B metabolism to control constitutive NF-Introduction
Cellular signal transduction pathways mediate responses to extracellular and intracellular signals, such as changing environmental and metabolic conditions, pathogen assault, and developmental cues. Many signaling pathways control the activity of transcription factors that regulate cognate target genes (Brivanlou and Darnell, 2002). For immediate early transcriptional responses (not requiring induced synthesis), such regulation may involve the reversible phosphorylation of the transcription factor to induce dimerization or nuclear translocation (e.g. the Stat, IRF, AP-1 transcription factor families). An alternate means of pathway activation involves stabilization of the transcriptional effector, as in the case of the genotoxic response regulator p53, the hypoxia response factor HIF-1
, or the developmentally regulated coactivator 
-catenin. Thus, signaling in response to stimulus involves alterations of the homeostatic rates of synthesis and degradation found in unstimulated cells.
In contrast, the cellular abundance of the transcription factor NF-
B does not change dramatically during signaling. NF-B is the critical mediator of cellular responses to a large number of physiological stimuli, including inflammatory cytokines, developmental signals, pathogens, and cellular stresses (Figure 1A) (Hoffmann and Baltimore, 2006). Although inflammatory signaling leads to transient NF-B activity that is dynamically regulated by feedback mechanisms, elevated constitutive levels of active NF-B are associated with chronic inflammatory diseases and many types of cancer (Karin, 2006).
Figure 1
Exploring the relative importance of IB degradation mechanisms by computational parameter sensitivity analysis. (A) Schematic of the NF-B signaling module and its physiological importance in the transduction of diverse inflammatory, developmental, and stress signals. (B) Illustration of the four IB degradation pathways within the NF-B signaling module. deg1 and deg4 are IKK-independent degradation rate constants for free and bound IB. r1 and r4 are IKK-dependent degradation rate constants for free and bound IB. (C) Computational simulation of NF-B activation over a 6-h time course. TNF stimulation begins at time 0, and is removed at 4 h. Mean activity in the first hour of stimulation and the second hour after removal of the stimulus (shaded in gray) were used to create the plots in (D–F) and (G). (D–G) Graphs showing the average nuclear NF-B (y axis) during the first hour (D, F) or during the second hour after 4 h (E, G) of TNF stimulation for different values (x axis) of the IKK-dependent (D, E) or -independent (F, G) degradation rate constants of free (blue line) and bound (red line) IB.
NF-B activity is inhibited by association with the inhibitor proteins, IB, IB, or IB
, which mask its nuclear localization sequence and inhibit its DNA-binding activity. The regulated metabolism of IB proteins—their synthesis and degradation—critically controls NF-B signaling (Ghosh et al, 1998). Synthesis of IB proteins is a highly regulated process, with at least two isoforms, IB and IB, being subject to NF-B-inducible synthesis, thereby providing negative feedback (Scott et al, 1993; Kearns et al, 2006). Stimulus-induced IB degradation is controlled by the IB kinase (IKK), which phosphorylates two N-terminal serines. This leads to IB polyubiquitination and degradation via the 26S proteasome, thus liberating NF-B for nuclear translocation (Ghosh et al, 1998; Yaron et al, 1998).
These processes were described in a mathematical model of the IKK-IB-NF-B signaling module to recapitulate NF-B activation in response to TNF stimulation (Hoffmann et al, 2002). Its construction relied on rate constants available in the literature from a diverse set of experiments. As no isoform-specific data were available, rate constants pertaining to IB and IB were assumed to be the same as those measured for IB. Although this model accurately recapitulates NF-B signaling in response to TNF, in the unstimulated state the estimated IB levels were found to be unexpectedly high (Lipniacki et al, 2004). In fact, the vast majority of IB were calculated to be in the free form, contradictory to experimental studies showing that free IB accounts for less than 15% of the total cellular IB (Rice and Ernst, 1993).
Despite our detailed understanding of stimulus-induced NF-B signaling, there is less clarity about the mechanisms mediating IB turnover in the absence of external stimulation. Early studies reported that basal turnover of IB, unlike its induced degradation, does not require the IKK-targeted serines, the C-terminal PEST domain, or poly-ubiquitination of IB (Krappmann et al, 1996), whereas others found robust C-terminal phosphorylation and poly-ubiquitination (Pando and Verma, 2000).
