Novel NEMO/IκB Kinase and NF-κB Target Genes at the Pre-B to Immature B Cell Transition*

The IκB kinase (IKK) signaling complex is responsible for activating NF-κB-dependent gene expression programs. Even though NF-κB-responsive genes are known to orchestrate stress-like responses, critical gaps in our knowledge remain about the global effects of NF-κB activation on cellular physiology. DNA microarrays were used to compare gene expression programs in a model system of 70Z/3 murine pre-B cellsversus their IKK signaling-defective 1.3E2 variant with lipopolysaccharide (LPS), interleukin-1 (IL-1), or a combination of LPS + phorbol 12-myristate 13-acetate under brief (2 h) or long term (12 h) stimulation. 70Z/3-1.3E2 cells lack expression of NEMO/IKKγ/IKKAP-1/FIP-3, an essential positive effector of the IKK complex. Some stimulated hits were known NF-κB target genes, but remarkably, the vast majority of the up-modulated genes and an unexpected class of repressed genes were all novel targets of this signaling pathway, encoding transcription factors, receptors, extracellular ligands, and intracellular signaling factors. Thirteen stimulated (B-ATF, Pim-2, MyD118,Pea-15/MAT1, CD82, CD40L,Wnt10a, Notch 1, R-ras,Rgs-16, PAC-1, ISG15, andCD36) and five repressed (CCR2,VpreB, λ5, SLPI, andCMAP/Cystatin7) genes, respectively, were bona fide NF-κB targets by virtue of their response to a transdominant IκBαSR (super repressor). MyD118 andISG15, although directly induced by LPS stimulation, were unaffected by IL-1, revealing the existence of direct NF-κB target genes, which are not co-induced by the LPS and IL-1 Toll-like receptors.

NF-B transcription factors are established nuclear regulators of gene expression programs culminating in a host of cellular stress-like responses that play important roles in an organism's acquired and innate immune responses (reviewed in Refs. [1][2][3][4]. A variety of extracellular and endogenous stimuli, including viral and bacterial infections, oxidative and DNA-damaging agents, hyperosmotic shock, chemotherapeutics, and pro-inflammatory cytokines all result in NF-B activation (1,(3)(4)(5). NF-B factors bind to DNA as heterodimers assembled from five known proteins (RelA, c-Rel, RelB, p50, and p52) with each subunit contacting one-half of a conserved 10-base pair consensus motif (GGGRNWTYCC) (1,5). NF-B is generally held in an inactive state, tethered in the cytoplasm to inhibitory factors termed inhibitors of NF-B (IBs) 1 (1,5). Activators of NF-B cause the specific phosphorylation of pairs of amino-terminal serines in the IBs, which mark them for ubiquitination and subsequent proteasomal destruction. NF-B then becomes available to activate its nuclear target genes (1,5). IB phosphorylation is mediated by a high molecular weight signalsome complex comprising at least two direct IB kinases (IKK␣ and IKK␤, also called IKK1/CHUK and IKK2) and a regulatory, docking/adapter protein (NEMO, NF-B essential modulator, also called IKK␥/IKKAP-1/FIP-3) (reviewed in Refs. 3,4,6). IKK␣ and IKK␤ are atypical serine/threonine kinases possessing an amino-terminal catalytic domain and two carboxyl-proximal interaction motifs resembling leucine zipper and helix-loop-helix domains (7)(8)(9)(10)(11)(12). The essential role of the IKK signalsome in NF-B activation has been demonstrated in mice lacking IKK␤ or IKK␣ (13)(14)(15)(16)(17)(18). NEMO is an essential non-catalytic, adapter/docking component of the IKK signalsome. Loss of NEMO in cultured cells resulted in a complete lack of signal-induced NF-B activation (19 -22). More importantly, murine embryos that were genetically null for NEMO, akin to IKK␤ KO mice, succumbed to severe liver apoptosis due to a virtually complete block in NF-B activation (23).
Major issues about the IKK signaling pathway remain unexplored, including the short versus long term effects of NF-B activation to program cellular gene expression on a genomic scale. Because NF-B regulates a variety of ubiquitous and cell-type-specific gene products in different cellular contexts, we elaborated the signal-induced, NEMO-dependent gene expression program in the context of the 70Z/3 murine pre-B lymphoma line. 70Z/3 pre-B cells recapitulate aspects of the pre-B to immature B cell transition in response to NF-B activation (24 -26). In response to LPS, they uniformly differentiate into an immature B lymphocyte-like state by activating the transcription of a pre-rearranged -light-chain allele (24 -27). By employing immunoselection against surface-bound IgM, Mains and Sibley (28) isolated spontaneously arising mutants of 70Z/3 that were completely unresponsive to LPS treatment, failing to express light chains (28,29). Molecular and biochemical analyses subsequently revealed that the 70Z/3-1.3E2 variant was defective in a crucial NF-B signaling step, making the cells refractory to all NF-B-activating stimuli, with the exception of anti-oxidant-insensitive pathways and the HTLV-Tax-1 gene product (19). More recently, Yamaoka et al. (22) showed that, unlike the 70Z/3 parental line, the 70Z/3-1.3E2 variant lacked NEMO protein expression but their wild type phenotype was rescued by NEMO. With high density oligonucleotide arrays, we have performed a genomic analysis of signal-induced, NEMO-dependent NF-B induction in 70Z/3 versus 1.3E2 cells. A large number of novel induced and repressed NEMO/IKK target genes were revealed. Experiments with an IB␣(SS/AA) super repressor revealed that many of these genes are novel and direct targets of NF-B.
