Two carboxyl-terminal activation regions of Epstein-Barr virus latent membrane protein 1 activate NF-kappaB through distinct signaling pathways in fibroblast cell lines.

Latent membrane protein 1 (LMP1), an Epstein-Barr virus transforming protein, is able to activate NF-kappaB through its carboxyl-terminal activation region 1 (CTAR1) and 2 (CTAR2), but the exact role of each domain is not fully understood. Here we show that LMP1 activates NF-kappaB in different NF-kappaB essential modulator (NEMO)-defective cell lines, but not in cells lacking both IkappaB kinase 1 (IKK1) and 2 (IKK2). Mutational studies reveal that CTAR1, but not CTAR2, mediates NEMO-independent NF-kappaB activation and that this process largely depends on IKK1. Retroviral expression of LMP1 mutants in cells lacking either functional NF-kappaB inducing kinase (NIK), NEMO, IKK1, or IKK2 further illustrates distinct signals from the two activation regions of LMP1 for persistent NF-kappaB activation. One originates in CTAR2, operates through the canonical NEMO-dependent pathway, and induces NFKB2 p100 production; the second signal originates in CTAR1, utilizes NIK and IKK1, and induces the processing of p100. Our results thus help clarify how two functional domains of LMP1 persistently activate NF-kappaB through distinct signaling pathways.

The NF-B family of transcription factors plays crucial roles in the immune, inflammatory and apoptotic responses (23). NF-B activation is induced by a variety of stimuli including TNF-␣, interleukin-1 (IL-1), lipopolysaccharide (LPS), doublestranded RNA, Tax of human T-cell leukemia virus type I (HTLV-I), and LMP1 (24). NF-B is composed of dimers of p50, p52, RelA, RelB, or c-Rel that are endogenously complexed to inhibitor proteins called IB and NF-B precursor proteins p105 or p100, which sequester NF-B in the cytoplasm. In response to various stimuli, inhibitory proteins are phosphorylated at specific serine residues and are rapidly processed by the proteasome after polyubiquitination. This exposes the nuclear localization signal of NF-B, leading to its nuclear translocation. The phosphorylation step is usually controlled by the IB kinase (IKK) complex, which contains two catalytic subunits, IKK1/␣ and IKK2/␤, and the regulatory subunit NF-B essential modulator (NEMO/IKK␥/IKKAP), HSP90, and Cdc37 (25)(26)(27)(28)(29)(30)(31)(32). Although IL-1-or TNF-␣-induced nuclear translocation of NF-B was not impaired in IKK1-deficient cells (33,34), it was severely impaired in IKK2-or NEMO-deficient cells (35)(36)(37)(38). Thus, activation of the IKK complex constitutes a converging regulatory step in the NF-B signaling pathway.
Recent studies have revealed two distinct pathways of NF-B activation. The canonical pathway is triggered by many inflammatory stimuli including TNF-␣, IL-1, LPS, and doublestranded RNA; depends on IKK2 and NEMO; and induces specific phosphorylation of IB proteins. The non-canonical pathway is triggered by a limited number of stimuli including lymphotoxin-␤ (LT-␤), B cell-activating factor (BAFF), and CD40 ligand that function in the development, organization, and proper function of lymphoid tissue (39 -44). This pathway involves the phosphorylation-dependent processing of NFKB2 p100 to p52, which requires IKK1 and functional NIK, resulting in the nuclear translocation of p52-RelB dimers. A previous study on HTLV-I Tax uncovered a unique mechanism whereby this viral protein persistently activates NF-B. Tax resides in the cytoplasm associated with the IKK complex and physically interacts with p100, thereby inducing its phosphorylation and processing through recruitment of the IKK complex to p100 (45). Although NIK, p52, and RelB were previously implicated in the LMP1 signaling to NF-B activation (16,18,20,21), the precise mechanism by which LMP1 persistently activates NF-B has remained elusive. Our initial study on the LMP1 signaling identified a novel NEMO-independent NF-B activation pathway that is mediated by CTAR1 and involves aberrant expression of p52 and nuclear translocation of RelB. We further provide genetic evidence of how NIK, IKK1, and IKK2 are involved in this process.
Cells-Rat-1 and 5R cells were described previously (29). IKK2 Ϫ/Ϫ and NEMO Ϫ/Ϫ fibroblasts were kindly provided by Dr. Manolis Pasparakis (EMBL, Rome, Italy), IKK1 Ϫ/Ϫ IKK2 Ϫ/Ϫ fibroblasts were by Dr. Inder M. Verma (Salk Institute, La Jolla, CA), IKK1 Ϫ/Ϫ fibroblasts were by Drs. Michael Karin and Véronique Baud (University of California, San Diego, CA), and NIK aly/aly fibroblasts were by Dr. Mitsuru Matsumoto (Tokushima University, Tokushima, Japan). To generate cells carrying a mutation that results in impaired NF-B activation, we mutagenized Rat-1 cells stably expressing the HTLV-1 Tax protein to constitutively activate NF-B and then subjected them to a lethal selection that employs the combination of NF-B-dependent expression of herpes simplex virus thymidine kinase and ganciclovir. Rat-1 cells were stably transfected with pIRES1neo-tax, pIg2bsrH, and pIg2tkH, and selected under G418, hygromycin, and blasticidin S. Isolated clones were tested for resistance to blasticidin S and susceptibility to ganciclovir. One clone, D2-19, was found to express Tax, FIG. 1. LMP1 activates NF-B in the absence of NEMO. A, schematic representation of primers and a probe used for RT-PCR Southern blot analysis. B, total RNA (1 g) was subjected to reverse transcription and subsequent PCR amplification of nemo with N0 and C0 primers.
