Gene Expression Profiling in Conjunction with Physiological Rescues of IKKα-null Cells with Wild Type or Mutant IKKα Reveals Distinct Classes of IKKα/NF-κB-dependent Genes*

Cellular responses to stress-like stimuli require the IκB kinase (IKK) signalsome (IKKα, IKKβ, and NEMO/IKKγ) to activate NF-κB-dependent genes. IKKβ and NEMO/IKKγ are required to release NF-κB p65/p50 heterodimers from IκBα, resulting in their nuclear migration and sequence-specific DNA binding; but IKKα was found to be dispensable for this initial phase of canonical NF-κB activation. Nevertheless, IKKα(-/-) mouse embryonic fibroblasts (MEFs) fail to express NF-κB targets in response to proinflammatory stimuli, uncovering a nuclear role for IKKα in NF-κB activation. However, it remains unknown whether the global defect in NF-κB-dependent gene expression of IKKα(-/-) cells is caused by the absence of IKKα kinase activity. We show by gene expression profiling that rescue of near physiological levels of wild type IKKα in IKKα(-/-) MEFs globally restores expression of their canonical NF-κB target genes. To prove that the kinase activity of IKKα was required on a genomic scale, the same physiological rescue was performed with a kinase-dead, ATP binding domain IKKα mutant (IKKα(K44M)). Remarkably, the IKKα(K44M) protein rescued ∼28% of these genes, albeit in a largely stimulus-independent manner with the notable exception of several genes that also acquired tumor necrosis factor-α responsiveness. Thus the IKKα-containing signalsome unexpectedly functions in the presence and absence of extracellular signals in both kinase-dependent and -independent modes to differentially modulate the expression of five distinct classes of IKKα/NF-κB-dependent genes.

The NF-B pathway is important for a host of cellular processes including its central role in responses to stress-like stimuli, the antiapoptotic cascade, the initiation and maintenance of immune responses, embryonic and adult tissue development, and cell cycle progression (for review, see Refs. [1][2][3][4][5][6][7][8][9][10][11]. In mammals the NF-B family of transcription factors is comprised of five subunits characterized by the presence of a conserved Rel homology DNA binding domain (for review, see Refs. 2, 12, and 13). The p65(RelA), c-Rel, and RelB NF-B subunits are fully functional transcriptional activators, whereas the p50 and p52 subunits lack a transcriptional activation domain (for review, see Refs. 12 and 13). NF-Bs function as specific heteroor homodimers that bind to a GGGRNWTYCC consensus DNA sequence found in the promoters or enhancers of NF-B target genes (for review, see Refs. 12 and 13). Transcriptional activating NF-B subunits are normally sequestered in the cytoplasm of unstimulated cells in a complex with one of the IB family proteins, which block their nuclear import and DNA binding activity. A large variety of extracellular activating stimuli induce the proteosomal dependent destruction of IBs thereby differentially freeing NF-Bs to bind DNA and activate the transcription of their genomic targets (for review, see Ref. 14). Stress-like inducers of NF-B (including proinflammatory cytokines such as TNF␣ 1 and IL-1) function in a classical monophasic capacity to drive the canonical NF-B activation pathway rapidly, which largely involves the activation of p65(RelA)/p50 DNA binding activity and transcriptional competence (for review, see Refs. 1, 2, 8, and 15). More recently, a distinct class of NF-B stimuli (exemplified by LT␤, BAFF, and CD40 ligand), which also contribute to the implementation of differentiation programs and the adaptive phase of immune responses, have been shown to function as biphasic activators initially acting via the rapid canonical pathway and subsequently feeding into a delayed noncanonical protein synthesisdependent route characterized by the activation of RelB/p52 heterodimers (for review, see Refs. 8, 9, 15, and 16).
With the exceptions of UV radiation and the effects of some * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. This research was supported in part by a National DNA-damaging agents (17,18), the release of NF-Bs from IBs is mediated by the cytoplasmic signalsome complex, which consists of two serine-threonine kinases (IKK␣, IKK␤) and NEMO/IKK␥, a regulatory/docking protein (for review, see Refs. 1, 7, 8, and 15). IKK␤ is essential for the phosphorylation of IBs on a pair of amino-terminal serines (residues 32 and 36 in IB␣) thereby targeting IB for ubiquitination and subsequent proteosomal destruction (for review, see Refs. 1, 7, and 8). In contrast, IKK␣ is not required for the phosphorylation of IBs via the canonical NF-B activation pathway in vivo, with the exception of receptor activator of NF-B (RANK) ligand signaling in mammary epithelial cells (19). Rather, IKK␣ plays an essential role in epidermal keratinocyte differentiation independent of both its kinase activity and NF-B activation and has also recently been found to play NF-B-dependent and -independent roles in tooth development (20 -22). With respect to its physiological role in NF-B signaling pathways, IKK␣ is instead essential for the activation of the noncanonical NF-B activation pathway, which requires neither IKK␤ nor NEMO/ IKK␥ (for review, see Refs. 8, 15, and 16). In this context, via NF-B-inducing kinase-dependent signaling, IKK␣ phosphorylates multiple serines of the p100 precursor of the p52 subunit, thereby inducing its proteosome-dependent processing into mature p52 subunits that are then freed to activate NF-B target genes as RelB/p52 heterodimers (8,9,23,24). We and other groups have also found that IKK␣ is required to activate the transcription of canonical NF-B target genes (25)(26)(27)(28)(29). The latter dependence on IKK␣ is independent of IB␣ destruction and instead appears to involve one or more nuclear targets perhaps including histone H3 (27,28) and the SMRT transcriptional corepressor (29) resulting in the de-repression of NF-B target genes.
