Cross-species Analysis Reveals Evolving and Conserved Features of the Nuclear Factor κB (NF-κB) Proteins*

Background: NF-κB regulates transcription via binding to DNA and interactions with cofactors. Results: NF-κB binding to DNA and cytosolic IκBs is conserved, whereas binding to nuclear IκBs has evolved. Conclusion: There is distinct evolutionary pressure on two NF-κB/IκB binding interfaces. Significance: The results provide inroads into IκB-specific modulation of NF-κB activity.


NF-B is a key regulator of immune gene expression in metazoans. It is currently unclear what changes occurred in NF-B during animal evolution and what features remained conserved. To address this question, we compared the biochemical and functional properties of NF-B proteins derived from human and the starlet sea anemone (Nematostella vectensis) in 1) a high-throughput assay of in vitro preferences for DNA sequences, 2) ChIP analysis of in vivo recruitment to the promoters of target genes, 3) a LUMIER-assisted examination of interactions with cofactors, and 4) a transactivation assay.
We observed a remarkable evolutionary conservation of the DNA binding preferences of the animal NF-B orthologs. We also show that NF-B dimerization properties, nuclear localization signals, and binding to cytosolic IBs are conserved. Surprisingly, the Bcl3-type nuclear IB proteins functionally pair up only with NF-B derived from their own species. The basis of the differential NF-B recognition by IB subfamilies is discussed.
Mammalian immunity is a constantly evolving system with multiple levels of redundancy and compensatory mechanisms allowing flexible but robust protection against pathogens. Recent advances in genomics and bioinformatics allow for systematic comparisons of the defense mechanisms in a wide selection of animal taxa. This is critical for an in-depth understanding of the molecular machinery underlying the host defenses against pathogens.
Cnidaria is a diverse phylum of basal animals, including sea anemones, jellyfish, and corals, the important builder species of marine ecosystems (1). As any other animals, cnidarians are attacked by various pathogens (2). It has been shown that a freshwater cnidarian Hydra produces antimicrobial peptides in response to pathogen-associated molecular patterns, which serve as an indicator of innate immunity (3).
The starlet sea anemone (Nematostella vectensis) is an emerging model system to study physiological processes in basal animals (4). Its recently published genome analysis revealed a remarkably large number of shared genes and genomic linkage groups between Nematostella and vertebrates (5). A number of homologs of key mammalian immune molecules were identified in Nematostella (6). Many of them have no paralogs present in the genome, which suggests their non-redundant function (7). For example, the N. vectensis (Nv) 4 genome contains just one NF-B gene, NvNF-B p50, with p50 and p52 being the closest human homologs (8).
NF-B is the key regulator of gene expression in immune processes (9,10). There are five NF-B subunits in mammals (NF-B1, NF-B2, RelA, cRel, and RelB), which regulate gene expression acting as homo-or heterodimers (10). The N-terminal part of NF-B molecules consists of a Rel homology domain (RHD), which binds to DNA, and a dimerization domain (IPT (Ig-like, plexins, transcription factors)). The C-terminal part of NF-B forms a transactivation domain in RelA, RelB, and cRel or, alternatively, is represented by an ankyrin repeat domain (ARD) in NF-B1 p105 and NF-B2 p100 (11). The latter proteins are also present in proteolytically processed forms, NF-B1 p50 and NF-B2 p52 (9,12). The processing requires a glycine-rich region, which is located between the N-terminal part and the ARD (13).
NF-B proteins are regulated by a family of cofactors called IBs (inhibitor of B proteins) (14). In mammals, NF-B is sequestered in the cytoplasm as an inactive complex bound to IB (15). There are three mammalian cytoplasmic IBs, IB␣, IB␤, and IB⑀ (14). Upon stimulation, IBs are degraded, and NF-B translocates to the cell nucleus to activate gene expression (9,15). There are also three mammalian nuclear IBs, Bcl3, IB, and IB␦, which are involved in the regulation of late NF-B-dependent genes (16). All IB proteins are structurally related to the ARDs of NF-B1/2 (14). Bcl3 and IB have an N-terminal transactivation domain, enabling these proteins, in addition to their inhibitory roles, to function as NF-B transcription coactivators for certain genes (17,18). Two IB genes were found in Nematostella, NvIB and NvBcl3, related to human cytoplasmic IB⑀ and nuclear Bcl3, respectively (19).
