Interaction of Epstein-Barr Virus Latent Membrane Protein 1 with SCF HOS/ (cid:1) -TrCP E3 Ubiquitin Ligase Regulates Extent of NF- (cid:2) B Activation*

The Epstein-Barr virus latent membrane protein 1 (LMP1) is pivotal in the transforming activity of this virus. We found that the common LMP1-95-8 variant interacts with Homologue of Slimb (HOS), a receptor for the SCF HOS/ (cid:1) TrCP ubiquitin-protein isopeptide ligase (E3) via one canonical and one cryptic HOS recognition site. These sites are mutated or deleted in the tumor-derived LMP1-Cao variant, which did not bind to HOS. Mutations within these sites on LMP1-95-8 abrogated HOS binding and increased transforming activity of LMP1. HOS did not regulate stability of LMP1-95-8 unless it was mutated to bear additional lysine residues near the cryptic motif. LMP1 proteins that could not bind to HOS exhibited an increased ability to induce I (cid:2) B degradation and NF- (cid:2) B-mediated transcription without further increase in activation of I (cid:2) B kinases. Expression of LMP1-95-8 reduced the levels of endogenous HOS available to interact with phosphorylated I (cid:2) B (cid:3) . Degradation of I (cid:2) B (cid:3) and dose dependence of NF- (cid:2) B activation by LMP1-95-8 were promoted by co-expression of HOS. Our data suggest that LMP1-95-8 is a pseudo-substrate of SCF HOS/ (cid:1) TrCP E3 ubiquitin ligase and that interaction between LMP1 and HOS restricts the extent of LMP1-induced NF- (cid:2) B signaling. We discuss the potential role of this mechanism in transforming and cytostatic effects of LMP1 variants in cells and Epstein-Barr virus-associ-ated tumors.

The latent membrane protein 1 (LMP1) 1 is a transmembrane signaling protein encoded by the BNLF1 gene of Epstein-Barr virus (EBV). LMP1 plays a pivotal role in immortalization and proliferation of human B-lymphocytes infected with EBV, which is associated with a wide spectrum of human malignancies (reviewed in Ref. 1). LMP1 resembles a viral oncogene by virtue of its capability to transform rodent cells in vitro and to induce tumorigenesis in transgenic mice (2)(3)(4). LMP1 elicits its effects by constitutively engaging multiple signaling pathways, including those that activate NF-B, jun/ AP-1, ATF2, signal transducers and activators of transcription, and Sma and Mad-related protein transcription factors, which, in turn, mediate pleiotropic events in cells infected with EBV or transfected with LMP1 vector (5)(6)(7)(8). Although the specific contributions of many signaling cascades to the oncogenic properties of LMP1 remain to be elucidated, activation of the NF-B pathway is linked to LMP1-induced immortalization of human primary B-lymphocytes (6,9,10) and tumorigenic transformation of cultured rodent cells (11,12).
NF-B, a dimeric transcription factor, activates transcription of a large number of target genes that play key roles in cell growth and death, as well as in immune and inflammatory responses. NF-B activity is frequently increased in human tumors; it is essential for growth and survival of tumor cells. Cells tightly regulate activation of NF-B by expression of a family of inhibitory proteins (IB). In the absence of stimuli, IB binds tightly to NF-B and masks the domains responsible for NF-B nuclear localization and DNA binding (13,14).
An increased capacity of some natural LMP1 isolates from tumors to induce NF-B activity is thought to be linked to their higher efficacy in transforming rodent cells in vitro compared with a canonical B95-8 variant of LMP1 (LMP1-95-8) (24,25). These highly transforming LMP1 proteins (including LMP1-Cao, 1510, and C15) bear numerous point mutations and a 30-bp deletion in the C-terminal domain (26 -28). Whereas the 30-bp deletion increases transforming activity of LMP1-95-8 (29) and is frequently found in EBV-associated tumors (30), the mechanisms underlying either increased tumorigenicity or augmented induction of NF-B by these LMP1 variants remain unknown.
Surprisingly, in transfected cells, when LMP1-95-8 is expressed at very high levels, it inhibits NF-B activity (31,32) and causes cytostatic and/or cytotoxic effects, indicating a role of the tight regulation of LMP1-95-8 levels in the maintenance of EBV latency (33)(34)(35). This inhibition is not observed upon expressing LMP1-Cao (33,36), despite the fact that LMP1-Cao protein is more stable than LMP1-95-8 and accumulates to a greater extent in cells (37).