In this report we have investigated the physiological requirement of the kinase activity of IKK␣ for the expression of NF-B-dependent genes on a genomic scale in IKK␣-null MEFs. Physiological expression of Wt. IKK␣ in IKK␣(Ϫ/Ϫ) MEFs by retroviral transduction resulted in the rescue of specific NF-B-dependent genes in the presence and absence of TNF␣ stimulation. Comparative microarray screens with NF-Bcompromised MEFs (p50(Ϫ/Ϫ) and Wt. ϩ IB␣(S32A,S36A)) revealed that the large majority of these IKK␣-rescued genes are either dependent on basal or TNF␣-inducible NF-B, thus demonstrating that 1) IKK␣ plays an essential role in controlling the expression of both signal-induced and basal NF-B-dependent genes, and 2) IKK␣ does not appear to influence the expression of a large number of genes outside the NF-B pathway. Comparable physiological rescue with a kinase-dead IKK␣ mutant protein (IKK␣(K44M)) showed that most of these genes are dependent on IKK␣ kinase activity for their stimulusdependent and -independent expression. However, the expression of up to 28% of these NF-B-dependent genes was also surprisingly rescued by the kinase-inactive IKK␣(K44M) mutant. Furthermore both wild type and mutant IKK␣ are also required for the basal levels of expression of specific NF-B-dependent genes. Thus, our findings collectively reveal that the levels of expression of different downstream NF-B-dependent genes are differentially codependent on catalytically active IKK␣ in the presence or absence of extracellular stimuli.
RNA Preparation-Total cellular RNAs were extracted from cell lysates with an RNeasy kit (Qiagen). Purified RNAs were converted to double-stranded cDNA with a Super Script Double Stranded cDNA synthesis kit (Invitrogen) and an oligo(dT) primer containing a T7 RNA polymerase promoter (GENSET). Biotin-labeled cRNAs were generated from the cDNA samples by in vitro transcription with T7 RNA polymerase (Enzo kit, Enzo Diagnostics). The labeled cRNAs were fragmented to an average size of 35-200 bases by incubation at 94°C for 35 min. Hybridization (16 h), washing, and staining protocols have been described (Affymetrix Gene Chip Expression Analysis Technical Manual; (34)).
DNA Microarrays and Clustering Analysis-We employed Affymetrix MG-U74Av2 chips that include 12,400 genes. Chips were stained with streptavidin-phycoerythrin (Molecular Probes) and scanned with a Hewlett-Packard Gene Array Scanner. DNA microarray chip data analysis was performed using MAS5.1 software (Affymetrix) and as described previously (26). Levels of gene expression were quantitated from the hybridization intensities of 16 pairs of perfectly matched and mismatched control probes (35) (Affymetrix, Inc.). The average of the differences (perfectly matched minus mismatched) for each gene-specific probe family was calculated and expressed as signal values. The software computes how the expression level of each transcript has changed between the base line and experimental samples (difference call/change call). A change call is a qualitative call that describes whether a transcript in an experimental array has changed compared with a base-line array. One array is designated as the experimental, and another array is designated as the base line. Wilcoxon's signed rank is used to generate a change p value. A change call is assigned based on analysis parameters. Change p values between 0.00 and 0.0025 are given an Increase call. Change p values between 0.0025 and 0.003 are given a marginal increase call. Change p values between 0.997 and 0.998 are given a marginal decrease call. Change p values between 0.998 and 1.00 are given a decrease call (Affymetrix User Manual). For a comparative chip file (such as TNF␣-stimulated IKK␣(Ϫ/Ϫ) MEF ϩ Wt. IKK␣ versus IKK␣(Ϫ/Ϫ) MEF ϩ EV, the experimental file (Wt. IKK␣ 2T) was compared with the base-line file (EV 2T). We employed the following stringent selection criteria to identify significant changes in gene expression: 1) a change call of increase or marginal increase in both samples, and 2) average -fold change values of 1.5 or greater (minimum of 1.3-fold each) in two independently stimulated samples of IKK␣(Ϫ/Ϫ) MEF ϩ Wt. IKK␣ 2T versus IKK␣(Ϫ/Ϫ) MEF ϩ EV 2T. The following additional criteria were employed to identify the spectrum of these genes that were also dependent upon NF-B for their expression: 1) a change call of either increase or marginal increase in Wt. MEF 2T versus Wt. MEF ϩ IB␣SR-Iresneomycin, or 2) a change call of either increase or marginal increase in Wt. MEF 2T versus p50(Ϫ/Ϫ) MEF 2T, and 3) a valid presence call in Wt. 2T screens. By this analysis we define NF-B dependence to represent genes whose expressions either directly (true direct targets of NF-B subunits) or indirectly (other downstream genes whose expressions are affected by the NF-B pathway) require NF-B. The TNF␣ inducibilities of genes rescued by either Wt. IKK␣ or IKK␣(K44M) were based on a combination of the following stringent criteria: 1) increase calls in duplicate 2T versus US microarray screens of IKK␣(Ϫ/Ϫ) MEFs rescued by Wt. IKK␣ or IKK␣(K44M), and/or 2) TaqMan real time PCR analysis performed at least in duplicate and an increase call in one of two screens. By these criteria, the vast majority of IKK␣-rescued genes that responded to TNF␣ did so with -fold change values of 2.0 and higher in duplicate screens with a minimum unstimulated (US) average -fold change value of 1.7.
Hierarchical clustering was performed with the Cluster program (available at rana.lbl.gov/) as described previously (36). Genes that have double increase calls and are induced Ͼ1.5-fold (average -fold values) in the duplicate primary Wt. IKK␣ versus EV-rescued IKK␣(Ϫ/Ϫ) MEFs (see above description) were selected. The signal values (equivalent to the quantities of mRNAs, see above) of the selected genes were mediancentered by subtracting the median observed value and normalized by genes to the magnitude (sum of the squares of the values) of a row vector to 1.0. The normalized data were clustered by average linkage clustering analysis of the y axis (genes) using an uncentered correlation similarity metric, as described in the program Cluster. Signal values of 50 or less were set to 50 before centering and normalization. The clustered data were visualized with the Treeview program (available at rana.lbl.gov/).