Previous reports showed NvNF-B binding to a human NF-B DNA site (20), suggesting conservation of NF-B function across evolution. In this study, we undertook a systematic functional comparison of human and Nematostella NF-B p50 proteins. Our high-throughput protein/DNA binding analysis confirmed the data obtained by Sullivan et al. (20) and demonstrated overall similar DNA binding specificities of the mammalian and Nematostella NF-B proteins. Nonetheless, we discovered significant differences in their interactions with cofactors. NF-B interaction with cytoplasmic IB molecules remained conserved during evolution, whereas its binding interface with co-regulators of the Bcl3 family has changed. This suggests that the interface between NF-B and its IB coactivators may be under evolutionary pressure from invading pathogens, for which blocking NF-B-mediated immune responses would be advantageous.

EXPERIMENTAL PROCEDURES
Bioinformatics-Sequence alignments were performed using the ClustalW 2.0 tool (European Bioinformatics Institute).
Plasmids and Reagents-nfkb1/2 Ϫ/Ϫ murine embryonic fibroblasts, kindly provided by Dr. Alexander Hoffmann (University of California), and HEK293ET cells were cultured in Dulbecco's modified Eagles medium (PAA Laboratories GmbH) supplemented with 10% fetal bovine serum (Invitrogen) and 1% penicillin/streptomycin (PAA Laboratories GmbH) at 37°C in 5% CO 2 and 95% humidity. Mouse monoclonal antibodies against the His tag (70796-3, Novagen) and against the Myc tag (clone 9E10, Santa Cruz Biotechnology), protein G-Sepharose (GE Healthcare), and FLAG peptide and anti-FLAG-agarose (Sigma) were used. Human and Nematostella NF-B p50 cDNAs was generated from human 293ET cell or Nematostella polyp total RNA, respectively, and subcloned into the pEAK8-Myc and pETM11-His vectors for mammalian and bacterial expression purposes. The cloned Nematostella p50 cDNA used in this study corresponds to the Ser-67 allele of the gene (20). NvBcl3 and NvIB was amplified from Nematostella polyp cDNA and subcloned into the pBent2 vector with an N-terminal FLAG tag. Deletion and point mutants of human and Nematostella p50 proteins were generated by PCR using wild-type cDNA templates and subcloned into the pEAK8 vector as fusion constructs labeled with Myc or Renilla luciferase tags at their N termini. A 700-bp-long fragment of the Nvbcl3 gene promoter, upstream of the first coding ATG, was amplified from Nematostella polyp genomic DNA and cloned into the pGL3-Basic vector (Promega) to drive expression of firefly luciferase. Renilla luciferase-tagged NF-B and IB␣ constructs were kindly provided by Stuart Bloor (MRC Laboratory of Molecular Biology, Cambridge, United Kingdom). All of the generated constructs were confirmed by sequencing. The pGL3-lcn2-luc reporter and pcDNA-FLAG-HsIB expression plasmids have been described previously (21). pNF-B-luc and pRL-TK were purchased from Clontech.
Electroporation, RNA Extraction, cDNA Synthesis, and Quantitative PCR (qPCR)-Murine embryonic fibroblasts (10 7 cells/cuvette) were electroporated with 5 g of pEAK8-Myc vector, 1 g of pmaxGFP (Lonza), and 4 g of a carrier DNA (10 g in total) using the Amaxa MEF1 Nucleofector kit (Lonza) according to the manufacturer's instructions. Three days after electroporation, cells were stimulated with 1 g/ml LPS (Alexis Biochemicals) for 1 and 4 h or left unstimulated before collection. Total RNA was extracted from cells using the Qiagen RNeasy mini kit (Qiagen). Cells were lysed in Buffer RLT (Qiagen), and cDNA synthesis was performed using the extracted RNA, an oligo(dT) primer, and SuperScript III reverse transcriptase (Invitrogen). The cDNA was PCR-amplified using EfficienSee FAST qPCR MasterMix Plus dTTP (Eurogentec). The TaqMan gene expression assays for mouse Hprt (housekeeper control) and Lcn2 were acquired from Applied Biosystems.
Transfection and Reporter Assays-293ET cells were transfected in 96-well plates using Lipofectamine 2000 (Invitrogen). The pEAK8-Myc-NF-B p50 (wild-type or mutant) and/or pBent2-FLAG (encoding IB proteins) expression construct (10 ng/well each) was cotransfected along with a given firefly luciferase plasmid (pNF-B-luc, pNGAL-luc, or pNvBcl3(Ϫ700)pro-luc) and the pRL-TK plasmid (10 ng/well each). In the experiments shown in Figs. 4B and 5A, a concentration range of the IB-encoding plasmids was used: 3 and 10 ng (Fig. 4B) and 2, 5, and 10 ng (Fig. 5A) per well. One day after transfection, cells were lysed, and luciferase activities were measured in lysates using the Dual-Glo luciferase assay kit (Promega). For stimulation experiments, cells were stimulated the next day after transfection with 10 ng/ml TNF-␣ (PeproTech) for 6 h before collection. Data are presented as means Ϯ S.D. from triplicate wells of a representative experiment.