Nested PCR products with Nemo5Ј and Nemo3Ј primers were fractionated on a 1% agarose gel, transferred to a nylon membrane, and identified by Southern blot analysis using a [␥-32 P]ATP-labeled oligonucleotide probe shown in A. As a control, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA was PCR-amplified and visualized by ethidium bromide staining. C, semiquantitative RT-PCR analysis of nemo mRNA. One microgram of total RNA was extracted from ϳ1 ϫ 10 5 Rat-1 cells and reverse-transcribed into minus strand cDNA in 50 l of reaction. PCR was performed for 35 cycles in a reaction mixture containing 5 l of the above synthesized cDNA or serially diluted rat nemo gene, ranging from 0.01 to 10 11 copies/sample, with primers N0 and C0. The PCR products were separated on a 1% agarose gel and visualized by ethidium bromide staining. D, 5 l of the above PCR products were subjected to nested PCR with primers Nemo5Ј and Nemo3Јfor 35 cycles. The PCR products were visualized as in C. E, LMP1 induces NF-B-dependent reporter gene activation in 5R cells. 5R cells were co-transfected with 0.5 g of Ig-ConAluciferase, 0.5 g of EF1-lacZ, and 0.1 g of pSG5 or pSG5-LMP1 by the calcium phosphate precipitation method. Forty-five hours after transfection, cells were either left untreated or stimulated with 15 g/ml LPS or 10 ng/ml TNF-␣ for 3 h. Cells were harvested at 48 h after transfection. Each raw luciferase activity was divided by ␤-galactosidase activity to correct the transfection efficiency. The result shown is representative of three independent experiments performed in duplicate. F, LMP1 induces NF-B-dependent reporter gene activation in 1.3E2 cells. Cells were transfected by the DEAE-dextran method with 2 g of Ig-ConAluciferase and 8 g of pSG5 or pSG5-LMP1. Cells were lysed 24 h later and subjected to luciferase assay. exhibit constitutive NF-B activity, be resistant to blasticidin S, and be killed completely in the presence of ganciclovir. D2-19 cells were subjected to four rounds of mutagenesis with 10 g/ml frameshift mutagen ICR191, followed by lethal selection in the presence of 1 g/ml ganciclovir. One cell clone, N1, did not significantly induce B-DNA binding activity in response to TNF-␣ or LPS. This clone was found to have lost Tax expression and resistance to G418. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine, and antibiotics (100 units/ml penicillin, 100 g/ml streptomycin) at 37°C in humidified atmosphere with 5% CO 2 .
PCR, Cloning of Rat nemo, and Southern Blotting-One microgram of total RNA was extracted from ϳ1 ϫ 10 5 Rat-1 cells and reversetranscribed into minus-strand cDNA in a 50-l reaction using Superscript II (Invitrogen) under the instructions from the manufacturer. PCR was performed in a reaction mixture containing 5 l of the above synthesized cDNA. DNA sequence homology between murine and human NEMO at the amino and carboxyl termini enabled us to design primers N2 (5Ј-GTGCAGCCCAGTGGTGGCCCAG) and C1 (5Ј-CTACTCTATGCACTC-CATGAC) and to amplify part of rat nemo cDNA by PCR. 5Ј-Rapid amplification of cDNA ends (RACE) was conducted using SMART TM RACE cDNA amplification kit (Clontech) under the instructions from the manufacturer. PCR products were purified by agarose gel electrophoresis, ligated to p-GEM-T Easy (Promega) vector, and sequenced. Based on the sequencing results, primers N0 (5Ј-CCTAGGAGCTCCGATTCTGC) and C0 (5Ј-GGAGCTGTCTACCCTAATAGGGG) were used to amplify the entire coding sequence of rat NEMO. Five l of the initial PCR reaction was subjected to nested PCR with primers Nemo5Ј (5Ј-GGATCCAGCAG-GCACCTCTGGAAGA) and Nemo3Ј (5Ј-CTCGAGCTACTCTATGCACTC-CATGACATGTATC). DNA fragments were subcloned into the BamHI and XhoI sites of pcDNA3HA, and pcDNA3HA-NEMO-Rat-1 and pcDNA3HA-NEMO-N1 were generated and sequenced. For PCR-Southern analysis, PCR was performed using cDNA with primers N0 and C0, followed by nested PCR with primers Nemo5Ј and Nemo3Ј. PCR products were electrophoresed on a 1% agarose gel, visualized by ethidium bromide staining, and transferred to Hybond-N ϩ nylon membrane (Amersham Biosciences) by alkaline blotting. Hybridization was carried out in a standard protocol, using a probe prepared by PCR with primers Southern F0 (5Ј-CAGATGCTGAGGGAACGCT) and Southern R0 (5Ј-AGTTCCCCCAGCAATGATGT). The blot was exposed to an x-ray film (MXJB-1; Eastman Kodak Co.) at Ϫ80°C. Glyceraldehyde-3-phosphate dehydrogenase cDNA was amplified with primers rat-gapdh-F (CGGT-GTCAACGGATTTGG) and rat-gapdh-R (GTAGGCCATGAGGTCCACC). For semiquantitative PCRs, 5 l of cDNA samples used above and serially diluted rat nemo plasmid DNA, ranging from 0.01 to 10 11 copies/sample, were used as template. The first step PCR was done for 35 cycles with primers N0 and C0, followed by the second step PCR for 35 cycles with primers Nemo5Ј and Nemo3Ј, using 5 l of the first PCR products.