TaqMan Real Time Quantitative PCR-TaqMan real time quantitative PCRs were performed as described previously (26,37). Data from TaqMan PCR analyses were normalized based on glyceraldehyde-3phosphate dehydrogenase mRNA copy numbers using rodent glyceraldehyde-3-phosphate dehydrogenase control reagents (Applied Biosystems). TaqMan probe sets were designed for the following genes using either Primer Express 1.5 or Bio-Rad Beacon Designer 2.0 software: ATF3, A20, ISG15, MyD118/GADD45␤, SAA3, and VCAM1. Murine IL-6 and RANTES TaqMan reagents were obtained from ABI. TaqMan real time PCR experiments were performed in an ABI PRISM 7700 sequence detector or in a Bio-Rad iCycler. DNA sequences for each of these probe sets are available upon request.
Western Blotting-Cell lysates were prepared in an isotonic lysis buffer containing 1% Nonidet P-40 supplemented with protease inhibitors. SDS-PAGE (10% gel) transfer to polyvinylidene difluoride membranes was performed as described previously, and membranes were probed with primary antibodies to either IKK␣ (Cell Signaling Technology) or NEMO (Santa Cruz Biotechnology) followed by an anti-rabbit-horseradish peroxidase-conjugated secondary antibody (Amersham Biosciences). Blots were developed using a Lumi-Light Plus kit (Roche Applied Science).
Immunostaining-Prior to immunostaining cells were maintained in 10-cm tissue culture-treated plates in their regular growth medium. Cells were trypsinized, resuspended in 12 ml of growth medium, and plated at 3 ml/well in 6-well plates containing 22-mm glass microscope coverslips (VWR) precoated with poly-L-lysine (Sigma). Cells were incubated overnight at 37°C with 5% CO 2 . The following day, cells were washed and then fixed in 50% methanol and 50% acetone for 10 min. These and all subsequent washes were with phosphate-buffered saline. The fixed cells were rehydrated in phosphate-buffered saline, and the coverslips were washed, blocked with phosphate-buffered saline containing 10% heat-denatured fetal bovine serum for 1 h, and washed again. In situ expression of retrovirally transduced Wt. IKK␣-HA and IKK␣(K44M)-HA proteins in IKK␣(Ϫ/Ϫ) cells were specifically detected with a primary anti-HA 12CA5 antibody (38). 12CA5 antibody was diluted in blocking solution and applied to the cells followed by a 1-h incubation at room temperature. After washing, alkaline phosphataseconjugated goat anti-mouse IgG (Jackson) secondary antibody, diluted 1:2,000 in block solution, was applied for a 1-h incubation at room temperature. The coverslips were washed and nitro blue tetrazolium/ 5-bromo-4-chloro-3-indolyl phosphate developing substrate applied (Roche Applied Science). Cells were visualized on a Nikon Diaphot phase-contrast microscope and photographed using a Nikon D1X digital camera.
NF-B DNA Binding-Activation of NF-B p65-dependent DNA binding activity was quantitated with an enzyme-linked immunosorbent assay-based kit (Active Motif, Inc.) (39). Nuclear extracts were prepared from unstimulated or 2-h TNF␣-stimulated WT. IKK␣-null and Wt. IKK␣ or IKK␣(K44M)-rescued IKK␣(Ϫ/Ϫ) MEFs and applied to 96-well plates containing an immobilized NF-B DNA binding consensus oligonucleotide according to the manufacturer's instructions (Active Motif, Inc.). DNA-bound NF-B was detected with a p65-specific primary antibody followed by the addition of a secondary antibody conjugated to horseradish peroxidase and absorbance quantitated at 450 nm with a microplate spectrophotometer. The specificity of NF-B DNA binding was confirmed by competitions with wild type and mutant NF-B binding sequences. Negative controls for stimulus-dependent NF-B nuclear localization and DNA binding activity also included nuclear extracts prepared from NEMO(Ϫ/Ϫ) and p65/p50(Ϫ/Ϫ) MEFs.

Physiological Rescue of IKK␣(Ϫ/Ϫ) MEFs with Wt. IKK␣ IKK␣(K44M) Proteins Does Not
Interfere with the Induction of NF-B DNA Binding Activity-By employing DNA microarray chip technology, we previously reported that the IKK␣ protein was as essential as the IKK␤ and NEMO/IKK␥ signalsome subunits for the genomic NF-B-dependent transcriptional response induced by TNF␣ or IL-1 stimulation. Our finding of a strict requirement for IKK␣ in the regulation of NF-B-dependent transcription in MEFs was controversial because earlier studies had shown that cells derived from IKK␣-null mice exhibited no significant defect in the stimulus-dependent induction of NF-B nuclear localization and DNA binding activity. However, in agreement with our observations, two other studies had shown that IKK␣ was required for the TNF␣-and IL-1-dependent transcriptional induction of IL-6 gene expression (25,40). Subsequent to these reports, chromatin immunoprecipitation experiments showed that the role of IKK␣ in engendering DNA-bound NF-B with transcriptional competence was associated with its TNF␣-dependent binding to the IB␣ gene promoter and with its ability to directly phosphorylate histone H3 on serine 10 in vitro (27,28) and more recently also to phosphorylate and thereby facilitate the release of the SMRT corepressor from specific NF-B target gene promoters (29).