Stimulations to activate the IKK signalsome pathway in 70Z/3 and 70Z/3-1.3E2 cells were performed by supplementing growth media with 15 g/ml LPS and 10 ng/ml PMA (both from Sigma Chemical Co.) for 12 h, or LPS alone for 2 or 12 h or 20 ng/ml recombinant murine IL-1 (Life Technologies, Inc.) for 2 or 12 h prior to isolating total cellular RNAs. In some experiments, cellular protein synthesis was inhibited by co-incubation with 100 M anisomysin (Sigma) to block translational initiation.
Probe Preparation, Chip Hybridization, and Data Analyses-Total cellular RNAs were extracted from 70Z/3 and 1.3E2 cells with Triazol reagent (Roche Molecular Biochemicals). Poly(A) ϩ RNAs were isolated from total RNAs of unstimulated and LPSϩPMA-stimulated cells with Oligotex (Qiagen). Purified RNAs were converted to double-stranded cDNA with a SuperScript kit (Life Technologies, Inc.) and an oligo-dT primer containing a T7 RNA polymerase promoter (Genset). Biotinlabeled cRNAs were generated from the cDNA samples by an 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 previously (Affymetrix Gene Chip Expression Analysis technical manual (33)). Affymetrix murine chips (mouse 11K set, subA and subB) were used for hybridization. Chips were stained with streptavidin-phycoerythrin (Molecular Probes) and read with a Hewlett-Packard GeneArray scanner.
DNA microarray chip data analysis was performed using GENE-CHIP 3.2 software (Affymetrix). The quantitation of each gene expression was obtained from the hybridization intensities of 20 perfectly matched and mismatched control probe pairs (34). The average of the differences (perfectly matched minus mismatched) for each gene-specific probe family was calculated. The software computes a variety of different parameters to determine if an RNA molecule is present or absent (Absolute Call) and whether each transcript's expression level has changed between the baseline and experimental samples (Difference Call). In this work, all chip files were scaled to a uniform intensity value (1500) for all probe sets. For a comparative chip file (such as stimulated Wt. versus stimulated Mut.), the experimental file (stimulated Wt.) was compared with the baseline file (stimulated Mut.). To minimize false positives, the following criteria were selected for significant changes for each primary screen: 1) the change in the average difference across all probe sets was Ͼ3-fold; 2) for induced genes, a difference call of "increase" or "marginal increase" should be present, and an absolute call of "presence" should be associated with the experimental file; 3) for suppressed genes, a difference call of "decrease" or "marginal decrease" should be present, and an absolute call of "presence" should be associated with the baseline file.
Hierarchical clustering was performed with the Cluster program (available at the Stanford Web site) as described previously (35). Genes that showed Ͼ3-fold changes in at least two of the Wt.ϩ/Mut.ϩ comparisons (i.e. IL-1, 2 h; IL-1, 12 h; LPS, 2 h; or LPS, 12 h) were subjected to clustering analysis for eight stimulated Wt. and Mut. samples (see Fig. 2). Genes that were detected as absent in all eight arrays were removed. The average difference values (representing the quantity of mRNA, see above) of the selected genes (360) were median-centered by subtracting the median-observed value, 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 through one cycle of K-means clustering (K ϭ 5) and then further clustered by average linkage clustering analysis of Y axis (genes) using an uncentered correlation similarity metric, as described in the program Cluster. Average difference values of 50 or less were set to 50 before median centering and normalization. The clustered data were visualized by the program TreeView (available at the Stanford Web site).
RT-PCRs and TaqMan Real-time Quantitative PCR-RT-PCRs were performed as previously described (36). To establish their relative qualities, serial dilutions of cDNAs were amplified with ␤-actin and GAPDH-specific primers for internal standardization. Similarly, linear response ranges were determined for each gene to semi-quantify their levels of expression as a function of LPS stimulation in 70Z/3 and 1.3E2 mutant cells. The sizes of PCR products corresponded to those expected for each gene. PCR primer pairs were 22-to 24-mers, and their nucleotide sequences are available from the authors upon request.
TaqMan Real-time quantitative PCR is based on a fluorogenic 5Јnuclease assay (37). The same total RNA samples that were used to prepare probes for microarray hybridization were treated with Dnase I followed by RNeasy Mini protocol for RNA cleanup (Qiagen). The Taq-Man probe consists of an oligonucleotide with a 5Ј-reporter dye (FAM) and a 3Ј-quencher dye (TAMRA). To measure the gene copy numbers of the target transcript, cloned plasmid DNA or mouse genomic DNA was serially diluted and used to produce a standard curve as described elsewhere (38). Data from TaqMan PCR analyses were normalized based on mRNA copy numbers of GAPDH using the TaqMan rodent GAPDH control reagents (Applied Biosystems).

RESULTS
The 70Z/3 pre-B line (Wt.) and its 1.3E2 NF-B signalingdefective mutant (Mut.) were initially exposed to a combination of 15 g/ml LPS and 100 ng/ml PMA (phorbol 12-myristate 13acetate) for ϳ12 h (equivalent to one to two cell generations) to simulate a condition of long term, constitutive NF-B activation. Employing DNA microarrays of 11,800 known cellular genes and expressed sequence tags (Affymetrix Mu11KsubA and Mu11KsubB Arrays), 1.3% of genes displayed greater than 3-fold increases whereas 0.9% revealed greater than 3-fold decreases in expression in comparisons of 12-h stimulated 70Z/3 wild type versus 1.3E2 mutant cells. Independent microarray screenings of both Wt. and Mut. cells that were either unstimulated or stimulated by 15 g/ml LPS or 20 ng/ml IL-1 were also performed. Genes affected 3-fold or more in the primary screening (Wt.ϩ/Mut.ϩ, 12-h LPSϩPMA) were confirmed in a 12-h LPS screen with only occasional variations. The -fold changes of these selected hits in various comparisons between Wt. and Mut. stimulated and unstimulated cells were visualized as a two-color image (Fig. 1). Most hits were also confirmed in independent microarray screens of stimulations for 12 h with IL-1 or LPS (Fig. 1), indicating that the different stimuli (LPSϩPMA, LPS, and IL-1) regulate these genes by a common mechanism.
