Identification of a Ligand-induced Transient Refractory Period in Nuclear Factor-κB Signaling*

In response to a variety of extracellular ligands, nuclear factor-κB (NF-κB) signaling regulates inflammation, cell proliferation, and apoptosis. It is likely that cells are not continuously exposed to stimulating ligands in vivo but rather experience transient pulses. To study the temporal regulation of NF-κB and its major regulator, inhibitor of NF-κBα (IκBα), in real time, we utilized a novel transcriptionally coupled IκBα-firefly luciferase fusion reporter and characterized the dynamics and responsiveness of IκBα processing upon a short 30-s pulse of tumor necrosis factor α (TNFα) or a continuous challenge of TNFα following a 30-s preconditioning pulse. Strikingly, a 30-s pulse of TNFα robustly activated inhibitor of NF-κB kinase (IKK), leading to IκBα degradation, NF-κB nuclear translocation, and strong transcriptional up-regulation of IκBα. Furthermore, we identified a transient refractory period (lasting up to 120 min) following preconditioning, during which the cells were not able to fully degrade IκBα upon a second TNFα challenge. Kinase assays of IKK activity revealed that regulation of IKK activity correlated in part with this transient refractory period. In contrast, experiments involving sequential exposure to TNFα and interleukin-1β indicated that receptor dynamics could not explain this phenomenon. Utilizing a well accepted computational model of NF-κB dynamics, we further identified an additional layer of regulation, downstream of IKK, that may govern the temporal capacity of cells to respond to a second proinflammatory insult. Overall, the data suggested that nuclear export of NF-κB·IκBα complexes represented another rate-limiting step that may impact this refractory period, thereby providing an additional regulatory mechanism.

In response to a variety of extracellular ligands, nuclear factor-B (NF-B) signaling regulates inflammation, cell proliferation, and apoptosis. It is likely that cells are not continuously exposed to stimulating ligands in vivo but rather experience transient pulses. To study the temporal regulation of NF-B and its major regulator, inhibitor of NF-B␣ (IB␣), in real time, we utilized a novel transcriptionally coupled IB␣-firefly luciferase fusion reporter and characterized the dynamics and responsiveness of IB␣ processing upon a short 30-s pulse of tumor necrosis factor ␣ (TNF␣) or a continuous challenge of TNF␣ following a 30-s preconditioning pulse. Strikingly, a 30-s pulse of TNF␣ robustly activated inhibitor of NF-B kinase (IKK), leading to IB␣ degradation, NF-B nuclear translocation, and strong transcriptional up-regulation of IB␣. Furthermore, we identified a transient refractory period (lasting up to 120 min) following preconditioning, during which the cells were not able to fully degrade IB␣ upon a second TNF␣ challenge. Kinase assays of IKK activity revealed that regulation of IKK activity correlated in part with this transient refractory period. In contrast, experiments involving sequential exposure to TNF␣ and interleukin-1␤ indicated that receptor dynamics could not explain this phenomenon. Utilizing a well accepted computational model of NF-B dynamics, we further identified an additional layer of regulation, downstream of IKK, that may govern the temporal capacity of cells to respond to a second proinflammatory insult. Overall, the data suggested that nuclear export of NF-B⅐IB␣ complexes represented another rate-limiting step that may impact this refractory period, thereby providing an additional regulatory mechanism.
Adequate resolution of an inflammatory reaction is as equally important as initiation. Persistent or fulminant responses can cause detrimental consequences both locally and systemically (1), and resolution of inflammation is important for both termination of an acute response as well as for preven-tion of destructive chronic responses. It is therefore not surprising that mechanisms aimed at rapid and specific initiation of proinflammatory reactions have co-evolved with mechanisms that provide timely termination of such processes. From a systems biology perspective, such "switchability" can be achieved by intracellular feedback loops that permit ligand-induced desensitization and resensitization of proinflammatory signaling cascades (2).
In this regard, recent studies have shown that nuclear factor-B (NF-B) 4 signaling plays a critical role in both initiation and resolution of inflammation (2,3). The transcription factor NF-B is a key regulator of innate and adaptive immune responses as well as a mediator of cell survival and proliferation (4). Improper regulation of NF-B contributes to induction and progression of a wide range of human disorders, including a variety of pathological inflammatory conditions, neurodegenerative diseases as well as many types of cancer (5,6). In resting cells, inactive NF-B is sequestered in the cytoplasm by binding to members of the inhibitor of NF-B (IB) family. Canonical activation of NF-B depends on IB kinase (IKK)-regulated proteasomal degradation of IB␣, an event that frees NF-B for nuclear translocation within minutes (4,7). Upon nuclear transport, NF-B regulates the transcription of a few hundred genes (8 -10) that can be divided into four major families (10,11): 1) proinflammatory genes (e.g. cyclooxygenase 2, interleukin-1 (IL-1), tumor necrosis factor ␣ (TNF␣), inducible nitricoxide synthase, intercellular adhesion molecule-1, E-selectin, etc.), 2) proproliferative genes (e.g. cyclin D, and cellular myelocytomatosis), 3) antiapoptotic genes (B-cell leukemia/lymphoma 2, B-cell leukemia/lymphoma extra long, X-linked inhibitors of apoptosis protein, and cellular inhibitors of apoptosis protein), and 4) autoinhibitory genes (e.g. A20, cylindromatosis, suppressor of cytokine signaling 1, and IB␣).
With respect to the last, other transcriptionally independent processes, aimed at autoinhibition of NF-B activity, do exist. Such mechanisms down-regulate NF-B signaling on a much shorter time frame (seconds to minutes). These include homologous receptor desensitization (12,13), asymmetric heterologous receptor desensitization (13,14), autocatalytic C-terminal IKK hyperphosphorylation (15), and protein phosphatase 2Cdependent dephosphorylation of IKK (16).