To determine whether the intrinsic defect of IKK␣(Ϫ/Ϫ) MEFs to express NF-B-dependent genes on a genomic scale was solely caused by the absence of a functional IKK␣ kinase, we employed retroviral transduction to rescue physiological levels of Wt. IKK␣ expression in a large population of IKK␣(Ϫ/Ϫ) MEFs. To determine whether IKK␣ kinase activity was required to rescue the expression of their NF-B-dependent genes, we also derived a similar population of IKK␣(Ϫ/Ϫ) MEFs expressing near physiological levels of a kinase-dead IKK␣ mutant (IKK␣(K44M)) (31,32). To rule out any effects of stable retroviral transduction, we also generated a matched negative control population of IKK␣(Ϫ/Ϫ) MEFs harboring the empty retroviral vector (IKK␣(Ϫ/Ϫ)-EV cells). Western blotting revealed that the levels of Wt. IKK␣ and IKK␣(K44M) expression in IKK␣(Ϫ/Ϫ)-rescued MEFs were similar to the expression of endogenous IKK␣ in wild type NF-B competent MEFs (see Fig. 1). Importantly, as shown in Fig. 2, in situ immunostaining also revealed uniform expression of either the Wt. IKK␣-HA or mutant IKK␣(K44M)-HA proteins in their respective stably transduced cell populations. These stable populations of IKK␣(Ϫ/Ϫ) cells expressing physiologically comparable levels of either Wt. IKK␣-HA or IKK␣(K44M)-HA were used in all subsequent experiments.
As discussed above, studies with IKK␣-null mice have proven that the loss of IKK␣ has no effect on the induction of NF-B DNA binding activity by proinflammatory cytokines (40 -42). Consequently, before proceeding to compare the effects of exogenous wild type and kinase-dead IKK␣ on the TNF␣-induced NF-B-dependent transcriptional response of IKK␣-null MEFs, it was necessary for us to verify that our retrovirally derived cell populations exhibited TNF␣-induced NF-B DNA binding activity comparable with that of their IKK␣(Ϫ/Ϫ) counterparts and Wt. MEFs. To this end, we employed a quantitative enzyme-linked immunosorbent assaybased DNA binding assay to directly compare levels of NF-B p65 subunit DNA binding activity induced by TNF␣ stimulation. Nuclear extracts of NEMO/IKK␥(Ϫ/Ϫ) and p65/p50(Ϫ/Ϫ) MEFs were employed as negative controls. As shown in MEFs. To validate the specificity of the DNA binding reactions, NF-B p65-dependent DNA binding was abolished in nuclear extracts of TNF␣-induced IKK␣(Ϫ/Ϫ) ϩ Wt. IKK␣ and IKK␣(Ϫ/Ϫ) ϩ IKK␣(K44M) cells by an excess of a wild type NF-B binding oligonucleotide but not by a mutant NF-B binding sequence (Fig. 3). Importantly, this experiment demonstrates that when expressed at near physiological levels, the IKK␣(K44M) mutant protein does not function in a general dominant-negative manner with respect to the nuclear localization and activation of NF-B DNA binding activity.

Physiological Rescue of IKK␣(Ϫ/Ϫ) MEFs with Wt. IKK␣ Is Sufficient to Restore the Global Expression of NF-B-dependent
Genes-To identify the cohort of TNF␣-responsive genes in IKK␣(Ϫ/Ϫ) MEFs, which are not expressed because of their IKK␣ deficiency, we compared DNA microarray screens of IKK␣(Ϫ/Ϫ) MEFs rescued with Wt. IKK␣ with IKK␣(Ϫ/Ϫ) cells expressing an empty retroviral vector (EV) as a matched neg-ative control. IKK␣(Ϫ/Ϫ) ϩ Wt. IKK␣ screens were performed in duplicate to rule out any gene-specific variations and were compared with TNF␣-stimulated IKK␣(Ϫ/Ϫ)-EV cells. 118 genes were rescued based on the stringent criteria of double increase calls (Affymetrix MAS 5.1) and average -fold change values of 1.5 or greater. Hierarchical clustering of the signal values of these 118 genes shows that two independent TNF␣ stimulations of the Wt. IKK␣-rescued IKK␣(Ϫ/Ϫ) cell population have very similar expression profiles (see first and second columns of Fig. 4). Some variations in the degrees of expression of specific genes were observed, but more importantly all of these genes are expressed at significantly higher levels in the two Wt. IKK␣-rescued samples compared with TNF-stimulated IKK␣(Ϫ/Ϫ) cells harboring the empty retroviral vector or parental IKK␣(Ϫ/Ϫ) cells.
To determine whether these IKK␣-dependent genes were also dependent on NF-B, we employed hierarchical clustering analysis to cross-compare these screens with others performed with two varieties of NF-B-compromised MEFs: 1) Wt. MEFs stably expressing an IB␣(S32A,S36A) super-repressor (26), and 2) NF-B p50(Ϫ/Ϫ) MEFs (43). Both cell lines were stimulated with TNF␣ for 2 h and compared with their wild type counterparts. Some variation in the degrees of NF-B dependence of subsets of genes in the p50(Ϫ/Ϫ) and Wt. ϩ IB␣SR screens were observed (see Fig. 4 Treeview image). This latter effect was most likely the result of a combination of the penetrance of the IB␣ super-represssor as well as the different thresholds of specific NF-B subunits required for the expression of their specific downstream target genes. Known genes and expressed sequence tags with significant decreases in expression in either the IB␣SR-expressing MEFs or in p50(Ϫ/Ϫ) MEFs compared with Wt. MEFs were judged to be dependent on NF-B for their activity.