Given that IKK/NF-B activation mediated by LPS, PMA, and IL-1 are all defective in the NEMO null 1.3E2 line, genes identified in the primary Wt.ϩ versus Mut.ϩ screens should represent direct and indirect targets of the IKK/NF-B pathway, provided that IKK/NF-B-independent NEMO signaling pathways do not exist. Even though NEMO/IKK␥ has been clearly established to function as an essential non-catalytic component of the IKK complex in vivo, this physiological role need not constitute its only cellular raison d'être. Because the IKK/NF-B pathway is latent in unstimulated cells, similar IKK/NF-B-dependent gene expression changes in stimulated versus unstimulated 70Z/3 Wt. cells would be anticipated. Consistently, the majority of the up-regulated and repressed genes identified in the Wt.ϩ/Mut.ϩ, LPSϩPMA comparison in Fig. 1 were also identified in another microarray screen comparing 12-h LPSϩPMA stimulated to unstimulated 70Z/3 wild type cells (see Wt.ϩ/Wt.Ϫ, LPSϩPMA in Fig. 1). As expected, genes affected in the Wt.ϩ versus Wt.Ϫ screen were not observed in a Mut.ϩ versus Mut.Ϫ screen (see Fig. 1). However, a fraction of the genes identified in the primary Wt.ϩ versus Mut.ϩ screen were inversely affected in the Mut.ϩ versus Mut.Ϫ screen (i.e. stimulated genes being repressed and repressed genes being stimulated) (see Fig. 1). The results indicate that these latter genes can only be effected by extracellular signals in the absence of NEMO or NF-B but not in the presence of NEMO or NF-B.
We classified the novel target genes into eight functional categories in Tables I and II. To ensure that genes identified in the primary Wt.ϩ versus Mut.ϩ screens have a higher proba-bility of belonging to the IKK/NF-B signaling pathway and not to an unknown, NEMO-dependent, IKK-independent pathway, we employed an additional selection criteria to assemble the relevant genes. Thus, all genes exhibiting inverse Mut.ϩ versus Mut.Ϫ effects of 2-fold or more in at least two independent screens were filtered out of Tables I and II. Table I displays the known genes identified by the initial screen of 70Z/3 Wt. versus 1.3E2 Mut. cells stimulated with LPSϩPMA for 12 h. In addition to co-stimulating with LPS and PMA for 12 h, we also performed chip screens of cells stimulated with LPS or IL-1 alone for 2 and 12 h. Most genes that were identified by the additional screenings are the same as those identified by the initial LPSϩPMA screen (Table I). Table II displays genes that were revealed by the 2-h LPS, 2-h IL-1, and 12-h IL-1 screens, which were not modulated more than 3-fold in the primary 12-h LPSϩPMA screen. Genes such as Etl-1, TNF-␣, Bcl-2, N-myc, PAC-1, PLA2, and 2B4 were only affected in the 2-h stimulation time (Table II). Genes presented in Tables I and II also showed minimal expression changes in a subsequent screen of Wt.Ϫ versus Mut.Ϫ cells ( Fig. 1 and Tables I and II), providing additional confidence that the activated IKK pathway targets them. Taken together, these results are consistent with most of the genes in Tables I and II being co-dependent on NEMO and the IKKs to activate the NF-B signaling pathway.
We used the hierarchical clustering method to identify gene expression patterns in Wt. and Mut. samples stimulated by  Genes are grouped based on their physiological functions. Within each group, stimulated genes are listed first followed by the repressed genes. Fold changes for all genes in four other independent screens are also listed. WtϪ and MutϪ represent unstimulated 70Z/3 and 1.3E2 cells. Genes were selected based on Ͼ3-fold changes in the primary screen of Wt.ϩ/Mut.ϩ (LPS ϩ PMA, 12 h) and a confirming Ͼ3-fold change in at least one other Wt.ϩ/Mut.ϩ screen (LPS, 2 h; LPS, 12 h; IL-1, 2 h; or IL-1, 12 h). Genes which showed significant changes in Mut.ϩ/Mut.Ϫ are not presented and remain under investigation. Redundancy hits (corresponding to different oligonucleotide regions of the same gene) are noted in the gene description column. ET61599 and ET62172 represent one of several hits of immunoglobulin light chains and heavy chains, respectively. Accessions with "ET" designations are derived from the TIGR database. Fructose-1,6-bisphosphatase (2 hits) either IL-1 or LPS for 2 or 12 h. As shown in Fig. 2, most genes are commonly up-regulated or repressed by IL-1 or LPS. The 2and 12-h screens showed a remarkable degree of correspondence (Fig. 2), indicating that the effect of the IKK pathway was sustained between 2 and 12 h for most genes. However, there were classes of genes whose expression was dramatically up-or down-regulated only in the LPS12-h sample (Fig. 2). A rapidly repressed class of genes is also evident that contains smaller subclusters of coordinately regulated genes (see pre-BCR components VpreB and 5, and the protease inhibitors SLPI and PN-1 in Fig. 2). Hierarchical clustering also revealed coordinately controlled classes of 2-h LPS and 2-h IL-1-specific genes (Fig. 2), representing immediate-early response genes, that were not detected after 12 h of exposure to either stimulus ( Fig.  2 and Table II). The presence of known NF-B targets among the stimulated 70Z/3 up-regulated genes (such as IB␣, the RANTES chemokine, lymphotoxin-␤, Ig light chain, macrophage inflammatory protein, NF-B P100, C4b binding protein, Bcl-2, and TNF-␣) (reviewed in Ref. 39) verify the efficacy of the screen. A number of the novel target genes were also revealed in independent microarray screens of genes in other cell types (monocytes, macrophages, and mature B cells), which are upor down-modulated by multiple NF-B-activating stimuli, indicating that a significant portion of these target genes are not unique to the 70Z/3 pre-B cell background (data not shown). A selected set of novel targets was chosen for TaqMan real-time quantitative RT-PCRs to verify the microarray data and to determine their mRNA copy numbers before and after stimulation (see 12 representative examples in Fig. 3). Each gene responded to 2 and 12 h of LPS and IL-1 signaling in agreement with the primary chip screen data (see Fig. 3).