Considering the complex nature of the inflammatory milieu, one would expect that stationary tissue-residing cells are exposed to a myriad of temporally distinct NF-B-stimulating cues. For instance, cells can be directly stimulated by pathogenderived products (e.g. lipopolysaccharide through TLR4 (tolllike receptor 4) receptors (17)), exposed to numerous soluble proinflammatory stimuli produced by circulating effector cells (e.g. cytokines, chemokines, etc.), and/or experience inflammation-induced oxidative stress (18). These signals can occur simultaneously or sequentially to one another. For example, systemic administration of bacterial lipopolysaccharide to mice was shown to induce transient production of TNF␣ (serum levels peaking at ϳ1.5 h and quickly returning to base line), but IL-1␤ production was delayed and prolonged (first detected at 2 h, but lasting Ͼ5-6 h) (19). Thus, cells co-expressing TLR4, IL-1, and TNF␣ receptors would sequentially interrogate signals arising from lipopolysaccharide, TNF␣, and IL-1␤, each of which could independently activate NF-B.
Central to any signaling desensitization mechanism is a refractory period during which cells cannot fully respond to a second insult (autologous or heterologous desensitization). Therefore, consideration of the dynamic pattern of stimulus exposure described above begs the immediate question of whether cells can instantly initiate an NF-B response to a second activating stimulus, and if not, when will such cells be able to remount a full response again? Specifically, are ligand-preconditioned cells capable of eliciting NF-B activation to the same extent as naive cells?
Little is known about the capacity of cells to activate NF-B in response to a second activating challenge, since the highly dynamic nature of this process presents many technical difficulties. These include low temporal resolution of conventional transcriptionally dependent NF-B reporter gene assays, low throughput, inability to acquire longitudinal data, and the semiquantitative nature of traditional biochemical assays (e.g. electrophoretic mobility shift assay, immunoblotting, etc.). Such limitations render these assays incapable of accurate analysis of the early, ligand-induced dynamic changes in the capacity of cells to elicit a response to a second challenge.
To efficiently address this question, we generated an improved, transcriptionally coupled version of a previously published genetically encoded IB␣-firefly luciferase (IB␣-FLuc) fusion reporter (20) in conjunction with dynamic, live cell bioluminescence imaging of cultured cells. We chose to focus on HepG2 human hepatoma cells as a model system, because 1) NF-B signaling has been extensively studied in these cells, 2) HepG2 cells have been shown to activate NF-B in response to a variety of proinflammatory ligands (21), 3) these cells can be easily transfected with readily available reagents, and, most importantly, 4) the pivotal role that NF-B signaling plays in hepatocytes to regulate inflammation, apoptosis, and carcinogenesis (22).
Using bioluminescence imaging of live cells in conjunction with a variety of biochemical assays, we demonstrate herein that a 30-s preconditioning exposure to TNF␣ is sufficient to robustly activate IKK, culminating in IB␣ degradation, NF-B nuclear translocation, and strong transcriptional up-regulation of IB␣. Furthermore, the capacity of preconditioned cells to degrade IB␣ in response to a second TNF␣ challenge is transiently refractory, regaining full responsiveness ϳ120 min later. Finally, both IKK regulation and possibly NF-B nuclear export, but not receptor dynamics, govern this transient refractory period. This study highlights the interlocking layers of NF-B regulation, ensuring efficient and timely propagation as well as termination of proinflammatory signals.
Cells and Transfections-HepG2 human hepatoma cells were from the American Type Culture Collection (Manassas, VA). Cells were cultured in DMEM supplemented with heatinactivated fetal bovine serum (10%) and L-glutamine (2 mM). Cell cultures were grown at 37°C in a humidified atmosphere of 5% CO 2 . HepG2 cells (10 5 ) were transiently transfected (Fugene 6; Roche Applied Science) with pB 5 3 IB␣-FLuc (200 ng/well) and plated in black-coated 24-well plates (In Vitro Systems GmbH, Gottingen, Germany). Cells were then allowed to recover for 48 h prior to imaging.
Dynamic Bioluminescence Live Cell Imaging-Prior to imaging, cells were washed with prewarmed phosphate-buffered saline (PBS, pH 7.4) and placed into 900 l of colorless HEPESbuffered DMEM, supplemented as above and with D-luciferin (150 g/ml). Cells were allowed to equilibrate for 1 h (37°C) before proceeding with ligand stimulation and imaging. Four different stimulation regimens were included in this study. 1) For continuous TNF␣ (C), TNF␣ (final concentration 20 ng/ml) or vehicle (colorless DMEM) was added (100 l) to D-luciferin-containing DMEM, and imaging was performed before and at the indicated time points after the addition of TNF␣. 2) For TNF␣ pulse (30 s, P), cells were pulsed for 30 s with TNF␣ (20 ng/ml) or vehicle, washed with PBS, returned to D-luciferincontaining DMEM, and imaged before and at the indicated time points after the pulse of TNF␣. 3) For TNF␣ precondition-ing (30-s pulse) followed by continuous TNF␣ challenge (P ϩ C), at t 0 cells were pulsed for 30 s with TNF␣ (20 ng/ml) or vehicle, washed with PBS, returned to D-luciferin-containing DMEM (900 l), and imaged before and at the indicated time points after the pulse of TNF␣. At t x , TNF␣ (final concentration 20 ng/ml) or vehicle (colorless DMEM) were again added (100 l), and imaging was performed before and at the indicated time points after the addition of TNF␣. 4) TNF␣ preconditioning (30-s pulse) followed by continuous IL-1␤ challenge (P ϩ C) was as in method 3, but continuous challenge was performed with IL-1␤ (10 ng/ml).
To analyze ligand-induced regulation of de novo reporter resynthesis, cells were pretreated with cycloheximide (100 g/ml) for 1 h before continuous stimulation with TNF␣ and bioluminescence imaging (as above).