The -fold change values of a representative set of 40 genes rescued by restoring Wt. IKK␣ expression in IKK␣-null MEFs are shown in Table I. The relative NF-B dependences of these genes are presented as positive -fold change values in one or both of the two microarray screens (Wt. MEF versus Wt. MEF ϩ IB␣SR or Wt. MEF versus p50(Ϫ/Ϫ)). Importantly, many of these genes were also rescued to levels of expression observed in Wt. MEFs. Comparisons of the relative mRNA expression levels of genes in duplicate TNF␣-stimulated and unstimulated samples revealed that the genes rescued by Wt. IKK␣ protein in IKK␣-null cells fell into two distinct stimulus-dependent and -independent classes (Table I, Table I and selected signal value comparisons in Fig. 5A). Among these 40 representative genes rescued by the Wt. IKK␣ protein, examples of signalindependent rescues by Wt. IKK␣ include ClastI/LR8, C1r, IFP35, Plf2, Plf3, Itm2b, NOV/CNN3, Decorin, Snx10, and IFITR2 (see genes highlighted in gray in Table I and also selected signal value comparisons in Fig. 5B). In agreement with these results, the expression of this same class of genes without extracellular stimulation was also found to be similarly reduced in Wt. MEFs harboring a constitutively expressed IB␣ super-repressor or in p50-null MEFs compared with their wild type counterparts (data not shown). Thus, these observations show that the Wt. IKK␣-containing signalsome is required for both the stimulus-dependent and basal levels of expression of NF-B-dependent genes. Because this class of NF-B-dependent genes required IKK␣ without TNF␣ stimulation, they define a novel class of IKK␣-dependent genes that are downstream of basally activated NF-B.
In addition, the vast majority of the 118 genes rescued by physiological restoration of IKK␣ in IKK␣-null MEFs were co-dependent on NF-B, based on their reduced or severely compromised expression in either Wt. MEFs expressing an IB␣ super-repressor or NF-B p50(Ϫ/Ϫ) MEFs ( Fig. 4 and Table I). Thus, our global expression results also show that IKK␣ is not likely required for the transcription of a large number of genes outside of the NF-B pathway. It also directly follows that despite the reported ability of IKK␣ to facilitate potentially more general aspects of chromatin activation by either phosphorylating serine 10 of histone H3 or the SMRT transcriptional corepressor (27)(28)(29), IKK␣ must somehow still remain preferentially targeted to NF-B-dependent genes.

A Kinase-inactive IKK␣ Mutant Rescues a Portion of the Genes Dependent on Wt. IKK␣ in IKK␣(Ϫ/Ϫ) MEFs-Because
the IKK␣ mechanism of action to activate NF-B-dependent transcription in the canonical pathway remains poorly understood, we next investigated whether the kinase activity of IKK␣ was essential for the expression of the 118 genes rescued by the Wt. IKK␣ protein in IKK␣(Ϫ/Ϫ) MEFs. Lysine 44 in the IKK␣ kinase domain is essential for its binding of ATP, and its mutation to methionine prevents ATP binding, thereby completely destroying IKK␣ kinase activity (32,44). To determine the contribution of IKK␣ kinase function for the rescue of NF-B-dependent genes, we used duplicate microarray analysis to determine the ability of a kinase-dead IKK␣(K44M) mutant to rescue IKK␣/NF-B-dependent targets when expressed at near physiological levels in the IKK␣(Ϫ/Ϫ) cells (see hierarchical Treeview comparisons of Wt. IKK␣ and IKK␣(K44M) expressing IKK␣(Ϫ/Ϫ) cells in the first and second columns and third and fourth columns, respectively, in Fig. 6). These results demonstrate that ϳ72% of the genes rescued by Wt. IKK␣ in IKK␣(Ϫ/Ϫ) cells were unaffected by the presence of comparable levels of an IKK␣(K44M) mutant protein. Surprisingly, ϳ28% (33 of the 118 IKK␣-dependent genes) were reproducibly rescued by the IKK␣(K44M) mutant. Most of these 33 IKK␣(K44M)-rescued genes can be visualized as two coclustered groups in the upper and lower portions of the hierarchical Treeview image presented in Fig. 6.
The -fold change values of 20 representative genes rescued by the IKK␣(K44M) mutant protein are presented in Table I. Comparable degrees of rescue of each of these genes were achieved with Wt. IKK␣ and the kinase-inactive IKK␣ mutant proteins. However, in contrast to the expression of genes restored by Wt. IKK␣, pairs of TNF␣-stimulated and -unstimulated microarray screens of IKK␣(K44M)-transduced IKK␣-    Fig. 9. Genes induced by TNF␣ are assigned a (ϩ) sign, and genes whose expressions were not significantly stimulated by TNF␣ were given a (Ϫ) sign. The expressions of 9 of this representative group of 40 genes were rescued independently of TNF␣ stimulation and are highlighted in gray. null cells revealed that only a small fraction of the genes rescued by the IKK␣(K44M) mutant responded to TNF␣ stimulation (see examples of A20, M/CSF-1 and VL30 highlighted in gray in Table II). Bar graphs of signal values (comparable with amounts of mRNAs) of three representative examples of the stimulus-dependent (A20, mVL30, and B94) and independent (Coagulation factor III, Sgk, and NDPP1) classes of IKK␣(K44M)-rescued genes are shown in Fig. 7. In addition to these two distinct classes of genes, a third class of IKK␣(K44M)-rescued genes only responded to TNF␣ stimulation after rescue by Wt. IKK␣ and not if rescued by the IKK␣(K44M) kinase-inactive mutant (see examples GADD45␤, ATF3, and JunB in Table II). Interestingly, compared with the NF-B-dependent genes solely rescued by the Wt. IKK␣ protein, a larger fraction of the IKK␣/NF-B-dependent genes rescued by the kinase-inactive IKK␣(K44M) mutant preferentially encoded proteins associated with NF-B autoregulation, growth arrest, apoptosis, proliferation, and survival (see representative genes in Table II and under "Discussion"). In addition, most of these NF-B-dependent genes, which are rescued by wild type and kinase-inactive IKK␣, depend on IKK␣ for their basal levels of expression in the absence of an extracellular stimulus. Thus, our global expression profiling analysis reveals the surprising result that different NF-B-dependent genes differentially require IKK␣ kinase activity for their basal and TNF␣-dependent expression, with the majority of genes encoding proteins associated with pro-inflammatory stress-like responses requiring IKK␣ kinase activity.