Transient Transfection and Luciferase assay-Transfection was carried out in six-well plates by the calcium phosphate precipitation method, FuGENE reagent (Roche Molecular Biochemicals), or DEAE- were co-transfected with 0.5 g of Ig-ConAluciferase and EF1-lacZ together with 0.1 g of pSG5 or pSG5-LMP1. Forty-five hours after transfection, cells were either left untreated or stimulated with 15 g/ml LPS or 10 ng/ml TNF-␣ for 3 h. Expression of wild type NEMO restores NF-B activation in response to TNF-␣ or LPS in N1 cells (right panel). N1 cells were transiently co-transfected with 0.25 g of Ig-ConAluciferase, 0.25 g of EF1-lacZ, and 1 g of pcDNA3HA or pcDNA3HA-NEMO-Rat-1 by the calcium phosphate precipitation method. Forty-five hours later, cells were either left untreated or stimulated with 15 g/ml LPS or 10 ng/ml TNF-␣ for 3 h. The luciferase activity was determined and normalized as described in the legend to Fig. 1E. Shown are results representative of three independent experiments done in duplicate. D, wild type NEMO enhances LMP1-induced NF-B activation in N1 cells. N1 cells were co-transfected with 0.5 g of Ig-ConAluciferase and EF1-lacZ, 0.1 g of pSG5 or pSG5-LMP1, and increasing amounts of plasmid encoding Rat-1 NEMO or N1 NEMO. The total amount of DNA (2.1 g) was kept constant by addition of pcDNA3HA vector. Cells were harvested at 48 h after transfection. Shown are results representative of three independent experiments done in duplicate. dextran method. Total amount of transfected plasmid DNA was kept constant with pcDNA3HA or pSG5. Where indicated, cells were stimulated with 15 g/ml LPS or 10 ng/ml TNF-␣ for 3 h before lysis. The luciferase activities were normalized on the basis of ␤-galactosidase activity. Experiments were repeated at least three times in duplicate.
Retrovirus Infection-Supernatants of transfected Plat-E cells (53) were recovered and passed through a 0.45-m filter every 12 h from 36 to 72 h after transfection, and either immediately used for infection of cells or frozen and stored at Ϫ80°C. For infection, cells on 100-mm dishes were exposed to 4 ml of virus supernatant in the presence of 10 g/ml Polybrene for 3 h. Rat-1 and 5R cells were harvested at 30 h after infection. Wild type, IKK1 Ϫ/Ϫ , IKK2 Ϫ/Ϫ , NEMO Ϫ/Ϫ , and NIK aly/aly fibroblasts were harvested at 36 h after infection. N1 cell clones expressing LMP1 were isolated through limiting dilution.
Preparation of Cell Extracts and Kinase Assay-Kinase assay was performed in parallel with immunoblot analysis following immunopre-cipitation. Cells were lysed in Buffer A (20 mM HEPES, pH 7.8, 0.15 mM EDTA, 0.15 mM EGTA, 10 mM KCl) supplemented with 1 g/ml aprotinin, 1 g/ml leupeptin, 0.57 mM phenylmethylsulfonyl fluoride, 100 M sodium vanadate, and 20 mM ␤-glycerol phosphate and left on ice for 15 min. Nonidet P-40 was added to 1%, and the cell suspension was incubated for 2 min on ice. After centrifugation at 14,000 rpm for 5 min, supernatants were used as cytoplasmic extracts and pellets were used for extraction of nuclear proteins. Gel filtration was performed with Superdex-200 as described previously (29). Immunoprecipitation was performed as described previously (29). After the last washing, one half of the beads were used to perform kinase assays and the remaining beads were used for immunoblot analysis. Kinase reactions were carried out at 30°C for 30 min in reaction mix (20 mM HEPES, pH 7.5, 10 mM MgCl 2 , 50 mM NaCl, 100 M sodium vanadate, 20 mM ␤-glycerol phosphate, 2 mM dithiothreitol, 20 nM ATP, [␥-32 P]dATP, and 1 g of GST-IB␣-(1-72)). Recombinant GST-IB␣ wild type and GST-IB␣
Immunoblot Analysis-Immunoblot analysis was performed under a standard protocol. Immunoreactive bands were visualized by ECL (NEN Life Sciences). When necessary, membranes were incubated in stripping buffer (62.5 mM Tris-HCl, pH 6.8, 100 mM 2-mercaptoethanol, 2% SDS) for 30 min at 50°C with constant agitation, washed, and reprobed with another antibody. The intensity of the p52 and p100 bands was determined by a computerized analysis (Image Gauge version 3.01, Fuji Film).