Protein Expression and Purification-Expression constructs for NF-B dimers used in this study were created as described (22). Briefly, pET vectors for expression in BL21(DE3) Escherichia coli (Merck) were used to produce His-tagged recombinant proteins. Proteins were overexpressed through induction with 0.2 mM isopropyl ␤-D-thiogalactopyranoside at 30°C for 5 h. Pellets of cells were harvested in nickel-nitrilotriacetic acid binding buffer with added EDTA-free protease inhibitor (Roche Applied Science) and pulse-sonicated for 2 min, and debris was removed by centrifugation at 16,000 ϫ g. NF-B proteins were purified by affinity chromatography in two steps: using first the nickel-nitrilotriacetic acid His-Bind resin system (Merck) and then biotinylated DNA oligonucleotides attached to streptavidin-agarose (Sigma). The bound proteins were eluted in the high-salt buffer (50 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 500 mM NaCl, 10% glycerol, and 0.01% Nonidet P-40) as described (22).
DNA Affinity Protein Purification-5Ј-Biotinylated DNA oligonucleotides containing the NF-B site from the 3Ј-UTR of the human TNF promoter (5Ј-biotin-(AGCT)GGGCATGGG-AATTTCCAACTCT-3Ј) or the control sequence (5Ј-biotin-(AGCT)GGGCATAAACCGGGTTAACTCT-3Ј) were used for NF-B purification as described (22). Briefly, the DNA oligonucleotides were immobilized on streptavidin-agarose (Pierce). Cell pellets were lysed in protein/DNA binding buffer (50 mM Tris-HCl (pH 7.4), 50 mM NaCl, 0.1% Triton X-100, 10% glycerol, and protease inhibitors) and subjected to a French press, and the soluble fractions (separated from insoluble debris by ultracentrifugation) were incubated for 2 h with the DNA sorbent at room temperature with shaking. The proteins were eluted in high-salt buffer.
Protein Binding Microarrays-We designed 8 ϫ 15K Agilent arrays using eArray as we have described in detail previously (23). Briefly, the canonical NF-B consensus binding sequence GGRRNNYYCC was expanded into the RGGRNNHHYYB 11-mer motif, which was processed using the principal coordinate method (24). The outcome was 803 DNA sequences that are representative of the "k-mer space" encompassed by the expanded motif. The resulting z-score was obtained using log 2transformed intensities, and the median of replicates was calculated for each probe within every array (supplemental Table  1). The binding affinities for each protein were calculated using three technical replicates. The Cy3 values from the double stranding were used to normalize the Cy5 values of the protein/ DNA binding. To approximate binding affinity values between proteins, these values were logged (log 2 ), and a z-score was created using the following formula: z-score ϭ (log 2 value Ϫ median of array)/S.D. of array. The z-scores were used as input for the MultiExperiment Viewer software (25,26) to create a heat map.
LUMIER-Luminescence-based mammalian interactome (LUMIER) mapping (27) was used to rapidly test protein/protein interactions between NF-B and IB proteins. Two putative interactors fused to either N-terminal FLAG or Renilla luciferase tags were coexpressed in 293ET cells. One day after transfection, the cells were lysed in immunoprecipitation buffer (10% glycerol, 150 mM NaCl, 0.1%. Triton X-100, 20 mM Tris-HCl (pH 7.4), 5 mM EDTA, and protease inhibitors), and the post-nuclear supernatants were incubated for 2 h with FLAGagarose. The beads were then washed four times with immunoprecipitation buffer, and the proteins were eluted for 30 min with FLAG peptide diluted to a final concentration of 150 g/ml in Renilla lysis buffer (Promega). The luciferase activity was measured in the eluates and total lysates using the Renilla luciferase assay system (Promega). The magnitude of luciferase activity correlates with the binding affinity within a pair of interactors. The data are labeled as -fold binding and are presented as the ratio of luciferase activity in eluates and lysates normalized against the control (empty vector).