TaqMan Real Time PCR Validations of Wt. IKK␣ and IKK␣(K44M)-rescued Genes-As an additional validation of the duplicate microarray screens, TaqMan real time PCR experiments were performed on eight NF-B target genes (IL-6, ISG15, RANTES, SAA3, VCAM1, GADD45B, A20, and ATF3). Importantly, TaqMan validations were performed at least in duplicate on a third independent set of unstimulated and 2 h (2T) TNF␣-stimulated samples, which were not employed in the duplicate microarray screens. Fig. 8 shows the absolute expression levels in the context of TNF␣ stimulation of each of these eight genes as mRNA copy numbers in IKK␣(Ϫ/Ϫ) MEFs expressing Wt. IKK␣, IKK␣(K44M), or an EV compared with Wt. MEFs. Each of these genes is expressed in Wt. MEFs and Wt. IKK␣-rescued IKK␣(Ϫ/Ϫ) MEFs above their low to negligible levels in IKK␣(Ϫ/Ϫ)ϩEV cells. Some variations in the degrees of the IKK␣-dependent rescues compared with wild type control cells are noted with some genes being expressed at higher levels in the wild type control and others expressed at higher levels in the Wt. IKK␣-rescued cells. Fig. 9 illustrates the TNF␣ dependences of the same eight genes. In agreement with the duplicate microarray screens, A20, ATF3, and MyD118/GADD45␤, were rescued by the IKK␣(K44M) mutant compared with IKK␣(Ϫ/Ϫ)ϩEV MEFs, whereas IL-6, ISG15, RANTES, SAA3, and VCAM1 failed to exhibit expression above background in IKK␣(K44M)-expressing IKK␣(Ϫ/Ϫ) cells. Of these three IKK␣(K44M)-rescued genes, only A20 responded significantly to TNF␣ stimulation. In agreement with our duplicate microarray screens, after their rescue by Wt. IKK␣ each of these eight genes was confirmed to be TNF␣-responsive. DISCUSSION Studies of IKK␣(Ϫ/Ϫ) and IKK␤(Ϫ/Ϫ) MEFs have definitively shown that IKK␤ is the in vivo IB␣ kinase and that IKK␣ is not needed for IB degradation, NF-B nuclear localization, nor for inducing NF-B DNA binding activity in response to proinflammatory NF-B stimuli such as TNF␣. However, IKK␣ functions in the canonical NF-B pathway to ensure or modulate the transcriptional competence of DNAbound NF-B. In support of this view, we have shown herein that restoration of IKK␣(Ϫ/Ϫ) MEFs with near physiological levels of a Wt. IKK␣ kinase globally and specifically activated NF-B-dependent genes in response to TNF␣ stimulation. In addition, these experiments also revealed a hitherto unknown requirement for IKK␣ to maintain the basal levels of expression of specific NF-B-dependent genes in the absence of an extracellular stress-like stimulus. Furthermore, the ability of a kinase-inactive IKK␣ mutant to rescue a portion of these NF-B target genes reveals that even though the IKK␣ protein is globally required for the expression of NF-B-dependent genes, its role as a functional kinase is also target gene-specific. In summary, our findings show that genes dependent on IKK␣ and NF-B can be formally divided into five distinct classes of responsive genes: 1) genes that require a functional Wt. IKK␣ kinase for their stimulus-dependent, NF-B-dependent expression; 2) genes that require a functional Wt. IKK␣ kinase for their stimulus-independent basal NF-B-dependent expression; 3) genes that require a functional Wt. IKK␣ kinase for their signal-dependent rescue but only require an IKK␣ protein for their basal, stimulus-independent expression; 4) genes that require the IKK␣ protein regardless of its kinase activity for their stimulus-dependent, NF-B-dependent expression; and 5) genes that require an IKK␣ protein regardless of its kinase activity for their stimulus-independent, basal NF-B-dependent expression.
Modifications of NF-B subunits and other post-translational nuclear processes are also necessary for the induction of NF-B target genes, and a number of reports have implicated IKK␣ in these events. Phosphorylation of serines 276 (in the Rel Homology domain/RHD) and serines 529 and 536 in the transcriptional activation domain (TAD) of the NF-B p65/ RelA subunit have been suggested to play activating roles, and a number of kinases have been directly implicated in this step including the IKKs (45)(46)(47)(48)(49)(50)(51)(52). Transactivation by p65/RelA in response to TNF␣ was localized to serine 529 within the p65/ RelA TAD and found to mediate its transcriptional activation independent of the ability of NF-B to bind DNA (89). Phosphorylation of serine 276 was also found to be involved in the activation of p65/RelA at least in part by controlling its association with the p300/CBP coactivator or the histone deacetylase-1 (53). IKK␤ and IKK␣ were both implicated as downstream effectors of Akt-dependent signaling targeted to serines 529 and 536 in the p65/RelA TAD, which in part appeared to involve the engagement of the CBP/p300 transcriptional coactivator (90,91). PTEN, a negative upstream effector of Akt, was also reported to inhibit TNF␣-induced NF-B activation (92,93), which was subsequently shown to occur solely at the level of p65 transactivation (94). Additionally, IL-1-and Akt-mediated NF-B activation was found to involve p65 TAD phosphorylations with a codependence on IKK␣ and IKK␤ (25). TNF␣induced phosphorylation of p65/RelA serine 536 was recently shown to be dependent on both IKK␣ and IKK␤ and mediated by TRAF2, TRAF5, and TAK1 signaling (54), and a requirement for IKK␣ in p65 serine 529 phosphorylation and NF-Bdependent transcriptional activation in response to LT␤ stimulation has also recently been reported (55). In addition, the activation of NF-B by the HTLV-Tax1 protein involves the specific phosphorylation of p65 serines 529 and 536, requiring IKK␣, but not IKK␤ (56). The IKK␣ mechanism of action in the canonical NF-B pathway has also been proposed to be purely nuclear in nature. In this context, IKK␣ has been shown to migrate into the nucleus (57) and associate with the promoters of NF-B-dependent genes upon TNF␣ stimulation (27,28). More recently mitogenic activation of the c-fos gene by epidermal growth factor-dependent signaling was found to require the constitutive and induced recruitment of p65/RelA and IKK␣, respectively, to the c-fos promoter (58). Because IKK␣ was also found to phosphorylate serine 10 of histone H3 in vitro, and the phosphorylation status of histone H3 in IKK␣(Ϫ/Ϫ) MEFs was enhanced by introduction of Wt. IKK␣ (27,28), these results suggested that IKK␣ might be functioning by enhancing the transcriptional accessibility of the chromatin of NF-B target genes and potentially a broader range of genes that also respond to TNF␣ stimulation. More recently, IKK␣ was also shown to be critical for the signal-dependent expression of the cIAP-2 and IL-8 NF-B-dependent genes in several cellular contexts by virtue of its required ability to phosphorylate the SMRT corepressor, thereby also inducing chromatin activation by facilitating the exchange of transcriptional corepressors for coactivators (29).
