fragments are used instead of the intact bivalent M5. In a striking parallel, IL-1␣-dependent activation of prostaglandin E 2 production depends on the bivalent M5 antibody in the C-127 cells, but is independent of this monoclonal antibody in the CHO-K1 cells. These results provide a strong correlation between the ability of IL-1 to cause the aggregation of IL-1 RI and the stimulation of a functional response.
IL-1␣ 1 and ␤ are polypeptide cytokines produced by a variety of cell types in response to injury and infection, and they play central roles in immune and inflammatory responses (1)(2)(3). They share similar biological activities and receptor binding properties, and their three-dimensional structures are closely related (4,5). A receptor antagonist polypeptide, IL-1ra, competes with IL-1␣ and ␤ for binding to receptors but does not stimulate a functional response (6). There are two known receptors for IL-1. IL-1 RI binds IL-1␣ or IL-1␤ to mediate the activation of T cells and hematopoietic cells, and to regulate the synthesis and secretion of acute-phase proteins, prostaglandins, and collagenase (7)(8)(9)(10). This receptor is found on almost all cell types and is often involved in immune and inflammatory responses (11,12). IL-1 RII is found on a more limited number of cell types, including B cells and fibroblasts, and may act as a "decoy" to regulate IL-1 RI-mediated responses (12,13).
IL-1 RI is a single polypeptide chain with an extracellular segment composed of three immunoglobulin-like domains that make up the IL-1 binding site (14). A single hydrophobic transmembrane sequence connects the extracellular region to an intracellular segment of 219 residues that has sequence homology to the Drosophila gene product Toll (15) but provides no clear prediction of the mechanism by which transmembrane signaling is mediated. Recently, Greenfeder et al. (16) have cloned a ϳ66-kDa polypeptide that interacts with IL-1 RI and exhibits significant sequence homology to this receptor polypeptide. Several recent studies have indicated that IL-1 activates a cascade of serine/threonine phosphorylation (17)(18)(19), but the earliest events in signal transduction stimulated by this receptor are still unknown.
A number of different cytokines and growth factor polypeptides have been shown to cause aggregation of their receptors to form dimers and sometimes higher order aggregates. One particular case with structural similarities to IL-1/IL-1 RI is the fibroblast growth factor (FGF) receptor system, in which the binding of FGF causes transphosphorylation of FGF receptors (20) and phosphorylation of downstream substrates on tyrosine residues (21), resulting in a cascade of signal transduction which leads to various physiological responses (22). FGF has a three-dimensional structure that is related to IL-1 ␣ and ␤ (23), and it binds to the external domain of the FGF receptor that contains three immunoglobulin domains, in analogy to IL-1 RI (24). The receptor for stem cell factor, c-kit, contains five immunoglobulin-like domains in its extracellular region, and recent studies have shown that binding of stem cell factor to the first three domains cause self-aggregation of the receptors that involves interactions dependent on the fourth immunoglobulin-like domain (25).
There are only limited data to suggest that IL-1 may cause the association of IL-1 RI with each other or with other polypeptides (14,16). One observation that is consistent with a functional role for ligand-induced receptor aggregation is the ability of the overexpressed mutant IL-1 RI lacking a cytoplasmic segment to inhibit IL-1-dependent activation of PGE 2 production by endogenous IL-1 RI in transfected CHO-K1 cells (26). In an analogous situation, co-transfection of excess inac-tive c-kit receptors with active receptors was shown to inhibit stem cell factor-dependent signaling in a process involving the co-dimerization of active and inactive receptors (27).
The present study used FRET monitored by flow cytometry to investigate whether IL-1 causes the aggregation of IL-1 RI at the cell surface. For these experiments, a noncompetitive, anti-IL1 RI mAb or its Fab fragments were labeled separately with the donor and the acceptor probes and bound simultaneously to transfected mouse IL-1 RI on either C-127 mouse mammary carcinoma cells or on CHO-K1 cells. By monitoring the ratio of acceptor emission to donor emission, we can readily detect sensitized acceptor emission and donor quenching that occurs upon addition of IL-1␣. Our results indicate that IL-1 binding leads to time-dependent aggregation of IL-1 RI, and that this aggregation is likely to play an important role in IL-1-dependent signal transduction.
