|Home | About | Journals | Submit | Contact Us | Français|
Inflammatory NF-κB/RelA activation is mediated by the three canonical inhibitors, IκBα, -β, and -ε. We report here that nfκb2/p100 forms two distinct inhibitory complexes with RelA, one of which mediates developmental NF-κB activation. Our genetic evidence confirms that p100 is required and sufficient as a fourth IκB protein for non-canonical NF-κB signaling downstream of NIK and IKK1. A mathematical model of the four-IκB-containing NF-κB signaling module accounts for NF-κB/RelA:p50 activation in response to inflammatory and developmental stimuli. By exploring signaling crosstalk between them, we find that the gene expression program induced by lymphotoxin β receptor (LTβR)- signaling is determined by the cellular history of exposure to inflammatory stimuli. Further integrated computational and experimental studies reveal that mutant cells with altered balances between canonical and non-canonical IκB proteins may exhibit inappropriate inflammatory gene expression in response to developmental signals. Our results have important implications for physiological and pathological scenarios in which inflammatory and developmental signals converge.
The transcription factor NF-κB plays critical roles in diverse physiological processes (Ghosh and Karin, 2002) and numerous human pathologies (Karin et al., 2004). The primary mediator of NF-κB transcriptional activity is a RelA:p50 heterodimer. Mouse knockout studies have confirmed that RelA is responsible for the expression of a large number of genes involved in inflammatory responses as well as in cell proliferation, cell adhesion and tissue remodeling (Hoffmann and Baltimore, 2006). NF-κB activity is inducible by a diverse range of stimuli (www.nf-kb.org); these include pathogen derived substances and inter-cellular mediators of inflammation, immune cell maturation and secondary lymphoid organ development. Understanding the mechanisms that regulate NF-κB activity is critical for developing therapeutic strategies for many human diseases (Ghosh and Karin, 2002; Hoffmann and Baltimore, 2006; Karin et al., 2004).
Induction of NF-κB/RelA activity in response to inflammatory stimuli does not require protein synthesis. Instead, early experiments distinguished between two activation mechanisms: a precursor processing mechanism, or regulation by a separate inhibitor protein. The detergent deoxycholate was shown to liberate fully active κB-site DNA binding activity in unstimulated cells, suggesting the existence of a separate inhibitor protein(s), termed IκB (Baeuerle and Baltimore, 1988). Three IκB proteins (IκBα, -β, or -ε) have been identified, which share the bona fide IκB properties of (1) binding NF-κB dimers thereby inhibiting their DNA binding activity and retaining them in a latent state and (2) allowing for NF-κB activation by undergoing stimulus-induced proteolysis. The “canonical” signaling pathway involves the stimulus-responsive phosphorylation of the IκBs by the NEMO and IKK2 containing kinase complex, which tags them for degradation via the ubiquitin-proteasome pathway. Several other IκB-like ankyrin-repeat containing NF-κB binding proteins have been reported to modulate nuclear NF-κB transcriptional activity on a subset of genes (Yamamoto et al., 2004) or, when over-expressed, prevent nuclear localization of NF-κB (Hatada et al., 1992; Inoue et al., 1992; Naumann et al., 1993). However, during inflammatory signaling the dynamic control of NF-κB nucleocytoplasmic localization is mediated by the three classical IκB proteins (Hoffmann et al.; Tergaonkar et al., 2005). A detailed molecular understanding of the biochemical events has allowed for the construction of a mathematical model that recapitulates the experimentally observed signaling behavior in response to inflammatory stimuli (Hoffmann et al., 2002; Kearns et al., 2006; Werner et al., 2005).
In contrast, a second NF-κB activation pathway is thought to regulate the activity of NF-κB/RelB dimers via a precursor processing mechanism. This signaling pathway is activated in response to developmental signals such as those transduced by lymphotoxin beta receptor (LTβR) and RANK, which are required for lymph node and osteoclast genesis and homeostasis (Senftleben et al., 2001; Weih and Caamano, 2003), or BAFFR, CD40 and CD27, which regulate B-cell survival and proliferation (Ramakrishnan et al., 2004; Zarnegar et al., 2004). The activation mechanism was shown to involve NF-κB-inducing kinase (NIK) and IKK1-dependent phosphorylation of the nfkb2 gene product p100 (Dejardin et al., 2002), which triggers its proteasomal processing. This partial proteolysis event removes a C-terminal ankyrin repeat domain (IκB-like domain), to generate the NF-κB protein p52 (Senftleben et al., 2001). Interestingly, only newly synthesized p100 was shown to undergo processing to generate nuclear p52-containing nuclear NF-κB activities (Mordmuller et al., 2003).