Wt. IKK␣ Kinase Restores the Expression of NF-B-dependent Genes in IKK␣-null MEFs-Our comparative genomic analysis of the abilities of Wt. IKK␣ and a kinase-dead IKK␣(K44M) ATP binding domain mutant indicate that IKK␣ kinase activity is required for the activation of most but not all genes whose expressions are dependent on NF-B. Indeed, we find that IKK␣ kinase activity is required for the induction of numerous TNF␣-responsive NF-B target genes (including RANTES, IB␣, IL-6, ISG-15, IFITR-1, mGBP-2, mGBP-3, Gro1, Fas ligand, Jun-B, caspase-11, ScyD1, serum amyloid A3, MMP3, MMP13, ScyB5/LIX, Ptx3, Schaflen 2, Met1, and   Fig. 7, and TaqMan real time PCR analyses of selected genes are in Fig. 9. Genes induced by TNF␣ are assigned a (ϩ) sign, and genes whose expression was not significantly stimulated by TNF␣ were given a (Ϫ). The IKK␣(K44M) mutant in a TNF␣-responsive manner comparable to that achieved by Wt. IKK␣ rescued three genes highlighted in gray. C/EBP␦), many of which we identified previously as targets of proinflammatory cytokine-mediated NF-B activation in Wt. MEFs but not in IKK␣-null MEFs (26).
The expression of the majority of IKK␣-rescued NF-B target genes in IKK␣-null MEFs was rescued by Wt. IKK␣ to within 2-fold of their levels in Wt. MEFs (see -fold change values of representative set of 40 genes in Table I) with a portion of NF-B targets being expressed at higher levels in Wt. MEFs. This latter observation is not that surprising given that long term inactive genes are known to differentially acquire the attributes of transcriptionally inaccessible chromatin states, which is unlikely to be fully overcome by IKK␣ reexpression. In addition, some NF-B-dependent genes were also expressed at higher levels in IKK␣-rescued IKK␣(Ϫ/Ϫ) cells compared with their wild type counterpart (see -fold change values for IFITR-1, Gro1, and caspase-11 in Wt. IKK␣-rescued IKK␣-null cells compared with their levels of expression in Wt. MEFs in Table I were stimulated with 20 ng/ml TNF␣ for 2 h., and RNAs were prepared and subjected to TaqMan real time PCR analysis. The levels of expression of eight representative genes (IL-6, ISG15, SAA3, RANTES, VCAM1, A20, ATF3, and GADD45␤) were quantitatively compared. Each bar represents data obtained at least in duplicate with the indicated standard deviations. All samples were normalized to a glyceraldehyde-3-phosphate dehydrogenase probe set, and mRNA copy numbers were determined compared with a genomic DNA standard for each probe set.
by Wt. IKK␣ in IKK␣-null fibroblasts did not appear to be dependent on NF-B for their expression employing the criteria of being unaffected by the inhibitory effects of the IB␣ superrepressor in Wt. MEFs nor by the loss of the NF-B p50 subunit. However, because genes such as MMP13, an NF-B target gene in other cellular contexts (59 -61), were among this latter gene subset (see Figs. 4 and 6 and data not shown), this supports our conclusion that the large majority of TNF␣-responsive, IKK␣-dependent genes in these cells are also NF-B-dependent.
Surprisingly, a number of NF-B-dependent genes were dependent on IKK␣ for their basal, stimulus-independent expression in IKK␣-compromised MEFs (21 of the 53 representative genes rescued by IKK␣ in Tables I and II). Furthermore, these same genes were also expressed at significantly lower levels in NF-B-compromised MEFs compared with Wt. MEFs (data not shown). Thus our findings show that the IKK␣-containing signalsome is also required to maintain the basal, intrinsic expression levels of specific NF-B-dependent genes. Interestingly, 14 of 21 of these stimulus-independent/Wt. IKK␣-rescued genes were also preferentially rescued by the IKK␣(K44M) mutant in the absence of stimulation (see Table II), also demonstrating that IKK␣ kinase activity is not required to restore their basal levels of NF-B-dependent gene expression in most but not all cases.