ChIP-A total of 10 7 HEK293ET cells were fixed by adding 1% formaldehyde (final concentration) for 5 min at room temperature. Nuclear extracts were subjected to 6ϫ 12-s pulses of sonication using a Vibra-Cell VCX130 processor (Sonics) at 20% amplitude. For immunoprecipitation reaction, nuclear extracts were precleared with a protein G-Sepharose bead slurry (GE Healthcare) for 2 h and then incubated with 2 g of 9E10 or isotype control antibodies overnight at 4°C with rotation. Immunocomplexes were collected with protein G-Sepharose beads for 30 min, rigorously washed, and eluted. Crosslinked protein-DNA complexes were reversed by incubation overnight at 65°C, and DNA fragments were purified using the QIAquick PCR purification kit (Qiagen). The immunoprecipitated DNA fragments were interrogated by real-time PCR using SYBR Premix Ex Taq II Master Mix (Takara Bio) and the indicated primers for the TNF promoter (5Ј-GGAAGCCAAGAC-TGAAACCAGCA and 5Ј-CCGGGAATTCACAGACCCC-ACT) and IL-10 promoter (5Ј-CCTGTGCCGGGAAACCTT-GATTGTGGC and 5Ј-GTCAGGAGCACCAGGCAACAGA-GCAGT) regions. Data were analyzed using Roto-Gene 6000 software (Corbett Life Science). All primer sets were tested for specificity and equal efficiency before use.

The DNA Binding Properties of NF-B Are Evolutionarily
Conserved-An amino acid sequence alignment of NF-B proteins from multiple mammalian species shows great conservation of their RHDs (8). We wished to systematically characterize the DNA binding preferences of NvNF-B p50 and its Homo sapiens (Hs) orthologs NF-B1 (HsNF-B p50) and NF-B2 (HsNF-B p52). The proteins were expressed in E. coli and purified by DNA affinity chromatography, and their functional activity was confirmed using specific and scrambled NF-Bbinding sequences in an oligonucleotide pulldown assay (Fig.  1A). Used as an additional control, a DNA-binding mutant of NvNF-B p50, FRY 3 AAA, failed to bind NF-B-specific DNA (data not shown).
Next, the binding of NvNF-B p50 to 803 11-mer sequences within the generalized NF-B consensus sequence RGGRNN-HHYYB flanked by four distinct flanking sequences was examined using double-stranded DNA microarrays essentially as described (23). We built a heat map of NvNF-B p50 binding by arranging variant 11-mer sequences in columns and color-coding the ranked binding affinities for these sequences from high (yellow) to low (blue) (Fig. 1B). The NvNF-B p50 binding profile was compared with those of HsNF-B p50, HsNF-B p52, and another mammalian homodimer, NF-B RelA. As expected, HsNF-B p50 and HsNF-B p52 had the most similar profiles to NvNF-B p50 (z-score correlation coefficients of 0.88 and 0.95, respectively), whereas the RelA binding profile was the most distant (z-score coefficient of 0.65) (Fig. 1C). The correlations between z-scores of the binding affinities of the selected NF-B dimers are shown in supplemental Table 1, and they were comparable with previously determined correlations within the HsNF-B family (23).
Despite a clear conservation of the 3Ј-pyrimidine-half DNA consensus sequences, i.e. a CC duplex at positions 9 and 10 (or on the complementary strand, a GG duplex at positions 2 and 3), we also observed differences in the DNA barcode when we compared the top 20 highest binding motifs of the individual NF-B proteins (Fig. 1D). The signature GGxxxxxxCC symmetry of the canonical NF-B motif observed in the case of human p52 and Nematostella p50 is lost in human p50, which has evolved to have a larger affinity for N 6 CCCC(T/C) sequences. The latter observation is supported by a previous study that reported the DNA binding preferences of mammalian NF-B dimers, including p50 (28). Nevertheless, taking into account the evolutionary distance of 600 million years between Nematostella and mammals (29), the in vitro DNA binding specificities of NF-B orthologs are remarkably similar. This is in contrast to noticeable differences between the DNA binding patterns of mammalian NF-B paralogs (human RelA versus NF-B1/2) (Fig. 1B).