The physiological role of LTβR is to transduce signals from hematopoietically derived lymphoid tissue inducer cells expressing membrane bound lymphotoxin (LTαβ) to mesenchymal or stromal cells to initiate critical steps in lymph node development (Mebius, 2003; Rennert et al., 1998). LTβR stimulation results in both RelA and RelB-containing NF-κB dimers (Muller and Siebenlist, 2003). The vascular-cell adhesion molecule 1 (VCAM1), a prominent RelA target gene, mediates early steps in lymph node genesis, the formation of lymph node anlagen during embryogenesis. Subsequently, homing of B-cells, whose interaction with stromal cells is critical for lymph node and splenic microarchitecture, requires lymphoid chemokines SLC/CCL21 and BLC/CXCL13, whose expression is thought to involve RelB (Bonizzi et al., 2004). The maturation of spleen and lymph nodes and continued influx and organization of lymphocytes in these secondary lymphoid organs during adulthood, is also dependent on LTβR signaling and the expression of the RelA target gene MadCAM (Browning et al., 2005). The phenotypes of knockout mice reflect the requirement of both RelA- and RelB-containing NF-κB activities. While RelA-deficiency results in a complete absence of lymph nodes in new born mice indicating an early organogenic defect (Alcamo et al., 2002), RelB appears to be required for their maintenance as RelB-deficient mice exhibit deterioration of nodes following birth (Weih et al., 2001).
Despite the importance of RelA activity in lymph node genesis, the molecular mechanism responsible for LTβR-induced RelA:p50 dimer activation has remained unclear. Both an IκB-dependent (Muller and Siebenlist, 2003) and an IκB-independent (Jiang et al., 2003) activation mechanism has been proposed, and for CD27 signaling the non-canonical signal transducer NIK has been implicated (Ramakrishnan et al., 2004). Here we report the existence of a fourth bona fide IκB protein that mediates RelA:p50 activation in response to non-canonical NF-κB signaling pathways. Reconstruction of the signaling mechanism in a mathematical model allowed us to demonstrate that this mechanism is sufficient to account for the experimental observables and explore signaling crosstalk in cells exposed to diverse NF-κB inducing stimuli.
Current pathway maps of LTβR signaling posit that IKK1-dependent processing of the nfκb2 gene product p100 to p52 results in the nuclear localization of a RelB:p52 dimer. However, the activation mechanism of RelA-containing dimers remains unclear and is generally thought to involve the IKK2-IκB-NF-κB signaling module that also allows for signal transduction of inflammatory stimuli (Figure 1A). As we previously recapitulated NF-κB/RelA activation in response to TNFR (Hoffmann et al., 2002) and TLR4 (Werner et al., 2005) stimulation in a mathematical model, we set out to examine the mechanism of NF-κB/RelA activation downstream of LTβR in a genetically tractable cell culture system to similarly reconstitute the underlying signaling mechanism in silico.
We stimulated mouse embryonic fibroblasts (MEF) with an agonistic antibody raised against LTβR that was shown to functionally complement for the genetic absence of LTαβ (Rennert et al., 1998). Using subsaturating concentrations of antibody that resulted in specific signaling events (Figure S1A-H), we detected nuclear NF-κB DNA binding activity (Figure 1B), that consisted of both RelA:p50 and RelB:p50 dimers (Figure 1C), as previously described (Derudder et al., 2003).
One of the critical steps in NF-κB/RelA activation in response to inflammatory stimuli is NEMO-dependent IKK phosphorylation and degradation of RelA-bound IκBs. Immunoprecipitated IKK activity could be detected in response to very low concentrations (0.01ng/ml) of TNF (Figure 1D, lanes 7–12), but, surprisingly, non-saturating LTβR stimulation conditions did not elicit any NEMO associated IKK activity (Figure 1D, lanes 18–22). Consistently, cytoplasmic immunoblotting revealed no decrease in IκB protein levels (Figure 1E, lanes 8–11), and no inducible IκBα phosphorylation (Figure S1I–J).
These observations led us to search for novel factors that regulate NF-κB/RelA dimers in response to non-canonical stimuli. To this end we utilized a cell line deficient in the three canonical IκB genes, IκBα, -β, and -ε (iκb−/−). Initial biochemical characterization revealed that despite the loss of these known NF-κB inhibitors only a small fraction of cellular RelA was constitutively nuclear, while the majority of RelA was still cytoplasmic (Figure 2A), consistent with observations in IκB knockdown cells (Tergaonkar et al., 2005). To investigate the molecular basis for cytoplasmic sequestration, we reconstituted a cell line also lacking the rela gene (iκb−/−rela−/−) with a tandem epitope tagged form of RelA (TAP-RelA) by retrovirus-mediated transduction (Figure S2A).
