Homodimer of Two F-box Proteins βTrCP1 or βTrCP2 Binds to IκBα for Signal-dependent Ubiquitination*

We found previously that overexpression of an F-box protein βTrCP1 and the structurally related βTrCP2 augments ubiquitination of phosphorylated IκBα (pIκBα) induced by tumor necrosis factor-α (TNF-α), but the relationship of the two homologous βTrCP proteins remains unknown. Herein we reveal that deletion mutants of βTrCP1 and βTrCP2 lacking the F-box domain suppressed ubiquitination and destruction of pIκBα as well as transcriptional activation of NF-κB. The ectopically expressed βTrCP1 and βTrCP2 formed both homodimer and heterodimer complexes without displaying the trimer complex. Dimerization of βTrCP1 and/or βTrCP2 takes place at their conserved NH2-terminal regions, termed a “D-domain” (for dimerization domain), located upstream of the F-box domain. The D-domain was necessary and sufficient for the dimer formation. Intriguingly, the βTrCP homodimer, but not the heterodimer, was selectively recruited to pIκBα induced by TNF-α. These results indicate that not only βTrCP1 but also βTrCP2 participates in the ubiquitination-dependent destruction of IκBα by forming SCFβTrCP1-βTrCP1 and SCFβTrCP2-βTrCP2 ubiquitin-ligase complexes.

Phosphorylation of two serine residues (Ser-32 and Ser-36) near the NH 2 terminus of IB␣, one of the best characterized members of the structurally and functionally related IB protein family, is essential for targeting IB␣ for signal-promoted destruction (4,5). This specific phosphorylation of IB␣ is catalyzed by an unusually large, multiprotein kinase complex, termed IB kinase, with an apparent molecular mass of 700 -900 kDa (6 -8). The phosphorylation of IB␣ by an IB kinase complex is necessary for its polyubiquitination at residues Lys-21 and Lys-22 (9,10). The ubiquitin (Ub) 1 -proteasome system then plays the next indispensable role for down-regulating IB␣ at the physiological level (6,11).
Ub is a landmark molecule in a post-translational protein modification system which plays a central role in intracellular protein breakdown (13)(14)(15). Protein ubiquitination is initiated by the formation of a high energy thioester bond between Ub and an E1 (Ub-activating enzyme) in a reaction that requires ATP hydrolysis. The activated Ub is then transferred to an E2 (Ub-conjugating enzyme). In some cases, E2 directly transfers Ub to target proteins, but the reaction often requires the participation of an E3 (Ub-ligating enzyme). Finally, a poly-Ub chain is formed by linking the COOH terminus of one Ub to a Lys residue within another Ub. The resultant poly-Ub chain acts as a degradation signal for proteolytic attack by the 26 S proteasome, a eukaryotic ATP-dependent 2-MDa protease complex (16,17). In this Ub-proteasome pathway, E3, the protein-Ub ligase, presumably plays the most important role in the selection of target proteins for degradation, because each distinct E3 binds the protein substrate with a degree of selectivity (14).
Accumulating evidence has revealed that activation of NF-B is induced by signal-dependent degradation of IB through the Ub pathway (14,15,18). An IB␣-Ub ligase (equivalent to IB␣-E3) is thought to be the key enzyme to control the level of IB. However, the long sought IB␣-E3 was identified only recently. The breakthrough for its discovery was a frequent report for Slimb, an F-box/WD40-repeat protein, which negatively regulates the Wingless pathway and Hedgehog pathway (19). Based on its structural feature, Slimb was postulated to be a component of the SCF (Skp1-Cdc53 or Cullins-F-box protein complex)-like Ub ligase complex which would participate in the degradation of IB␣ as well as ␤-catenin (18). To date, SCF is known to be a large multisubunit Ub-ligase that has been identified as a ubiquitination factor acting on a number of regulatory proteins such as cyclin-dependent kinase inhibitors Sic1 (20,21), p21 (22), and p27 (23)(24)(25), cyclin Cln1 (26), and cyclin D (22,27), E2F-1 (28), and CD4 (29). Growing in vivo and in vitro evidence demonstrates that IB␣-E3 is indeed an SCF-like complex, consisting of Skp1, Cullin-1 (abbreviated Cul-1), and ␤TrCP (30 -34). Intriguingly, the same IB␣-E3 has been found to catalyze the ubiquitination of ␤-catenin (30,(35)(36)(37).