To compare the efficiency of in vivo binding of HsNF-B1 p50 and NvNF-B p50 to the NF-B-regulated human gene promoters, NF-B proteins were cloned into a mammalian expression vector containing a Myc tag and expressed in the human cell line HEK293ET. The cell lysates were subjected to ChIP using anti-Myc (clone 9E10) or IgG control antibodies, followed by qPCR with specific primer sets. HsNF-B1 p50 and NvNF-B p50 proteins showed a similar degree of recruitment to the human TNF and IL-10 promoters (Fig.  1E). Therefore, both DNA binding specificity and genomic FIGURE 1. DNA binding profiles of Nematostella and human NF-B dimers. A, NF-B proteins were expressed in E. coli, and the soluble fractions of bacterial lysates (Input; separated from insoluble debris by centrifugation) were incubated for 2 h at room temperature with streptavidin-agarose attached to biotinylated DNA oligonucleotides containing NF-B-binding (from the 3Ј-region of the TNF gene) or control sequences. The bound proteins were eluted with high-salt buffer, and the eluates were subjected to SDS-PAGE, followed by Coomassie Blue staining. B, a heat map of the binding profiles based on the microarray analysis of four NF-B dimers (presented in columns) was generated using MultiExperiment Viewer (25,26). The z-scores (supplemental Table 1) of each individual protein were used as input for MultiExperiment Viewer. Within the heat map, probes that contain the 803 11-mer sequences and represent k-mer space given by the consensus sequence RGGRNNHHYYB can be found as rows. A color gradient reflects the binding affinity z-scores of NF-B dimers for a probe, where high-affinity probes (positive values) are shown in yellow, z-scores near zero are shown in black, and low-affinity probes (negative values) are shown in blue (see side bar). Hierarchical clustering was used to describe relationships between binding profiles of the different dimers (Euclidean distance correlation and complete linkage analysis). C, pairwise comparisons of the DNA binding profiles of Nematostella and human NF-B dimers based on the array analysis. The value at the bottom of each graph is the correlation coefficient for the pair in the graph. These graphs were built for the z-score data sets. z-scores were obtained using log 2 -transformed intensities, and the median of replicates were calculated for each probe within every array. D, DNA barcodes of the NF-B proteins based on the top 20 highest binding motifs for each individual protein were created using WebLogo online software. E, recruitment of human and Nematostella NF-B p50 to human gene promoters. Plasmids encoding Myc-tagged HsNF-B p50 and NvNF-B p50 proteins were transfected into human 293ET cells. One day after transfection, cells were lysed, and the lysates were subjected to ChIP using anti-Myc (clone 9E10) or IgG control antibodies. The NF-B recruitment to gene promoters was analyzed by qPCR of the precipitated DNA using specific primers to the human TNF and IL-10 gene promoters. The data are presented as -fold change over a negative IgG control. APRIL

NF-B Interactions with IB␣ Are Evolutionarily Conserved-
The NF-B dimerization region is well conserved among various animal NF-B orthologs (8) and includes a nuclear localization sequence (NLS) of NF-B, which is recognized by IB proteins (30). To examine whether NvNF-B p50 is able to form dimers with different human NF-B subunits and IB␣, we employed a LUMIER assay (27), which allows measurement of potential binding between a pair of proteins. NvNF-B and HsNF-B were tagged with FLAG, whereas all of the human NF-B subunits and IB␣ carried a Renilla luciferase tag. The individual pairs of proteins were expressed in HEK293ET cells and immunoprecipitated from cell lysates using FLAG-agarose. Both NvNF-B and HsNF-B interacted strongly with IB␣ and all Rel proteins, whereas a negative control protein, FLAG-GFP, did not (Fig. 2). Thus, the dimerization properties of NF-B and its interaction with cytoplasmic IBs are evolutionarily conserved.
HsIB Discriminates between Human and Nematostella NF-B-The mammalian NF-B p50 subunit lacks a transactivation domain and requires a cofactor like HsIB to drive transcription of its dedicated genes (17). We used the LUMIER assay to compare HsIB binding of HsNF-B p50 and NvNF-B p50. FLAG-tagged HsIB was expressed along with the Renilla luciferase-tagged NF-B proteins. FLAG-RelA was used as a confirmed binding partner of p50. Interestingly, HsNF-B, but not NvNF-B, bound to HsIB, whereas both proteins bound well to RelA (Fig. 3A).
It is not entirely known how NF-B interacts with HsIB, with the C terminus of p50 being proposed to play role (31). To clarify which region of HsNF-B interacts with HsIB, we con-structed several deletion mutants of HsNF-B p50 and tested them in the LUMIER assay for their affinity to HsIB (Fig. 3B). The DNA-binding domain of HsNF-B p50 was dispensable for its interactions with HsIB, but so was its C terminus, as HsNF-B mutant N380 (amino acids 1-380) could still bind to HsIB (Fig. 3B). However, HsNF-B p50 mutant N355 (amino acids 1-355) failed to interact with HsNF-B (Fig. 3B). The alignment of the region encompassing amino acids 355-380 of human and Nematostella NF-B p50 demonstrated that it is only 40% identical between the species and encompasses the conserved NLS motif (amino acids KRxK, underlined in Fig.  3C). To further dissect the contribution of the NF-B p50 evolving versus conserved amino acids in recognition of HsIB, we used NLS-free mutants of HsNF-B p50 and NvNF-B p50 in the LUMIER binding assay and found that the NLS is important for NF-B binding to HsIB, as well as to HsIB␣, which was used as a control (Fig. 3D). Therefore, we concluded that the sequences outside of the NLS are responsible for differential recognition of HsIB by human and Nematostella NF-B p50.