Tandem affinity purification of epitope tagged RelA in non-denaturing conditions followed by silver staining of purified proteins separated on SDS-PAGE as well as mass-spectrometric analysis revealed nfkb1 p50 and nfkb2 p100 as the two major interacting proteins (Figure 2B). Co-immunoprecipitation analysis confirmed that p100 was bound to RelA not only in IκB-deficient (Figure 2C and Figure S2B, right panels) but also wild type extracts (Figure 2C and Figure S2B, left panels). Interestingly, a comparison of immunoprecipitates of RelA and RelB revealed that p100 was distributed evenly between RelB- and RelA-bound complexes in wild type cells (Figure 2D).
We employed the detergent deoxycholate (DOC) to disrupt the interaction of endogenous IκB proteins and NF-κB (Baeuerle and Baltimore, 1988), and reveal the DNA binding activity (Figure 2E, compare lanes 1-2) of a RelA:p50 dimer (Figure S2C, top panel) in cytoplasmic extracts prepared from unstimulated wild type MEF. As expected, prior immunodepletion of IκBα from the extract removed the majority of latent NF-κB DNA-binding activity (lane 3). However, immunodepletion of p100 also resulted in a significant loss in DOC-inducible DNA binding activity, while immunodepletion of p105 did not (lanes 4-5). These results suggest that a fraction of cytoplasmic RelA:p50 dimer is not bound to canonical IκB proteins, but to p100. Strikingly, treating iκbα−/−β−/−ε−/− cytoplasmic extracts with DOC liberated latent NF-κB DNA binding activity composed of RelA and p50 (Figure 2E, lanes 6-7 and Figure S2C, bottom panel). Prior immunodepletion of p100 specifically eliminated almost all of the DOC-induced DNA binding activity (Figure 2E, lanes 8-9).
These biochemical results prompted us to examine the RelA:p50-p100 interaction genetically. Lentiviral delivery of small hairpin RNA (shRNA) to “knockdown” of nfkb2 expression in IκB-deficient cells resulted in more than 90% reduction of p100/p52 protein levels (Figure 2F, left panel). Importantly, EMSA with nuclear extracts prepared from these iκb−/−/p100KD cells showed a ~5-fold increase in the NF-κB activity over controls (Figure 2F, right panel, and S2E). Supershift analysis confirmed that the majority of this DNA binding activity consisted of RelA:p50 dimers (Figure 2G, data not shown). In sum, our results demonstrated that, analogous to the canonical IκB proteins, p100 is able to sequester RelA:p50 dimers in the cytoplasm and inhibit their DNA binding activity in a DOC-sensitive manner.
We sought to examine the stimulus-responsive behavior of p100 bound to RelA:p50 dimers. By immunoprecipitating RelA in non-denaturing conditions from cellular extracts, we found that LTβR stimulation led to a decrease in RelA-associated p100 protein (Figure 3A), suggesting that p100 may mediate NF-κB/RelA activation in response to LTβR stimulation. Using a genetic approach, we found that LTβR stimulation of nfκb2−/− cells were not only deficient in RelB activation as reported (Derudder et al., 2003; Lo et al., 2006), but that NF-κB/RelA activation was also abrogated (Figure 3B, top panel, lanes 1-4). Reconstitution of nfkb2−/− MEF with retrovirally-expressed p100 restored LTβR-inducible RelA and RelB DNA binding activities and RelA translocation (Figure 3B, lanes 5-8) concomitant with p100 degradation (Figure S3A). Interestingly, reconstitution of nfkb2−/− cells with the processed p52 protein did not restore signal inducible RelA nuclear activity indicating the requirement for the IκB-like ankyrin repeat domain of p100 (Figure 3B, lanes 13-16). Further, a mutant form of p100 lacking the C-terminal signal responsive phosphorylation sites also did not functionally complement the knockout cells (Figure 3B, lanes 9-12). Strikingly, expression in nfkb2−/− MEFs of the C-terminal portion of p100 containing the ankyrin repeat domain (IκBδ) was sufficient for at least partial RelA:p50 activation upon LTβR ligation (Figure 3B, lanes 17-20).