IKK␣ Does Not Always Function as a Kinase to Ensure the Expression of NF-B-dependent Genes-We also employed duplicate microarray screens to compare the abilities of Wt. IKK␣ or a kinase-dead IKK␣(K44M) mutant protein to rescue NF-B-dependent gene expression in IKK␣-null fibroblasts on a genomic scale. These findings, in combination with quantitative TaqMan real time PCR assays on selected genes, reveal that even though IKK␣ kinase activity is required to activate the majority of NF-B-dependent genes, it is not functioning directly as a kinase for the expression of ϳ28% of the NF-Bdependent genes in this cell background. Interestingly, most of the latter subset of IKK␣(K44M)-rescued genes were unresponsive to TNF␣, with the exception of several genes (including M/CSF-1, A20, mVL30, and B94) (see Table II, Figs. 7 and 9). In contrast to their TNF␣-independent rescues by the IKK␣(K44M) mutant, GADD45␤/MyD118, ATF3, and JunB were rescued by Wt. IKK␣ in a TNF␣-dependent manner (Table II and Fig. 9). It is also noteworthy in this context that our quantitative TaqMan analysis revealed that Wt. IKK␣ compared with IKK␣(K44M) also rescued higher levels of GADD45␤, ATF3, and A20 expression (Figs. 8 and 9). Taken together these observations show that the nature of the Wt. IKK␣-versus IKK␣(K44M)-mediated rescues of GADD45␤, ATF3, JunB, and perhaps even A20 are intrinsically different from each other, revealing that their dependences on the IKK␣ protein occurs at more than one regulatory level.
Of particular interest, a larger proportion of the IKK␣-rescued genes, which were also rescued by the kinase-inactive IKK␣(K44M) mutant, encode proteins with functional properties in NF-B autoregulation, cellular proliferation, growth arrest, apoptosis, or cellular survival (M/CSF-1, PLF2, PLF3, GADD45␤/MyD118, A20, Sequestosome/p62, NDPP1/CARD 8, JunB, and ATF3). GADD45␤/MyD118 was first described to function as an effector of myeloid cell differentiation and as a member of a class of cell cycle checkpoint protein arresting cellular growth in response to DNA damage or in association with terminal cellular differentiation (62,63). In contrast and more recently, GADD45␤ has also been shown to be an NF-Bdependent antiapoptotic effector in response to TNF␣, where it appears to act by directly blocking MKK7/JNKK2 activity thereby leading to the suppression of the c-Jun NH 2 -terminal kinase (JNK) cascade in response to TNF␣ stimulation (64,65)and also by suppressing Fas/CD95/APO-1-induced caspase activation in response to CD40 triggering in B lymphocytes (66). Our results show that the induction of GADD45␤ by TNF␣ requires IKK␣ kinase activity for its NF-B-dependent expression. However, because we also find that a kinase-inactive IKK␣(K44M) mutant restored unstimulated levels of GADD45␤ expression in IKK␣-null MEFs, our results also indicate that other intrinsic properties of the IKK␣ protein contribute to the expression of GADD45␤ under different physiological circumstances, perhaps leading to other functional properties ascribed to GADD45␤/MyD118. Sequestosome/p62, an atypical protein kinase C-interacting protein, interacts with RIP (receptor-interacting protein), linking it to TNF␣-mediated NF-B induction with its inhibition or down-modulation also interfering with IL-1 and TRAF 6-dependent NF-B activation (67). A20 is a TNF␣-and IL-1-induced zinc finger protein, which has been reported to act as a negative effector of NF-B (68 -70), mediated by RIP and TRAF2 signaling (70). A20 was also recently shown to attenuate TNF␣-mediated NF-B acti- vation via the cooperative action of its two intrinsic ubiquitinediting domains resulting in the polyubiquitination of RIP, an essential mediator of the TNF receptor 1 signaling complex, thereby targeting RIP proteasomal destruction (71). Akin to GADD45␤, A20 has also been shown to block TNF␣-dependent apoptosis at least in part by preventing TRADD and RIP recruitment to TNFRI (72,73). NDPP1 a novel member of the caspase-associated recruitment domain (CARD) family of proteins, has also been reported either to promote or suppress apoptotic responses in specific cell types and to block TNF␣induced NF-B activation (74,75). ATF3, a member of the ATF/CREB family of leucine zipper transcription factors (also known as LRF-1/TI-241) (76,77) and a stress-induced transcriptional repressor (78 -80), has recently been shown to be dependent on both the NF-B and JNK signaling pathways for its expression in response to TNF␣ and nitric oxide (81). ATF3 has also been shown to play roles in either protection against or induction of apoptotic responses, dependent on the nature of the signal and cellular context (79,82,83). Finally, Proliferin 2 and Proliferin 3/mitogen-regulated protein 3, which were also rescued by both Wt. IKK␣ and IKK␣(K44M) in a stimulusindependent manner, are members of the prolactin growth hormone family of proliferins whose functional properties have been associated with cellular proliferation and migration, wound healing, and angiogenesis (84 -88). Because an IKK␣(K44M) kinase-inactive mutant was found to be incapable of phosphorylating serine 10 of histone H3 (28), our observations also indicate that not all genes whose expressions are codependent on IKK␣ and NF-B require IKK␣ kinase activity for histone H3 phosphorylation.
In summary, our global expression profiling analysis shows that IKK␣ functions in both kinase-dependent and -independent modes, thereby revealing the existence of five distinct classes of genes codependent on IKK␣ and NF-B in mouse embryonic fibroblasts. In addition to the importance of a Wt. IKK␣-containing signalsome for the activities of many stimulus-dependent target genes of NF-B, the Wt. IKK␣ protein also plays a role in the regulation of a distinct class of NF-B-dependent genes requiring only basal levels of active NF-B for their regulated expression. We envision that the IKK␣ kinase-independent mode of action to ensure the expression of a subset of NF-B-dependent genes may be attributable to a novel regulatory/docking-like property of the IKK␣ protein, which facilitates the recruitment of other regulatory factors required for the expression of specific downstream targets of the NF-B pathway.