Preparation of Fluorescent M5 mAb Derivatives-The FITC derivative of M5 mAb was prepared as follows: 25 mM FITC in Me 2 SO added to M5 mAb (1 mg/ml) in PBS/EDTA, pH 7.7 (10 mM sodium phosphate, 100 mM NaCl, 1 mM EDTA) to give a molar ratio of 100:1 FITC:M5 mAb. This reaction solution was incubated at room temperature in the dark for 36 h. The sample was then microcentrifuged at 9,000 ϫ g to remove trace amounts of aggregated protein or dye, and the supernatant was exhaustively dialyzed in PBS/EDTA at pH 7.4 to remove unconjugated FITC. Molar ratios of coupling were estimated to be between 5.6 and 6.2 to 1 FITC:M5 in four preparations as based on extinction coefficients of 74,000 M Ϫ1 cm Ϫ1 (495 nm) for FITC, 25,000 M Ϫ1 cm Ϫ1 (280 nm) for FITC, and 210,000 M Ϫ1 cm Ϫ1 (280 nm) for M5 mAb.
The Cy3 derivative of M5 was prepared as follows: M5 mAb (1 mg/ml) was dialyzed overnight in borate-buffered saline (BBS: 0.16 M sodium chloride, 0.2 M sodium borate, pH 9.1). One vial of the Cy3 dye was dissolved in 650 l of BBS, and 160 l of this were added to 0.2 ml of the mAb solution and incubated at room temperature in the dark for 45 min. Glycine was added at a final concentration of 8 mM to quench the reaction, and the sample was then microcentrifuged and exhaustively dialyzed in PBS/EDTA, pH 7.4, to remove unconjugated dye. Molar ratios of coupling were estimated to be 5.9 or 8.8 to 1 Cy3:M5 in two preparations as based on extinction coefficients of 130,000 M Ϫ1 cm Ϫ1 (552 nm) for Cy3, 6,500 M Ϫ1 cm Ϫ1 (280 nm) for Cy3, and 210,000 M Ϫ1 cm Ϫ1 (280 nm) for M5 mAb.
Preparation of Labeled M5 Fab Fragments-FITC-M5 or Cy3-M5 were dialyzed overnight in PBS/EDTA at pH 8.0, then dithiothreitol was added to a final concentration of 1 mM. Papain, 1 mg/ml in PBS/ EDTA, pH 8.0, was activated by incubation at 37°C for 10 min in the presence of 1 mM dithiothreitol, then added to the mAb solution at 2.5% (w/w) and incubated in the dark at 37°C for 3.5 h. N-Ethylmaleimide was then added to a final concentration of 3 mM to alkylate the dithiothreitol and inactivate the papain. After an additional 60 min at 37°C, the solution was dialyzed overnight in PBS/EDTA, pH 8.0. Gel permeation chromatography was performed with a Superose 12 column (Phar-macia Biotech Inc.) connected to a Waters high performance liquid chromatography system equipped with a 280-nm detector. Molecular weight standards (Bio-Rad), and unlabeled M5 mAbs were analyzed under the same conditions to provide calibration standards. SDS-polyacrylamide gel electrophoresis was also performed on the modified Fab fragments. No intact FITC-M5 was detected by either of these criteria. Gel permeation chromatography was used to remove a small amount of Cy3-M5 digestion products larger than Fab fragments. In some preparations, M5 was digested with papain prior to modification with Cy3. Gel permeation chromatography, and SDS-polyacrylamide gel electrophoresis showed no significant contamination of these Cy3-M5 Fab fragments with intact antibody or larger fragments.
Binding of Fluorescently Labeled M5 and M5 Fab Fragments to Cells-Binding experiments were carried out by incubating cells (3 ϫ 10 6 cells/ml) with different concentrations of fluorescent M5 or Fab fragments (0.5 nM to 70 nM) for 50 min at 22°C. A Coulter Epics Profile flow cytometer (Coulter Electronics, Hialeah, FL) was used to determine the mean fluorescence intensity on cells (as described below). Nonspecific binding and cell autofluorescence were assessed by addition of fluorescently labeled ligands after blocking the receptors with a 20-fold excess of unlabeled M5 mAb. Mean fluorescence intensity of these control cells was usually less than 20% of the total fluorescence in specifically labeled cells. Most of this background signal is due to cell autofluorescence.
