RelA Control of IκBα Phosphorylation A POSITIVE FEEDBACK LOOP FOR HIGH AFFINITY NF-κB COMPLEXES

NF-κB-IκB complex formation regulates the level and specificity of NF-κB activity. Quantitative analyses showed that RelA-NF-κB-induced IκBα binding is regulated through inhibitor retention and phosphorylation. RelA caused an increase in IκBα phosphorylation and in degradation, which was enhanced monotonically with inhibitor concentration. In vivo analysis demonstrated the RelA-induced IκBα/RelA interactions to be specific, saturable, and phosphorylation-dependent. In addition, it showed that phosphorylation regulates both the level and affinity of the complexes and demonstrated an increased average affinity to coincide with reduction in the level of complexes during cytokine-induced pathway activation. The data show that RelA regulation of NF-κB-IκBα complex formation is IκBα phosphorylation-dependent and that IκBα/NF-κB binding is dynamic and determined by concentration of the subunits. In addition, they suggest that regulation of both complex levels and affinities through phosphorylation, with effects on the system steady state, participate in selective activation of the NF-κB pathway.

NF-B transcription factors are central to inflammatory and immune responses (1)(2)(3). They are composed of homo-and heterodimers of a family of at least five related subunits, characterized by a ϳ300-amino acid Rel homology domain (4). Activation is controlled by IB (inhibitor of NF-B) proteins, the degradation of which is a critical step in activation of NF-B (5). The IB family includes seven proteins containing multiple ankyrin-like repeats, which mediate interaction with the Rel homology regions of NF-B dimers. IB␣, the most extensively studied protein in this family, interacts primarily with p50/ RelA and p50/c-Rel heterodimers but also with c-Rel and RelA homodimers (4,6). IB␣ inhibits NF-B activation by masking its nuclear localization signals and thereby sequestering the two proteins in the form of complexes in the cytoplasm (7).
Induction of the pathway is mediated by a significant number of stimuli, including cytokines, lipopolysaccharides, cellular stress, and cell adhesion (4,8,9). During activation, NF-B is transported into the nucleus and activates transcription of genes bearing cognate binding motifs (4). Transport is made possible by dissociation of complexes following phosphorylation of IB on serines 32 and 36 by IB kinases, ubiquitination, and proteosome degradation (10 -13). The significance of IB/ NF-B interactions as a regulatory step in pathway activation has been extensively documented (6, 14 -16).
Here we report the use of the fusion protein IB␣EGFP 1 to examine regulation of the function of IB␣ by NF-B. RelA regulation of IB␣ and of IB␣/NF-B interactions were analyzed in single living cells, using confocal microscopy, showing that all aspects of IB␣ function are critically dependent on that of NF-B and that phosphorylation of the inhibitor constitutes the basis for interaction and selection during pathway activation.

EXPERIMENTAL PROCEDURES
Plasmids-The plasmid pIB␣EGFP encodes a red-shifted variant of green fluorescent protein fused to the carboxyl terminus of IB␣, constructed by cloning the IB␣ cDNA from pIB␣ ctag (kind gift from Prof. Ronald Hay) (17) digested with BamHI plus EcoRI into pEGFP-N2 (Clontech) digested with BglII plus EcoRI and in-filling of a SmaI plus EcoRI digest for alignment of the reading frame. Plasmid pIB␣EYFP was constructed by subcloning the same IB␣ fragment into pEYFP-N1 (Clontech), digested with BglII plus EcoRI as described (18). Plasmid pECFPrelA was constructed by cloning the RelA cDNA from pBluescript-relA (19), digested with HindIII plus BamHI, into similarly prepared pECFP-C1 (Clontech). Plasmid pECFPp50 was constructed by cloning the p50 cDNA from pRSV-NFB1(p50) (19) digested with Hin-dIII plus BglII into pECFP-C1 (Clontech) and in-filling of the XhoI site. Site-directed mutagenesis of pIB␣EGFP (Ser 3 Ala at positions 32 and 36) was performed using the Muta-Gene Phagemid kit (Bio-Rad). Mutagenesis of pECFPrelA to obtain nonfluorescent fusion proteins was performed by deletion of residues 65-67 (Thr, Trp, and Gly) within the ECFP N terminus, critical for fluorescence, using the QuikChange TM site-directed mutagenesis kit (Stratagene). Oligonucleotides were synthesized from Life Technologies. All constructs were confirmed by sequence analysis.