In human cells, LMP1-95-8 protein undergoes rapid degradation, which depends on the ubiquitin-proteasome pathway (38). However, E3 ubiquitin ligase(s) responsible for ubiquitination of LMP1 have not been identified as yet. We have noted that the C-terminal domain of LMP1-95-8 contains a potential ␤-TrCP recognition motif (DSGHES) and that this motif is mutated in LMP1-Cao. In this study, we investigated a potential role of SCF HOS/␤TrCP E3 ubiquitin ligases in degradation of LMP1. We found that LMP1-95-8, but not LMP1-Cao, interacted with ␤-TrCP2/HOS. Although SCF HOS ligase did not regulate LMP1 stability, the association between LMP1-95-8 and the SCF HOS complex affected the availability of HOS to interact with phosphorylated IB as well as IB degradation and NF-B activation. These data provide an insight into mechanisms underlying differences between LMP1-95-8 and highly tumorigenic LMP1 variants in their ability to activate NF-B signaling.
Cells, Transfection, and Cell Lysate Preparation-293T cells were grown in Dulbecco's modified Eagle's medium in the presence of 10% fetal bovine serum and antibiotics at 37°C and 5% CO 2 . Rat1a cells were maintained in Dulbecco's modified Eagle's medium in the presence of 10% calf serum and antibiotics under similar conditions. Transfections were performed according to the calcium phosphate procedure or with the aid of LipofectAMINE Plus (Invitrogen). Cells were harvested and lysed with RIPA buffer (0.5% sodium deoxycholate, 1% Triton X-100, and 0.1% SDS in phosphate-buffered saline) supplemented with inhibitors of phosphatases (20 mM NaF and 1 mM Na 3 VO 4 ) and a "mixture" of protease inhibitors (Sigma). Lysates were cleared by centrifugation at 16,000 ϫ g for 30 min at 4°C, and protein concentrations were measured using Bradford reagent (Pierce). pRSV-␤-gal expression vector was added to the transfection mixture to normalize the transfection efficiency on the basis of ␤-galactosidase activity (measured with a Promega kit). Expression of recombinant SCF HOS complex components (HOS-HA, Cullin1, Skp1, and Roc1) in 293T cells and their purification were carried out as described previously (45).
Immunotechniques-Monoclonal antibodies against HA tag (Roche Applied Science), LMP1 (CS1-4, Dako), IKK␥ (Pharmingen), ␤-catenin (Transduction Laboratories), IKK␣ and IB␣ (Santa Cruz Biotechnology), and ␣-tubulin (Sigma) as well as Cullin1 polyclonal antibody (NeoMarkers) were purchased. Polyclonal antibody against HOS (HOS-N) was described earlier (46). Pulse-chase analysis was performed as described previously (40). Briefly, 293T cells grown on 100-mm dishes and transfected with indicated plasmids were starved in medium lacking methionine and cysteine, metabolically labeled with a 35 S-labeled methionine/cysteine mixture (PerkinElmer Life Sciences), and chased with medium containing 2 mM unlabeled methionine and cysteine for various time points. Cells were harvested, and LMP1 proteins were immunopurified by CS1-4 antibody, separated by SDS-PAGE, and analyzed by autoradiography. Results were quantified using a Storm 860 Imager (Amersham Biosciences). An IKK immunokinase assay was carried out with IKK␥ antibody as described elsewhere (47). Immunoprecipitation and immunoblotting procedures were described earlier. Data were analyzed using Scion Image Software (version Beta 4.0.2).
Luciferase Assays-Luciferase assays in the cells plated in 24-well plates and co-transfected with either Bor jun2-driven luciferase reporter constructs and pRSV-␤-gal plasmid were performed using the appropriate kit (Promega) as described earlier (40). Luciferase activity was normalized with respect to ␤-galactosidase activity.
Foci Formation Assay-Rat1a cells were transfected with LMP1 expressing constructs (0.5 g) and pBABE-puro (3.5 g) and grown for 14 days in the presence of puromycin (2 g/ml, Sigma). The appearance of foci of transformed cells characterized by lack of contact inhibition was observed 14 days after transfection, and the number of foci was scored in a double blind manner.