FRET Measurements-A Coulter Epics Profile flow cytometer was used to monitor the fluorescence of M5 derivatives bound to IL-1 RI on transfected CHO-K1 or C-127 cells. Excitation was provided by an air-cooled argon ion laser with spectral lines at 458, 488, and 514 nm (the intensities at 458 or 514 nm are each about 20% that of 488-nm line). Emission was measured after passing through a 457-515-nm laser blocking glass interference filter, followed by a 550-nm dichroic short pass glass filter. The transmitted light traveled through a 525-nm bandpass glass filter and was detected as 1 (FITC fluorescence). The reflected light traveled through a 590-nm long pass glass filter and was detected as 2 (largely Cy3 fluorescence).
Cells were harvested and resuspended at 3 ϫ 10 6 cells/ml in HBS as described above, and equimolar donor-labeled and acceptor-labeled M5 mAb or M5 Fab fragments were combined and added to the cell suspension. After 50 min at 22°C, the mixture was divided into two or more samples, and labeled cells were monitored in the flow cytometer as a function of time. After monitoring the cell-associated fluorescence to establish a baseline, either IL-1 or IL-1ra was added to one sample, and no addition was made to a control sample. Other samples routinely measured in parallel were cells incubated with donor-and acceptorlabeled M5 (or Fab) after preincubation with 20-fold excess unlabeled M5 (nonspecific control), and cells incubated with donor-labeled M5 only.
FRET Data Analysis-IL-1-dependent FRET between donor and acceptor M5 derivatives on the dually labeled samples was monitored as an increase in the ratio of (sensitized) acceptor emission to (quenched) donor emission, measured as 1 FITC / 1 total . These fluorescence values were determined from the following equations: where 1 and 2 are the flow cytometry measurements described above. total refers to measurements with the dually labeled samples, and the bkgd values refer to measurements with the nonspecific control samples. The second term in Equation 2 corrects for FITC fluorescence that is observed at the 2 wavelengths in the dually labeled sample; in this equation the ratio 2 FITC / 1 FITC was determined from the sample labeled with donor only. Equation 1 does not require a similar correction because no significant Cy3 fluorescence is observed at the 1 wavelengths (data not shown).
For most experiments the acceptor to donor fluorescence ratio, 2 Cy3 / 1 FITC , observed with the dually labeled sample was normalized by dividing this value for the liganded (IL-1 or ILra) cells by the value for the nonliganded cells at each parallel time point (see Fig. 1).
Determination of PGE 2 Release-Cells were plated at a density of 1.25 ϫ 10 5 cells/2 ml/well in 6-well tissue culture plates in growth medium. Two days later, when the cells became confluent, the medium was removed, and cells were washed twice with HBS. Then cells were incubated with or without saturating amounts of M5 mAb (15 nM) or M5 Fab fragment (30 nM) for 1 h at 37°C before addition of IL-1␣, and the incubation was continued for 3 h at 37°C. Aliquots of supernatants (100 l) were collected and assayed for PGE 2 release by a radioimmunoassay using 125 I-PGE 2 , anti-PGE 2 , and a standard curve established with synthetic PGE 2 (Dupont NEN).

RESULTS
The rat mAb M5 has been previously shown to bind to murine IL-1 RI at a site in the extracellular region that is distinct from the IL-1 binding site (29). This mAb appears to cause the dimerization of these receptors at the cell surface (32), but fails to cause any significant cellular response, and does not interfere with the ability of IL-1␣ to stimulate these responses in EL-4 cells (29). In initial experiments using flow cytometry, we established that the FITC and Cy3 derivatives of M5 mAb bind to transfected murine IL-1 RI on CHO-K1 cells and on C-127 cells with the properties expected from previous results with 125 I-derivative of M5 mAb (29) (data not shown). In addition, fluorescence microscopy and steady state fluorescence spectroscopy were used to look for endocytosis-mediated quenching due to acidification in endosomes, and we established that Ͼ95% of IL-1 RI labeled by these derivatives remains at the surface for at least 1 h at 37°C in the presence or absence of IL-1 ligands (data not shown). Because of these properties, we were able to bind donor-and acceptor-labeled M5 mAb simultaneously to CHO-K1 cells to monitor time-dependent changes in the ratio of acceptor emission to donor emission ( 2 Cy3 / 1 FITC ) as a sensitive indicator of FRET at the cell surface. In this experiment, IL-1-dependent aggregation of its receptors might be expected to bring donor and acceptor probes into closer proximity that could be monitored with FRET. This process is revealed by an increase in sensitized acceptor emission and a concomitant decrease in donor emission (33).