IB␣ Responsiveness Assays-HepG2 cells, transfected with pB 5 3 IB␣-FLuc (as above) or HeLa cells, stably expressing pCMV3 IB␣-FLuc (20), were plated in 4 wells of a 6-well plate (one plate per time point) and grown for 48 h. At t 0 , all wells were washed with prewarmed PBS, pulsed for 30 s with TNF␣ (20 ng/ml, 1 ml) or vehicle (PBS), washed again with PBS, returned to regular medium (1 ml), and placed in a 37°C incubator. This procedure was defined as TNF␣ preconditioning (P). At t x , two wells were treated (continuously) with TNF␣ (20 ng/ml), and two wells were treated with vehicle only (PBS). This procedure was defined as TNF␣ challenge (C). Following this TNF␣ challenge, cells were returned to the incubator. At t x ϩ 25 min (time of maximal IB␣ degradation (20); see Fig. 2A for schematic timeline), cells were harvested (by scraping) in reporter lysis buffer (Promega, Madison, WI). Cell lysates were normalized for protein content by a BCA protein assay (Promega), aliquoted, and frozen (Ϫ80°C) for in vitro bioluminescence and Western blot analyses (see below). For in vitro bioluminescence assays, lysates (10 l, in triplicate) were mixed with luciferase assay buffer (190 l; 25 mM HEPES, 154 mM NaCl, 5.4 mM MgSO 4 , 10 mM dithiothreitol, 5 mM ATP, 150 g/ml D-luciferin, pH 8.0) in a 96-well plate immediately prior to imaging. Assay plates were imaged using the IVIS 100 bioluminescence imager (acquisition time, 10 s; binning, 4; field of view, 10 cm; f/stop, 1; filter, open).
Western Blot Analyses-Whole cell lysates were resolved by 10 or 7.5% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and probed for the indicated proteins using standard immunoblotting techniques. Primary antibodies against total human IB␣, ␤-actin, and IKK␣ were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Anti-phospho-IB␣ (Ser-32/36) was from Cell Signaling Technologies (Danvers, MA). Secondary horseradish peroxidase-labeled antimouse and anti-rabbit IgG antibodies were from GE Healthcare Biosciences.
IKK Kinase Assay (IKK-KA)-IKK-KA reactions were carried out as per Werner et al. (23) and quantified in a medium throughput manner as per Hastie et al. (24). Briefly, HepG2 cells were grown in 10 cm tissue culture dishes to confluence. Cells were then washed in PBS (once) and treated with 20 ng/ml TNF␣ using three different treatment regimens: P, C or P ϩ C (see above). To capture the full IKK activity profiles of cells treated with continuous (C) or pulse (P) regimens, cytosolic extracts were prepared at t ϭ 0 (before) or 5, 10, 15, 30, 60, 120, or 240 min post-TNF␣ treatment. To capture maximal IKK activity of P ϩ C-treated cells, cytosolic extracts were prepared 10 min post-TNF␣ challenge (given at 10, 30, 60, 120, and 240 min post-preconditioning). Cells were harvested by removing media, washing in ice-cold PBS ϩ EDTA (1 mM), scraping, and pelleting at 2000 ϫ g. To prepare cytosolic extracts, cell pellets were resuspended in 200 l of cytosolic extract buffer (10 mM HEPES-KOH, pH 7.9, 250 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 0.2% Tween 20, 2 mM dithiothreitol, 20 mM ␤-glycerophosphate, 10 mM NaF, and 0.1 mM Na 3 VO 4 supplemented with complete protease inhibitor mixture), incubated on ice (2 min), vortexed (1 min), and pelleted at 2000 ϫ g. Supernatants were collected, normalized for protein content by a Bradford assay (Pierce), and stored at Ϫ80°C. To immunoprecipitate IKK complexes, cytoplasmic extracts (100 l) were incubated with anti-IKK␥ antibody (15 l, overnight, 4°C with rotation) and then with Protein G 4FF bead slurry (20 l, 50% (v/v)). Beads were pelleted at 4600 rpm and washed twice with cytosolic extract buffer (500 l) and once with Kinase Buffer (500 l; 20 mM HEPES, pH 7.7, 20 mM ␤-glycerophosphate, 100 mM NaCl, 0.1 mM Na 3 VO 4 , 10 mM MgCl 2 , 2 mM dithiothreitol supplemented with complete protease inhibitor mixture). For the IKK kinase reaction, beads were incubated for 30 min at 30°C in Kinase Buffer (20 l) containing 20 M ATP, 10 Ci of [ 32 P]ATP, and 0.5 g of GST-IB␣-(1-54). Beads were removed by centrifugation (4600 rpm), and 15 l of each reaction supernatant was spotted onto a 1-cm 2 square of P81 phosphocellulose paper (Millipore, Billerica, MA) and immediately immersed in phosphoric acid (75 mM) for 5 min. Phosphoric acid washes were performed two more times, and papers were rinsed in acetone and then allowed to dry. Each paper was transferred to a scintillation vial, and radioactivity was determined on a ␤ counter (Beckman Coulter, Fullerton, CA). Blank and no-lysate controls were subtracted from the experimental samples. Data were represented as fold-initial (untreated controls).
Calculating Ligand-dependent IKK Responsiveness-IKK responsiveness profiles (i.e. the net kinase capacity of IKK in response to a second challenge of TNF␣, as a function of time after initial 30-s preconditioning) were calculated numerically from IKK-KA data using the following formula, where PC x ϩ 10 is IKK activity of preconditioned plus challenged cells, as recorded 10 min postchallenge. P x ϩ 10 is the residual IKK activity of preconditioned but unchallenged cells at this exact time point. C 0 and PC 0 are initial IKK activities of challenged but unpreconditioned and fully preconditioned and challenged cells, respectively. C 10 is the maximal IKK activity of challenged but unpreconditioned cells (recorded 10 min postchallenge). Note that although all parameter units in the nominator and denominator are in counts/min, IKK responsiveness is dimensionless, similar to IB␣ responsiveness. Computational Simulations-To simulate the dynamics of major regulators on the IKK-NF-B axis, we used a well established computational model generated by Hoffmann et al. (25) and refined by Werner et al. (23). Briefly, an experimentally or hypothetically derived IKK activity profile was fed into the program as an input. Embedded in the model were 24 components, 70 reactions, and 70 parameters or rate constants for these reactions. Differential equations were solved numerically using Matlab 7.0 (Mathworks, Natick, MA) with subroutine Ode15. Interpolated and extrapolated (0 -360 min at 5-min intervals) IKK activity profiles were calculated (Origin version 7.5, Orig-inLab, Northampton, MA) from experimental IKK-KA data (see above). To fit the model, initial steady-state IKK activity (i.e. intracellular concentration of active IKK) was set to be 1 nM. To computationally simulate IB␣ dynamics of cells challenged at different times after initial preconditioning, when assuming no upstream IKK or receptor regulation, we used hypothetical IKK activity profiles as inputs, derived from superimposing experimentally acquired IKK activity profiles of 30-s pulsed and continuously treated cells at increasing intervals (30, 60, 120, and 240 min; see Fig. 4, black lines).