Tissue Culture and Transfection-Human gingival fibroblasts (HGFQ1 and HGFQ2, transfer 10 -19) and HeLa cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum (Invitrogen) and 5 mM sodium pyruvate, 100 g/ml penicillin and streptomycin and kept at 37°C in a 95% O 2 , 5% CO 2 atmosphere. Cells were plated 24 h prior to transfection and transiently transfected using calcium phosphate co-precipitation with glycerol shock (60 s, 15% glycerol in phosphate-buffered saline) 4 h after transfection, as described previously (20,21). Control experiments demonstrated that transfection of mediators of the NF-B pathway, under these conditions, had no effect on endogenous cytokine production, which was significant for pathway activation (20 -22). Stimulation with IL-1 (1 nM) kindly provided by Dr. Steve Poole (National Institute of Biological Standards and Control) was carried out 24 h after transfection, for various times, as indicated.
Confocal Microscopy-EGFP fusion proteins were visualized using a confocal laser-scanning microscope (Molecular Dynamics, CLSM 2010) fitted with a 37°C stage incubator and coupled to a Nikon Diaphot 300 microscope and a Silicon Graphics work station. Laser power was set to 10 mW, band selection to 488 nm, photo multiplier tube voltage to 750, and varying laser attenuation to maintain pixel density below 200, within the linear range of the instrument. Emission scans were done with a ϫ 60 Plan Apo oil immersion objective (numerical aperture of 1.4) and a 50-m aperture generating an optical section of 0.54 m and using a 530-nm band pass filter. To quantitate the cytoplasmic and the nuclear fusion protein level, transfected cells were scanned horizontally through the nucleus, and images were analyzed using NIH Image. Relative fluorescence was calculated by measuring mean intensity of representative areas of nucleus or cytoplasm and normalized by dividing by the attenuation and further by 2.41 for consistency with previous results obtained with a photo multiplier tube of 666 V (20).
Western Analysis-The relative levels of endogenous and exogenous IB␣ were determined by Western analyses as described (20) using cells transfected with ctag-containing, or fluocrome-tagged (pEGFP, pEYFP, pECFP, pIB␣EYFP, or pECFPrelA) NF-B signaling components to make it possible to distinguish endogenous and exogenous proteins. Cell extracts were prepared by lysis in 1ϫ Laemmli sample buffer (23), resolved on 12.5% SDS-PAGE, and transferred to polyvinylidene difluoride membrane (Amersham Biosciences) with ␤-actin used as an internal control. Total and phosphorylated IB␣ were detected using primary rabbit polyclonal anti-IB␣ and anti-phospho-IB␣ (1:5,000; BioLab), ECFPrelA using goat anti-RelA IgG (1:10,000; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), EGFP and its variants using rabbit anti-GFP IgG (Santa Cruz Biotechnology), and ␤-actin using mouse anti-actin IgG (Sigma), followed by horseradish peroxidase-conjugated secondary antibodies (1:2000), and visualized by ECL (Amersham Biosciences) according to the manufacturer's instructions. Quantitation of exogenous and endogenous expression levels of the various IB␣ and RelA proteins was done of cells transfected with constructs containing the untagged inserts and those transfected with pIB␣EGFP or pEGFPrelA for confocal microscopy and with pIB␣EYFP or pECFPrelA for fluorescence resonance energy transfer (FRET) analysis as well as with empty vectors (pEGFP, pEYFP, and pECFP). Band intensity was compared with a range of concentrations of purified EGFP, and data were analyzed using NIH Image. These experiments validated the comparative use of these various constructs and showed that expression levels of untagged proteins RelA and p50 following transfection, relative to levels of the corresponding fluorescence-tagged fusion protein were 80 Ϯ 6 and 120 Ϯ 15%, respectively. In addition, they demonstrated that the tagged and untagged NF-B fusions had the same effects on IL-1-induced IB␣ turnover.