As shown in Fig. 1A (छ), addition of a saturating amount of IL-1␣ to CHO-mu1c cells prelabeled with equimolar amounts of FITC-M5 and Cy3-M5 results in a time-dependent increase in the ratio of 2 Cy3 / 1 FITC . In a control sample, labeled cells monitored during the same time period in the absence of IL-1␣ show a small increase in this ratio over time that generally exhibits a constant slope. This upward drift in acceptor/donor fluorescence ratio in the absence of added ligands is seen to varying extents in different experiments (see below), and the time-dependent changes due to ligand addition are therefore represented as shown in Fig. 1B (ࡗ) as normalized fluorescence ratios by dividing for each time point the measured ratio for the samples with added ligands (छ) by the corresponding ratio for the control sample (Ⅺ). As seen in Fig. 1B, the timedependent increase in the normalized ratio becomes maximal by about 60 min following the addition of IL-1␣, with a halftime of about 20 min. As summarized in Table I,  Also shown in Fig. 1 (E, q) is the effect of addition of IL-1ra to the same labeled cells. In this case, a much smaller increase in the normalized fluorescence ratio is observed. The difference between the fluorescence ratio for IL-1ra-treated cells and control cells (Fig. 1A) is probably not statistically significant at any particular time point, but similar small increases over time were observed in three out of three experiments with this ligand (Table I, line 2). These initial observations indicated that, in the presence of M5 mAb, IL-1␣ causes a time-dependent increase in FRET, and that this effect is much smaller with the functionally inactive IL-1ra. In support of these conclusions, control experiments in which cells were labeled singly with FITC-M5 or Cy3-M5 showed no consistent change in either FITC emission or Cy3 emission in response to IL-1␣ (data not shown). Thus, the time-dependent increase in the normalized fluorescence ratio due to the addition of IL-1␣ (Fig.  1B) represents a significant amount of FRET that indicates IL-1-dependent co-aggregation of donor and acceptor-labeled IL-1 RI which correlates with the stimulatory capability of this ligand compared to IL-1ra.
Because of the potential for receptor dimerization by the M5 mAb, labeled Fab fragments of M5 were prepared and used to investigate whether IL-1␣-dependent aggregation detected by FRET was influenced by M5-mediated dimerization. Fig. 2 (Ⅺ) shows that the addition of IL-1␣ causes a time-dependent increase in the normalized fluorescence ratio for CHO-mu1c cells labeled with FITC-M5 Fab and Cy3-M5 Fab. Similar to the results with M5-labeled cells in Fig. 1, the time course of the increase in FRET is relatively slow, reaching a maximal value by 80 min, with half-time of ϳ40 min. The value for the maximum increase in the normalized ratio of acceptor emission/ donor emission in this experiment (ϳ1.5) is greater than the value observed with the M5 derivatives in Fig. 1, suggesting that the extent of aggregation observed with the M5-Fab fragments is at least as great as that observed with the bivalent M5 labels. The maximal values observed for the normalized fluorescence ratios due to IL-1␣-dependent FRET were generally found to be very reproducible for a particular set of donor-and acceptor-labeled M5 derivatives, but some variation in this maximal value is observed with different preparations of these derivatives, and with different ratios of donor-labeled and acceptor-labeled M5 derivatives (data not shown). In some experiments we examined the IL-1␣ dose-dependence of FRET using donor and acceptor-labeled Fab fragments of M5. A concentration of 0.5 nM IL-1␣ is sufficient to occupy ϳ50% of the IL-1 RI (data not shown), and this concentration results in a maximal value of FRET that is ϳ80 -90% of that observed with saturat- Also shown in Fig. 2 is the time course for the normalized fluorescence ratio of CHO-mu1c cells labeled with M5 Fab, following the addition of IL-1ra (छ) in the same experiment. In this situation, no significant FRET was detected due to IL-1ra addition, and similar observations were made in two other experiments (Table I, line 4). These results suggest that the small amount of FRET in response to IL-1ra on CHO-K1 cells labeled with bivalent M5 (Fig. 1) may depend on the ability of M5 to dimerize IL-1 RI which could facilitate a small amount of further aggregation by IL-1ra.