By distinguishing between NF-B-bound and free IB pools using an IB interaction mutant, the half-life of bound IB was found to be five-fold longer than that of free IB in unstimulated cells (Pando and Verma, 2000). However, free IB is a poorer substrate for IKK than NF-B-bound IB (Zandi et al, 1998), although it is routinely used as a substrate to measure IKK activity in vitro. Free IB turnover was proposed to involve casein kinase 2 (CK2)-mediated phosphorylation of the C-terminal domain and ubiquitination (Schwarz et al, 1996; Bren et al, 2000), but others suggested that CK2 is involved in inducible degradation of NF-B-bound IB (Kato et al, 2003), or that ubiquitination was not required (Krappmann et al, 1996; Alvarez-Castelao and Castano, 2005).
Given these contradictory results in the literature, the lack of data on two of the three IB isoforms, and the poor fit of computational simulations of the NF-B signaling module in cells not exposed to TNF, we generated genetic tools—mouse knockout cell lines—to isolate cleanly the endogenous-free and -bound IB protein pools and probe their degradation with kinase knockouts and pharmacological inhibitors. In addition, we used computational modeling (i) to identify which constitutive degradation rate constants play a critical role in determining stimulus responsiveness, (ii) to determine new biochemical rate constants based on our experimental results, (iii) to confirm the validity of the new parameters by simulating the cellular steady state, and (iv) to reveal the control of IB degradation by NF-B as a cross-regulatory mechanism.
Results and discussion
IKK-dependent and -independent degradation of I
Bs determine NF-B signaling
Four degradation rate constants govern the in vivo half-life of IB proteins (Figure 1B). An IB molecule can exist in either the free or NF-B-bound form. Both forms may be degraded in an IKK-dependent manner (we denote the IB rate constants of these processes r1 and r4, respectively), but are also subject to constitutive degradation in an IKK-independent manner (with the rate constants denoted as deg1 and deg4). These mechanisms are described as first-order rate constants in our mathematical model of NF-B signaling (Hoffmann et al, 2002).
To explore the functional significance of each IB degradation rate constant in NF-B signal transduction, we performed simulations of TNF signaling after altering one of the four rate constants (simultaneously for the IB, , and isoforms) with a parameter multiplier ranging from 0.01 to 100. For each parameter multiplier, we calculated the average nuclear NF-B level in response to TNF during the early phase (during the first hour of stimulation) and the later attenuation phase (during the second hour after a 4 h stimulation) (Figure 1C). By plotting the calculated NF-B activity against its parameter multiplier, we can interpret the sensitivity of the system to each rate constant for two critical features of the NF-B response to a transient TNF stimulus: activation and attenuation of NF-B activity.
We first examined the impact of changes in IKK-dependent IB degradation rate constants on NF-B activation. During the first hour of TNF stimulation, the amount of nuclear NF-B calculated by the model is fairly insensitive to even drastic changes in the IKK-dependent degradation rate of free IB (Figure 1D, blue line). In contrast, slowing down the IKK-induced degradation of NF-B-bound IB severely dampens NF-B activity (Figure 1D, red line). During the attenuation phase, the amount of nuclear NF-B predicted by the model was similarly found to be insensitive to changes in IKK-dependent degradation of free IB (Figure 1E, blue line), but slowing the IKK-dependent degradation rate of bound IB results in a loss of attenuation (Figure 1E, red line).
The IKK-independent IB degradation rates control the stimulus-independent turnover of IB proteins, and thus maintain a resting state equilibrium of IB levels. Examining whether these IKK-independent degradation rates play a role in determining the cellular responsiveness to inflammatory stimuli revealed that during the first hour of TNF stimulation the signaling module is dramatically more sensitive to the basal turnover rate of free IB (Figure 1F, blue line) than of bound IB (Figure 1F, red line). Furthermore, our simulations predicted that a more stable free IB results in a loss of attenuation, whereas the basal turnover rate of the bound IBs had no effect (Figure 1G).