Immunofluorescence-Immunofluorescent staining of transfected cells was carried out as described previously (20). Briefly, incubation with a rabbit polyclonal IB␣ antiserum (1 g/ml, 1 h, room temperature; Santa Cruz Biotechnology) of cells transfected with IB␣EGFP was carried out following fixation in methanol (Ϫ20°C, 5 min) and incubation in blocking serum (5%, 1 h) and prior to incubation with biotinylated secondary antibody (2 g/ml, overnight at 4°C; Santa Cruz Biotechnology) and streptavidin/Texas Red (0.2 g/ml, 15 min; Amersham Biosciences). Data (red fluorescence) were acquired using 750 V with excitation at 568 nm and emission at 590 nm. Excitation and emission of green fluorescence from the fusion protein was done at 488 and 530 nm, respectively, as above. Green and red fluorescence was plotted for individual cells over a range of transfection levels. The level of GFP fluorescence was corrected for minor reduction in intensity (15%) induced by methanol fixation.
Immunoprecipitation-Biochemical analysis of complex formation was carried out by immunoprecipitation using an anti-RelA antibody (Santa Cruz Biotechnology), followed by Western analysis, as above, of cells transfected with wild type and mutant pIB␣EYFP, alone or together with pCMV:relA, before and after IL-1 stimulation. Thus, HeLa cells seeded at 3 ϫ 10 6 cells/10-cm plate were transfected with the indicated cDNAs, totaling 4.4 g/plate, and after 24 h they were left unstimulated or treated with IL-1␤ for 8 and 40 min and lysed. Immunoprecipitation of cell lysates, as above, was performed by incubating with 5 g of anti-RelA antibody and protein G conjugated to Sepharose beads (overnight, 4°C; Sigma). Following resuspension in lysis buffer, precipitates were separated by gel electrophoresis (20% SDS, Tris-HCl) and transferred to polyvinylidene difluoride membranes for Western analysis, as above.
FRET Analysis-Images of ECFP, EYFP, and FRET were obtained through a ϫ 60 Plan Apo oil immersion objective (numerical aperture of 1.4) using a Nikon Diaphot 300 microscope and recorded by a 12-bit Hamamatsu digital camera C4742-95 driven by OpenLab software (Improvision). Complex formation was determined through measuring close association of fluorophores ECFP and EYFP by excitation of the donor, resulting in an increase in acceptor emission fluorescence and/or a decrease in donor fluorescence due to FRET (24 -26) using a series of filter sets (Omega Optical): XF114 for ECFP (440DF21 excitation, 455DRLP dichroic, 480DF30 emission), XF104 for EYFP (500DF25 excitation, 525DRLP dichroic, 545DF35 emission), and XF88 for FRET (440DF21 excitation, 455DRLP dichroic, 545DF35 emission). Relative fluorescence was calculated for the various emission settings by measuring the mean intensity of representative areas of nucleus or cytoplasm and normalized by dividing by attenuation, as above. All images were corrected for background, and FRET images were further processed by subtracting overspill of ECFP (50.7%) and EYFP (30.4%) fluorescence assessed using vectors containing ECFP or EYFP only. The accuracy of the FRET signal was further confirmed by inhibition of emission at 545 nm after photobleaching the acceptor-EYFP at 500-nm excitation. In addition, control experiments included demonstration of a correlation between the reduction in yellow fluorescence and increase in cyan intensity.