To examine whether nfkb2 p100 is sufficient for NF-κB/RelA activation or the canonical IκB proteins are also required, iκbα−/−β−/−ε−/− cells were examined in stimulation timecourses. While the elevated nuclear RelA:p50 activity in these cells was not further inducible by TNF (Figure 3C, lanes 1-4), LTβR stimulation resulted in strong induction of both nuclear RelA and RelB proteins and DNA binding activities with kinetics that were similar to those observed in wild type cells (Figure 3C, lanes 5–8, 3D, S3B, S3C). Activation of NF-κB binding activities in iκbα−/− β−/−ε−/− cells was accompanied by loss in cellular and RelA-associated p100 but not p105 protein (Figure S3D-E). Finally, we examined LTβR signaling in iκb−/−/p100KD cells, and found that p100 deficiency resulted in abrogation of LTβR responsive NF-κB (Figure 3E). Collectively, these analyses provide biochemical and genetic evidence for the requirement and sufficiency of p100 as a stimulus-specific regulator of RelA dimers in response to non-canonical signaling triggered by LTβR stimulation.
Unlike canonical IκBs, nfkb2 p100 is not only capable of interacting with RelA:p50 dimers in “trans” via its IκBδ domain as revealed by DOC treated extracts analyzed by EMSA (Figure 2E), it can also dimerize with other NF-κB proteins via its RHD domain. p100 is thus capable of forming complexes with NF-κB of distinct molecular architectures. To characterize the architecture of the complex(es) relevant for non-canonical signaling, we mapped DOC sensitive and insensitive interactions between RelA and p100 during LTβR signaling (Figure 4A). In resting cells (lanes 1–6), we found that most but not all p100 could be DOC-stripped off immunoprecipitated RelA proteins (3rd panel from top). In contrast, in LTβR-stimulated cells (lanes 7–12), no DOC-sensitive p100 could be detected in the RelA immunoprecipitate, while the DOC resistant fraction remained (3rd panel). Efficient stripping of IκBα and retention of p50 in the RelA immunoprecipitate confirmed the specificity of the DOC strip conditions. Similar analyses of canonical IκB-deficient cells (Figure S4) also revealed that DOC-sensitive p100-RelA complexes are responsive to LTβR-signaling.
Given that the IκBδ domain is genetically sufficient for LTβR induction of RelA:p50 dimer (Figure 3B), we examined biochemically the fate of the p100 RHD domain during non-canonical signaling. To this end, cellular proteins were pulse labeled with 35S-methionine before adding the receptor agonist. Immunoprecipitates obtained with an N-terminal p100 RHD antibody revealed 35S-labeled p100 prior to stimulation (Figure 4B, lane 1). Interestingly, LTβR stimulation led to the conversion of this pre-existing p100 to a polypeptide (lane 2) of the same molecular weight as the p52 generated de novo during LTβR signaling (lane 5). However, while de novo generated p52 protein participates in the LTβR induced NF-κB DNA binding activities (Figure S1C), the p52 protein generated from pre-existing p100 complexes does not (Figure 1C, S3C).
To summarize our data we propose the following molecular architectures of p100 complexes (Figure 4C). LTβR unresponsive complexes of p100 include the previously identified “self-inhibited” p100:RelA, p100:RelB, p100:p50 and p100:p52 dimers in which the IκBδ domain of p100 folds back onto the dimer to inhibit its DNA binding activity in cis (Beinke and Ley, 2004; Mercurio et al., 1993; Naumann et al., 1993). In addition, two p100 proteins may form a necessarily asymmetric dimer, analogous to two p105 proteins (Moorthy and Ghosh, 2003; Moorthy et al., 2006). In that complex, one of the IκBδ domains would inhibit the DNA binding activity of the p52 homodimer, while the other IκBδ domain would be available to inhibit a RelA:p50 or a RelB:p50 dimer. Our results suggest that this second IκBδ domain is sensitive to LTβR signaling to undergo proteolysis leaving the C-terminal RHD domain intact. The remaining p100:p52 complex would be self-inhibited and therefore incapable of participating in NF-κB DNA binding activity. However, during the later phase of LTβR signaling de novo synthesized p52 dimerizes with Rel proteins via the RHD and contributes to nuclear DNA binding activity as previously observed (Dejardin et al., 2002; Muller and Siebenlist, 2003).
Based on these proposed functional roles of distinct p100 complexes, we would predict that the amount of p52 protein associated with p100 increases during the early phase of LTβR signaling. Indeed, immunoprecipitation of p100 with an antibody specific for the C-terminal IκBδ domain yields more co-precipitating p52 protein at the 5hr timepoint of LTβR signaling than in resting cells (Figure 4D).