Measurement of Substrate-free HOS Levels-IB␣-derived phosphorylated peptide (KKERLLDDRHDpSGLDpSMKDE, where pS is phosphoserine) and its non-phosphorylated counterpart were synthesized by the Protein Chemistry Core Laboratory of Baylor College of Medicine (Houston, TX). Peptides were covalently coupled with beads using an AminoLink kit (Pierce), and the beads were washed and incubated with 1 mg of lysates from 293T cells transfected with LMP1 constructs for 1 h at 2°C. After extensive washes with ice-cold phosphate-buffered saline supplemented with Nonidet P-40 (0.1%), the proteins were eluted with Laemmli sample buffer, separated by SDS-PAGE, and analyzed by immunoblotting with HOS-N antibody.

LMP1-95-8 but Not LMP1-Cao Interacts with HOS-
We sought to investigate whether LMP1 is a target for SCF ␤-TrCP/HOS E3 ubiquitin ligases. Both ␤-TrCP1 and HOS recognize the DpSGXXpS motif (22) and play a redundant role in ubiquitination of IB and ␤-catenin (48). Because a similar motif is found within the cytoplasmic C-terminal domain of LMP1-95-9 (Fig. 1C), we chose to focus our studies on interaction of LMP1 with HOS, which, unlike ␤-TrCP1, is primarily localized in the cytoplasm (49,50).
Interaction of HOS with LMP1-95-8 Requires Both Canonical and Cryptic HOS Recognition Sites-We next substituted Gly-212 in LMP1-95-8 for serine. Intriguingly, this mutation alone did not affect the ability of the LMP1 protein to interact with HOS ( Fig. 2A, lanes 3 and 4), indicating that LMP1-95-8 may contain additional sites for HOS binding that are altered in LMP1-Cao. Although the DpSGXXpS motif is known to mediate HOS binding in IB and ␤-catenin, it has been shown that motifs with more than three amino acids between the phosphoserines may also be recognized by ␤-TrCP proteins in precursors of NF-B transcription factors (51) and ATF4 transcription factors (50). Another DSG sequence containing Ser-350 residue within LMP1-95-8 is absent in LMP1-Cao as a part of 30-bp deletion frequently found in LMP1 isolates from tumors (30). Moreover, distally located serine 366 is substituted for threonine in LMP1-Cao protein (Fig. 2B), and mutation of Ser-366 is one of the most frequently observed among natural LMP1 isolates (see "Discussion").
We further assessed a potential role of Ser-350 and Ser-366 in HOS binding. Neither S350A nor S366T individual mutants lacked the ability to interact with HOS ( Fig. 2A, lanes 5 and 6). However, double mutations of LMP1-95-8 such as G212S/ S350A or G212S/S366T as well as the triple amino acid substitution (Triple, G212S/S350A/S366T) dramatically inhibited the ability of LMP1 to interact with HOS ( Fig. 2A, lanes 7-9). These results suggest the existence of a cryptic HOS recognition site that requires the integrity of serines 350 (within the DSG sequence) and 366 in LMP1-95-8, which are deleted or mutated in LMP1-Cao. In addition, these data indicate that both canonical and cryptic HOS recognition sites are required for binding of HOS to LMP1-95-8 and provide a plausible explanation for the inability of LMP1-Cao, in which both sites are altered, to interact with HOS.
HOS Is Not Required for LMP1-95-8 Degradation-As LMP1-Cao protein, which cannot bind HOS (Fig. 1B), is more stable than LMP1-95-8 (37), we were prompted to investigate the role of HOS in the regulation of LMP1 protein stability. To this end, we compared the rate of proteolysis of LMP1-95-8 with its Triple mutant, which exhibited no binding to HOS ( Fig.  2A). Contrary to our expectations, pulse-chase analysis of LMP1 proteins showed that the half-life of the Triple mutant was no longer than that of LMP1-95-8 (Fig. 3A). This result indicates that abrogation of HOS binding is not sufficient for stabilization of LMP1 proteins. Furthermore, LMP1-95-8 protein was not stabilized by co-expression of the dominant negative mutant HOS ⌬F (Fig. 3A). This mutant, which was shown to abrogate degradation of known HOS/␤-TrCP substrates (40,41), was evidently expressed under the conditions of our experiments, resulting in accumulation of ␤-catenin (Fig. 3B). In some experiments we observed a modest stabilization of Triple-LMP1 protein in cells overexpressing HOS; the significance of these observations remains unclear.