Binding Analyses-The level of IB/RelA binding was measured by FRET following co-transfection with IB␣EYFP (yellow fluorescence) and ECFPrelA (cyan fluorescence), before and after IL-1 stimulation (30 min) in four independent experiments. The FRET signal from single transfected cells expressing a range of IB␣ or S32A/S36A-phosphodeficient IB␣ mutant (ligand) concentrations was quantitated, as reported for assessment of protein/protein interactions in a variety of systems in live cells (27)(28)(29)(30). Analysis was limited to including cells expressing levels of fusion protein of up to 8 -10 times that of the endogenous levels, corresponding to fluorescence of 2.5-3, to ascertain reliable cytokine responses (20,21). Readings were ranked according to yellow fluorescence (IB␣EYFP), and to increase accuracy, they were pooled in consecutive groups of three for data analysis. The specificity of the interaction was demonstrated by competition of the IB␣EYFP/ ECFPrelA binding, carried out by co-transfecting with increasing levels of pCMV:relA or pCMV:p50, or using a mutant pECFPrelA lacking the cyan fluorescent site. Saturated binding and apparent dissociation constants were estimated by nonlinear least squares fitting of a noncooperative one-site equilibrium-binding model of the FRET versus IB␣ (yellow fluorescence). All data were analyzed using MLAB for the Macintosh (Civilized Software, Silver Spring, MD).
Cellular concentrations of the various fusion proteins were determined by Western analysis of transfected cells as above and were included in calculations of binding affinity (K a ) using the equation, where Bound (molecules/cell) represents the level of bound IB␣/cell determined by FRET; Receptors/cell or Binding protein is RelA concentration (molecules/cell); and [Ligand] is IB␣ concentration (molecules/cell). Controls included calculations using various size pools for data analysis (groups of 3-6), as well as using larger data sets allowing up to 12 individual cell readings for each IB␣ concentration, all demonstrating the same results as regards specific binding sites and affinity values. In addition, the single cell analysis was assessed by determining the R 2 values. These were found to be of the same order as that obtained from conventional binding analysis, varying by 15 Ϯ 5%, demonstrating a high level of accuracy, and further revealing an underlying Gaussian distribution as regards cell/cell variation.

RESULTS
Kinetic analysis of single cell data obtained by confocal microscopy demonstrated a pronounced enhancement of IB␣EGFP degradation by co-transfection with RelA alone or together with p50, increasing IL-1-induced turnover from 60 to 90 -95% (Fig. 1, A and B). In contrast, levels of a serine 32/36 phosphorylation-deficient IB␣ mutant (31)(32)(33) were unaffected by IL-1 stimulation, either when transfected alone or following co-transfection with NF-B subunits. In addition, there was a positive correlation between the magnitude of the effect and the initial cytoplasmic levels of IB␣ (Fig. 1C).
The RelA-mediated effect on turnover was confirmed by Western analysis, which, in addition, demonstrated an enhanced phosphorylation of IB␣EGFP in IL-1-stimulated cells following co-transfection (Fig. 2, A and B). The RelA-mediated effect on phosphorylation of the inhibitor protein resulted in a 2-fold increase in non-IL-1-stimulated levels and caused a 50 -100% enhancement of the peak value at 5-10 min of IL-1 stimulation. In addition, these experiments showed similar but less pronounced effects on the endogenous protein.
The dependence of the RelA-induced effects on IB␣ concentration, demonstrated by single cell analysis, prompted quantitation of inhibitor levels. Immunocytochemical staining of transfected cells and monitoring of red and green fluorescence demonstrated a linear relationship between total and transfected levels of the protein over a wide range of concentrations (Fig. 3, A and B). A mean transfection level of 1 fluorescent unit corresponded to an increase in cytoplasmic IB␣ to about 3-fold the average endogenous levels (Fig. 3B). Quantitation by Western analysis of cells transfected with pIB␣EYFP and pECF-PrelA showed average concentrations of 400,000 and 370,000 molecules/cell for IB␣EYFP and ECFPrelA (Fig. 3C), respectively, and revealed endogenous IB␣ levels to be about 135,000 molecules/cell, 2.5-fold higher than that estimated for the average levels of endogenous RelA, as expected (20,22), with all values close to the ranges demonstrated for the endogenous proteins (22).