Next, we sought to examine the functional consequences of the differential HsIB/p50 binding. We coexpressed HsIB with human or Nematostella NF-B p50 and a luciferase reporter driving the expression of the HsIB-dependent gene NGAL/lcn2 in human 293ET cells (Fig. 3E). HsNF-B p50 coinduced greater reporter activation compared with NvNF-B p50. Finally, we used a genetic complementation test in nfkb1/ 2 Ϫ/Ϫ murine embryonic fibroblasts (32) to examine expression of the endogenous lcn2 gene. Introduction of human p50, but not its Nematostella counterpart or the human p50 mutant lacking the C-terminal HsIB-binding region, could restore LPS-inducible expression of lcn2 in these cells (Fig. 3F).
In summary, these data indicate that HsIB can discriminate between human and Nematostella NF-B proteins. Thus, we hypothesized that the IB/p50 interaction interface might have undergone changes during evolution.
NvBcl3 Discriminates between Human and Nematostella NF-B-To test our hypothesis, we first examined binding of the Nematostella homolog of HsIB to both HsNF-B p50 and NvNF-B p50. Two IB-related proteins, NvIB and NvBcl3, have so far been identified in Nematostella (8,19). NvBcl3 is phylogenetically related to mammalian Bcl3 and HsIB and appears to be localized in both the cytosol and nucleus (8,19). NvIB is similar to mammalian cytoplasmic IBs: it has an N-terminal IB kinase phosphorylation motif, and it was shown to sequester NvNF-B in the cytoplasm when ectopically expressed in mammalian cells (8,19). We tested the binding of NvIB and NvBcl3 to HsNF-B p50 and NvNF-B p50 by LUMIER. NvIB strongly bound to both human and Nematostella NF-B p50, whereas NvBcl3 bound only to NvNF-B (Fig.  4A). Moreover, the expression of NvIB, but not NvBcl3, inhibited the TNF-␣-induced pNF-B-luc reporter in human HEK293ET cells (Fig. 4B), confirming the fact that NvIB, but not NvBcl3, can bind human NF-B.
To better understand the molecular basis of the differential recognition of NF-B by cytosolic and nuclear IB proteins, we considered crystallographic and molecular modeling studies of complexes between IB and NF-B (30). First, they showed that

. Conservation of NF-B protein/protein interactions. FLAG-and
Renilla luciferase-tagged proteins were expressed as pairs in 293ET cells. The cell lysates were subjected to immunoprecipitation using FLAG-agarose, the proteins bound to the sorbent were eluted with FLAG peptide, and the luciferase activities were measured in the eluates and total lysates. The data are shown as the ratio of luciferase activity in eluates and lysates normalized against the control (empty vector).
the IB␣ DGD motif interacts with the NF-B p50 NLS (30). We performed an amino acid sequence alignment of the DGD region for all human and Nematostella IB proteins and found that it is conserved in IB and Bcl3-like proteins (Fig. 4C). This supports our finding that the NLS is crucial for NF-B binding by both groups of IBs (Fig. 3D). Second, they predicted that loops located between ankyrin repeats 4, 5, 6, and 7 of IBs are involved in NF-B binding. For example, the loop located between ankyrin repeats 5 and 6 of human Bcl3 contains several residues contacting the NF-B p50 dimer (30). We generated an amino acid sequence alignment of this region for all human and Nematostella IB proteins and found that there is only a little conservation in the predicted residues contacting NF-B (Fig. 4D). It is therefore possible that the dynamic evolution of the indicated IB loops is responsible for the differential recognition of NF-B by cytosolic and nuclear IB proteins.