Activation of RelB dimers in the non-canonical signaling pathway was shown to depend on signal responsive NIK dependent activation of IKK1 (Beinke and Ley, 2004; Senftleben et al., 2001; Xiao et al., 2001). We asked if IKK1 is also required for activation of RelA dimers in response to LTβR signaling. Consistent with previous analyses (Derudder et al., 2003), ikk1−/−MEF did not show processing of p100 upon LTβR activation, but this could be restored upon retroviral reconstitution (Figure S5A). While TNF activation of NF-κB was unaffected (Figure 5A, bottom panel), we did not detect inducible NF-κB activity containing either RelA or RelB in IKK1 deficient cells in response to LTβR signaling (top panel). Similarly, MEF derived from NIK knockout mice or from the alymphoplasia mutant mouse strain, which contains an inactivating mutation in NIK (Shinkura et al., 1999), showed not only a complete abrogation of RelB but also RelA dimer activation upon LTβR stimulation (Figure 5B, S5B). Furthermore, shRNA mediated knockdown of NIK by lentiviral vectors in IκBα/β/ε-deficient cells (Figure S5C) dramatically reduced RelA-activation in response to LTβR engagement (Figure 5C). These results support the notion that non-canonical signaling via NIK and IKK1 regulates the degradation of nfkb2 p100 to effect NF-κB/RelA activation.
A hallmark of canonical NF-κB signaling is the requirement for NEMO in inflammatory stimulus responsive degradation of canonical IκB proteins. However, p100 processing to p52 upon LTβR activation was found to be intact in nemo−/− MEFs (Dejardin et al., 2002; Muller and Siebenlist, 2003). Our EMSA analysis showed that nemo−/− MEFs fail to induce NF-κB activity upon TNF treatment (Figure S5D), but we found strong induction of NF-κB DNA binding activity in response to LTβR treatment (Figure 5D). DNA binding analysis performed in the presence of RelB or RelA supershift antibody revealed that both RelA:p50 and RelB:p50 activities remained inducible in the absence of NEMO (Figure 5E, S5E).
Our results suggest a new model for the activation mechanism of NF-κB/RelA dimers in response LTβR signaling. While LTβR signaling was proposed to bifurcate into the NEMO-IKK2-IκB-RelA and the IKK1-p100-RelB axes, our new understanding of the pathway suggests that p100 is associated with pre-existing RelA:p50 and RelB:p50 dimers (Figure 5F). NIK and IKK1 regulate the signal-responsive proteolysis of p100/IκBδ, which results in nuclear RelA:p50 and RelB:p50 DNA binding activity in a NEMO independent manner. Continued synthesis and processing of p100 to p52 may enhance the level of p52 containing NF-κB DNA binding complexes at later time points (Dejardin et al., 2002; Muller and Siebenlist, 2003).
By investigating the mechanisms responsible for RelA dimer activation upon LTβR stimulation, we identified p100 as a stimulus-selective signal transducer that satisfies two criteria of bona fide IκB proteins: the ability (1) to sequester latent NF-κB dimers and (2) to release the bound dimer via stimulus-induced proteolysis. While p100/IκBδ is a bona fide IκB protein, it differs from the canonical IκB proteins IκBα, -β, -ε in its ability to sequester not only RelA- but also RelB-containing dimers, and in its responsiveness to non-canonical (e.g. LTβR, BAFF) rather than canonical (e.g. and TNF, TLR) signaling pathways. Thus, we propose to, distinguish between the canonical IκB proteins IκBα, -β, and -ε, and the non-canonical IκB protein p100/IκBδ.
Our new understanding expands the NF-κB signaling module to include nfkb2 p100/IκBδ as a fourth IκB that controls NF-κB/RelA activity in response to non-canonical, IKK1-inducing stimuli (Figure 6A). Studies of the dynamic behavior of the canonical IκB proteins have resulted in important insights about NF-κB signaling and stimulus-specific gene expression (Hoffmann et al., 2002; Kearns et al., 2006; Werner et al., 2005) in response to inflammatory stimuli. To investigate how the dynamics of p100 regulation affect RelA activation in response to non-canonical signaling, we described the synthesis, degradation and molecular interactions of p100 with ordinary differential equations (see supplementary information) and included them in a mathematical model that already recapitulated TNF and LPS signaling. The resulting model (version 3.0) includes 98 biochemical reactions, which were parameterized based on published measurements, our own measurements, and fitting procedures. For p100 in particular, we measured mRNA and protein levels and their halflife in resting cells and in response to NF-κB inducing stimuli (V. Shih, J.K., S.B., A.H., data not shown). Transcriptional and translational synthesis and degradation rate constants were fitted to these measurements.