In addition, although SCF HOS could be successfully bound to LMP1-95-8 (Fig. 1A), we failed to ubiquitinate this protein in vitro under conditions that enabled efficient ubiquitination of IB␣. 2 Taken together, these results suggest that despite the ability of LMP1-95-8 to interact with SCF HOS E3 ubiquitin ligase, LMP1-95-8 is not a genuine substrate for this E3 ligase. These findings are in line with conclusions of another group that the C-terminal domain of LMP1-Cao does not contribute to its increased stability (37).
LMP1-95-8 Is a Pseudo-substrate of SCF HOS Because of the Absence of Ubiquitin-Acceptor Lysines-The recently solved SCF ␤-TrCP1 structure predicts that efficiency of ubiquitination by this ligase depends on the location of the ubiquitin-accepting lysine(s) relative to the destruction motif (52). LMP1-95-8 contains a single lysine residue (Lys-330) that is not essential for its ubiquitination (38). Lack of convenient ubiquitin conjugation site(s) in proximity to HOS-binding sequences similar to Lys-21/22 within IB␣ (23) or Lys-19 within ␤-catenin (52) may explain why LMP1-95-8 is not a bona fide substrate for SCF HOS . To investigate this possibility, we substituted two amino acids positioned at a similar spacing from the cryptic 2 W. Tang, V. Spiegelman, and S. Fuchs, unpublished data. HOS recognition site within LMP-95-8 with two lysine residues and compared the rates of degradation of LMP-95-8 and of this 95-8-KK mutant (L339K/M340K). Pulse-chase analysis showed that introduction of the lysines noticeably decreased the halflife of the mutant LMP1 protein (Fig. 4, A and D). Furthermore, inhibition of HOS activities by co-expressing HOS ⌬F protein stabilized 95-8-KK (Fig. 4, B and D). This result indicates that introduction of lysines in proximity to the HOS cryptic recognition site creates a protein whose degradation is regulated by SCF HOS .
To confirm the role of HOS in controlling stability of 95-8-KK, we introduced the L339K/M340K substitution mutation into the background of the Triple mutant, which cannot bind HOS (Fig. 2A). The resulting Triple-KK mutant protein exhibited a half-life similar to that of LMP1-95-8, and the stability of Triple-KK was not increased by co-expression of HOS ⌬F (Fig. 4,  B and D).
To corroborate our hypothesis that an increase in the rate of ubiquitination of 95-8-KK results in destabilization of this protein, we carried out in vivo ubiquitination experiments. Cotransfection of LMP1-95-8 with hexahistidine-tagged ubiquitin in 293T cells followed by purification of ubiquitinated proteins under stringent denaturing conditions (44) revealed that LMP1 is covalently conjugated with ubiquitin in vivo under these conditions (Fig. 4C). Compared with the LMP1-95-8 protein, the 95-8-KK mutant exhibited a noticeably higher extent of ubiquitination (Fig. 4C, lane 5 versus lane 3), and this increase was inhibited by co-expression of HOS ⌬N (lane 6 versus lane 5). Mutation of HOS recognition sites in the Triple-KK protein abrogated the effect of introducing lysines (Fig. 4C, lane 7). In addition, the dominant negative HOS ⌬F mutant did not inhibit ubiquitination of either LMP1-95-8 or Triple-KK proteins. These results indicate that introduction of lysines into an LMP1 protein capable of binding to HOS converts the protein into a substrate for HOS-dependent ubiquitination. In all, our evidence suggests that LMP1-95-8 is a pseudo-substrate of the SCF HOS E3 ubiquitin ligase. Although this ligase is capable of binding to LMP1-95-8, it does not undergo ubiquitination, most likely because of the lack of conveniently located lysine residues that would serve as ubiquitin acceptor sites.