Monitoring the ratio of acceptor emission to donor emission as shown in Figs. 1 and 2 using equimolar amounts of donor and acceptor-labeled M5 or its Fab fragments provided the most sensitive and reliable indication of FRET in these experiments, but this method of analysis does not readily yield the efficiency of energy transfer since the amount of sensitized acceptor emission is dependent on several parameters that are difficult to measure directly in this situation (33). In attempts to quantify the energy transfer efficiency, several experiments were carried out in which cells were labeled with a mixture of FITC-M5 Fab and Cy3-M5 Fab in a molar ratio of 0.6:1, to maximize detection of energy transfer by donor quenching. With this ratio, the donor signal is reduced, but there is an increased probability that the donor-labeled Fab will be adjacent to an acceptor-labeled IL-1 RI during IL-1 dependent aggregation. In these experiments, addition of IL-1␣ caused a maximum quenching of 10 and 11% of the donor emission in two separate experiments after 70 min of incubation (data not shown). Using the simplest model of FRET between single donor-acceptor pairs (34), in one limit we assume that all of the donor FITC probes bound to aggregated IL-1 RI are within 10 Å of a Cy3 acceptor probe in these measurements. In this case the estimated R 0 value of 55 Å for the FITC/Cy3 donor-acceptor pair predicts that E Ϸ 1.0 for these donors, and thus at least 10% of these are co-aggregated with acceptor-labeled IL-1 RI. In the other limit, all of the labeled IL-1 RI donor are assumed to be adjacent to acceptor-labeled IL-1 RI due to complete aggregation, so that the value of R 0 estimated and the FRET efficiency measured lead to an average of ϳ70 Å between donors and acceptors bound to aggregated IL-1 RI.
In an effort to understand the molecular basis for IL-1-dependent FRET, we investigated the temperature dependence of this process. Fig. 3 (Ⅺ) shows an experiment in which CHO-K1 cells were initially labeled with FITC-M5 Fab and Cy3-M5 Fab fragments at 4°C, then the acceptor/donor fluorescence ratio was monitored before and after the addition of IL-1␣ at 4°C. Under these conditions, no significant change in the normalized fluorescence ratio was detected following addition of IL-1␣ for a period of 80 min. In several other experiments, a small amount of FRET could be detected following extended incubation at 4°C, but this was always substantially less than the FRET observed with a parallel sample at 22°C (data not shown). Following the incubation with IL-1 at 4°C, the temperature of the cells was raised to 22°C and a substantial increase in the normalized fluorescence ratio was observed (Fig. 3, Ⅺ). The maximal value of this ratio occurred after 50 min of incubation at 22°C, and was similar in magnitude to that obtained with a separate sample of the labeled cells that were treated with IL-1␣ at 22°C from the outset of the experiment (Fig. 3, E). These latter cells were cooled to 4°C after 80 min of incubation at 22°C, and we continued to monitor the fluorescence ratio for a additional 80 min. As seen in Fig. 3 (E), only a small decline in the normalized fluorescence ratio was observed during this extended time period at 4°C. These results are representative of three separate experiments and indicate that the IL-1-dependent aggregation process detected by FRET is highly temperature dependent, but, once it has occurred, it remains stable during incubation of the cells at the nonpermissive temperature. In several other experiments, we  found that IL-1␣-dependent FRET occurs at 37°C to a similar extent as at 22°C, and the time course for this process is similar (data not shown).
To determine whether interaction of the cytoplasmic segment of IL-1 RI with each other or with other cellular components are necessary for IL-1-dependent aggregation detected by FRET, we carried out energy transfer experiments with a mutant that is lacking the C-terminal 194 residues out of 219 in the cytoplasmic segment (26). CHO-extn cells containing these mutant receptors were initially labeled with donor-and acceptor-labeled M5 mAb, but we found that in this situation, there was a large time-dependent increase in the ratio acceptor emission to donor emission even in the absence of IL-1␣, suggesting that the bivalent M5 mAb might be causing efficient aggregation of these mutant receptors (data not shown). When labeled M5 Fab fragments were used instead of the intact M5 mAb, the ratio of acceptor emission to donor emission was found to be nearly constant in the absence of IL-1 (data not shown). Fig. 4 (छ) shows the results from an experiment in which IL-1␣ caused a time-dependent increase in the normalized fluorescence ratio that is somewhat less than that for the wild-type receptor (Ⅺ) measured in the same experiment. In four separate experiments, we observed a wider variation in the amount of energy transfer with the truncated receptor than with the wild-type receptor, but in all four experiments some energy transfer was observed with this mutant. These results suggest that the cytoplasmic segment of IL-1 RI is not essential for IL-1-dependent aggregation, but that it may regulate the aggregation process.