The resulting in silico model of the NF-κB signaling module predicts NF-κB/RelA:p50 activity in response to either canonical or non-canonical stimuli, as well as the abundance of the 31 model components including the IκB proteins, free or complexed to IKK or NF-κB, in either the cytoplasm or the nucleus (Figure S6, and Supplementary Information). Simulations of LPS and TNF pulse stimulations show the dynamics of NF-κB activities and cellular canonical IκB levels (Figure 6B, left and center columns) that match previous measurements (Werner et al., 2005). Unlike the canonical IκB proteins, NF-κB p100 is not degraded in response to IKK2 mediated inflammatory signals (Figure 6B, bottom row). In contrast, IKK1-mediated LTβR signaling results in p100 degradation while the canonical IκB proteins are unaffected (Figure 6B, right column). Most importantly, the model recapitulates our measurements of slow induction of nuclear NF-κB/RelA activity that reaches a maximum at 5–15 hours (Figure S1B).
Given that nfkb2 p100, like IκBα and IκBε (Kearns et al., 2006), is an NF-κB target gene, we used computational simulations to explore feedback regulation and the dynamics of p100 protein levels in a variety of stimulation regimes. Remarkably, we found that a one-hour pulse of TNF stimulation resulted in an elevation of p100 protein of about four fold that persisted for more than 20 hrs (Figure 7A, top graph). In contrast, graphing the sum of all canonical IκB proteins revealed the expected transient trough and rapid recovery. In sum, the model predicted that the fraction of p100 associated RelA protein at the 20hr timepoint to be increased three-to-four fold (Figure 7A, bottom graph).
We examined these predictions experimentally. Western blotting of total cellular IκB protein confirmed that p100 protein is induced by one-hour TNF pulse stimulation, while the levels of canonical IκB proteins were not discernibly different (Figure 7B). By quantitating IκB-NF-κB interactions with the DOC-EMSA assay coupled to immunodepletions, we found that TNF-primed cells contained much more p100-NF-κB complexes than unstimulated/naïve cells (Figure 7C, bottom panel). These data confirm the model’s prediction that priming of cells can shift the balance of latent NF-κB associated with canonical or non-canonical IκB proteins.
This shift in the IκB homeostasis may have functional consequences. Computational simulations of LTβR-signaling in naïve and TNF-primed cells led to the prediction that primed cells may respond with a stronger NF-κB activation profile (Figure 7D). This prediction is remarkable because in the simulated stimulation regimen canonical signaling has long ceased when cells are exposed to agonist antibody, and it thus not only constitutes an example of signaling crosstalk but of cellular memory. We examined the prediction experimentally. Our EMSA and nuclear Western results confirmed that two to three fold more NF-κB/RelA activity is induced by LTβR stimulation in cells that were previously primed with TNF (Figure 7E and and7F7F).
We investigated the gene expression effects of this elevated NF-κB response using quantitative multiplex RNAse protection assays. While naïve MEF did not show appreciable inflammatory gene activation in response to the LTβR stimulation, primed MEF not only exhibited induction of the IκBα gene, a hallmark of the canonical NF-κB response, but we also found significant mRNA levels of the inflammatory cytokine TNF induced by LTβR signaling (Figure 7G). These results constitute qualitative changes in stimulus-responsive gene expression programs. Our computational simulations of the NF-κB signaling module revealed that the homeostatic control of IκB signal transducers integrates the history of cellular exposure to NF-κB inducing stimuli. Our experimental results confirmed the validity of the proposed model and resultant cross-regulatory mechanism between canonical and non-canonical NF-κB inducing signals, and revealed that its functional consequences for stimulus-specific gene expression programs may be profound.
The homeostasis of canonical and non-canonical IκB proteins may not only be subject to physiological stimuli, but altered homeostasis due to genetic aberrations may be the underlying cause for pathological misregulation of stimulus-specific gene expression. To examine this possibility and further test the validity of our mathematical model, we used computational simulations to predict the relative amounts of p100 protein in a panel of cells deficient for one or more canonical IκB proteins. For example, cells deficient in IκBα were predicted to have two fold more p100 protein than wild type cells (Figure 8A). Similarly, doubly-deficient and triply-deficient cells are predicted to show further increases in p100 except for IκBβ/ε-deficient cells, in which p100 levels were predicted to be near wild type. Western blotting for p100 with extracts derived from these different cell lines confirmed this prediction (Figure 8B); in particular, we observed no significant increases in p100 levels in IκBβ/ε-deficient cells while immunoblot signals were elevated in other knockouts as predicted.