Novel Target Genes Directly Induced or Repressed by NEMOdependent Signaling-Thirty-five up-regulated and repressed genes (32 novel and 3 known NF-B targets) were selected for a combination of semi-quantitative RT-PCR and quantitative TaqMan real-time PCRs to validate their direct or indirect NEMO target status and to compare their expression time courses as a function of NEMO activation by different inducers. As internal reference controls, RT-PCR reactions were performed for Gapdh and ␤-Actin with limiting amounts of cDNA templates, prepared from total cellular RNAs of stimulated 70Z/3 and 1.3E2 cells (see GAPDH and ␤-Actin results for LPSϩPMA-stimulated cells in Fig. 4, A and B, and Gapdh for IL-1-stimulated cells in Fig. 4C). All but five (Lef-1, Stat1, Gas2, PtpN8, and Mkp-3) of these 32 novel genes were confirmed as direct NEMO targets of LPSϩPMA signaling, because they were modulated independently of de novo protein synthesis (Fig. 4B)

cells were initially obtained expressing the IB␣SR-IRES-Neo cassette (70Z/3-IBIN cells). To increase the penetrance of the super repressor in cells, the 70Z/3 IBIN population was sequentially infected with the IB␣SR-IRES-Puro virus (70Z/3-INIP cells).
70Z/3, 70Z/3-1.3E2, 70Z/3-IBIN, and 70Z/3-INIP cells were stimulated with LPS for 2 h, and each of the 27 novel genes exhibiting direct NEMO responses by RT-PCR in Fig. 4B were similarly re-evaluated. Expression of the IB␣SR protein in LPS-stimulated 70Z/3-IBIN and 70Z/3-INIP populations were verified by Western blotting (data not shown). Similar results were obtained with 70Z/3 cells harboring either one or two copies of IB␣SR. As shown in Fig. 5A, 18 genes responded to LPS stimulation in 70Z/3-IBIN and 70Z/3-1.3E2 cells in a similar fashion, substantiating their NF-B target status. EBI-3, which was previously shown to be an NF-B-dependent gene target in response to the EBV (Epstein-Barr virus) LMP-1 (latent membrane protein-1) gene product in human lymphoblastoid cells (42), was included as a positive control in Fig. 5A. Fig. 6). Unlike the RT-PCRs, the quantitative TaqMan real-time analyses also permitted a comparison of the relative mRNA copy numbers for each IKK/NF-B regulated gene (see Figs. 3 and 6). Individual genes were affected to varying degrees by the IB␣SR, revealing differential dependencies on NF-B for their regulated expression. Nevertheless, the NEMO null 1.3E2 cells exhibited greater penetrance over the 70Z/3-INIP populations in a number of cases. This was not unexpected, because the expression of the IB␣SR is likely to be somewhat heterogeneous, and its ubiquitination and proteasomal destruction are not completely abolished by its dual serine to alanine mutations (31,40,41). The remaining nine genes, which were direct targets of NEMO in Fig. 4B (CCR7, Mapk/Erk1, PN-1, Mirf5, CycD2, CycD3, HexII, BID, and Ich-3), were not significantly affected in either 70Z/3-IBIN or -INIP cells implying that: 1) they are regulated by low levels of NF-B that escaped IB␣SR sequestration or 2) they represent a class of NEMO-dependent, NF-B-independent target genes. Additional experiments are in progress to address these two possibilities.

of these 19 genes (including EBI-3) were selected for quantitative TaqMan real-time PCR with two independent populations of 70Z/3-INIP cells yielding similar results (see representative examples in
Because these screens were performed in the context of 70Z/3 pre-B cells, a number of these NF-B target genes might be expected to be more specific to the pre-B to immature B cell stages of B cell development. Nine of nineteen bona fide NF-B target genes were also affected in a similar fashion by an IPTG-inducible IB␣SR in response to CD40 ligand engagement of the CH12 mature B cell line (see RT-PCRs for R-ras, Pim-2, MyD118, Pea-15/MAT1, CD36, CD40L, Rgs16, CD82/ KAI1, and EBI-3 in Fig. 5B).

Distinct Categories of NEMO-responsive Genes Reveal Differential Modes of Gene Expression Programming-Many genes
were up-or down-modulated in a sustained fashion throughout the 2-to 12-h stimulation regimens, but brief and long term stimulations also revealed transiently responsive categories, immediate-early genes, and a delayed responder class (see hierarchical clustering in Fig. 2 and Tables I and II). Genes including N-myc, PAC-1, PLA2, and the NF-B target genes Bcl-2 and TNF-␣ were more significantly up-modulated within 2 h of stimulation (Table II). In contrast, Lef-1, Gas2, PtpN8,

FIG. 2. Hierarchical cluster image showing gene expression patterns of NEMO/IKK targeted genes under different stimuli (LPS and IL-1) for two time periods (2 and 12 h).