Immunofluorescence Microscopy-HepG2 cells were seeded into 35-mm glass bottom culture dishes (MatTek Corp., Ashland, MA) and grown to ϳ40% confluence. Cells were pulsed for 30 s with TNF␣ as above and fixed at the indicated time points (by washing once with PBS, followed by fixation (4% paraformaldehyde for at least 15 min) and permeabilization (ice-cold methanol, 10 min at Ϫ20°C)). Cells were washed in PBS, blocked in 5% normal goat serum in 0.3% Triton X-100, PBS (1 h), and then incubated with anti-p65 antibody (Santa Cruz Biotechnology; 1:200 in 0.3% Triton X-100, PBS at 4°C, overnight with rocking). Cells were next incubated with Alexa-Fluor 635-conjugated goat anti-rabbit antibody (Invitrogen; 1:200 in 0.3% Triton X-100, PBS, 90 min, at room temperature with rocking). Cells were washed three times with PBS before being mounted with VECTASHIELD Mounting medium (Vector Laboratories; Burlingame, CA). Confocal images were captured using the ϫ40 objective (water immersion) on a Zeiss Axiovert 200 (Zeiss, Thornwood, NY) laser-scanning microscope equipped with the appropriate filter sets and analyzed using a Zeiss LSM Image Browser and Adobe Photoshop CS2.

RESULTS
Real Time Bioluminescence Imaging of pB 5 3 IB␣-FLucexpressing Cells Recapitulated IKK-induced Dynamics of Endogenous IB␣-To monitor ligand-induced IB␣ rapid dynamics as well as physiologic transcriptionally coupled behavior, we modified our previous IB␣-FLuc fusion reporter (20) to be driven by a synthetic promoter composed of five tandem B response elements (TGGGGACTTTCCGC) followed by a minimal TATA-box. We hypothesized that this reporter would allow quantitative measurements of IKK-induced degradation as well as NF-B-induced resynthesis and post-translational stabilization of IB␣ from intact living cells (Fig. 1A). To validate use of this reporter, HepG2 cells were transiently transfected with a plasmid encoding the reporter and allowed to recover for 2 days before stimulation with a continuous or 30-s pulse of TNF␣ (20 ng/ml) to induce IKK activation. Upon the addition of TNF␣, a rapid and dramatic decrease in bioluminescence was observed when readouts were normalized to untreated controls (20) under both continuous (C) and 30-s pulse (P) regimens (Fig. 1, B and C). This decrease in normalized bioluminescence, reflecting IKK-induced reporter degradation was followed by a sharp increase in bioluminescence, reflecting NF-B-dependent reporter resynthesis, reaching maximum values at ϳ120 min and then gradually declining toward base line. Note that the rate at which IB␣ levels return to base line is steeper under continuous TNF␣ treatment compared with the 30-s pulse, providing evidence for reactivation of ligand-induced IB␣ degradation during continuous stimulation (23). The magnitude of the initial decrease in bioluminescence was greater in continuously treated cells than in 30-s pulsed cells (70% versus 40% of initial decrease, respectively), indicating that a 30-s pulse of TNF␣ leads to ϳ50% depletion of the IB␣⅐NF-B pool compared with continuous TNF␣ exposure (Fig. 1C, 120 min). These data suggested that 1) this reporter construct could report on both IKKinduced IB␣ degradation and successive resynthesis of IB␣, 2) a 30-s pulse of TNF␣ at a saturating concentration (20 mg/ml) elicited robust IKK activity, culminating in IB␣ degradation and a full IB␣ transcriptional up-regulation, and 3) with the current B 5 synthetic promoter system, there was a nonlinear relationship between IB␣ degradation and NF-Bdependent resynthesis of IB␣ (i.e. saturation of IB␣ resynthesis even at submaximal IB␣ degradation levels).
Strikingly, Western blot analysis revealed that endogenous IB␣ behaved exactly as the reporter under both C and P conditions, recapitulating the degree of degradation, recovery, and return to base line (Fig. 1D). Pretreating pB 5 3 IB␣-FLucexpressing HepG2 cells with cycloheximide did not affect degradation of IB␣-FLuc but abolished signal recovery, indicating that this phase was totally dependent upon transcription and translation of new IB␣-FLuc (Fig. 1E).
TNF␣ Preconditioning Induces a Transient Refractory Period of IB␣ Processing-Upon a proinflammatory insult in vivo, effector cells (e.g. circulating macrophages) release TNF␣ and other activating cytokines in a temporally and spatially discrete manner. As a consequence, stationary target cells (e.g. epithelial cells, endothelial cells, hepatocytes, etc.) will sense a stochastic rise in the levels of such proinflammatory ligands. In such a dynamic environment, as ligand-secreting cells continuously migrate to sites of inflammation, it is anticipated that over time, target cells will experience multiple pulses of activating ligands.
We therefore aimed to elucidate the effects of such ligand pulses on the capacity of hepatocytes to respond to a subsequent challenge of the same ligand. Having shown that 1) pB 5 3 IB␣-FLuc provided an accurate readout of IB␣ processing in intact cells and that 2) a 30-s pulse was sufficient to induce robust IKK activity, we next sought to investigate whether a short 30-s preconditioning pulse with TNF␣ had a substantial effect on the capacity of cells to process IB␣ upon a subsequent continuous TNF␣ challenge.