The Ability of LMP1-95-8 to Interact with HOS Affects Transforming and Signaling Activities of LMP1-We compared the ability of LMP1-95-8 and LMP1-Cao to induce in vitro transformation in Rat1a fibroblasts. Consistent with data published previously (25,29), transforming activity of LMP1-Cao was significantly higher than that of LMP1-95-8 (Fig. 5A). As LMP1mediated transformation is known to depend on gene dosage (2), we initially attributed higher efficiency of LMP1-Cao solely to its greater stability (37) and abundance (Fig. 1B). However, the Triple mutant of LMP1-95-8, which was expressed at levels comparable with those of LMP1-95-8 ( Fig. 2A), was also significantly more active in transforming Rat1a cells than LMP1-95-8 (Fig.  5A). This finding indicates that sequences within the canonical and cryptic degradation motifs of LMP1-95-8, which determine its ability to interact with HOS, may play a role in the outcome of LMP1-mediated transformation.
LMP1 expression initiates a number of signaling pathways, including induction of stress-activated protein kinases (reviewed in Refs. 7 and 8), that phosphorylate c-Jun and ATF2 transcription factors and activate the jun2 regulatory element within the promoter of the c-jun gene (43). We used jun2luciferase reporter to compare the ability of LMP1-95-8, Triple mutant, and LMP1-Cao to induce this signal transduction pathway. All three LMP1 constructs induced comparable levels of jun2-driven luciferase in a dose-dependent manner (Fig. 5B). This result suggest that the ability of LMP1 to induce stressactivated protein kinase signaling is unlikely to confer variability in transforming activities on LMP1-95-8 and its Triple mutant or LMP1-Cao.
Activation of NF-B-dependent transcription is a hallmark of LMP1-mediated signaling that plays a key role in LMP1mediated cell transformation (6,9,10). Whereas low doses of the LMP1-95-8 construct readily induced the activity of Bdriven luciferase reporter, no further activation was achieved by higher doses (Fig. 5C). Moreover, although increasing concentrations of LMP1-95-8 produced less NF-B activation compared with low doses, LMP1-Cao activated NF-B in a dosedependent manner (Fig. 5C). These results are consistent with earlier observations of Johnson et al. (36) and more recent evidence from other groups (31,32) confirming the biphasic curve of NF-B activation mediated by LMP1-95-8. Interestingly, the Triple mutant activated NF-B in a dose-dependent manner similar to LMP1-Cao (Fig. 5C). This result indicates that mutations in HOS recognition sites of LMP1-95-8 abrogate its ability to inhibit NF-B activation at high doses. Taken together, these data imply that binding of HOS to LMP1 may play a role in regulating the extent of LMP1-induced activation of NF-B signaling without affecting the stress-activated protein kinase pathway.
LMP1 Proteins Lacking HOS Binding Accelerate IB␣ Degradation without Additional Induction of IB Kinases-Activation of NF-B by LMP1 relies on induction of the IB kinases (17), which phosphorylate IB and mark it for ubiquitination and proteasomal degradation (13). To identify the mechanisms underlying the differences in NF-B activation by LMP1-95-8 and its Triple mutant or LMP1-Cao, we performed immuno-kinase assays to investigate the ability of high doses of LMP1 proteins to induce IKK activity. Expression of all three LMP1 proteins activated the endogenous IKK complex to a similar extent (Fig. 6A). This result indicates that an advanced ability of Triple or LMP1-Cao to activate NF-B (compared with LMP1-95-8, Fig. 5C) can hardly be attributed to differences in induction of IKK activity.