Inhibition of NF-B Activity by Cnidarian IB Proteins-NvIB and NvBcl3 have been shown previously to inhibit Nvp50-mediated activity of the synthetic pNF-B-luc reporter (19). We extended these observations (Fig. 5A) to NF-B-dependent Nematostella gene expression by analyzing the activity of a reporter construct driven by a 700-nucleotide upstream region of the predicted NvNF-B target gene, Nvbcl3 (Fig. 5B). NvNF-B induced pNvBcl3(Ϫ700)-luc reporter activity, FIGURE 3. Human IB discriminates between human and Nematostella NF-B. A, Renilla luciferase-tagged NF-B p50 proteins coexpressed with either FLAG-tagged RelA or HsIB. The protein complexes were pulled out from cell lysates using FLAG-agarose, the bound proteins were eluted with FLAG peptide, and the luciferase activities in the lysates and eluates were measured. The data are shown as -fold binding compared with the control (bacterial alkaline phosphatase). B, the deletions mutants of human NF-B p50, N-terminally tagged with Renilla luciferase (left), were coexpressed with either bacterial alkaline phosphatase (control) or FLAG-tagged HsIB. The protein/protein interactions were analyzed as described for A. Rel, RHD. C, the HsIB-binding region of HsNF-B p50 was aligned with the corresponding region of NvNF-B p50 using ClustalW software. The NLS motif is underlined. D, LUMIER assay of selected NF-B/IB interactions. Renilla luciferase-tagged wild-type human and Nematostella NF-B p50 or NLS mutant versions were coexpressed with either IRF3 (control) or FLAG-tagged IB proteins in 293ET cells. The protein complexes were pulled out from cell lysates using FLAG-agarose, the bound proteins were eluted with FLAG peptide, and the luciferase activities in the lysates and eluates were measured. The data are shown as -fold binding compared with the control (IRF3). E, HsNF-B p50 or NvNF-B p50 was coexpressed with or without HsIB and the NGAL/lcn2 promoter-based luciferase reporter. The luciferase activities were measured in cell lysates. The data are shown as -fold induction over the control (empty vector) as means Ϯ S.D. of triplicate experiments. F, plasmids expressing either GFP or NF-B proteins were delivered into nfkb1/2 Ϫ/Ϫ murine embryonic fibroblasts using electroporation. Two days later, cells were stimulated with 1 g/ml LPS for 1 and 4 h and subjected to RNA extraction. Induction of lcn2 mRNA synthesis was measured by qPCR. The data are shown as means Ϯ S.D. of a representative experiment.
whereas NvIB or NvBcl3 attenuated this induction (Fig. 5B). These data suggest that inhibition of NF-B activity by IB proteins is not a recent property but that it originated before the separation of cnidarian and triploblastic animal lineages.

DISCUSSION
NF-B is a key regulator of gene expression in inflammation and other immune responses, but little is known about the evolution of the NF-B system. In this study, we compared the functional properties of human and Nematostella NF-B proteins and found that their DNA binding specificities in vitro and in vivo are largely conserved and that evolutionarily conserved immune genes have been regulated by NF-B since, at least, the origin of cnidarians. We have shown that both human and Nematostella NF-B can interact equally well with all mamma-lian NF-B subunits and inhibitors HsIB␣ and NvIB. However, the binding of NF-B to the Bcl3 family of proteins is not conserved: HsIB and NvBcl3 can discriminate between human and Nematostella NF-B and bind only the transcription factor from their own species.
The similarity in the functional properties of NF-B proteins reported in this study is a direct result of evolutionary conservation of the amino acid sequence of NF-B (8). For example, the RFRYXCE region of the RHD, responsible for p50 binding to DNA (33,34), is highly conserved between Nematostella and mammalian NF-B proteins (8). Still, paralog formation in the vertebrate NF-B lineage appears to have led to a specialization of DNA recognition among different NF-B subunits within the limits imposed by structural constraints of the conserved elements of the RHD fold. Although human p50 and p52 subunits are most closely related to Nematostella p50 and have similar DNA binding specificities, a subunit more distantly related to them, NF-B RelA, shows divergent DNA sequence preferences. Thus, the heterodimerization of NF-B paralogs along with the dynamic evolution of the NF-B/DNA binding barcode might have stimulated complexity and flexibility in the NF-B transcriptome in vertebrates.