Average-linkage hierarchical clustering method was applied to cluster the mRNA expression values of IKK/NEMO target genes (see "Experimental Procedures"). The annotated genes are NF-B targets confirmed by literature or RT-PCR/Taqman results, except for CyclinD2, D3, N-myc, and PN-1. and Mkp-3 were repressed and STAT1 was stimulated by 12 h but not within 2 h of LPS or IL-1 stimulation (Table I and RT-PCR analyses not shown). Thus, LPS and IL-1 signaling also established a delayed program of gene expression with indirect dependence on NEMO. Interestingly, expression levels of several direct NEMO/NF-B targets continued to rise throughout the 12-h time course, suggesting that prolonged LPS and IL-1 signaling could maintain NF-B activity.
After 2 h of stimulation, six NF-B target genes (CD36, EBI-3, CD40L, UCRP/ISG15, and Rgs16) were induced in the 70Z/3 wild type line and were virtually undetectable in the 1.3E2 mutant, indicating that a NEMO-regulated signalsome was absolutely essential for their activation (see Figs. 3 and 4). Levels of each activated gene remained fairly constant throughout the time course. In the CH12LX mature B cell line (43), which constitutively expresses nuclear NF-B, considerably higher levels of MyD118, EBI-3, CD40L, and Rgs16 expression were observed. The expression of these genes was further induced in response to NF-B-activating stimuli (30) (see Fig. 5B). These observations would be consistent with the notion that these genes turn on at the pre-B/immature B cell transition at the time of NF-B activation.
Unexpected Classes of Novel NEMO/NF-B-dependent Repressed Target Genes-The NEMO-dependent NF-B signaling pathway is well known to orchestrate the transcriptional activation of genes encoding mediators of stress-like responses, although its potential ability to repress gene expression has not been examined. Surprisingly, our global screen for novel target genes, which were regulated by the NEMO signaling pathway, revealed several classes of repressed genes. Akin to many of the up-modulated targets, signal-induced repression of a portion of these repressed genes was also found to be independent of de novo protein synthesis, implying a direct, NEMO-dependent repression phenomenon. However, unlike the induced class of target genes, repression was usually incomplete or partial. In a number of cases, repressed targets were down-modulated in unstimulated wild type versus mutant cells and further downmodulated by NEMO-dependent signal transduction. Some genes were only repressed after prolonged stimulation implying an indirect link to the NEMO-initiated expression program. In addition, expressions of some repressed genes also exhibited more complex regulation being partially up-modulated by LPS or IL-1 stimulation of 1.3E2 cells, suggesting that they could be positively influenced by other signaling pathways in the absence of NEMO. Repression of CCR2/Mcp-1R, CMAP/Cysta-tin7, SLPI, VpreB, and 5 in 70Z/3 Wt. cells were shown to be dependent on NF-B, as their levels of expression in the presence of an IB␣SR rose to those observed in the 1.3E2 line. The mechanisms of NF-B-mediated repression in each of these examples remain to be determined. Conceivably, some control elements of these genes may preferentially bind to dimers of p50 and/or p52 NF-B subunits, excluding the binding of the p65 and c-Rel transactivation competent subunits. Alternatively, this phenomenon could be caused by physiological competition between NF-B and other transcriptional activators for the same limiting pool of the p300 or CREB-binding protein transcriptional co-activators (44) or other limiting co-factors necessary for their optimal transcriptional activity. Indeed such negative, transcriptional cross-talk between NF-B and p53 has recently been described (44).
Not All Direct NF-B Targets Are Induced by Both LPS and IL-1 Signaling-Interestingly, RT-PCR and quantitative Taq-Man PCR analyses confirmed that MyD118 and UCRP/ISG15 were strongly induced by LPS but unaffected by IL-1 (Figs. 3 4B and data not shown). UCRP/ISG15 and MyD118 were also tightly clustered in hierarchical expression analysis (Fig. 2), and both were also found to be direct targets of NF-B (Fig. 5A). Similarly, the RANTES chemokine (Scya5), an established NF-B target gene (45), which was dramatically induced by stimulation with LPS or LPSϩPMA, was only marginally affected in response to IL-1 (see Scya5 in Table I). MyD118 is a mediator of growth suppression and apoptosis in differentiating cells (46). UCRP/ISG15, a novel 15-kDa ubiquitin homologue, is activated as part of the cellular response to viral infection, type I interferons, and TNF-␣ (47,48).
It is important to note in this context that the Toll family receptors, responding to LPS and IL-1, both employ MyD88 as a common adapter protein. MyD88 funnels signals to other adapters and associated kinases, including IRAK, TRAF6, ECSIT, and MEKK-1, which bifurcate to activate NF-B via the NEMO/IKK complex and other mediators of mitogenic and stress responses via parallel MAPK/ERK cascades (49 -51). Indeed MyD88 KO mice presented a severe deficit in their physiological responses to IL-1 and LPS and a marked delay in the activation of NF-B and MAP kinases (52). Our findings demonstrate that LPS signaling can lead to direct activation of some NF-B target genes that do not respond to IL-1. These results imply that unlike the IL-1 Toll-like receptor, the LPS Toll-like-receptor 4 can also directly respond via an MyD88independent mechanism leading to IKK activation. Alternatively, MyD118, UCRP/ISG15, and Scya5 could also belong to a unique class of MyD88-dependent NF-B target genes requir- ing the coordinate, post-translational activation of other transcription factors, which respond to LPS but not IL-1. In contrast, all immediate-early IL-1 target genes revealed in this genomic screen also responded to LPS signaling in agreement with published reports that MyD88 knockout mice were completely defective in NF-B activation by IL-1 signaling.