We next examined degradation of endogenous IB␣ in cells transfected with high doses of LMP1 constructs and treated with cycloheximide to prevent de novo protein synthesis. In the absence of LMP1 proteins, hardly any IB␣ degradation was observed within 20 min of cycloheximide treatment, whereas noticeable degradation was detected in cells expressing LMP1-95-8. At the same time, more robust IB␣ degradation was observed in cells expressing either Triple or LMP1-Cao proteins (Fig. 6B). More efficient degradation of IB in cells expressing LMP1 proteins that cannot bind to HOS (Triple and LMP1-Cao) may explain why high doses of these proteins are better activators of NF-B than LMP1-95-8 (Fig. 5C). Because all tested LMP1 proteins activated IKK to a similar extent (Fig.  6A), a limited ability of high doses of LMP1-95-8 to induce IB␣ degradation (Fig. 6B) and NF-B activation (Fig. 5C) must be attributed to a mechanism that acts downstream of IKK activation and IB phosphorylation. (41,46,53,54). The function of ␤-TrCP proteins can easily be disrupted by forced expression of viral Vpu protein, which acts as a pseudo-substrate and squelches ␤-TrCP to prevent its binding to phosphorylated IB (55,56). We sought to investigate whether LMP1 is capable of squelching HOS in a manner similar to that of Vpu. We assessed the levels of cellular HOS, which is not bound to its substrates and hence is capable of associating with immobilized IB␣-derived peptide phosphorylated at Ser-32/36. We found that whereas ϳ10% of total HOS levels from the cells transfected with empty vector were pulled down with this peptide (Fig. 7A, panel I, lane 1), no HOS binding was detected with peptide in which Ser-32/36 was not phosphorylated (Mock*). Expression of LMP1-Cao or Triple moderately reduced levels of substratefree cellular HOS (Fig. 7A, panel I, lanes 3 and 4 versus lane 1). This reduction was not unexpected because both Triple and LMP1-Cao efficiently induced IKK (Fig 6A) and therefore could be predicted to increase intracellular levels of HOS substrate (phosphorylated IB). Under the same conditions, expression of LMP1-95-8 strikingly decreased the levels of substrate-free HOS below the limits of detection of our assay (Fig. 7A, panel  I, lane 2 versus lane 1). As IKK activity was induced by all tested LMP1 proteins to a comparable extent (Fig. 6A) and neither of the tested LMP1 proteins affected the overall levels of HOS (Fig. 7A, panel II), our results suggest that expression of LMP1-95-8 decreases substrate-free HOS levels in cells via HOS-LMP1-95-8 association.

LMP1-95-8 Is Capable of Squelching Cellular HOS and Restricting NF-B Activation-We previously found that an overall abundance of ␤-TrCP/HOS proteins in cells is important for regulating activity of SCF ␤-TrCP/HOS E3 ubiquitin ligases and NF-B activation
HOS is required for phosphorylation-dependent IB degradation and NF-B activities (40,57). It is conceivable that interaction of exogenously overexpressed LMP1-95-8 with endogenous HOS may reduce availability of HOS for IB␣, resulting in delayed IB␣ degradation (Fig. 6B) and limited NF-B activation (Fig. 5C). To examine this possibility, we attempted to reverse the effects of high doses of LMP1-95-8 on IB degradation and NF-B activation by co-expressing HOS. Expression of HOS alone did not stimulate either IB␣ degradation (Fig. 7B) or NF-B activation (data not shown). However, co-expression of HOS with LMP1-95-8 was sufficient to FIG. 5. Transforming and signaling activity of LMP1 proteins. A, rate of transformed foci formation in Rat1a cells transfected with indicated LMP1 constructs and puromycin-expressing vector (pBABEpuro) and grown in the presence of puromycin (2 g/ml) for 2 weeks. The graph depicts the average % of LMP1-95-8 foci-forming efficiency (100%) calculated from three independent experiments (each in triplicate). In t test: *, p Ͻ 0.05 versus LMP1-95-8. B, transcriptional activation of the jun2 element measured by activity of the luciferase reporter in 293T cells co-transfected with various amounts of the LMP1 constructs (no LMP1, black bars; 20 ng, white bars; 40 ng, light gray bars; 80 ng, dark gray bars). The graph depicts the average fold increase of luciferase activity (normalized with respect to ␤-galactosidase activity) calculated from three independent experiments (each in triplicate). C, transcriptional activation of NF-B measured by activity of the corresponding luciferase reporter in 293T cells co-transfected with various amounts of the LMP1 constructs and measured and calculated as described in B. *, p Ͻ 0.01 versus 20 ng of LMP1-95-8; **, p Ͻ 0.001 versus 80 ng of LMP1-95-8. accelerate the rate of IB␣ degradation by LMP1-95-8 (Fig. 7B) and to restore the dose dependence of LMP1-95-8-mediated NF-B activation (Fig. 7C). In all, these data suggest that LMP1-95-8 (but not Triple or LMP1-Cao) is capable of squelching endogenous cellular HOS, thus preventing its interaction with phosphorylated IB␣ and, in turn, inhibiting the following IB␣ degradation and NF-B activation.