LMP1
Activates NF-B in NEMO-deficient Cells-We used the 5R cell line to examine whether LMP1 can activate NF-B in a NEMO-independent manner. We demonstrated previously that 5R cells did not express any detectable NEMO protein by Western blot analysis (29), but its mRNA expression was not explored. For this purpose, the entire coding region of the nemo cDNA expressed in Rat-1 cells was cloned by reverse transcription-polymerase chain reaction (RT-PCR), using primers corresponding to regions highly conserved between the human and murine nemo genes (Fig. 1A). Southern blot analysis of RT-PCR products failed to detect nemo mRNA in 5R cells (Fig. 1B). Semiquantitative PCR analysis revealed more than 1000 copies of nemo cDNA in the sample prepared from 100 Rat-1 cells, whereas none was detected in 100 5R cells (Fig. 1C). Aliquots of the PCR-amplified material were further analyzed for NEMO cDNA by nested PCR. Although 10 copies of nemo cDNA were amplified in the control samples, none was amplified from 5R cells (Fig. 1D). These results indicate 5R cells do not express nemo mRNA, and the assay sensitivity is 10 copies of NEMO cDNA in 100 cells. Fig. 1E shows that stimulation of 5R cells with TNF-␣ or LPS does not induce any NF-B-dependent reporter gene activation as previously reported (29), whereas LMP1 expression in 5R cells resulted in a significant activation that was, however, weaker than that observed in Rat-1 cells (see Fig. 5B). We also tested a previously characterized NEMO-deficient pre-B cell line 1.3E2 (29,50) for NF-B activation by LMP1 (Fig. 1F). Transient expression of LMP1 in 1.3E2 and its parental cell line 70Z/3 caused strong NF-B-dependent reporter gene activation. These results indicate that LMP1 can activate NF-B in a NEMOindependent manner.
5R cells stably express the Tax protein of HTLV-I, which might influence NF-B activation by LMP1. Although Tax cannot bind to the IKK complex (data not shown) or activate NF-B in 5R cells, its presence hampered further studies on the functional importance of NEMO in NF-B activation by other stimuli. We thus established through mutagenesis a subline of Rat-1 designated as N1 that does not express Tax and is defective in NF-B activation by TNF-␣ or LPS (Fig. 2, B and C). Sequencing of the nemo cDNA expressed in N1 cells displayed a frameshift mutation that resulted in a large carboxyl-terminal deletion of the NEMO polypeptide beyond amino acid 151 ( Fig. 2A). Western blot analysis with polyclonal antiserum that recognizes an amino-terminal region (amino acids 23-39) of NEMO failed to detect this protein (Fig. 2B). Thus, N1 cells are predicted to express a very short NH 2 -terminal part of NEMO (N1 NEMO) at an undetectable level. Complementation of N1 cells with a tiny amount of wild type NEMO restored NF-B activation in response to TNF-␣ or LPS (Fig. 2C), but expression of cloned N1 NEMO did not (data not shown), indicating that the mutation in nemo represents the cause of defective NF-B signaling in N1 cells. Notably, LMP1 expression elevated NF-B-dependent reporter gene expression in N1 cells (Fig. 2C), which was further enhanced when supplied with wild type NEMO, but not with N1 NEMO (Fig. 2D). A dose-dependent decline of the reporter gene activation in cells complemented with wild type NEMO could be explained by overexpression of this scaffold protein that might exceed the stoichiometric amount for signaling. This enhancement with a small amount of wild type NEMO reflects NEMO-dependent part of NF-B activation by LMP1 and suggests that the limited degree of NF-B activation by transient LMP1 expression in N1 cells results not solely from poor expression of N1 NEMO, but from its functional defect.

LMP1 Causes IKK Activation and Nuclear Translocation of NF-B in Mutant
Cells-To investigate the mechanism of NEMO-independent NF-B activation by LMP1, 5R cells were infected with retrovirus capable of expressing LMP1 and subjected to in vitro kinase assay and EMSA. Retroviral expression of LMP1 in Rat-1 and 5R cells caused similar enhancement of IKK1-associated kinase activity determined in vitro on recombinant IB␣ protein, when the IKK complex was immunoprecipitated with an IKK1-specific antibody at 30 h after infection (Fig. 3A, left upper panel). Phosphorylation was not detected with recombinant IB␣ protein mutated on critical serine residues 32 and 36 that IKK1 or IKK2 phosphorylates. Control immunoblot studies revealed that equivalent amounts of IKK1 and IKK2 were precipitated from Rat-1 and 5R cells, and that NEMO was pulled down only from Rat-1 cells. The IKK1associated kinase activity in 5R cells expressing LMP1 eluted from a gel filtration column (Superdex-200) mainly in the molecular mass range between 440 and 232 kDa (Fig. 3B), where IKK1 was detected by immunoblotting. Thus the previously reported smaller IKK complex in 5R cells lacking NEMO (29) was indeed activated following LMP1 expression. EMSAs detected NF-B DNA binding activity not only in Rat-1, but also FIG . 5. CTAR1 mediates NEMO-independent NF-B activation. A, structure of wild type and mutant LMP1 proteins. CTAR1 is located between amino acid residues 194 and 232. CTAR2 is between 351 and 386. The hollow circles indicate amino acid substitutions. B, CTAR1, but not CTAR2, LMP1 activates NF-B in 5R cells. Rat-1 and 5R cells were co-transfected with 0.25 g of Ig-ConAluciferase and EF1-lacZ together with 1 g of pSG5, pSG5-LMP1, pSG5-LMP1.Y384G, pSG5-LMP1.349⌬, pSG5-LMP1.AAA, or pSG5-LMP1.