Confocal microscopy further demonstrated retention of IB␣ in the cytoplasm by RelA (Fig. 4). Thus, an increase in nuclear localization of tagged (Fig. 4A, 1-4) or untagged (Fig 5A, 5 and  6) inhibitor at higher expression levels (Fig. 4A, 3 and 5) was not observed in the presence of a comparable amount of the NF-B subunit (Fig. 4A, 4 and 6). In addition, quantitation over a wide range of IB␣EGFP expression levels showed the effect to correlate with increasing levels of (pCMV) relA in a concentrationdependent manner (Fig. 4, B and C).
The RelA/IB␣ interdependence was further analyzed by simultaneous observations of RelA and IB␣ using cyan (ECFP) and yellow (EYFP) variants of GFP, respectively (pECFPrelA and pIB␣EYFP) (Fig. 5A). These experiments demonstrated that in the unstimulated state, both subunits were located in the cytoplasm. Quantitation showed that, similarly to experiments using EGFP containing constructs (20), IL-1 stimulation caused a successive increase in nuclear RelA but had no appreciable effect on cytoplasmic levels (Fig. 5B). Similarly, the rate of reduction of cytoplasmic inhibitor levels during activation was the same as that using the green fluorochrome, (compare Figs. 5B and 2B) and coincided with a reduction in the much lower levels of nuclear IB␣. The levels of RelA-IB␣ complexes were assessed by measuring FRET using ECFPrelA as a donor and IB␣EYFP as the acceptor (34). FRET was observed in cells expressing both fusion proteins demonstrating in vivo association of IB␣EYFP and ECFPrelA (Fig. 5A, 6 and 9), not seen in control cells, transfected with constructs containing EYFP and ECFP only (Fig. 5A, 3). A reduction in FRET signal (compare 6 and 9) following cytokine stimulation correlated with nuclear translocation of RelA (compare 4 and 7) and a decrease in cytoplasmic IB␣ (compare 5 and 8). Both the level of FRET and of IB␣ showed a steady state at about 50% after 30 min of IL-1 stimulation (Fig. 5B). The addition of RelA caused a concentration-dependent increase in the level of complexes over a range of inhibitor concentrations, reaching successively higher levels of saturation (Fig. 5C).
Analysis by immunoprecipitation of transfected cells demonstrated that RelA caused an increase in IB␣-containing complexes, both in control and IL-1-stimulated cells. Using a phosphospecific antibody, the experiments similarly demonstrated a significant increase in the level of complexed phosphorylated inhibitor in the presence of RelA (Fig. 6A). Transfecting the Ser 32 /Ser 36 -phosphodeficient mutant, however, resulted in a significant reduction in IB␣/RelA interactions, in both control and IL-1-stimulated cultures (Fig. 6B). Quantitation further demonstrated a correlation between effects on complex levels (1.96 Ϯ 0.4) and phosphorylation (2.35 Ϯ 0.58) using tagged or untagged RelA, before and after stimulation, with unaffected low levels using the phosphodeficient IB␣EGFP mutant (Fig. 6C).
The NF-B/IB␣ interaction was further characterized using FRET by measuring signals from transfected cells expressing a range of IB␣ levels and modeling the data by noncooperative binding of ligand (IB␣) to a single class of sites (RelA). Specificity of the interaction was demonstrated by competition by increasing amounts of pCMVrelA or pCMV p50, or of a non-

FIG. 1. NF-B enhances IL-1-stimulated inhibitor degradation at high levels of IB␣.
A, series of confocal images of fibroblasts, co-transfected with equal amounts of pIB␣EGFP and pCMV:relA, obtained at various times after the addition of IL-1␤ as indicated. Bar, 20 m. B, average IB␣EGFP levels were determined at various times during IL-1␤ stimulation and transfected with pIB␣EGFP or a mutant pIB␣EGFP, lacking phosphorylation sites at serines 32 and 36, alone or co-transfected with pCMV:relA, or with pCMV:relA and pCMV:p50. C, the percentage degradation of the IB␣EGFP fusion protein following transfection with pIB␣EGFP alone or together with pCMV:relA or together with pCMV:relA and pCMV:p50, following 15-or 60-min stimulation with IL-1␤, was determined for individual cells and plotted against initial fluorescence. A best fit logarithmic curve to the data is shown.
fluorescing mutant of ECFPrelA (ECFPrelAmut) (Fig. 7A). This showed a dose-dependent inhibition of the FRET signal, which corresponded to about 70% at 4-fold overexpression of either of the RelA constructs, whereas co-transfection with p50 had no effect. Further, in agreement with biochemical data, shown in Fig. 6C, using the phosphorylation site-deficient IB␣-mutant resulted in about a 50% reduction in specific RelA binding (Fig. 7A).