DISCUSSION
In this study, we found that LMP1-95-8 is a pseudo-substrate for SCF HOS E3 ubiquitin ligase and that LMP1-95-8 is capable of squelching HOS and restricting the extent of NF-B signaling. LMP1-95-8 binds to HOS in a manner that depends on the integrity of both canonical and cryptic degradation motifs, which are located in the C-terminal domain of LMP1-95-8 ( Figs. 1 and 2). This binding does not affect protein stability of LMP1-95-8, most likely because LMP1-95-8 lacks ubiquitinacceptor lysines located within reach of the SCF HOS E3 ligase catalytic center (Figs. 3 and 4). When expressed at high doses in cells, LMP1-95-8 is capable of squelching HOS and preventing its interaction with phosphorylated IB (Fig. 7). This squelching is likely to contribute to a limited ability of LMP1-95-8 to induce degradation of IB (Fig. 6) and activation of NF-B (Fig. 5) in comparison with LMP1-Cao protein, which does not bind to HOS.
Technical problems with the quality of a few available antibody reagents against ␤-TrCP1 precluded us from conclusively assessing whether LMP1 is also capable of squelching this protein. 2 Nevertheless, considering the important role of ␤-TrCP1 in ubiquitination of IB (48,58) together with our data showing that LMP1-95-8 is capable of inhibiting IB degradation (Fig. 6B), it is likely that the squelching effect of LMP1-95-8 on HOS can be likely extrapolated to HOS-related ␤-TrCP1 protein. In fact, a similar mechanism of squelching ␤-TrCP1 was reported for the DSGXXS motif-containing Vpu protein of the human immunodeficiency virus, type 1 (18). During the course of the human immunodeficiency virus infec-FIG. 6. Induction of IB kinase activity and IB␣ degradation by LMP1 proteins. A, phosphorylation of GST-IB␣ recombinant protein by endogenous IKK complexes immunoprecipitated with IKK␥ antibody from 500 g of lysates from the 293T cells transfected with 2 g of the indicated LMP1. Autoradiograph of 32 P-labeled GST-IB␣ is shown (upper panel). Aliquots of immunoprecipitated material were also analyzed by immunoblotting with IKK␣ antibody (lower panel). Control for LMP1 protein expression is shown in Fig. 7A. B, degradation of endogenous IB␣ after treatment of 293T cells (transfected with 2 g of indicated LMP1 constructs or empty vector (Mock) and 0.4 g of ␤-galactosidase expressing vector) with cycloheximide (50 g/ml for 20 min) assessed by immunoblotting with IB␣ antibody (upper panel). Loading was normalized with respect to ␤-galactosidase activity and verified by immunoblotting with ␣-tubulin antibody (lower panel). Control for LMP1 protein expression in the absence of cycloheximide (CHX) is shown in Fig. 7A. tion Vpu prevents ␤-TrCP proteins from interacting with phosphorylated IB, thus inhibiting NF-B activity and contributing to the human immunodeficiency virus type 1-induced apoptosis (55,56).
␤-TrCP1 and HOS are known to bind to phosphorylated substrates (22), and it remains to be investigated whether serines 211, 215, 350, and 366 within LMP1-95-8 are phosphorylated in cells. If such phosphorylation is indeed essential for the interaction of LMP1 with HOS, it is expected that the ability of LMP1 to squelch endogenous HOS and restrict the extent of NF-B signaling in various types of cells will depend on the specific activities of the relevant protein kinase(s) in these cells.
The interaction of LMP1-95-8 with HOS is mediated by both the canonical DSGXXS motif and a more distal sequence within the C terminus of LMP1-95-8, and mutation of both sites is needed to abolish HOS-LMP1 binding (Fig. 2). At present, it is not clear how the non-canonical motif would bind HOS/␤-TrCP. One possibility is that this cryptic motif is formed by phosphorylated serines 350 and 366, whereas two proline residues between those serines may enable the twisting of the polypeptide chain, which would allow presenting putative phosphoserines to HOS/␤-TrCP (Fig. 2B). Seven WD40 domains of ␤-TrCP form a funnel-like structure with a channel that accommodates phosphorylated serines, which constitute a destruction motif within a substrate (52). The existence of additional twists of polypeptide chain between Ser-350 and Ser-366 within the non-canonical DSG(X) 14 S motif of LMP1 (shaded in Fig. 2B) may contribute to positioning of both phosphoserines to fit this channel. The existence of an extended cryptic degradation motif in LMP1 seems unusual, although ␤-TrCP/HOS substrates harboring spacers longer than three amino acids between phosphoserines have been reported (50,59).