Novel Effectors of Development and Cellular Differentiation-The IKK complex modulated gene products that represent regulators of cell fate determination and differentiation. Important extracellular factors or receptors orchestrating the acquisition of novel cell fates in embryonic development like Wnt10a (53) and Notch1 (54) were up-regulated with a direct dependence on NF-B activity (Figs. 5A and 6). Interestingly, Jagged 1, a Notch 1 receptor ligand, has also recently been shown to be a direct target of NF-B (55). Lef-1, an important transcriptional regulator of early T and B cell development (56), which responds to the Wnt signaling pathway regulating cell fate and proliferation (57), was repressed in response to prolonged IKK activation. Furthermore, the VpreB and 5 genes, encoding the proteins of the pre-B cell receptor that are required during the early phase of B lymphoid cell development, were coordinately repressed in response to both brief and long term LPS stimulation. Their levels of partial repression were similar after 2-or 12-h LPS or IL-1 stimulation, suggesting the recruitment of other regulatory factors or post-transcriptional effects occurs during the 12-h time course. Notably, the partial repression of both pre-B cell receptor genes was abrogated by enforced expression of an IB␣(S32A/S36A) super repressor, identifying them as direct NF-B repression targets. VpreB and 5 are positively regulated by early B cell factor and the E47 helix-loop-helix protein during early stages of B cell development, but their expressions are extinguished between the pre-B to immature B cell transition. However, factors involved in their coordinate, developmentally regulated repression have not been identified. Our results suggest that NF-B activation could play a direct, albeit partial, role in programming the disappearance of early B cell regulators, which would also be consistent with NF-B acquiring constitutive activity during later stages of B cell differentiation. Experiments with normal, differentiating pro-B and pre-B cells will be necessary to address the physiological significance of this interesting phenomenon. In contrast, the RelB NF-B subunit was positively autoregulated (Table I), in agreement with the appearance of constitutively elevated levels of RelB in maturing B cells (58). The B-ATF activator was also directly induced by NF-B activation (Fig. 5A). B-ATF encodes a member of the ATF/CREB family that is up-regulated in Epstein-Barr virusstimulated mature human B cells and likely functions as a cell-type, tissue-specific modulator of the AP-1 transcription factor complex (59).
Novel Immunomodulatory Genes-CD40L induces NF-B responses upon engaging the CD40 receptor, a member of the TNF receptor superfamily (reviewed in Ref. 39). Our results now show that CD40 ligand is itself a new direct target of IKK/NF-B induction. In fact, CD40L gene expression was virtually undetectable in unstimulated 70Z/3 and stimulated 1.3E2 mutant cells and was dramatically induced by LPS and IL-1 stimulation in an NF-B-dependent fashion, making its expression critically dependent on NF-B signaling in this cellular context. CD40L was also an NF-B target in the CH12 mature B cell background (Fig. 5B). CD40L/CD40 binding is critical for the growth, survival, and differentiation of maturing B lymphocytes (60). Future experiments will be directed to determining the importance of NF-B control for functional CD40L expression by immunoregulatory cells.
Two chemokine receptor genes (CCR7/EBI-1 and CCR2/ Mcp-1R) were respectively identified as novel stimulated and repressed NEMO signaling targets, with CCR2 repression also being dependent on NF-B activation. CCR7/EBI-1 was originally identified as a B cell-specific gene that is rapidly induced by EBV infection and also by a conditionally activated EBNA2 protein, a known NF-B activator (61). CCR7/EBI-1 encodes a seven transmembrane-spanning G protein-coupled chemokine receptor that has recently been shown to be essential for the physiologically appropriate seeding of mature dendritic cells (DCs) and resting T and B cells within the micro-environments of secondary lymphoid organs (62). Resting T cells and mature, antigen-presenting DCs up-regulate CCR7/EBI-1 to facilitate their migration to the periarteriolar lymphoid sheaths of secondary lymphoid organs (reviewed in Refs. [63][64][65]. Antigenbearing DCs emigrate from the epithelia, where they first encounter and become activated by foreign antigen as immature DCs. Interestingly, maturing, activated DCs produce large quantities of pro-inflammatory response chemokines like MIP-1, MCP-1, IL-8, and RANTES (all products by known NF-B target genes) at the site of antigen encounter. These chemokines are thought to maintain the recruitment of immature DCs expressing cognate receptors (such as CCR1, CCR2, CXCR1, and CCR5) (64). Interestingly, NF-B-activating, proinflammatory stimuli like IL-1, TNF-␣, and LPS facilitate DC maturation by down-modulating the expression of inflammatory response receptors (like CCR2/MCP-1R), while up-modulating the expression of CCR7/EBI-1 (64,65). Our results in the context of 70Z/3 pre-B cells suggest that coordinate activation of CCR7 and repression of CCR2 transcription could in part be controlled by the NEMO signaling pathway. EBI-3 was first identified as an EBV (Epstein-Barr Virus)induced gene in lymphoblastoid cells and encodes an immunomodulatory, 34-kDa secreted glycoprotein with homology to the p40 subunit of interleukin 12 and the ciliary neurotrophic factor receptor (66). EBI-3 protein associates with the p35 IL-12 subunit in vivo suggesting that it may function as a modulator of cell-mediated immune responses (66). EBI-3 was also shown to be up-modulated by the EBV latent membrane protein-1 (LMP-1), dependent on NF-B activation (42). Akin to CD40L, quantitative TaqMan PCR revealed that EBI-3 expression was also critically dependent on NF-B activation by either LPS or IL-1 NEMO-dependent signaling in the 70Z/3 pre-B line (see Figs. 3 and 6).