HepG2 cells transiently expressing pB 5 3 IB␣-FLuc were given a 30-s pulse of TNF␣ (20 ng/ml) or vehicle at t 0 , washed, replaced in media containing D-luciferin and repeatedly imaged (every 5 min) prior to a TNF␣ challenge. At t 30 , t 60 , t 120 , or t 240 (min) after pulsing, cells were then challenged with a second continuous concentration of TNF␣ (20 ng/ml), and live cell imaging was continued up to 360 min. To compare the processing dynamics of IB␣-FLuc in naive (unpreconditioned) cells with that of preconditioned cells, the resulting biolumines-cence profiles of preconditioned cells ( Fig. 2A, black lines) were plotted along with the bioluminescence profile of unpreconditioned cells (i.e. only treated with continuous TNF␣ at t 0 , red line, Fig. 2A). The different graphs represent the differential dynamics of IB␣-FLuc processing as the preconditioning pulse-challenge (P-C) intervals temporally increased (0 -240 min).
We observed that challenging preconditioned cells with a continuous exposure to TNF␣ near the time that they had achieved maximal degradation from the preconditioning pulse (i.e. 30 min postpreconditioning) resulted in a small amount of additional IB␣ degradation. As the interval between preconditioning and challenge increased, the magnitude of challenge-induced IB␣ degradation also increased. These data suggested that the TNF␣-NF-B system possessed a built-in refractory period following TNF␣ treatment that prevented cells from fully responding to a second exposure to ligand. To quantify this phenomenon independent of confounding factors that may affect dynamic bioluminescence readouts (e.g. D-luciferin, ATP, O 2 , or pH dynamics) and to verify its existence for endogenous IB␣, we performed a similar experiment, but instead of live cell imaging, we harvested whole cell lysates at t x ϩ 25 min (time of maximal IB␣ degradation after a ligand challenge given at t x (Fig. 1C); for a schematic timeline, see Fig. 2B). IB␣-FLuc reporter levels in these lysates were analyzed by bioluminescence imaging (upon the addition of saturating D-luciferin and ATP), and endogenous IB␣ levels were determined by Western blot analysis and semiquantitative densitometric analysis (Fig. 2C). From these data, we were then able to calculate responsiveness levels for both IB␣ and IB␣-FLuc as a function of time after TNF␣ preconditioning. Responsiveness at each challenge time was calculated by determining the magnitude of IB␣ degradation induced by TNF␣ challenge divided by the magnitude of IB␣ degradation in unpreconditioned cells from the same plate. Specifically, the ratio at t x ϩ 25 min of IB␣ in preconditioned cells challenged with TNF␣ over preconditioned cells challenged with vehicle was divided by the ratio at t x ϩ 25 min of IB␣ in unpreconditioned cells challenged with TNF␣ over unpreconditioned cells challenged with vehicle, the latter ratio representing the maximal possible response. We observed a strong correlation (r ϭ 0.95) between levels of responsiveness for endogenous IB␣ and IB␣-FLuc (Table 1). Consistent with our earlier observations derived from live cell dynamic bioluminescence imaging experiments ( Fig. 2A), we observed that at 30 min postpreconditioning, cells were approximately half as responsive as naive (i.e. unpreconditioned) cells to a TNF␣ challenge and had gained full responsiveness by 120 min. Thus, a transient refractory period seemed to exist from 30 to 120 min post-TNF␣ preconditioning that rendered the cells unable to fully respond (as measured via IB␣ degradation) to a second challenge of TNF␣, and beyond this period, the cells were able to mount a full response to a second TNF␣ challenge. Notably, similar experiments performed with HeLa cells stably expressing pCMV3 IB␣-FLuc (HeLa IB␣-FLuc (20)) yielded almost identical results (data not shown), suggesting that 1) the TNF␣-induced transient refractory period was not limited to hepatocytes and 2) this effect was independent of both NF-B-induced IB␣ transcription and the initial levels of IB␣-FLuc (substantially higher in HeLa IB␣-FLuc (20)).
The Ligand-induced Transient Refractory Period for IB␣ Processing Correlated in Part with Temporal Down-regulation of IKK but Not Receptor Dynamics-Hypothetically, this loss and regain of the capacity of cells to process IB␣ can be explained by 1) internalization or shedding of TNF␣ receptors, followed by their recycling to the cell membrane (26,27), 2) transient down-regulation of IKK activity as previously reported (15,28), or, alternatively, 3) by a yet unknown mechanism of regulation, downstream of IKK. We therefore sought to establish the relative contributions of receptor dynamics and IKK regulation to this refractory period.
To determine the extent of receptor dynamics in governing the observed loss and regain of IB␣ processing, we took advantage of a discovery, made 20 years ago (14), that IL-1␤ induces transient downregulation of TNF␣ receptors but not vice versa (i.e. TNF␣ has no effect on either the affinity or the number of IL-1␤ surface receptors), as tested in a variety of cell lines and primary cells. Hence, we aimed to determine IB␣ responsiveness to an IL-1␤ challenge as a function of time after TNF␣ preconditioning in HepG2 cells. Cells expressing pB 5 3 IB␣-FLuc were treated with a 30-s pulse of TNF␣ (20 ng/ml), followed by a continuous challenge with IL-1␤, initiated at increasing P-C intervals (0 -240 min). IB␣ processing was analyzed

TABLE 1 Percentage responsiveness of IB␣ processing
Quantification of IB␣-FLuc and IB␣ responsiveness to a second continuous challenge of TNF␣ at the indicated interval following a 30-s preconditioning pulse of TNF␣ was determined from the bioluminescence imaging and Western blot data shown in Fig. 2C. Responsiveness at each challenge time was calculated by determining the percentage of challenge-specific IB␣ degradation divided by the percentage of IB␣ degradation in unpreconditioned cells from the same plate. The responsiveness of IKK was determined by an IKK kinase assay. by live cell dynamic bioluminescence imaging (Fig. 3A). Using this experimental setup, we again observed a transient refractory period (from 30 to 120 min post-TNF␣ preconditioning) during which HepG2 cells exhibited decreased IB␣ responsiveness. The magnitude of the ligand-induced degradation increased as the interval to the IL-1␤ challenge increased, becoming fully responsive again by 120 min (Fig. 3A). These data suggested that even in the absence of ligand-induced receptor desensitization or cross-regulation, the capacity of cells to process IB␣ was compromised within the first 2 h after a short TNF␣ stimulation. We next aimed at deciphering whether transient down-regulation of IKK activity could explain the observed loss and regain in IB␣ responsiveness. We therefore performed a series of IKK kinase assays in order to directly measure the temporal activity profile of IKK, a central junction of the TNF␣ and IL-1␤ pathways that integrates signals from a myriad of upstream regulators (e.g. TNF receptor-associated factors, mitogen-activated protein/extracellular signaling-regulated kinase kinase kinase, TGF␤-activated kinase-binding protein, TGF␤-activated kinase, NF-B-inducing kinase, receptor-interacting protein, A20, protein kinase C, etc. (2,7,29)). HepG2 cells were treated with TNF␣ (20 ng/ml) either as a 30-s pulse or continuously. At the indicated time points, cells were harvested, and IKK complexes were immunoprecipitated and assayed for their capacity to phosphorylate exogenous GST-IB␣-(1-54) (23). We found that for both 30-s pulses and continuous TNF␣ exposure, temporal profiles of IKK activity were almost identical, with both peaking at 10 min. However, consistent with our earlier findings that continuous TNF␣ treatment elicits greater IB␣ degradation than a 30-s pulse (Fig. 1C), continuous TNF␣ treatment exhibited slightly elevated and more sustained levels of IKK activity compared with pulsed TNF␣ treatment (Fig.  3B). Importantly, Western blot analysis showed that IKK complex levels (as determined by IKK␣ protein) did not change over the experimental time course (Fig. 3C), confirming that the increase in net kinase activity was due specifically to IKK activation.