It is also unclear why threonine in the 366 position would prevent recognition of non-canonical sequence by HOS/␤-TrCP. We observed that S366T substitution alone caused a larger decrease in HOS binding than G212S mutation ( Fig. 2A), indicating that Ser-366 may exert phosphorylation-independent effects on interactions between LMP1 and HOS. Further studies will identify structural determinants allowing formation of the HOS recognition sites on LMP1 as well as determine whether the efficiency of HOS binding may differ between canonical and cryptic sites.
LMP1-95-8 contains a single lysine (Lys-330) that plays no role in LMP1 ubiquitination (38). Recent structural studies indicate that SCF E3 ligases complexed with a phosphorylated substrate, and recruited E2 ubiquitin-conjugating enzymes exhibit a fairly rigid conformation, suggesting that ubiquitinaccepting lysine residues within a substrate should be optimally aligned toward E2 (60 -62). This positioning model predicts an optimal spacing between a degradation motif and a lysine residue as a key characteristic of a bona fide SCF substrate. Indeed, efficient ubiquitination of ␤-catenin-derived phosphopeptide by SCF ␤-TrCP1 requires such specific spacing (52). Our data showing that LMP1-95-8 stability is not regulated by SCF HOS E3 ubiquitin ligase unless lysine residues are introduced at optimal distance from non-canonical HOS recognition sites within LMP1 are in agreement with this positioning model.
Expressing high levels of LMP1-95-8, but not of LMP1-Cao, is known to limit NF-B activation and elicit cytostatic effects (31,32). Whereas squelching of HOS by LMP1-95-8 is apparently involved in restricting the extent of NF-B signaling, further studies are required to determine whether binding of HOS by LMP1 plays a role in mediating the cytostatic effects of LMP1. Our data indicate that squelching HOS/␤-TrCP by LMP1-95-8 decreases the rate of IB degradation (Figs. 6 and  7). On the other hand, such squelching may also diminish NF-B signaling by preventing appropriate processing of NF-B precursors that depend on SCF HOS/␤-TrCP (50,51). Furthermore, in EBV-infected cells, depletion of free HOS/␤-TrCP proteins by LMP1 may also impede proteolysis of other substrates of HOS/␤-TrCP, including such key regulators of the cell cycle as Wee1A 3 and Emi1. Emi1 stabilization in cells (63) and in ␤-TrCP1 knockout mice (48) may lead to defects in cell division. It is conceivable that HOS squelching mediated by degradation motifs within the C-terminal domain of LMP1-95-8 may contribute to its cytostatic and cytotoxic effects in concert with the mechanisms conferred by the transmembrane domain of LMP1-95-8 (31,32,35). These mechanisms are thought to play key roles in regulating the levels of LMP1 tolerated by infected cells as well as in supporting EBV persistence within infected cells (64).
Although future studies are needed to determine whether the ability of LMP1-95-8 to squelch endogenous HOS plays a role in limiting tumorigenic properties of this LMP1 variant, extensive epidemiological data refer to LMP1 mutations expected to abolish the ability of LMP1 to interact with HOS. In addition to 30-bp deletions, these mutations may include D349A substitution (65). Similar substitution within the DG-SXXS motif in IB␣ abrogated the ability of IB␣ to recruit ␤-TrCP/HOS (66). Furthermore, substitutions of Ser-366 (such as S366T, S366A, and S366N) are observed more frequently in LMP1 isolates from EBV-associated tumors than in LMP1 from peripheral blood lymphocytes from apparently healthy individuals (30,39,65,67,68). G212S mutation may not be as common as substitution of Ser-366 (65, 67). We did not find this substitution in many German or Russian LMP1 isolates that contained the S366T mutation. On the other hand, none of the described cases has revealed the presence of the LMP1-95-8 identical isolate in tumor tissue of any EBV-associated disease (39). 4 These data imply that, in tumorigenesis, LMP1 is under selective pressure to diminish its ability to interact with HOS and thus to abandon restrictions on activating NF-B signaling. The existence of such selection would predispose EBVcontaining tumor cells to accumulate mutated LMP1 proteins, and this phenomenon is indeed frequently observed in EBVassociated tumors that predominantly contain LMP1 variants with either Cao-like deletions or mutations in positions 212 and 366 (30,69,70). Further detailed studies will determine the role of LMP1 mutations that affect the LMP1-HOS interaction in the pathogenesis of EBV-associated malignancies.