UCRP/ISG15 is a ubiquitin-like polypeptide that is transcriptionally induced by IFN␤-dependent antiviral responses and TNF-␣ (47,48). We found that NF-B activation was essential for UCRP/ISG15 expression by 70Z/3 pre-B cells (Figs. 5A and 6) and that it was also strongly induced by LPS stimulation of murine RAW 264 macrophages (data not shown). UCRP/ISG15 (ubiquitin cross-reactive protein) is produced by different cell types and secreted by human monocytes and lymphocytes displaying the properties of an immune cell modulatory factor. ISG15 was reported to stimulate the T cell-dependent expansion of B cell-depleted populations of CD56ϩ NK (natural killer) cells, to induce IFN␥ production by T cells and NK cytolytic activity against tumor cell targets, indicating that it enhances lymphokine-activated killer-like activity (67,68). It has been proposed to augment and target the effects of IFN␣ or IFN␤ (67,68).
Novel Arbitrators of Cellular Growth and Survival-A number of the novel target genes attest to the attributes of NF-B activation to promote cellular growth and survival. Surprisingly, the N-myc and Pim-2 proto-oncogenes were two of the most dramatically stimulated genes. N-myc was induced to the highest degree (more than 70-fold) by LPS or IL-1 signaling in both Wt.ϩ/Mut.ϩ and Wt.ϩ/Wt.Ϫ comparisons implying that, akin to its founding family member c-myc, it too is likely to be regulated by IKK/NF-B signaling. N-myc functions as an important transcriptional regulator of cell cycle progression, cellular growth, and differentiation and has recently been shown to functionally substitute for c-myc in vivo (69). Unlike N-myc's transient expression profile, Pim-2 levels were only modestly increased within 2 h of LPS or IL-1 stimulation but climbed to a maximum of 46-fold after 12 h of exposure to LPSϩPMA, at which point N-Myc had turned off (see Fig. 3 and Table II). Pim-2 and the related Pim-1 gene encode labile, cytoplasmic serine/threonine kinases that were discovered by virtue of their ectopic activation via Moloney Murine Leukemia proviral insertional mutagenesis in T cell lymphomas (70). Pim-1 and Pim-2 are highly expressed in mitogenically activated hematopoietic cells and are induced by a variety of cytokines that also induce NF-B (71). Notably, Pim-1 and Pim-2 collaborate with c-myc to induce neonatal pre-B cell leukemia in doubly transgenic mice (71). Moreover, Pim-1 and Pim-2 have recently been shown to be targets of gp-130-mediated STAT3 signaling (72). These signals then cooperate with c-myc to facilitate cell cycle progression by promoting cell survival and inhibiting the induction of apoptotic pathways (72). Pea-15/MAT1 also responded in a direct fashion to NF-B activation. Pea-15/MAT1 is a microtubule-associated phosphoprotein containing a death effector domain (73,74), which has been reported to elicit cellular survival or cell cycle progression pathways (73)(74)(75). Recent work indicates that Pea-15 binds to FADD and caspase-8, apical effectors of the TNF apoptotic pathway (73). In addition, astrocytes of Pea-15 null mice have been recently shown to be more susceptible to TNF-induced cellular death, implicating Pea-15 as a cellular survival factor (73). Interestingly, Bcl-2, a known survival factor and NF-B target gene, and the glycolytic enzyme Hexokinase II were also among the induced genes. The latter two proteins are converging antiapoptotic effectors that independently antagonize the opening of mitochondrial pores, thereby preventing the cytoplasmic release of apoptotic amplifiers like cytochrome c (76). Finally, a gene encoding a glucocorticoid-induced leucine zipper (GILZ) factor was repressed by NEMO signaling (Fig. 1 and Table I). GILZ is preferentially expressed in lymphoid cells and was reported to selectively protect T cells from anti-CD3-induced apoptosis in conjunction with reduced Fas and FasL expression (77).
NEMO Regulates Arbitrators of Cell Cycle Arrest, Apoptosis, and Metastatic Potential-The MyD118, BID, and Ich-3 genes encode proteins that cause cell cycle arrest or growth suppression; and all were induced in direct response to NEMO-dependent signaling (see Fig. 4, B and C). These polypeptides join a growing list of direct or indirect mediators of cell cycle arrest and cellular death that are positive targets of NEMO signaling and NF-B activation, including Fas-ligand, CD95/Fas, p53, Bcl-Xs, c-Myc, and IB␣ (reviewed in Ref. 4). MyD118 was a confirmed NF-B target in 70Z/3 and CH12 cells and is a GADD (growth arrest DNA damage) family member, which regulates growth arrest and apoptosis in differentiating hematopoietic and non-hematopoietic cells by P53-dependent and -independent pathways (46). BID encodes a novel death agonist that heterodimerizes with either agonists (BAX) or antagonists (Bcl-2) of cell death responses (78). BID counters the protective effects of Bcl-2 by enhancing mitochondrial permeability to release cytochrome c and expression of BID, without another death stimulus, thereby induces ICE-like proteases and apoptosis (78). Ich-3/Caspase11, a member of the ICE/CED-3 family of cell death genes, induces apoptosis that can be counteracted by Bcl-2 (79). Ich-3 null mice are resistant to LPS-induced endotoxicity and are also deficient in IL-1␣ and IL-1␤, because Ich-3 is essential for ICE activation (79).