Responsiveness of IB␣
IKK-KA data were also collected from preconditioned cells, 10 min post-challenge (at the time of maximal IKK activity; see Fig. 3B) at increasing P-C intervals (0 -240 min). Using these data together with the IKK activity profiles generated for 30-s pulse and continuous TNF␣ treatment regimens (Fig. 3B), we were able to calculate the net capacity of IKK to phosphorylate IB␣ as a function of time after TNF␣ preconditioning (i.e. IKK responsiveness (Table 1); see "Experimental Procedures" for details on this calculation). Based on this calculation, we noted that the capacity of IKK to respond to a second challenge of TNF␣ was significantly compromised at 30 min post-TNF␣ preconditioning and then gradually increased, reaching ϳ75% responsiveness by 120 min. Up to 240 min, IKK activity did not fully recover to initial levels, consistent with other reports indicating that upon TNF␣ stimulation, IKK activity rapidly and transiently declines due to autocatalytic C-terminal hyperphosphorylation (15) and protein phosphatase 2C-dependent dephosphorylation (16), followed by late NF-B-dependent down-regulation, a process attributed, in part, to A20, an IKKinhibitory protein (29). Hence, these data suggested that 1) the observed ligand-induced transient refractory period of IB␣ processing (Figs. 2 and 3 and Table 1) correlated only in part with ligand-induced transient down-regulation of IKK activity and that 2) the level to which cells are able to degrade IB␣ was not linear with the capacity of IKK to phosphorylate IB␣ (i.e. full IB␣ responsiveness was observed as soon as 120 min post-TNF␣ preconditioning (Figs. 2 and  3A), a time point where IKK responsiveness was still compromised (Table 1)). These data indicated that either submaximal IKK activity could now fully support ligand-induced IB␣ degradation following the refractory period or that additional ligand-responsive elements existed that converged on IB␣ to induce a full response.

Computational Modeling of NF-B Signaling Suggested an
Additional Layer of Regulation, Downstream of IKK, Governing the Observed Refractory Period for IB␣ Processing-The NF-B pathway provides an excellent example of a complex signaling system employing numerous temporally distinct autoregulatory mechanisms and negative feedback loops. IKK enzymatic activity, which is both endogenously and exogenously regulated, controls the degradation of its own substrate (IB␣), which is later strongly up-regulated in an NF-B-dependent manner (Fig. 1A). Rapid changes in substrate availability, conformation, and subcellular localization imply that alternative mechanisms of regulation might exist other than changes in enzymatic activity. Although a ligand-induced transient refractory period of IB␣ processing could be explained in part by down-regulation of IKK activity, we were intrigued to examine whether an alternative regulatory mechanism, based on substrate (IB␣) dynamics, might exist to complement or "back up" IKK regulation. Obviously, inhibition of IKK was not a viable option for analyzing downstream regulation, since such inhibition will result in complete loss of responsiveness in the absence or presence of preconditioning. We therefore decided to undertake a computational approach and explore IB␣ dynamics in silico, assuming no down-regulation of IKK activity. We used a well accepted computational model that used experimentally or hypothetically driven IKK activity profiles as inputs and, in return, calculated ligand-induced dynamics of 24 different subpopulations of mediators on the IKK-NF-B axis.
As a first step, to test the robustness of the model, we sought to compare our IB␣-FLuc bioluminescence imaging data for 30-s pulsing and continuous TNF␣ treatments (Fig. 1C) with the dynamics of IB␣, as predicted by the computational model. To accomplish this, we used as inputs the IKK activity profiles generated for 30-s pulse and continuous TNF␣ treatment regimens (Fig. 4A, left; see "Experimental Procedures" for details on numerical processing of the raw data to fit the model). The dynamics of six different free and complexed IB␣ subpopulations could be predicted by the model (i.e. free IB␣ cyt , IB␣⅐IKK cyt , IB␣⅐NF-B cyt , IB␣⅐IKK⅐NF-B cyt , free IB␣ nuc , and IB␣⅐NF-B nuc ). Since live cell bioluminescence imaging of IB␣-FLuc could not distinguish between these populations, we summed up the predicted concentrations of all IB␣ subpopulations and plotted the predicted total IB␣ levels as a function of time (Fig. 4A, right). For both treatment regimens, we noted an excellent correlation between the predicted profiles of IB␣ and the experimentally generated profiles of IB␣-FLuc (Fig. 1C). The timing and extent of IB␣ degradation as well as the overall dynamic behavior were highly similar. However, differences in the amplitude and timing of resynthesis (experimental: ϳ8-fold initial at ϳ120 min; computational: 1.2-1.5-fold initial at ϳ90 min) were observed and could be explained by dynamic differences between the endogenous IB␣ promoter and the synthetic B 5 -TATA promoter driving IB␣-FLuc (i.e. differences in binding affinity and cooperativity toward NF-B).