In sum, our computational simulations revealed that two of the four possible degradation pathways play a particularly important role in controlling NF-B signaling. Whereas much is known about the stimulus-responsive IKK-mediated degradation pathway, the IKK-independent degradation mechanism of free IB has received surprisingly little experimental attention. Given the importance of these degradation rate constants in our computational analysis, we set out to examine them in more detail experimentally.
NF-B regulation of IB protein turnover and synthesis
To measure experimentally in vivo degradation rate constants for NF-B-bound IB proteins, we used the ribosomal inhibitor cycloheximide (CHX) to reduce the synthesis of new IB proteins (by 85%; Supplementary Figure S1A), and examined the amount of nuclear NF-B DNA-binding activity via electrophoretic mobility shift assay (EMSA). Treatment of wild-type MEFs with CHX over a 60 h time course induced nuclear NF-B activity that corresponds to 25–35% of peak TNF-induced NF-B activity (Figure 2A, Supplementary Figure S1B). To determine the relative contributions of each NF-B-bound IB isoform (, , and ) to CHX-mediated NF-B activation, we used a panel of IB double-knockout MEFs, which contain only one IB isoform. These cells were previously used to determine the degradation rate constants for each IB isoform by TNF-induced NF-B activation, which revealed that upon IKK activation, IB was degraded most rapidly, followed by IB, and then IB (Hoffmann et al, 2002). Interestingly, we find the same trend in stimulus-independent degradation, where IB is the most stable and IB is the least stable (Figure 2B and Supplementary Figure S1C).
Figure 2
Experimental studies of degradation pathways of NF-B-bound and -free IB proteins. (A) NF-B activity as measured by EMSA of nuclear extracts from wild-type cells treated with 10 
g/ml CHX or 1 ng/ml TNF for indicated times. (B) NF-B activity as measured by EMSA of nuclear extracts from ikb-/--/-, ikb-/--/-, or ikb-/--/- cells treated with 10 g/ml CHX or 1 ng/ml TNF. (C) Western blot for IB in CHX-treated wild-type and nfkb-/- cells. The first two lanes show ib-/--/--/- extract and ib-/--/--/- extract mixed with 5% wild-type extract to show the protein level of IB in the nfkb-/- cells was approximately 5% that in the wild-type cells at time zero. (D) Cytoplasmic extracts of wild-type cells were immunoprecipitated with IKK
antibody and subject to an in vitro kinase assay. In the 'mock' lane, no antibody was added during the IP. (E) NF-B activity as measured by EMSA of nuclear extracts from ikk-/--/- or wild-type MEFs treated with 10 g/ml CHX. (F) Western blot for IB of protein extracts from TNF (1 ng/ml)-treated wild-type or rela-/-crel-/-nfkb1-/- cells. (G) Western blots for IB of protein extracts from TNF-treated wild-type cells (top panel) in the presence or absence of the IKK-inhibitor sc-514. Bottom panels show Western blots for IB of protein extracts from CHX (10 g/ml)-treated cells in the presence or absence of sc-514. (H) Western blots for IB of protein extracts from wild-type cells treated with TNF (1 ng/ml) with or without the presence of the proteasome inhibitor MG132 (top panel). Western blots for IB of protein extracts from nfkb-/- cells treated with 10 g/ml CHX, 10 M MG132, or both.
To investigate the stability of the unbound, or 'free', IB proteins in resting cells, we generated crel-/-rela-/-nfkb1-/- (termed 'nfkb-/-') MEFs deficient in the three NF-B proteins known to interact with the classical IB proteins: RelA, c-Rel, and p50. Western blots revealed a dramatic reduction in the amount of total IB protein level in these cells compared to wild type (Figure 2C, compare lanes 3 and 6). A dilution series of wild-type protein extract with ikb-/--/--/- extract showed that the amount of IB in the nfkb-/- cells was approximately one-twentieth the amount in wild-type cells, and that this ratio is probably even lower for IB and IB (Supplementary Figure S2A). No decrease in IB levels was detected in MEFs deficient in the NF-B proteins RelB and nf
b2 p52, which are non-canonical NF-B proteins that do not bind canonical IB proteins.