To relate the FRET results to a functional response mediated by IL-1 RI, we measured the production of PGE 2 by the CHO-mu1c cells in response to IL-1␣. As shown in Fig. 5A, IL-1␣ stimulates the production of PGE 2 by about 10-fold over that in the unstimulated cells, and neither M5 Fab fragments nor the bivalent M5 mAb had any significant effect on PGE 2 secretion in the presence or absence of IL-1␣. These results are consistent with previous measurements obtained with EL-4 cells in the presence and absence of M5 mAb (29). We also examined the PGE 2 secretion response in a separate IL-1 RI transfected cell line, C-127 mouse mammary cells, which express substan-tially more receptors on their cell surface than the CHO-mu1c cells (data not shown). As shown in Fig. 5B, these cells exhibit virtually no response to IL-1␣ by itself or in the presence of the Fab fragment of M5, but they show a significant response to IL-1␣ in the presence of the bivalent M5 mAb. As with the CHO-K1 cells, M5 in the absence of IL-1␣ did not stimulate PGE 2 production.
These results suggest that M5 mAb can facilitate a functional response to IL-1␣ in the C-127 cells. In order to investigate whether this facilitation is related to an effect on IL-1-dependent receptor aggregation, we monitored FRET in the C127 cells as described above for the CHO-K1 cells. As shown in Fig.  6A (Ⅺ), C-127 cells labeled with M5 mAb exhibit a time-dependent increase in the normalized fluorescence ratio in response to IL-1␣ that is generally smaller in magnitude than that observed with the CHO-K1 cells, but which occurs on a somewhat faster time scale. As for the CHO-mu1c cells, IL-1ra causes a much smaller amount of FRET under these conditions (छ), even though it binds and occupies most of the receptors (data not shown). In a separate experiment, we compared the ability of IL-1␣ to cause a time-dependent increase in FRET in C-127 cells labeled with either bivalent M5 derivatives or with M5 Fab fragments. As shown in Fig. 6B, IL-1␣ causes little or no increase in the normalized fluorescence ratio for cells labeled with the monovalent Fab fragments (Ⅺ), even though the expected amount of FRET is observed with the cells labeled with bivalent M5 (E). This clear difference was observed in two separate experiments and is in striking contrast to the results with CHO-mu1c cells described above, in which cells labeled with the Fab fragments showed at least as much IL-1␣-dependent FRET as the same cells labeled with the bivalent M5 derivatives. These M5 valency-dependent differences in IL-1dependent FRET between the two cell lines parallel the differential requirements of these cell lines for IL-1␣ stimulated PGE 2 production in the presence and absence of M5 mAb. The results, taken together, provide a strong correlation between the ability to detect IL-1␣-dependent aggregation of receptors by FRET and the ability of IL-1␣ to stimulate a functional response.

DISCUSSION
The molecular mechanism by which IL-1 binding to IL-1 RI causes transmembrane signaling that leads to the activation of pro-inflammatory cellular responses has been difficult to ascertain (14). Recent studies indicate that IL-1, like tumor necrosis factor-␣, activates a stress-sensitive cascade of mitogen-activated protein-kinase-related enzymes (35,36), but the earliest events that follow IL-1 binding are largely unknown. Other recent results indicate that IL-1 binding stimulates serine/ threonine phosphorylation of a IL-1 RI-associated 65-kDa substrate (37), and this phosphorylation may play an important role in the initiation of the kinase cascade. Our present results indicate that binding of IL-1␣ causes co-aggregation of IL-1 RI at the cell surface, and that this aggregation process is highly correlated with the activation of at least one functional response, the production of PGE 2 .
Using flow cytometric FRET we have established a sensitive and straightforward method to detect IL-1-dependent aggregation of IL-1 RI labeled with monoclonal anti-IL-1 RI or their Fab fragments. This method can be applied to other cytokine or growth factor receptor system, provided that a noncompetitive mAb specific for the receptor of interest is available. As demonstrated by our results, this method can readily detect ligand-dependent receptor aggregation when fewer than 10 4 receptors/ cell are present, as for the transfected CHO cells in some of our experiments (data not shown). These experiments were carried out with a simple analytical flow cytometer that simultaneously excites both donor (488 nm) and acceptor (514 nm, at lower intensity), and the detection of changes in the ratio of (sensitized) acceptor emission to (quenched) donor emission provides a sensitive means of detecting small changes in FRET that are readily observed upon addition of ligand.