In addition, these experiments showed that IB␣/RelA interaction (FRET) is dynamic and dependent on IB␣ concentration (EYFP fluorescence), demonstrating saturation before and after incubation with IL-1 (Fig. 7B). A pronounced fall in the FRET signal after 30 min of IL-1 stimulation corresponded to a reduction in complex concentration to 38 Ϯ 8%, correlating with the decrease determined by immunoprecipitation (see Fig.  6). In comparison, using the phosphonegative mutant resulted in a much decreased level of complexes, which were unaffected by IL-1 stimulation (Fig. 7B). Characterization of the interaction by Scatchard analysis demonstrated a pronounced (4.2 Ϯ 1.0-fold) increase in affinity (K a ) of RelA binding to the wild type inhibitor following stimulation with the cytokine, relative to that calculated for control cultures (Fig. 7C). Quantitation of fusion protein levels, as in Fig. 3C, gave estimated affinity constants (K a ) for these interactions in the range of 10 7 to 5 ϫ 10 7 M Ϫ1 in unstimulated cells and between 5 ϫ 10 7 and 5 ϫ 10 8 M Ϫ1 following IL-1 stimulation, respectively. In comparison, the affinity of RelA for the phosphonegative mutant was only 25% of that for the wild type and, in addition, was unaffected by cytokine activation (Fig. 7C). DISCUSSION The data presented in this report are in broad agreement with the widely accepted model based on NF-B/IB interdependence and regulation through IB␣ phosphorylation. The real time GFP measurements on single cells, in addition, show that the IB␣/NF-B interaction is specific, saturable, and dynamic; that the RelA-induced increase in binding relies on trafficking; and that regulation of complex formation and of the affinity of the interaction are IB␣ phosphorylation-dependent.
Our data underscore the reciprocal nature of NF-B/IB regulation, coupled with a system in which free components successively accumulate in the nucleus with increasing levels, and bound complexes largely remain in the cytoplasm, in the resting state (20). This is reflected in the inverse correlation between the inhibitor cytoplasmic concentration and the ratio of cytoplasmic to nuclear levels and by the cytoplasmic accumulation of IB␣ in the presence of excess RelA. Regulation is probably influenced by the level of the endogenous proteins and of de novo synthesis through autoregulation (35), since the low transfection levels of the GFP fusions, corresponding to 2-3-

FIG. 2. RelA enhances IL-1-induced degradation and phosphorylation of both endogenous and exogenous IB␣.
A, mock-transfected fibroblasts (pCMV empty vector) and fibroblasts transfected with IB␣EGFP alone (pIB␣EGFP) or together with RelA (pIB␣EGFP ϩ pCMV: relA) were stimulated with IL-1 (10 Ϫ9 M) for the indicated times. Cell extracts were separated on a 12.5% SDS gel, and levels of endogenous IB␣ were determined by Western blotting using primary anti-rabbit polyclonal IB␣ IgG. Quantitation of the level of the inhibitor was done by scanning of autoradiographs. One representative experiment of three is shown. Data are expressed relative to the initial values for each condition. B, cells transfected with pIB␣EGFP alone or co-transfected with equal amounts of pIB␣EGFP and pCMV: relA as above were stimulated with IL-1␤ for the times indicated. The level of phosphorylation on serine 32 of endogenous and exogenous IB␣ was determined by Western blotting, using primary anti-rabbit polyclonal phospho-IB␣, and quantitated, as above. Data from one of two experiments are shown and expressed relative to initial levels of the endogenous protein.
fold that of the endogenous protein, are well within the physiological range of variation (22). Thus, the successive increase in nuclear levels at higher concentrations of IB-EGFP probably reflects titering out of endogenous NF-B. Further, competition for the endogenous pool of NF-B could account for the enhanced level of breakdown of IB␣EGFP and the limited effect of co-transfection at low levels of exogenous inhibitor (22).