⌬187-351 by the calcium phosphate precipitation method. Cells were harvested and lysed 36 h after transfection. The luciferase activity was normalized as described in the legend to Fig. 1E. The results shown are representative of independent experiments carried out three times in duplicate. Expression of each LMP1 protein was verified by immunoblot analysis of cell extracts. C, LMP1 requires IB kinase to activate NF-B. Wild type (wt), IKK1 Ϫ/Ϫ , IKK2 Ϫ/Ϫ , and IKK1 Ϫ/Ϫ /IKK2 Ϫ/Ϫ fibroblasts were co-transfected with 0.5 g of Ig-ConAluciferase and EF1-lacZ together with 0.1 g of pSG5 or pSG5-LMP1. Forty-five hours after transfection, cells were either left untreated or stimulated with 15 g/ml LPS or 10 ng/ml TNF for 3 h. The luciferase activity was determined and normalized as described in the legend to Fig.  1E. Experiments were done in duplicate and repeated three times. Results were essentially reproducible. in 5R cells expressing LMP1 (Fig. 3A, right panel). N1 cell clones expressing retrovirally transduced LMP1 showed elevated IKK activity and NF-B DNA binding activity as well (Fig. 3C). The weak induction of DNA binding in the absence of functional NEMO could be the result of the lack of canonical NF-B activation that can enhance expression of p100 and RelB (57,58). In supershift assays with nuclear extracts at 30 h after infection (Fig. 3D), LMP1-induced DNA-NF-B complexes in Rat-1 cells were mostly supershifted by anti-p50 serum. Anti-RelA serum supershifted the slowly migrating (filled arrowhead) rather than the faster migrating (open arrowhead) part of DNA-binding complexes seen in the control lane, whereas anti-RelB serum did the opposite. We failed to detect supershift of rat p52 with the anti-p52 antibody used in this study, which efficiently supershifted human p52 (59) (data not shown). LMP1-induced DNA-NF-B complexes in 5R and N1 cells were generally weak in binding, but showed supershifts similar to those for Rat-1 cells except that RelA was less efficiently supershifted for 5R and N1 cells. Thus, the absence of functional NEMO greatly attenuates, but does not abolish NF-B activation by LMP1.

LMP1-induced NEMO-independent NF-B Activation Involves Aberrant Expression of p52-
The presence of RelB in the DNA binding activity led us to ask whether LMP1 expression causes enhanced processing of p100, because RelB was reported to preferentially form cytoplasmic complexes with p100 (60). Western blot analysis detected remarkably elevated expression of p52 and p100 in Rat-1 cells harvested 30 h after infection with retrovirus capable of expressing LMP1 (Fig. 4A,  upper panel). In 5R cells infected and harvested in the same way or N1 cell clones stably expressing LMP1, the amounts of p52, but not of p100, increased, probably because the lack of NEMO in these cells impeded NF-B-dependent p100 induction. Band intensities relative to that for p100 in Rat-1 cells infected with the control virus are shown in the bottom panel. These results indicated that LMP1-induced p52 generation did not necessarily require NEMO. LMP1-induced generation of p52 was further revealed by transient expression of LMP1 and human p100 in Rat-1 and N1 cells (Fig. 4B), where LMP1 expression enhanced generation of p52 from exogenous p100 regardless of NEMO expression.
CTAR1, but Not CTAR2, Directs NEMO-independent NF-B Activation-Prior studies identified two functional subdomains for NF-B activation, CTAR1 and CTAR2, in the carboxylterminal cytoplasmic domain of LMP1 (14,15). It is known that TNF receptor 1-associated death domain (TRADD) and receptor-interacting protein are recruited to CTAR2 and that TRAFs are involved in both CTAR1 and CTAR2 signaling (12,19,20). However, the mechanism by which signals from these separate subdomains converge on and regulate components of the IKK complex has not been fully explained. To dissect relative contribution of CTAR1 and CTAR2 to the NEMO-dependent and -independent NF-B activation, Rat-1 and 5R cells were subjected to NF-B reporter assays upon transient expression of wild type or mutant forms of LMP1 (Fig. 5). Expression of each construct was examined by Western blotting, although expression of ⌬187-351 lacking the epitope of our LMP1 antibody could not be verified. Mutants AAA and ⌬187-351 transduce NF-B activating signals through CTAR2 but are defective for functional CTAR1. Mutants Y384G and 349⌬ transduce NF-B activating signals through CTAR1 but are defective for functional CTAR2 (Fig. 5A). Consistent with previous reports, either single amino acid substitution or a deletion of CTAR2 (Y384G and 349⌬) impaired NF-B activation in Rat-1 cells, whereas mutations in CTAR1 affected NF-B activation marginally (Fig. 5B, upper graph). In contrast, mutations in CTAR1 (AAA and ⌬187-351) ablated NEMO-independent NF-B activation in 5R cells despite comparable expression of the mutant AAA, whereas this activation was virtually unaffected by CTAR2 mutations (Y384G and 349⌬) (Fig. 5B, bottom  graph). These results clearly indicate that CTAR1, but not CTAR2, directs NEMO-independent NF-B activation and that CTAR2 requires NEMO for NF-B activation.