The time course of FRET that we detect in response to IL-1␣ binding is slow relative to the time course of IL-1␣ binding. Under the conditions of our FRET experiments, FITC-IL-1 saturates the IL-1 RI receptors and attains a steady state within about one minute of addition (data not shown). In contrast, the maximum amount of receptor aggregation detected by FRET typically requires ϳ60 min for the CHO-mu1c cells at 22°C, with a half-time of 20 -30 min in most experiments. This indicates that there is a rate-limiting step subsequent to binding that is necessary for receptor aggregation to occur. In the presence of bivalent M5, the IL-1-dependent aggregation process is significantly faster in the C-127 cells, suggesting that the rate-limiting step is sensitive to some difference between these two cell lines. This rate-determining difference is unlikely to be lateral mobility, as IL-1 RI actually diffuses faster in CHO-K1 cells than in C-127 cells, as measured by fluorescence photo- bleaching recovery using labeled M5. 2 It is possible that differences in the stoichiometry of IL-1 RI and the newly discovered 66-kDa receptor-associated polypeptide (16) might affect the kinetics of IL-1 RI aggregation.
Our measurements do not distinguish between the formation of receptor dimers and larger aggregates due to IL-1␣ binding, but under conditions of maximal FRET, IL-1-dependent formation of patches of aggregated IL-1 RI are not detectable by confocal fluorescence microscopy, suggesting that aggregation is limited to substantially less than 1000 receptors per aggregate (data not shown). The strong temperature dependence for aggregation that we observe suggests that the rate-limiting step in the aggregation process has an activation energy barrier with a large temperature coefficient. It is reminiscent of the temperature dependencies reported for epidermal growth factor-dependent epidermal growth factor receptor aggregation (38) and decreased rotational diffusion (39), both observed in plasma membrane preparations. Although the molecular basis for this temperature dependence remains to be determined, it provides a useful experimental strategy for investigating the relationship between IL-1␣-dependent aggregation and other changes in IL-1 RI that can be monitored by physical and biochemical methods.
Our FRET measurements provide direct evidence for IL-1␣dependent association of IL-1 RI with each other and allow us to examine whether this process is related to the initiation of signaling by this ligand. As summarized in Table I, the ability to detect a functional response as represented by PGE 2 secretion that is elicited for a particular combination of cells, ligand, and mAb is highly correlated with the ability to detect a substantial amount of FRET. Minimal aggregation detected by FRET in response to IL-1ra or in response to IL-1␣ on C-127 cells in the absence of bivalent M5 is apparently insufficient to cause productive signaling, while larger amounts of FRET (Table I, ϩϩ 3 ϩϩϩϩϩ) detectable at 22°C correlate with productive signaling at 37°C. The ability to detect substantial FRET with the mutant IL-1 RI that lacks most of the cytoplasmic segment of this receptor indicates that aggregation is not a consequence of signals generated by the liganded receptor, since this mutant fails to cause detectable signaling (26). This result further indicates that receptor-receptor interactions, if they occur during aggregation detected by FRET, are probably mediated by the extracellular or transmembrane segments of IL-1 RI, although we cannot rule out the possibility that interactions involving the cytoplasmic segment play some role in the aggregation process with wild-type receptors. It is also formally possible that receptor-bound IL-1␣ could directly interact with each other to mediate the aggregation process, but there is currently no evidence for self-aggregation of IL-1␣.
It is notable that aggregation as detected by FRET occurs as readily in response to IL-1␣ at 22°C as it does at 37°C, although stimulated PGE 2 production is detectable only at the higher temperature (data not shown). This indicates that other biochemical processes occurring subsequent to IL-1-dependent receptor aggregation have a different temperature dependence and are also necessary for this functional response. The difference in the magnitude of stimulated PGE 2 production (and FRET) between the CHO-mu1c cells and the C-127 cells further indicates that other biochemical events and/or components are involved and can determine the extent of the functional response. These results, taken together, indicate that IL-1-dependent aggregation of IL-1 RI is probably a necessary but, by itself, an insufficient step in the sequence of events leading to IL-1 RI-mediated signaling. The molecular mechanism by which IL-1 mediates the aggregation of its receptor remains to be determined, but the FRET method we have developed should provide a means by which to further understand the structural basis and biochemical consequences of this process.