We next generated hypothetical IKK profiles representing IKK activities from preconditioned/challenged cells, assuming no upstream receptor or IKK regulation (i.e. experimentally derived challenge-specific IKK activity was overlaid on top of experimentally derived precondition-specific residual IKK activity). These hypothetical IKK activity profiles (Fig. 4, B-E, left, each generated with a different P-C interval) were used as inputs for computing total IB␣ dynamics (Fig. 4, B-E, right). Surprisingly, the computational model predicted that even in the absence of receptor dynamics or IKK regulation, IB␣ processing would be transiently compromised (compare, for example, the second, challenge-induced degradation phase at 120 or 240 min with the ones at 30 or 60 min). These data suggested that although IKK down-regulation partially correlated with the ligand-induced transient refractory period for IB␣ processing, an additional regulatory mechanism was present downstream of IKK. Importantly, IB␣ availability per se was not FIGURE 4. Computational simulation of IB␣ responsiveness in the absence of upstream receptor or IKK regulation. A, interpolated and extrapolated (0 -360 min, at 5-min intervals) IKK activity profiles (right) of cells treated continuously (C, green curve) or by a 30-s pulse (P, blue curve) of TNF␣ (20 ng/ml) were used as inputs to computationally simulate total IB␣ dynamics (right). B-E, left panels, hypothetical IKK activity profiles of preconditioned cells, challenged at the indicated times (denoted by black arrowheads) with a second, continuous dose of TNF␣ were generated by superimposing the continuous TNF␣-induced IKK profiles at increasing intervals after the 30-s pulse TNF␣-induced IKK profile. For generating these hypothetical profiles, we assume no preconditioning-induced receptor or IKK regulation. Right panels, the hypothetical IKK profiles were used as inputs into the model to predict IB␣ dynamics. Note that challenge-induced IB␣ degradation (initiated at the red arrowhead) is recovered in a time-dependent manner. sufficient to explain changes in IB␣ responsiveness, because, as confirmed experimentally and computationally, at 60 min post-preconditioning, the IB␣ concentration had already recovered, whereas degradation potential was still low (compare Fig. 2, A and C, Table 1, and Fig. 4C).
Nuclear Export of IB␣-NF-B complexes may also control the capacity of cells to process IB␣. Having demonstrated experimentally the phenomenon of a ligand-induced transient refractory period for IB␣ processing and after dissecting biochemically and computationally the origins of this observation, we next sought to more closely examine the components of the computational model in order to identify candidates, downstream of IKK, capable of regulating IB␣ responsiveness. While examining the rate constants of a variety of reactions used by the model, we noticed that free versus NF-B-bound IB␣ differed tremendously in their capacity to associate with IKK (1.35 versus 11.1 M Ϫ1 min Ϫ1 , respectively) and to be degraded in an IKK-dependant manner (0.12 versus 0.00006 min Ϫ1 , respectively). These differences in IKK association and ligand-induced degradation were experimentally established by Zandi et al. (30).
This led us to put forward the following model (Fig. 5A). 1) free IB␣ and NF-B-bound IB␣ represent "protected" and "unprotected" populations with respect to ligand-induced, IKKdependent proteasomal degradation. 2) Under steady-state conditions, there is a stoichiometric excess of IB␣ over NF-B in the cytoplasm (ϳ0.7 NF-B per IB␣ according to the model). This may explain our observations that even at saturating concentrations of TNF␣ or IL-1␤, IB␣ degradation never exceeded 70 -80% of initial level (e.g. Fig. 1C). 3) Upon ligand stimulation, NF-B-bound IB␣ is degraded, NF-B translocates to the nucleus, and IB␣ is resynthesized. 4) At this point, although IB␣ is highly abundant, its capacity to be degraded in response to a second stimulus is still severely compromised, because NF-B is in the nucleus. 5) IB␣ can freely shuttle between the cytoplasm and the nucleus, pulling NF-B molecules (that lack nuclear export signals (31)) back to the cytoplasm. This step may be the rate-limiting step for acquisition of full responsiveness. 6) Newly synthesized IB␣ molecules uncomplexed with NF-B are rapidly degraded (32), and only after all NF-B molecules are recovered back to the cytosol and the NF-B-bound IB␣ over free IB␣ ratio returns to prestimulation levels (ϳ0.7) are cells able to mount a full response again.
To experimentally examine the nuclear export hypothesis, we sought to analyze ligand-induced changes in cytoplasmic IB␣⅐NF-B complexes. However, the computational model predicted that ligand-induced changes of cytoplasmic IB␣⅐NF-B and total cytoplasmic NF-B were essentially the same (i.e. at any given time, virtually all cytoplasmic NF-B was bound to IB␣; Fig. 5B), suggesting that monitoring cytoplasmic total NF-B was an excellent approximation for following cytoplasmic IB␣⅐NF-B complexes. We therefore pulsed HepG2 cells for 30 s with TNF␣ (20 ng/ml), and at various times after stimulation, we fixed, permeabilized, and immunostained the cells for p65 NF-B (Fig. 5C). We found that upon a 30-s TNF␣ pulse, p65 rapidly translocated to the nucleus (maximal by 30 min) but by 60 -120 min was back in the cytoplasm. The excellent temporal correlation between the levels of cytoplasmic NF-B (as derived computationally or experimentally; Fig. 5, B and C, respectively) and the competence of cells to degrade IB␣ in response to a proinflammatory ligand (i.e. Table 1) strongly suggested that nuclear transport of NF-B provided a potential alternative mechanism to transiently desensitize IB␣ processing (refractory period), in addition to the mechanism of IKK down-regulation (Fig. 3, B and C, and Table 1).

DISCUSSION
Ligand-induced desensitization is a common theme in many biological systems (13), thereby allowing cells to mount an appropriate response independently of ligand exposure time. Thus, prolonged exposures will not result in excessive responses, but instead, cells are enabled to build up a downstream response while being unable to perceive a second activating cue. Desensitization and resensitization are traditionally perceived to be linked to receptor dynamics (internalization, shedding, and recycling); however, any mediator or regulator along a signaling pathway can be hypothetically desensitized, therefore transiently blocking signal transduction (13).