Strikingly, the level of IB mRNA in the nfkb-/- cells was only two-fold lower than in wild type, with even smaller differences in IB and IB mRNA levels (Supplementary Figure S2B), suggesting that differential protein stability may account for different IB protein levels in wild-type and nfkb-/- cells. Indeed, treating nfkb-/- cells with CHX resulted in rapid decreases of IB protein, whereas it remains stable in the wild-type cells beyond 2 h (Figure 2C). These results suggest that NF-B has a regulatory role not only in controlling IB transcription, but also in stabilizing IB proteins.
We next investigated whether the dramatically different half-life of free and bound IB proteins may be due to different mechanisms governing their degradation.
IKK phosphorylation is a key mediator of the stimulus-induced degradation of NF-B-bound IB proteins, yet it is unclear if and how IKK may participate in the basal degradation of bound IB. We first performed a kinase assay to examine the IKK activity of immunoprecipitated IKK complex from wild-type MEFs. Surprisingly, even in resting cells, a substantial amount of basal activity associated with the IKK complex was detectable (Figure 2D). In cells lacking the IKK catalytic subunits, IKK and IKK, no activation of NF-B upon CHX treatment (Figure 2E) was observed, indicating that IKK-dependent phosphorylation is required for the basal turnover of NF-B-bound IB proteins.
We sought to determine if IKK activity is involved in the turnover of free IB proteins as well. IP-IKK kinase assays determined that the basal and inducible IKK activities are intact in the nfkb-/- cell line (Supplementary Figure S3A). Treatment of wild-type cells with TNF led to the rapid degradation of IB, but did not affect the levels of IB in the nfkb-/- cells (Figure 2F and Supplementary Figure S3B), suggesting that the inducible IKK activity is not involved in the degradation of free IB. Further, the IKK inhibitor, sc-514, which diminishes the TNF-induced degradation of IB in wild-type cells, did not have an effect on the basal turnover of free IB in nfkb-/- cells (Figure 2G). In contrast, the proteasome inhibitor MG132 prevented not only TNF-induced degradation of IB in wild-type cells, but also led to accumulation of free IB in nfkb-/- cells, and prevented its degradation when cells were cotreated with CHX (Figure 2H).
Based on our new biochemical data, we revised the parameter values governing degradation of IB proteins within the NF-B signaling module and incorporated these into our mathematical model (now termed model 1.1). Half-lives for free IB proteins were determined to be 5–10 min (Figures 2H and Supplementary Figure S2, allowing us to calculate the respective first order rate constants (deg1-3). Our data highly constrained IKK-independent degradation rate constants (deg4-6) for NF-B-bound IB proteins (Figure 2E). While previous studies suggested that NF-B stabilized free IB degradation by a factor of 5 (Pando and Verma, 2000), our new measurements (Table I) indicate an NF-B effect of 2000-fold with respect to the IKK-independent degradation of IB proteins. This large discrepancy likely lies in the facts that (i) we have used a clean genetic system to isolate free endogenous IB from NF-B proteins, and have thus obtained a much faster degradation rate for free IB and (ii) we have determined that degradation of NF-B-bound IB proteins can only occur through an IKK-involving mechanism and have thus drastically decreased the IKK-independent degradation rate of bound IB.
Table 1:
Rate constants for IKK-independent I
B degradation: a role for NF-B
![Table 1 - Rate constants for IKK-independent I|[kappa]|B degradation: a role for NF-|[kappa]|B](/common/images/table_thumb.gif)
Full tableFigures & Tables index
After incorporation of the new rate constants in Table I, we performed model fitting as described previously (Hoffmann et al, 2002) to obtain new degradation rate constants for IKK-induced degradation of NF-B-bound IB proteins (r4-6; Supplementary Table SI). As IKK-mediated phosphorylation of IB is five-fold more efficient when NF-B is present (Zandi et al, 1998), we divided the newly determined r4-6 by 5 to determine IKK-induced degradation of free IB proteins (Supplementary Table SI, see Supplementary information for rate constant derivations). Including eight-fold differential IKK association rate constants (Zandi et al, 1998), the combined NF-B effect on IKK-mediated degradation of free and bound IB proteins is almost 50-fold.