Altered balances in the canonical and non-canonical IκB proteins may be expected to lead to alterations in the LTβR responsiveness. Computational simulations indeed predicted that NF-κB/RelA responses to agonist receptor would be highest in IκB triple knockout and IκBα/ε-deficient cells, intermediate in IκBα- and IκBα/β-deficient cells and relatively low in wild type and IκBβ/ε-deficient cells (Figure 8C). The results from experimental analysis using EMSA of nuclear extracts prepared from LTβR stimulated cells of each respective genotype were in overall agreement with these simulations (Figure 8D). These results suggested that genetic alterations in cells that constitutively alter the homeostasis of canonical and non-canonical IκB proteins affect the stimulus-specific activation of NF-κB. To determine whether these changes are functionally relevant, we examined the expression of known NF-κB response genes. Indeed, we found that cells that showed hyper-responsive NF-κB/RelA activation to LTβR stimulation, also exhibited inflammatory gene expression, such as those encoding iNOS, MIP-2, IP-10 or KC, GM-CSF, not normally seen in wild type MEF (Figure 8E). These effects were not merely quantitative but qualitative. Of the six target genes we examined we found a range of responsiveness that suggests that NF-κB target genes may have different threshold requirements for NF-κB activity that allow for differential regulation.
Our results offer a significant revision of our understanding of NF-κB signaling. By examining the mechanism for NF-κB/RelA activation in response to LTβR signaling, we identified nfκb2 p100 as the pertinent signal transducer that functions as a bona fide IκB molecule. We propose that as a feedback regulator of the RelA:p50 transcriptional activator the non-canonical IκB protein p100 represents an integrator of canonical and non-canonical signals. The resulting signaling crosstalk between inflammatory and developmental stimuli suggests that the canonical and non-canonical NF-κB signaling pathways (previously described as separate) ought to be considered as being contained within a single signaling module. The “four IκB” computational model that integrates IKK1 and IKK2 mediated signals represents a first attempt in that direction. Future efforts will have to quantitatively address the regulation of RelB complexes and the generation of p52-containing dimers.
Within their physiological environment, cells receive multiple signals, and thus the potential for signaling crosstalk is significant. Relevant to lymphnode development, constitutive TNF signaling has been observed in the spleen (Schneider et al., 2004). Indeed, the in vitro generation of fibroblast reticular cells from LN tissue requires both TNFR and LTβR stimulation (Katakai et al., 2004), while abrogation of TNF signaling in vivo results in the loss of lymphnode architecture (Rennert et al., 1998). Fundamentally, all cells are continuously subject to low or varying amounts of TLR/TNFR signaling, albeit these conditions are not usually reproduced in sterile in vitro cell culture conditions. Recent in vivo studies suggest that naïve T cell responses are regulated by integrating signals from memory and regulatory T cells via p100 (Ishimaru et al., 2006). Remarkably, nfκb2 p100 mediated crosstalk does not have to occur coincidently but integrates the history of cellular exposure to inflammatory signals.
Conversely, our studies demonstrate that alterations in the homeostasis of IκB proteins can result in inappropriate expression of inflammatory genes in response to developmental/non-canonical LTβR signaling. We utilized defined genetic mutants with altered IκB protein expression patterns. These cells may serve as model systems for disease-associated misregulation. In this context, it is of interest to note that a significant fraction of malignant Reed Sternberg cells in Hodgkin’s lymphoma has defective IκBα genes (Krappmann et al., 1999). Our findings suggest that pathogenic misregulation of gene expression in Reed Sternberg cells may in part be the result of non-canonical developmental signals impinging on an NF-κB signaling module with an altered homeostasis of IκB proteins. We suggest that the integration of multiple signals via signaling crosstalk represents an opportunity for therapeutic intervention, which may be more effectively exploited through the combined tools of computational modeling and biochemical analysis.
Primary and 3T3 immortalized MEF were generated from E12.5–14.5 embryos and maintained as previously described (Hoffmann et al., 2003). Generation of iκbα−/−β−/−ε−/− and iκbα−/−β−/−ε−/−rela−/−cell line will be described elsewhere. Agonistic LTβR monoclonal antibody raised in hamster (AF.H6) was obtained from Jeff Browning (Biogen Inc). Two other previously used Rat anti-LTβR antibodies, 3C8 and 4H8 (Dejardin et al., 2002) were also used (Figure S1F). Recombinant murine TNF was from Roche. RelA/p65 (sc-372), RelB (sc-226), cRel (sc-71), IκBα (sc-371), IκBβ (sc-946), NIK (H-248) antibodies were from Santa Cruz Biotechnology. Phospho-IκBα (ser 32, 36) monoclonal antibody was from Cell Signaling. N-terminal p52 (1495), p50 (1157 and 1263) as well as C-terminal anti-sera for p100 (1310) and p105 (1140) were a generous gift from Nancy Rice. Anti-serum against TAF20 was used to test nuclear fractionation. Sc-372G antibody was used for RelA immunoprecipitation and sc-226 cross-linked to protein-A-sepharose beads was used for RelB immunoprecipitate. Supershift antibodies have been described previously (Hoffmann et al., 2003).