Recently we reported that the putative IB␣-E3 present in HeLa cell cytosol is recruited to be specifically associated with in vivo phosphorylated IB␣ produced by tumor necrosis factor-␣ (TNF-␣) stimulation and recombinant IB␣ previously phosphorylated in vitro by IB kinase, causing polyubiquitination of IB␣ (38). We also found that the SCF complex acting as an IB␣-E3 contains not only ␤TrCP1 but also the structurally related ␤TrCP2, both of which augment ubiquitination of IB␣ when they are overexpressed in the HeLa cells (39). In the present study, we demonstrate that two homologous F-box/ WD40-repeat proteins, ␤TrCP1 and ␤TrCP2, are assembled with Skp1 and Cul-1 to form SCF complexes for ubiquitination of IB␣, which would lead to signal-dependent activation of NF-B. In addition, ectopically expressed ␤TrCP1 or ␤TrCP2 resulted in homo-and heterodimer complexes through their NH 2 -terminal regions, and each homodimer, but not heterodimer, was found to be recruited to phosphorylated IB␣ (pIB␣) in response to external signals. The possible roles of distinct SCF complexes, composed of two structurally related ␤TrCP proteins during the signal-dependent ubiquitination of IB␣, are discussed.
Reporter Assays-Mouse L929 cells were maintained in minimal essential medium containing 5% fetal calf serum and antibiotics. In the experiment, 2 ϫ 10 5 cells were transfected by the DEAE-dextran method with a total of 2.5 g of plasmid DNA containing 0.5 g of H2-Luc reporter plasmid and various amounts of effector plasmid (pBOS-␤TrCP1, pBOS-␤TrCP2, pBOS-␤TrCP1⌬F, pBOS-␤TrCP2⌬F) with an empty vector to adjust to 2 g. Cells were split into two wells at 24 h post-transfection and then stimulated with or without mouse TNF-␣ (1 ng/ml) for 1 h after an additional 24 h. Cells were lysed, and luciferase activity was measured with luciferase assay reagent (Promega) using a luminometer (Berthold). The relative fold activation of luciferase activities upon TNF-␣ treatment was calculated to normalize transfection efficiency.

RESULTS
Roles of ␤TrCP1 and ␤TrCP2 in Ubiquitination and Destruction of pIB␣ Induced by TNF-␣-Recently we reported that overexpression of ␤TrCP1 or ␤TrCP2 caused marked augmentation for ubiquitination of pIB␣ in response to TNF-␣ stimulation (39). To extend these observations, we examined the effect of deletions of the F-box domain of both ␤TrCP1 and ␤TrCP2 (abbreviated ␤TrCP1⌬F and ␤TrCP2⌬F, respectively; for these mutant structures, see Fig. 1, upper panel) for IB␣ ubiquitination. As shown in Fig. 1 (lower panel), treatment of TNF-␣ caused rapid loss of IB␣ detected by Western blotting as a function of time after TNF-␣ stimulation in HEK293 cells. In contrast, pretreatment of proteasome inhibitor MG132 (equivalent to Suc-LLL-CHO) prior to TNF-␣ stimulation caused a marked increase in phosphorylated pIB␣, in turn, accompanied by a decrease in its unphosphorylated form. This compound also resulted in abnormal accumulation of multiple polyubiquitinated bands of endogenous IB␣ within 5 min, and its sustained effect was observed up to 20 min ( Fig. 1, lower panel). In contrast, in the absence of MG132, the degradation of pIB␣ seemed to be so rapid that the ubiquitinated forms could not be detected to any visual extent. However, ectopic expressions of ␤TrCP1⌬F or ␤TrCP2⌬F strongly suppressed the accumulation of ubiquitinated forms of pIB␣ in a dose-dependent manner. These dominant-negative effects of ␤TrCP1⌬F or ␤TrCP2⌬F strongly indicate that both ␤TrCP1 and ␤TrCP2 direct the ubiquitination of pIB␣ in vivo.
We examined next whether ␤TrCP1 and ␤TrCP2 are involved in regulating instability, that is, the destruction of pIB␣ in vivo using these F-box deletion mutants. As shown in Fig. 2, the level of IB␣ decreased rapidly within 20 min after TNF-␣ stimulation (see mock transfection). However, the immediate loss of IB␣ caused by the treatment of TNF-␣ was markedly prevented, depending on the ␤TrCP1⌬F and ␤TrCP2⌬F added (top panel), even if MG132 had not been pretreated (compared with Fig. 1). Importantly, ectopic expressions of ␤TrCP1⌬F and ␤TrCP2⌬F also caused considerable accumulation of pIB␣ (top panel). The transfections of ␤TrCP1⌬F and ␤TrCP2⌬F resulted in the accumulation of these cellular levels in proportion to the amounts of transfected plasmids (middle panel), whereas the level of Skp1 as an internal control was unchanged.