Our results emphasize that NF-B determines the degradation mechanism of IB proteins. When bound to NF-B, IB turnover is slow and dependent on the basal activity of IKK. In contrast, when not bound to NF-B, IB degradation is rapid and independent of IKK activity.
Cross-regulation between IB proteins via half-life control by NF-B
We compared the steady-state levels of IB predicted for unstimulated cells by the previous version of the model (referred to as model 1.0) (Hoffmann et al, 2002) and the new version of the model that incorporates the new rate constants of Tables I and Supplementary Table I (model 1.1) (Figure 3A, white and gray bars, respectively). The new degradation rate constants result in predictions of a much smaller pool of free IB protein, as well as less total cellular IB protein. The simulation results produced with the new model (1.1) are therefore in much better agreement with experimental observations (Rice and Ernst, 1993) than those with the previous model. Although a previous study (Lipniacki et al, 2004) lowered the IB synthesis rate to correct the model-predicted ratio of free to bound IB protein in the steady state, our new data indicate that the rapid free IB degradation necessitates a high synthesis rate.
Figure 3
An improved model of the homeostatic NF-B signaling module. (A) Model calculations of IB, , and protein levels (nM) in unstimulated cells for total IB, free IB, or NF-B-bound IB. Model 1.0 predictions are white bars, model 1.1 predictions are gray bars. (B) Model calculations of NF-B activity (nM) in unstimulated wild-type and ikb-/- cells predicted by model version 1.0 (white bars), version 1.1 (gray bars), and version 1.2 (black bars). (C) NF-B activity in untreated cells as measured by EMSA of nuclear extracts from the ib-/- cell genotype labeled above each lane. The last three lanes are controls for the NF-B band and are nuclear extracts from wild-type cells treated with TNF (1 ng/ml). (D) RNase protection assay showing levels of IB mRNA in untreated wild-type cells, ib-/- cells with empty vector control, or ib-/- cells expressing a retroviral ib transgene. GAPDH is used as a loading control. (E) Western blots for IB and IB in resting cells. The cell genotype is listed above each lane.
Next, we examined the consequences of the new degradation parameters on constitutive NF-B activity in a series of IB knockout cell lines. The previous model predicts that the removal of IB results in high levels of nuclear NF-B activity in unstimulated cells (Figure 3B), which does not match with our experimental observations (Figure 3C). The differential degradation rates of bound and unbound IB protein may result in molecular compensation among the IB isoforms; upon deletion of a single IB isoform, the newly available NF-B may act to stabilize the remaining IB isoforms, resulting in the cytoplasmic retention of NF-B. Indeed, model 1.1 predicts a lower level of NF-B activity in unstimulated knockout cells than version 1.0 (Figure 3B, compare white and gray bars). EMSA results (Figure 3C) confirm the new predictions, indicating that functional IB compensation via differential half-life control indeed exists.
In the case of ib-/--/- cells where IB is the only isoform present, our model predicts a markedly higher level of NF-B activity than seen experimentally. However, we have recently characterized an NF-B-inducible IB mRNA synthesis mechanism (Kearns et al, 2006). Incorporation of this feedback mechanism into the model (referred to as model 1.2) indeed lowers the predicted basal NF-B levels (Figure 3B, black bars) to levels that are in good agreement with the EMSA results. We measured IB mRNA levels and found that they are indeed upregulated in ib-/- cells compared to wild type (Figure 3D), resulting in higher IB protein levels (Figure 3E). To determine whether this effect was the result of homeostatic regulation within the NF-B signaling module, we used a retroviral transgene to reconstitute IB expression in ib-/- cells. Indeed, we found that IB upregulation was reversible, confirming that even in resting cells constitutive NF-B activity plays a role in transcriptional regulation of its inhibitors to controls its own steady-state activity.