Whole cell-extracts were prepared in RIPA buffer and normalized before immunoblot analysis. For immunoprecipitation-Western analysis, whole cell lysate from 106 cells was prepared in buffer containing 20% glycerol, 0.2mM EDTA, 0.5% NP-40 and 150mM NaCl. Band intensities were quantitated by phosphor-imager and normalized against the respective actin band. To evaluate the DOC sensitivity of RelA-p100 interactions, RelA immunoprecipitates obtained from cytosolic extract were washed with buffer containing 10mM HEPES-KOH, pH 7.9, 150mM KCl, 1mM EDTA, 0.1% NP-40 containing 0%, 0.3% or 0.8% deoxycholate. Washes and SDS sample buffer eluted immunopellets were immunoblotted.
In vivo pulse labeling of MEFs was done with 100 μCi/ml trans 35S-Met label (MP Biomedicals, Inc.) using the indicated timecourses. p52/p100 were immunoprecipated (N-terminal antiserum 1495) from RIPA lysate. Proteins were resolved in 8% SDS-PAGE and visualized by autoradiography.
IKK1 and p100, p1001–774 or p52 expressing retroviral constructs were gifts from M. Karin and A. Rabson. pBabe.IκBδ.puro was constructed by PCR amplifying c-terminal region of p100 encoding amino acids (483–934) using pBabe.p100.puro as template. Retroviral constructs were co-transfected with pCL.Eco into 293T-cells and 42hrs post-transfection filtered supernatant was used to infect MEF. Transduced cells were selected for 72h with 2.5-5μg/ml puromycin hydrochloride (Sigma).
Lentiviral constructs expressing shRNAs against nfkb2 and nik genes and control oligos were packaged into viruses as described previously (Tergaonkar et al., 2005). Sequences for specific oligo used in knockdown studies are available upon request. Four different control shRNA were used (Figure S2D) to ensure functional specificity.
Murine RelA gene was cloned into pRAV-Flag-TAP bicistronic retroviral vector that expresses EGFP from an IRES promoter (Knuesel et al., 2003). Tagged-RelA gene was transduced into iκbα−/−β−/−ε−/−rela−/− cell-line and stable expressors were obtained by GFP-mediated fluorescence cell sorting. IgG binding domain tagged p65 and interacting proteins were immunoprecipitated from the cell-extract (from ~108 cells) using IgG-Sepharose 6 Fast Flow resin (Amersham Biosciences). Flag-RelA was eluted from the immunopellet by TEV-protease cleavage (Invitrogen), and re-precipitated with anti-Flag M2 affinity gel resin (Sigma). RelA complexes were eluted with 1mg/ml Flag peptide (Sigma), were analyzed by SDS-PAGE. Gel slices were cut and in-gel trypsin digested, and analyzed by LC-MS/MS at the Scripps Research Institute Center for Mass spectrometry and searched against MASCOT/BLAST.
EMSAs were done as described with a κB-site containing probe that binds to both RelA and RelB dimer with similar affinities (Lo et al., 2006). Super-shift assays were done with antibodies as described before (Hoffmann et al., 2003). For DOC sensitivity, 2μg of cytosolic extract was treated with final 0.8% DOC for 30min and subsequently subjected to EMSA analysis. Where specified, cytosolic extracts were immunodepleted for indicated protein prior to DOC treatment.
Total RNA was extracted from stimulated fibroblast using Trizol reagent. RPA was performed with 5μg RNA using Riboquant probe set (BD-Bioscience) according to manufacturers’ protocols.
The previously described mathematical model (Werner et al., 2005) was extended to include nfκb2 p100 and non-canonical signaling via IKK1 as shown in Figure S6 and described in detail in the Supplemental Data, which can be found online at http://www.cell.com/cgi/content/full…. . Simulations were done in Matlab version 2006a (Mathworks) using the built-in ode15s solver at default settings. The Matlab model is available at http://signalingsystems.ucsd.edu, and an SBML model version will be available at http://jjj.biochem.sun.ac.za/database/index.html.
We thank M. Karin, D. Shultz, and A. B. Rabson for plasmids, J. Browning for LTβR agonist antibody, Santa Cruz Biotechnology for a variety of antibodies, and N. Rice and M. Ernst for antisera against p105/p50 and p100/p52. We also thank C. Lynch and V. Shih for help with experiments and anonymous reviewers for constructive criticism. V.T. was supported by the Leukemia and Lymphoma Society of America. The study was supported by the following grants: CA69381, NIH AI22068 (C.W.), American Heart Association 0330064N, NIH AI61549 (C.A.B.), NIH GM071573 (A.H.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.