NF-B has been shown to control immune gene expression in both Drosophila and mammals (9,35). With the overall conservation of NF-B/DNA binding across the animal kingdom, it is logical to speculate that Nematostella NF-B may regulate immune genes in its own host also. A previous study predicted potential gene targets of NF-B in Nematostella and showed that NvIB gene promoter activity can be induced by the ectopic coexpression of NvNF-B in human cells (19). Here, we cloned the promoter region of another IB gene, Nvbcl3, whose mammalian paralog expression is known to be NF-B-depen- The Renilla luciferase-tagged human and Nematostella NF-B p50 proteins were coexpressed with one of the FLAG-tagged IB proteins in 293ET cells. The protein complexes were pulled out from cell lysates using FLAG-agarose, the bound proteins were eluted with FLAG peptide, and the luciferase activities in the lysates and eluates were measured. The data are shown as -fold binding compared with the control (bacterial alkaline phosphatase). B, 293ET cells seeded in 96-well plates were transfected with 3 or 10 ng (lo and hi, respectively)/well NvBcl3-or NvIB-encoding plasmid and 10 ng/well pNF-B-luc reporter plasmid. One day later, the cells were stimulated for 6 h with 10 ng/ml TNF-␣ before collection. The luciferase activities were measured in cell lysates. The data are shown as -fold induction over the control (empty vector) as means Ϯ S.D. of a triplicate experiment. C and D, the amino acid sequence alignments of human and Nematostella IB proteins. Sequences were aligned using ClustalW software, and the ankyrin repeats (AR) were labeled according to a previously used annotation (30). Conserved residues are shown in boldface, and the predicted loop residues of HsBcl3 contacting HsNF-B p50 are underlined. FIGURE 5. Nematostella NF-B activity is suppressed by Nematostella IB proteins. A and B, 293ET cells seeded in 96-well plates were transfected for 24 h with either 10 ng/well pNF-B-luc (A) or 10 ng of Nvbcl3 gene promoterbased (B) luciferase reporter plasmid and increasing amounts of the NvBcl3or NvIB-expressing constructs (2, 5, and 10 ng/well in A and 10 ng/well in B) or control plasmid (empty vector). One day later, the cells were lysed, and the luciferase activities were measured. The data are shown as -fold induction over the control (empty vector) as means Ϯ S.D. of a triplicate experiment. dent (18). As expected, Nematostella NF-B triggered the Nvbcl3 promoter-driven luciferase expression. This indicates that evolutionarily conserved immune genes, such as IBs, have been regulated by NF-B since, at least, the origin of cnidarians.
It is significant that the entire amino acid sequence of NF-B is not conserved. The 355-380-amino acid region next to the IPT dimerization domain shows only a partial similarity between human and Nematostella p50 proteins. This region is known to be a binding hub for IB proteins (30). Here, we have demonstrated that it is recognized differentially by the two subfamilies of IBs. Both human and Nematostella IB␣ proteins bind to human and Nematostella NF-B equally well and inhibit the transcriptional activity of both proteins. In contrast, NvBcl3 can bind only to Nematostella NF-B, but not human NF-B, whereas HsIB fails to recognize the cnidarian NF-B protein. The selectivity of the Bcl3 family proteins in NF-B binding is reflected in their function. In association with human NF-B, HsIB has been shown to drive lcn2 gene transcription (36). We have demonstrated that ectopic HsIB triggers less lcn2 reporter activity or lcn2 gene expression when coexpressed with Nematostella NF-B than when it is paired with human NF-B. Conversely, NvBcl3 can inhibit only NvNF-B-induced reporter activation but fails to suppress TNF-␣-induced NF-B activation in human cells.
What is the molecular basis of the differential recognition of NF-B by the two IB subfamilies? The in silico modeling suggests that loops of the ARDs, which show great sequence diversity, confer the specificity of NF-B/IB interactions (30). However, these predictions alone do not explain why cytoplasmic IBs bind strongly to both human and Nematostella NF-B proteins, whereas the Bcl3-type IBs are more selective and weaker binders. We hypothesize that the cytoplasmic IBs bind to NF-B proteins via conserved elements, whereas the Bcl3type IBs bind via the variable ARD loops, but more work is needed to address this issue experimentally.
The interface between NF-B and its IB family coactivators may be under evolutionary pressure from invading pathogens keen to hamper NF-B-mediated immune responses. Interestingly, a recent study has shown that a measles virus encodes an IB-like protein, which blocks the NF-B response by retaining RelA in the cytoplasm (37). Another example is an insect virus encoding eight proteins orthologous to Drosophila IB called Cactus, which have been shown to block the midgut melanotic response during infection (38). Therefore, pathogen-derived IB molecules can compete with NF-B coactivators, i.e. the Bcl3-type proteins, to suppress the host immunity. On the other hand, the host IB inhibitors do not interfere with a pathogen's agenda, and therefore, there is less evolutionary pressure on them.
In summary, in addition to demonstrating a remarkable conservation of the NF-B/DNA binding, dimerization properties, and interactions with cytosolic IBs, we have uncovered a previously unknown phenomenon of the evolving interaction interface between NF-B and nuclear IBs. It is conceivable that a stronger evolutionary pressure on the NF-B interaction with coactivator nuclear IB molecules has been exerted by the pathogens trying to evade the host immunity.