CTAR1 and CTAR2 Differentially Utilize Components of the IKK Complex for NF-B Activation-The NEMO-independent NF-B activation by LMP1 raises a question of whether it is mediated by certain activity unrelated to IKK1 and/or IKK2. To address this point, we transiently expressed LMP1 in fibroblasts derived from either wild type (wt) or mutant mice genetically deficient for IKK1 (IKK1 Ϫ/Ϫ ), IKK2 (IKK2 Ϫ/Ϫ ), or both IKK1 and IKK2 (IKK1 Ϫ/Ϫ /IKK2 Ϫ/Ϫ ) (Fig. 5C). TNF-␣ and LPS induced significant NF-B reporter gene activation in wild type fibroblasts, whereas these stimuli failed to do so in the absence of IKK1 or IKK2. Unlike TNF-␣ and LPS, LMP1 was able to induce NF-B reporter gene activation in IKK1 Ϫ/Ϫ and IKK2 Ϫ/Ϫ cells, but failed to do so in IKK1 Ϫ/Ϫ /IKK2 Ϫ/Ϫ cells, indicating that LMP1 requires either IKK1 or IKK2, but not necessarily both of them for NF-B activation. We next examined how wild type and mutant LMP1 proteins induce NF-B DNA binding and the processing of p100 in fibroblasts deficient for IKK1 or IKK2 (Fig. 6A). Cells were infected with a high titer retrovirus capable of expressing LMP1 and examined for DNA binding and p52 generation at 36 h after FIG. 7. CTAR1-mediated NF-B activation requires functional NIK activity. Wild type, NIK aly/aly , and NEMO Ϫ/Ϫ fibroblasts were infected with retrovirus capable of expressing control vector, wild type LMP1, Y384G, or AAA, and harvested for preparation of nuclear extract and whole cell extract at 36 h after infection. A, nuclear extracts (5 g) were subjected to EMSA with the KBF1 probe. Whole cell extracts (30 g) were run on a 9% SDS-polyacrylamide gel and subjected to immunoblot analysis with anti-p52 antibody (sc-7386). The same membrane was used for detection of LMP1 and actin. NS, nonspecific band. Intensities of the p100 and p52 bands were determined by computerized image analysis and shown in the bottom panel as relative band intensity to p100 in wild type cells infected with the control virus. B, supershift assays were performed by pre-incubating nuclear extracts prepared at 36 h after infection for 30 min with pre-immune (P.I.), anti-p50 (p50), anti-RelA (RelA), anti-RelB (RelB), or anti-p52 (p52) serum. The arrowheads and asterisks denote the positions of shifted p52 complexes and nonspecific bands, respectively. The bottom panel shows the same supershift results for NIK aly/aly and NEMO Ϫ/Ϫ fibroblasts after a longer exposure.
infection. Expression of LMP1 and equivalent loading of samples were verified by Western blotting. In wild type cells, both Y384G and AAA induced DNA binding activity. The steady state levels of p100 were elevated in wild type cells expressing wild type LMP1 or AAA, whereas p100 was poorly detected in wild type cells expressing Y384G. The p52 generation was found in wild type cells expressing wild type LMP1 or Y384G, but not AAA despite its p100 induction. The elevated amounts of p100 are likely to result from accumulation of p100 because of enhanced production of p100 through activation of the NEMO-dependent canonical IKK complex. The impaired generation of p52 by AAA is consistent with its failure to activate NF-B in the absence of NEMO (Fig. 5B). The NEMO-independent NF-B activation by Y384G (Fig. 5B) suggests that the low amount of p100 in wild type cells expressing Y384G could partly be the result of processing of p100 to p52. The amounts of p100 and p52 relative to that of p100 detected in wild type cells infected with the control virus are shown in the bottom panel of Fig. 6A. In the absence of IKK1, induction of DNA binding by wild type LMP1 or Y384G was reduced, whereas that of AAA was barely affected. The generation of p52 was generally diminished in IKK1 Ϫ/Ϫ cells, whereas p100 apparently accumulated most likely because of the lack of the IKK1-dependent p100 processing. In IKK2-deficient cells, induced DNA binding activities were further weakened with that by AAA being most profoundly affected. The p52 generation by wild type LMP1 or Y384G was also impaired, but not lost. The relatively weak induction of DNA binding and p52 generation in IKK2-deficient cells is reminiscent of those in NEMO-deficient 5R and N1 cells (Fig. 4A).
We further analyzed DNA-binding NF-B components induced by retroviral expression of LMP1 in wild type, IKK1 Ϫ/Ϫ , and IKK2 Ϫ/Ϫ fibroblasts (Fig. 6B). Supershift assays revealed that LMP1 induced DNA-binding complexes composed of p50, RelA, RelB, and p52 in cells from wild type mice. The absence of IKK1 diminished RelB and p52 DNA binding under detectable level and reduced RelA binding. Combined with the poor processing of p100 in IKK1 Ϫ/Ϫ cells (Fig. 6A), this suggests that IKK1 is important for LMP1-induced RelA and RelB activation. In IKK2-deficient cells, RelA and RelB still remained detectable in LMP1-induced DNA-binding complexes, but p52 was undetectable, although a weak p52 generation could be detected by immunoblotting (Fig. 6A). Taken together, our results indicate that IKK1 is important for CTAR1-mediated NF-B activation and that IKK2 plays important roles in CTAR1-and CTAR2-mediated NF-B activation.
In NEMO Ϫ/Ϫ fibroblasts (Fig. 7A), retroviral expression of wild type LMP1 or Y384G induced NF-B DNA binding quite similar to each other, whereas AAA expression failed to induce DNA binding beyond that seen in the control infection. NEMO Ϫ/Ϫ cells showed a relatively high basal level of p52, which was further elevated by expression of wild type LMP1 or Y384G, but not by AAA expression. The bottom panel of Fig. 7A shows the intensities of p100 and p52 bands relative to that for p100 in wild type cells infected with the control virus. These results are consistent with the failure of AAA to activate NF-B in 5R cells (Fig. 5B). In supershift assays (Fig. 7B), p50, RelB, and trace amounts of p52 and RelA were detected in LMP1induced DNA-binding complexes in NEMO Ϫ/Ϫ cells.