In this work, we demonstrated that although cells can efficiently activate NF-B in response to a TNF␣ exposure as short as 30 s, such stimulation was followed by a refractory period during which the capacity of cells to respond to a second homologous or heterologous stimulus was severely compromised. We further found that this transient refractory period correlated in part with a temporal down-regulation of IKK activity but not with receptor desensitization. Computational modeling enabled us to identify an additional layer of regulation, downstream of IKK, controlling the capacity of cells to respond to a second challenge. Ligand-induced dynamic changes in substrate (IB␣) availability, conformation, and subcellular localization form the basis for this mechanism. Further analysis led us to conclude that nuclear export of NF-B may be a rate-limiting step in controlling IB␣ homeostatic metabolism, a term recently coined by O'Dea et al. (33).
Our study highlights the multifaceted regulation of NF-B signaling (Fig. 6) and sheds light on the refractory nature of IB␣ processing as a route to transiently desensitize NF-B activity upon subsequent rounds of stimulation. Rapid and transient deactivation of IKK activity as well as temporal reduction in its capacity to respond to a subsequent challenge (IKK responsiveness) seems to play a crucial role in this process. Previous studies indicated that both the amplitude and the timing of IKK activation affect not only the intensity of NF-B-dependent transcription but also the specificity of the transcriptional response (23,34). This indicated that besides resolution of the inflammatory response and induction of a refractory period (temporally preventing subsequent rounds of IB␣ degradation upon restimulation), rapid down-regulation of IKK activity (28) plays a pivotal role in determining the type of elicited transcriptional program.
In addition to IKK regulation, our work demonstrated that nuclear export of IB␣⅐NF-B complexes may have also regu-lated IB␣ responsiveness (Figs. 4 and 5). This suggested that NF-B positively controls IB␣ both transcriptionally and post-translationally. Such double-layered feedback regulation ensures that NF-B transcriptional activity will fully resume only after reconstitution of the cytosolic pool of NF-B. Two other IB isoforms, IB␤ and IB⑀, are degraded more slowly under both TNF␣-induced and unstimulated conditions (23,25) and have been implicated in dampening IB␣-mediated oscillations of NF-B activity (25,35,36). IB⑀ has been shown to be highly NF-B-inducible in mouse embryo fibroblasts and to contribute to nuclear export of NF-B but only at times greater than 3 h poststimulation (37). Thus, it is seems unlikely that IB⑀ contributes substantially to the export of NF-B over the 2 h of the refractory period observed in the present study. It may be interesting to determine whether similar transient refractory periods exist for processing of other IB isoforms.
Although TNF␣-induced resynthesis of endogenous IB␣ peaks at ϳ60 -90 min after onset of stimulation (as validated both experimentally and computationally ( Fig. 1D and 4A, respectively)), maximum levels of newly synthesized IB␣-FLuc reporter were observed ϳ120 min after TNF␣ stimulation (Figs. 1, C and E, and 2A). This discrepancy may be explained by differences likely to be present in affinity and cooperativity of binding of NF-B to endogenous versus synthetic promoters (the endogenous promoter contains three distant B sites, whereas the synthetic promoter contains five tandem high affinity B sites). Nevertheless, since both endogenous IB␣ and IB␣-FLuc exhibit similar half-life times (20), differences in the timing of resynthesis cannot be explained by differences in turnover rate. Following the peak of IB␣ resynthesis, both endogenous IB␣ and our IB␣-Fluc reporter begin returning to base-line levels faster under continuous TNF␣ treatment, suggesting that ligand-induced reactivation of IB␣ degradation is occurring under continuous TNF␣ exposure, as expected (23).
In the present and previous studies (20), we demonstrated that dynamic bioluminescence imaging of IB␣-FLuc reporters in live cells provides robust and accurate readouts of ligand-induced IB␣ dynamics. In effect, real time bioluminescence imaging was equivalent to performing continuous on-line Western blots of IB␣ at 5-min intervals. An analogous transcriptionally coupled reporter (kB 5 3 IB␣-EGFP) was generated by Nelson et al. (35) for monitoring IB␣ dynamics in single cells by live cell fluorescence microscopy. Although such a system provides the means to monitor ligand-induced translocations and oscillations in IB␣ levels, temporal resolution of this reporter is limited by the long maturation time of EGFP (Ͼ1 h) (38,39). This notion and the fact that Nelson et al. (35,36) co-overexpressed p65-red fluorescent protein may explain the vast difference between the observed period of IB␣-EGFP oscillations (ϳ300 min) and the period of endogenous IB␣ oscillations, as predicted computationally (ϳ90 -120 min) (25). Although longer term IB␣ oscillatory behavior was not the focus of the present study, we did observe single oscillations within ϳ150 -180 min. Because FLuc is active immediately upon translation, our reporter should afford greater temporal resolution, enabling accurate readouts of IB␣ dynamics and oscillations in live cells for such studies as well as the multistimulation protocols as described herein.
Of note, a previous study aimed at analysis of IB␣ stabilization indicated a role for p38 in IB␣ stabilization and, in some cell lines, in prevention of sequential degradation of IB␣ upon concurrent exposure to TNF␣ following continuous pretreatment with IL-1␤ (40). However, since IL-1␤ has been shown to induce rapid and dramatic down-regulation of TNF␣ receptors (but not vice versa) (14), inhibition of TNF␣-induced IB␣ processing, as observed by Place et al. (40), could be attributed directly to receptor dynamics rather than IB␣ stabilization. This confounding factor highlights the importance of asymmetric receptor cross-desensitization, a phenomenon that remains poorly understood but has far reaching physiological consequences.
In conclusion, TNF␣ preconditioning protocols and dynamic imaging revealed a transient suppression of the capacity of cells to process IB␣. This refractory period for IB␣ processing was controlled both by IKK activity and NF-B distribution. In particular, the data suggested that nuclear export of NF-B may provide additional rate-limiting regulation governing the refractory period machinery. These regulatory mechanisms provide a "molecular timer" controlling the ampli-  tude, timing, and specificity of the NF-B-mediated transcriptional program.