Gas2 was also identified among the NEMO-dependent repressed genes, but only after long term LPS stimulation. Gas2, a component of the microfilament system, up-regulates at growth arrest but down-regulates upon cell cycle re-entry (80). Interestingly, in vivo cleavage of Gas2 is believed to lead to some of the microfilament and cell shape changes characteristic of apoptosis; and Gas2 is a known death substrate for Caspase-3 (80,81).
CD82/KAI1, a member of the tetraspan transmembrane 4 superfamily, was a confirmed NF-B target in 70Z/3 and CH12 cells. CD82/KAI1 expression was reported to diminish during the progression of a variety of epithelial malignancies (82,83) and appears to function akin to a tumor suppressor by inhibiting pulmonary metastases in experimental metastasis models of prostate cancer and melanoma (83). It has recently been shown to associate with the EGF receptor and to suppress EGF-induced lamellipodial extensions and cell migration by desensitizing EGF-induced signaling (84).
NEMO and NF-B Effects on Cell Cycle Regulators and Other NF-B-independent Signaling Molecules-A number of surprising cross-talk connections were revealed between the NEMO/NF-B activation pathway, components of independent signal transduction pathways, and regulators of cell cycle progression. Cyclin D2, known to be up-modulated in B lymphoid malignancies and preferentially expressed during B cell maturation (85), was a direct target of the NEMO signalsome (Fig. 4, B and C). TaqMan real-time PCRs also revealed that Cyclin D2 expression was absolutely dependent on NEMO expression (data not shown). In sharp contrast, Cyclin D3 (86) was directly repressed, indicating that the NEMO/IKK signaling pathway can have opposing direct effects on regulators of the G 1 /S transition. Interestingly, two pivotal members of independent signaling pathways, Mapk/Erk1 (87) and Stat1 (88) were upregulated as direct and delayed NEMO-dependent responses, respectively. R-ras, a G protein highly related to the H-Ras proto-oncogene (89), was directly induced by NF-B in 70Z/3 and CH12 cells. Rgs16, a negative regulator of G proteincoupled receptor signaling induced in response to bacterial infection (90,91), behaved in a similar fashion to R-Ras. In addition, Rlf (RalGDS-like factor), a candidate effector of the Ras and Rap1A GTPases (92), and Rgs2, a selective negative modulator of Gq signaling (90), were stimulated and repressed NEMO targets, respectively. PAC-1, a nuclear dual specificity phosphatase with specificity toward ERK and p38 MAPKs and previously shown to be transiently induced in response to ERK signaling in activated B cells (93), was directly induced by NF-B (Figs. 4B, 4C, and 5A). In contrast, Ptpn8 tyrosine phosphatase (94) and Mkp-3 dual specificity protein phosphatase (95) were repressed as part of the delayed response to NEMO-dependent signal transduction. A tyrosine kinase receptor and effectors of intracellular calcium levels were also among NEMO's immediate-early-responsive genes (see "Signal transduction" category in Table II). Each of these selected examples points to the surprising, unpredicted potential of NEMO and NF-B to alter the outcomes of diverse intracellular signaling pathways.
Differential Regulation of Pro-and Anti-inflammatory Mediators-CD36, a class B scavenger receptor of Ox-LDL (oxidized low density lipoprotein) (96), was dramatically up-regulated in response to NEMO-dependent signaling and was also a confirmed direct NF-B target in 70Z/3 cells and CH12 cells. Ox-LDL accumulates in cells of atherosclerotic lesions, playing an important role in foam cell development (96). Interestingly, Ox-LDL has also been reported to increase CD36 expression itself (97). In addition, CD36-mediated uptake of Ox-LDL has recently been shown to contribute to the expression of a variety of inflammatory response cytokines and to also increase NF-B activation (98). However, this is the first direct link between NF-B signaling and CD36 induction.
In contrast, several protease inhibitors were coordinately repressed in stimulated 70Z/3 cells, including CMAP/Cystatin 7/Cystatin F (99), PN-1 (protease nexin-1) (100), and SLPI (secretory leukocyte protease inhibitor) (101). CMAP/Cystatin 7, a cystatin-like mediator of liver metastasis that presumably functions as a protease inhibitor akin to related cystatin proteins (99), was directly repressed by NF-B activation. PN-1, a serapin class anti-protease and inhibitor of urokinase-type plasminogen activator (100), was a NEMO repression target in 70Z/3 cells (Fig. 4, B and C) and an NF-B repression target in CH12 cells (data not shown). SLPI is a potent inhibitor of elastase and cathepsin G serine proteases that suppresses the chronic, destructive phase of inflammatory reactions and syndromes characterized by overt tissue destruction. SLPI was a direct NF-B repression target in the 70Z/3 line (Figs. 5A and  6). SLPI has also been reported to suppress NF-B activity by specifically increasing IB␤ levels in inflamed lung tissue (102). These results suggest that therapeutic inhibition of NF-B activation may not only impede the initiation of inflammatory responses but might also ameliorate the more deleterious consequences of chronic inflammatory reactions and diseases by preventing the down-modulation of protective serine proteases. Counter regulation of protease inhibitors in response to brief or prolonged NF-B activation will be necessary in other cell types to assess the generality of these observations.
In conclusion, we have described the results of the first genomic screen for genes modulated directly or indirectly by activation of the NEMO/IKK␥ signaling pathway. More than 100 gene targets were identified in the context of a NEMO/ IKK␥ null, murine pre-B cell line stimulated by LPS or IL-1, and many of them represent novel NF-B target genes. Our findings allow the establishment and exploration of many new physiological liaisons between NF-B signaling and cellular gene expression programming.