NIK Is Involved in CTAR1-mediated NF-B Activation-We finally asked how NIK is involved in NF-B activation by LMP1 mutants, using fibroblasts from mutant mice expressing NIK with the aly mutation (NIK aly/aly ) that was reported to disrupt the NIK-IKK1 interaction (21) (Fig. 7). EMSA revealed that wild type LMP1 and AAA induced similar retarded bands lacking the lower part observed in wild type cells and that Y384G-induced DNA binding was severely impaired, but not completely lost. This NIK mutation did not profoundly affect the p100 induction by wild type LMP1 or AAA, but virtually abolished p52 generation. Supershift assays (Fig. 7B) revealed that LMP1-induced NF-B DNA-binding complexes in NIK aly/aly fibroblasts contained primarily p50 and RelA and a tiny amount of RelB, but not a detectable level of p52. Thus, functional NIK activity is important for the CTAR1-mediated p52 generation and induction of DNA binding activity. DISCUSSION Results of the present study illustrate two signaling cascades for NF-B activation by LMP1 (Fig. 8). One originates in CTAR1, utilizes NIK and IKK1, and induces the processing of p100, whereas the other originates in CTAR2, requires NEMO for induction of DNA binding, and does not involve the processing of p100. The former cascade represents the non-canonical pathway of NF-B activation recently described for LT-␤ receptor, BAFF receptor, and CD40 (39 -44). This pathway is characterized by involvement of NIK, IKK1, but not NEMO, as well as by the processing of p100 and nuclear translocation of RelB. The CTAR2 cascade is compatible with the canonical pathway of NF-B activation in that it requires NEMO and does not involve the processing of p100.
A question may arise as to how biologically important this NEMO-independent NF-B activation by LMP1 is, when NEMO is available. The absence of NEMO indeed reduced CTAR2 triggers the canonical NF-B activation pathway that requires NEMO, resulting in enhanced expression of NFKB2 p100. CTAR1 stimulates the non-canonical pathway that involves NIK and IKK1, promoting the processing of p100 to p52 and nuclear translocation of RelBcontaining dimers. LMP1-induced NF-B DNA binding (Fig. 3A), but this could partly result from impaired induction of p100 (Figs. 4A and 7A) and possibly of RelB as a result of lack of the NEMO-dependent canonical NF-B activation. Moreover, IKK1-associated kinase activities induced by LMP1 in Rat-1 and 5R were similar, at least 30 h after retrovirus infection. Thus the reduction of DNA binding activities in NEMO-defective cell lines may not represent the actual contribution of the NEMO-independent pathway to LMP1-induced NF-B activation in the presence of NEMO. On the other hand, in cells carrying the aly mutation of NIK, where the processing of p100 through activation of CD40, BAFF receptor, or LT-␤ receptor were previously shown to be abolished, the DNA-binding complexes induced by LMP1 were markedly reduced and seemed indistinguishable from those by the AAA mutant (Fig. 7A, middle panel). This mutant, in contrast, does not work in NEMO Ϫ/Ϫ cells (Fig. 7A, right panel), indicating that the AAA actions represent the canonical part of NF-B activation by LMP1. In wild type murine fibroblasts, induction of DNA binding by AAA was ϳ10 -20% of that by wild type LMP1 at 30 h after retrovirus infection. Collectively, these observations suggest that the non-canonical pathway of NF-B activation by LMP1 is not a minor one, at least in our experimental conditions. Indeed, induction of DNA binding by Y384G in wild type fibroblasts was robust. However, Y384G induction of DNA binding in NIK aly/aly or IKK1 Ϫ/Ϫ fibroblasts was not completely lost, suggesting that this mutant retains a weak ability to bypass the non-canonical pathway, and therefore is not really appropriate for evaluating the solo contribution of the non-canonical pathway. Our reporter gene experiments revealedthatAAAand⌬187-351activateaclassicalBenhancerdependent transcription more efficiently than Y384G and 349⌬, but it remains to be clarified whether this reporter gene responds in a similar efficiency to alternative NF-B components such as those containing p52 or RelB.
NIK Is Involved in the CTAR1 Signaling to IKK1-Prior studies revealed that TRAF2 is involved in CTAR1-and CTAR2mediated NF-B activation, where CTAR1 interacts with TRAF2 directly, but CTAR2 does through TRADD (61). TRAF2 is known to interact with NIK. In addition, a dominant negative form of NIK was reported to suppress NF-B activation by LMP1 (20), but a dominant negative form of aly NIK did not suppress LMP1induced NF-B activation (21). Thus, signals from CTAR1 and CTAR2 were thought to converge on NIK and/or Tpl-2/Cot (22), which were implicated in the downstream signaling cascade of LMP1 to the IKK complex, but the contribution of these kinases to LMP1-induced NF-B activation remained to be examined in knockout cells. Our present study, using a series of genetically manipulated cells, has shown for the first time different signaling modules involved in IKK activation by CTAR1 and CTAR2. The CTAR2 signaling to NF-B is not significantly affected by the aly type mutation of NIK and appears to share a similar signaling mechanism with TNF receptor 1 in terms of NEMO requirement. In contrast, the CTAR1 signaling to NF-B is severely impaired in cells expressing the aly type NIK, but does not necessarily require NEMO. A recent report showed that the TRAF2/3/1binding site of CD40 played an important role in its NIK activitydependent p100 processing (43). This region contains a PXQXT consensus motif for TRAF binding that can also be found in CTAR1 (62), suggesting that the TRAF2/3/1-binding site of CD40 and CTAR1 share a similar signaling mechanism in the noncanonical pathway.