Enhanced cytoplasmic retention of IB␣, such as that induced by RelA, has significant effects on pathway regulation (36 -41). 2 This is supported by mathematical modeling, which shows that the B system is dissipative and far from equilibrium and is at a steady state determined by dissociation of complexes, intercompartmental trafficking, and synthesis and breakdown of components (21). 3 Further, simple kinetics (Briggs-Haldane) show that saturable NF-B/IB␣ binding, under steady state conditions, can occur with apparent affinity constants much lower than those observed with pure compo-nents in vitro, in agreement with the lower affinities measured in our study, compared with those observed using Biocore (42).
The involvement of basal cycling in regulation (21), indicated by the slightly delayed decrease in FRET (47%) compared with that of the inhibitor fluorescence (56%) (significant at 5%), suggests a reassociation of released ECFPrelA with free IB␣EYFP prior to nuclear entry. The biological significance of transport is not likely to enhance the level of activation, since the kinetics of inhibitor reduction in the two compartments is identical (60 -70%). Rather, it may constitute the basis for recruiting specific subsets of NF-B intermediates for distinct facets of pathway regulation, such as complexes activated subsequent to nuclear translocation (36). The significance of selectivity during pathway activation is supported by the burgeoning evidence of diversity and complexity of NF-B induction (2,36,41,(43)(44)(45)(46)(47). This further supports the notion of a system where perturbations in concentrations of signaling components and in the system steady state, such as that induced by in- Enhanced inhibitor phosphorylation in the presence of exogenous RelA is probably to some extent due to an increase in NF-B-IB␣ complexes and reflects preferential phosphorylation of complexed inhibitor by IB kinases (48). Conversely, the pronounced reduction in RelA/IB␣ binding at steady state, observed with the 32/36 mutant, demonstrated dependence of complex formation and affinity on IB␣ phosphorylation, as found for interactions through the PEST domain of the protein (49). In the absence of a direct involvement of the IB␣ signalreceiving domain in NF-B interaction (42), the effects on specific binding, observed using the mutants, could be explained by changes in inhibitor conformation in vivo. A decrease in binding of IB␣ proteins lacking the phosphorylation sites necessary for proteosome degradation (6,31,32), is consistent with a critical role for RelA in controlling IB␣ turnover. Of the three components of the classical B complex, RelA is present at the lowest concentration (22). Linking IB␣ turnover to RelA levels will therefore optimize signals through the IB kinases. In addition, ongoing studies suggest that further amplification through positive feedback is induced by increased activation of upstream kinases in the presence of excess RelA. 4 This is in agreement with a system based on reutilizing a limited amount of RelA through repeated cycles of IB␣ binding, phosphorylation, and degradation and with continuous enhanced effects with increasing RelA concentrations.
Alterations in the level and affinity of IB␣-NF-B complexes correlated with the rate of inhibitor degradation. With only 20% of NF-B free of IB␣ (20), a 4-fold change in complex affinity is likely to have a significant role in regulation of inhibitor turnover, subsequent to effects on basal shuttling and reduction in free NF-B. The identical results obtained in the blocking experiments, using tagged and untagged RelA, show that the lower affinities we observe in vivo compared with that measured with purified proteins (41) are not artifacts of GFP tagging (50). The increase in NF-B/IB␣ affinity correlates with the enhanced cytoplasmic FRET/IB␣EYFP ratio following IL-1 stimulation, suggesting a selective process. This is also supported by the lack of a simple inverse correlation between the affinity and the level of complexes and indicates that low affinity complexes are specifically targeted for degradation during cytokine activation. The biological significance of this selection could be to facilitate a rapid release in response to a transient signal through the low affinity complexes, characterized by a high dissociation rate. The heterogeneity of NF-B-IB␣ complex affinity may thus constitute a buffer to incoming signals.