Homeostatic control via distinct IB degradation pathways
Our analysis of the NF-B signaling module in unstimulated cells reveals a highly dynamic homeostatic state that is controlled by multiple synthesis and degradation mechanisms of the regulatory IB proteins. As such we find that NF-B itself has two roles in regulating its own basal activity. NF-B binding to IB proteins removes them from this rapid degradation pathway, and sensitizes them to a slow degradation mechanism that is dependent on basal IKK activity. Second, constitutive NF-B activity also impacts transcription rates of IB and IB, thus providing for negative feedback even in the absence of an external stimulus.
Our studies identify the free IB protein degradation pathway as a major determinant of constitutive NF-B and of stimulus responsiveness of the NF-B signaling module. Given this hitherto unappreciated importance, determining the enzymatic and potentially regulatory mechanisms of the free IB degradation pathway is critical for understanding the regulation of NF-B in diverse physiological and pathological settings.
Owing to the dynamic nature of the IB-NF-B equilibrium, the majority of newly synthesized IB is likely degraded before ever binding NF-B. However, this is not unlike other signal transduction pathways that consume significant cellular resources for the maintenance of a dynamic homeostatic state. For example, the transcription factors p53, HIF-1, and -catenin are continually synthesized and degraded. Upon signaling, the respective degradation pathways are inhibited to allow for their nuclear accumulation and function (Ivan et al, 2001; Jaakkola et al, 2001; Moon, 2005). How may this energy-consuming process of maintaining a dynamic homeostasis benefit the cell? Future computational studies may suggest that homeostatic control of the NF-B signaling module confers sensitivity to signals but ensures a very steady low equilibrium activity that is less likely to drift (D Barken, unpublished results). In addition, combined computational and experimental studies may demonstrate that such a dynamic equilibrium state sensitizes the signaling pathway to metabolic changes, such that stress conditions constitute an input signal that results in cellular responses (Ellen L O'Dea, unpublished results).
Materials and methods
Cells and reagents
Primary and 3T3 immortalized MEF were generated from E12.5–14.5 embryos and maintained as described previously (Hoffmann et al, 2002). rela-/-crel-/-nfkb1-/- MEFs were generated from E12.5-timed matings of rela+/-crel-/-nfkb1-/- mice and ib-/--/--/- cells will be described elsewhere. ikk1-/-ikk2-/- cells were a kind gift from Inder Verma. ib-/- MEF lines reconstituted with pBabe-IB and empty vector control were a generous gift from Erika Mathes. Recombinant murine TNF was from Roche; CHX, sc-514, and MG132 from Sigma. RelA/p65 (sc-372), RelB (sc-226), cRel (sc-71), IB (sc-371), IB (sc-946), and IB (sc-7156) antibodies were from Santa Cruz Biotechnology. Trans35S-methionine label was from MP Biomedicals.
Biochemical analysis
Whole-cell extracts were prepared in RIPA buffer and equivalent protein amounts subjected to immunoblot analysis using ECL-plus (Amersham/GE Healthcare). Nuclear extracts were prepared and used for electrophoretic EMSA as described (Hoffmann et al, 2002). Immunoprecipitation kinase assay performed as in Werner et al (2005). Signals were quantified using a phosphorimager (Molecular Dynamics) and ImageQuant software version 5.2 (GE Healthcare). Dilution series with knockout extracts assured that Western blot signals were in the linear range. Total cellular RNA was isolated with Trizol reagent (Invitrogen) and used for RNase protection assay as described in Kearns et al (2006).
Cells were labeled with 200 Ci/ml 35S-methionine label for indicated times. Whole-cell extracts were prepared in RIPA buffer and dried on filter paper. 35S-Met incorporation was measured by scintillation count and CHX-treated cells versus untreated cells were compared to measure the percentage of translational inhibition.
Computational modeling
The mathematical model of the IKK-IB-NF-B signaling module was described in Hoffmann et al (2002). This model (version 1.0) was used to generate Figure 1. Model version 1.1 includes the parameter values shown in Table I and baseline level of IKK of 1 nM. Simulations were performed in Matlab and Excel as described previously (Hoffmann et al, 2002) with extended equilibration times. A complete list of the parameter values can be found in the Supplementary information. Graphs were generated in Excel. The Matlab code file is available upon request, and the SBML code is available at the MSB website.