Distinct Contribution of the IKK Components to CTAR1-and CTAR2-mediated Signaling Pathways-Our results in fibroblasts derived from aly-NIK or NEMO knockout mice clearly discriminate between two signaling pathways from CTAR1 and CTAR2, whereas studies in IKK1-or IKK2-deficient cells sug-gest that the canonical and non-canonical NF-B pathways triggered by LMP1 cross on the IKK complex and influence each other for IB regulation and nuclear translocation of NF-B components (Fig. 6). Regarding IKK2, not only AAAinduced DNA binding, but also wild type-and Y384G-induced p52 generation and DNA binding, were markedly reduced in cells lacking IKK2, indicating that IKK2 plays a pivotal role in regulating IB proteins and nuclear NF-B components. It should be noted that the absence of NEMO completely abolished AAA-mediated enhancement of DNA binding activity (Fig. 7A), whereas there remained residual AAA-induced DNA binding activity in IKK2-deficient cells (Fig. 6A), suggesting that NEMO and IKK1 partially mediated CTAR2 signaling in IKK2-deficient cells. On the other hand, the absence of IKK1 resulted in an obvious difference between these signaling pathways. The remarkable reduction of Y384G-induced DNA binding activity in IKK1-deficient cells (Fig. 6A) indicates an important role for IKK1 in CTAR1 signaling, whereas IKK1 appears not to be required for AAA induction of DNA binding activity. Unlike LMP1, Tax has a unique mechanism of NF-B activation, in which it physically associates with and activates the IKK complex as well as directly binds to p100, thus recruiting p100 to the activated IKK complex to induce the phosphorylation-dependent processing of p100. Because NEMO mediates the Tax-IKK interaction, this process absolutely requires NEMO, but does not depend on the functional NIK activity (45). Thus, induction of the p100 processing by a membrane protein LMP1 differs from that by Tax of HTLV-I.
Regulation of NF-B and IB Components by LMP1-We have demonstrated that LMP1 induces the generation of p52 and nuclear translocation of RelB in the absence of NEMO (Figs. 3D and 4A) and that CTAR1, but not CTAR2, is responsible for the NEMO-independent NF-B activation (Figs. 5B and 7A). Besides, IKK1-associated phosphorylating activity was demonstrated in NEMO-deficient 5R cells expressing LMP1, which existed in smaller IKK complexes migrating in a gel filtration column at ϳ300 kDa, as shown previously (29). Thus, it is reasonable to assume that IKK1 activated by CTAR1 could phosphorylate p100 and trigger the processing. On the other hand, expression of CTAR2 LMP1 caused an increase in the steady state levels of p100 in wild type, NIK aly/aly , and IKK1-deficient cells, but this was not obvious in IKK2-deficient cells. Similar lack of p100 induction was observed in NEMOdeficient fibroblasts (Figs. 4A and 7A), suggesting a role for the canonical IKK complex in LMP1 induction of p100 expression.
LMP1 differs from CD40 in that it binds TRADD through the CTAR2 domain as well as in that LMP1 permanently activates NF-B, whereas CD40-mediated activation does not persist for days. Moreover, LMP1 is thought to form a higher order clustering than a trimer without ligand binding, and thus can more efficiently activate NF-B compared with CD40 (63). Although a model of temporal regulation of the two activation pathways was proposed for CD40-or LT-␤ receptor-mediated NF-B activation (41,43), the two pathways stimulated by CTAR1 and CTAR2 are likely to co-operate simultaneously for persistent NF-B activation in cells stably expressing LMP1. Indeed, both CTAR1 and CTAR2 domains were reported to be required for proliferation of primary B cells or transformation of rodent fibroblasts (64 -67).
Constitutive NF-B Activation and Human Malignancies-Shortly after stimulation of the canonical pathway, production of p100 is up-regulated, which usually results in cytoplasmic sequestration of active NF-B components. Because p100 is not sensitive to the canonical IKK activity, constitutive NF-B activity could be based on disruption of the powerful IB activity of p100. This idea is supported by aberrant p52 expression in a wide variety of human malignancies with constitutive NF-B activity, including breast cancer and lymphoid malignancies with chromosomal re-arrangements in the nfkb2 locus (68 -70). Probably, the IKK1-dependent processing of p100 contributes not only to the disruption of the powerful IB function of p100, but also to the p52/RelB-mediated induction of a set of genes that play important roles in cell proliferation (71). In fact, IKK1 is required for induction of the cyclin D1 gene transcription in mammary epithelial cells during pregnancy (72). In addition, recently reported common sites of retroviral insertional mutagenesis that induces a high incidence of hematopoietic malignancies include the nfkb2 locus (73). However, the fact that a homozygous deletion of the carboxyl-terminal ankyrin repeats of NFKB2 led to increased nuclear B-binding activity containing p52, resulting in gastric hyperplasia and lymphoproliferative disorders, but not any malignancy suggests that constitutive processing of p100 is not sufficient for cancer development. Further studies will be needed to clarify the role of p52-RelB complex in the development of human malignancies. Characterization of the IKK activity specifically phosphorylating p100 may also facilitate understanding the molecular mechanisms of persistent NF-B activation in a wide variety of human malignancies.