In summary, the data show that translocation, phosphorylation, and complex formation are all critical for regulation of IB␣ steady state levels by NF-B. They demonstrate that FIG. 5. ECFPrelA regulates IB␣-EYFP complex formation in living cells. A, fibroblasts were co-transfected with control constructs, pECFP-C1 plus pEYFP-C1 (1-3), or pECFPrelA plus pIB␣EYFP (4 -9). Simultaneous analysis of RelA localization (4 and 7), IB␣ degradation (5 and 8), and complex formation (6 and 9) during stimulation was carried out by recording cyan, yellow, and FRET images, respectively. In the unstimulated cell (4 -6), both fusion proteins are located in the cytoplasm. 30-min stimulation with 1 nM IL-1 (7-9) causes an increase in nu- IB␣/NF-B binding is dynamic, saturable at steady state, and phosphorylation-dependent. They also show that both the extent and the affinity of the IB␣/NF-B interaction play a role in system regulation, affecting both basal levels and those induced in response to incoming signals. The multifaceted effects of RelA on the concentration-dependent limitation of inhibitor turnover thus constitute a complex feedback loop, with effects on the system steady state and with expected significance for the specificity and level of NF-B activity.
FIG. 6. RelA induces phosphorylation-dependent complex formation. A, cells transfected with IB␣EYFP alone or co-transfected with an ECFPrelA-containing construct were stimulated with IL-1 for 0 or 8 min. Extracts were subjected to immunoprecipitation using an anti-RelA antibody and separated on a 12.5% SDS gel, and the levels of RelA, IB␣, and phospho-IB␣ for the various conditions were determined by Western analysis, as described under "Experimental Procedures." Similarly, the level of IB␣EYFP in the total lysate was assessed by Western analysis. One of two experiments is shown. B, cells were transfected with wild type IB␣EYFP or with a mutant lacking phosphorylation sites Ser 32 and Ser 36 , together with ECFPrelA, and stimulated with IL-1 for 0 or 8 min, as above. Complexes were immunoprecipitated following extraction by using an anti-RelA antibody, separated on a 12.5% gel, and subjected to Western analysis, as in A. One of three experiments is shown. C, levels of total and phosphorylated complexes were determined for cells transfected with untagged or fluorescence-tagged IB␣ or with a fluorescence-tagged mutant IB␣ lacking phosphorylation sites serine 32 and 36, together with untagged or tagged RelA as indicated, and stimulated with IL-1 for 0, 8, or 40 min. Quantitation was carried out following immunoprecipitation, gel electrophoresis, and Western analysis, as above, by scanning autoradiograms. Left panel, the level of phosphorylated complexes at 0 and 8 min. Data represent the average of three experiments and are expressed relative to total IB␣ in the lysate. Right panel, the total level of complexes at 0 and 40 min. One representative experiment of three is shown, with data expressed relative to total RelA. One of two experiments for each condition is shown. B, an analysis of binding of IB␣EYFP with ECFPrelA before and after 30min stimulation with IL-1␤. Each data point represents the average of readings from three cells. Left panel, binding before and after IL-1 stimulation in cells co-transfected with the wild type IB␣EYFP construct and ECFPrelA. Right panel, binding before and after IL-1 stimulation, in cells co-transfected with the phosphorylation site-deficient IB␣EYFP mutant and ECFPrelA. One representative experiment of four and two, respectively is shown for the conditions. ‚, control cultures; E, following 30 min of IL-1 stimulation. C, relative level of affinity of RelA interactions with the wild type (Wt) and the phosphodeficient mutant (Mut) of IB␣, respectively, in control and in cells stimulated with IL-1 for 30 min, as indicated. Scatchard analysis of experiments as in B were carried out using mlab, and data are expressed relative to the affinity constant (K a ) in unstimulated cells transfected with equal amounts of ECFPrelA and wild type IB␣EYFP. Shown are the averages of four (wild type) and two (mutant) experiments, respectively.