To extend further that ␤TrCP1 and ␤TrCP2 are involved in IB␣ degradation, the release of NF-B was measured by NF-B-dependent transcription using a reporter plasmid that harbors NF-B binding sites upstream of the luciferase gene. As shown in Fig. 3, ectopic expression of ␤TrCP1⌬F and ␤TrCP2⌬F significantly inhibited TNF-␣-induced NF-B activation in a dose-dependent manner, in agreement with stabilized IB␣ (Fig. 2). Full-length ␤TrCP1 and ␤TrCP2 restored the transcriptional activity inhibited by ␤TrCP1⌬F or ␤TrCP2⌬F, although they did not significantly elevate the level of NF-B activations beyond that of cells transfected by an empty vector. Paradoxically, overproduction of ␤TrCP resulted in rather partial inhibition of NF-B activation, probably because of the nature of the F-box/WD40-repeat proteins that dimerize each other with a different partner before they incorporate into the SCF complex (data not shown; see "Discussion"). Taken together, it was clearly concluded that not only ␤TrCP1, which had been reported by many groups (31)(32)(33)(34), but also ␤TrCP2, indicated recently by us (39), are involved in the signal-dependent proteolysis of IB␣.
Homo-and Heterocomplexes of ␤TrCP1 and ␤TrCP2 Which Are Associated with Skp1 and Cul-1-It is interesting to determine how the two homologous ␤TrCP1 and ␤TrCP2 interact with pIB␣ as an SCF complex. To test this, two different epitopes (FLAG and T7 epitopes) were tagged to the COOH terminus or NH 2 terminus of ␤TrCP1 and ␤TrCP2; then their interaction in transfected HEK293 cells was assessed when treated with or without TNF-␣. Immunoprecipitation of FLAGtagged ␤TrCP1 or ␤TrCP2 resulted in specific coimmunoprecipitation of phosphorylated IB␣ only when the cells were stimulated by TNF-␣, indicating that these FLAG-tagged proteins function in cells (Fig. 4, right). As shown in Fig. 4, T7tagged ␤TrCP1 and ␤TrCP2 were evident in immunoprecipitates from both TNF-␣ treated and untreated cells, indicating that ␤TrCP1 and ␤TrCP2 form not only homocomplexes but also heterocomplexes, irrespective of TNF-␣ stimulation (Fig.  4). Skp1 and Cul-1, common partners of SCF complexes, were also evident in ␤TrCP immunoprecipitates from TNF-␣-untreated cells, indicating that neither complex formation nor SCF complex assembly is regulated by TNF-␣ stimulation or substrate binding. Each homocomplex and heterocomplex seems to form a functional SCF complex because SCF components in cell lysates are efficiently coimmunoprecipitated (left panel).
During analysis of the SCF assembly, we noticed that the anti-Cul-1 antibody exhibits two distinct bands in the immunoprecipitates of ␤TrCP1 or ␤TrCP2 in the TNF-␣-treated cell FIG. 1. Effect of ␤TrCP1⌬F and ␤TrCP2⌬F on TNF-␣-induced ubiquitination of IB␣. HEK293 cells were transfected with pcDNA3-FLAG-␤TrCP1⌬F, pcDNA3-FLAG-␤TrCP2⌬F (0.08 and 2 g/ well of a six-well plate), or pcDNA3 (mock, 2 g) and were treated for 1 h with or without MG132 (50 M) prior to stimulation for the indicated times with TNF-␣ (1 ng/ml). Cell lysates were prepared in the presence of 2 mM N-ethylmaleimide and 0.05% SDS to inhibit isopeptidase activities and were analyzed by immunoblotting with anti-IB␣ antibody. Multiple ubiquitinated IB␣ bands are designated (Ub) n -IB␣. The deletion mutants of ␤TrCP1 and ␤TrCP2 used in this study are represented schematically (upper panel). wt, wild-type. extracts but exhibits a single form in those from unstimulated cells (Fig. 4, bottom panel). The molecular size of Cul-1 deduced from the nucleotide sequence is approximately 90 kDa, which apparently corresponds to the smaller band of the two Cul-1 bands detected, indicating that Cul-1 may be modified in response to TNF-␣ stimulation or binding to pIB␣. Previously, it was reported that Cdc53, a yeast homolog of Cul-1, is modified by the Ub-like protein, Rub1 (40). In addition, we recently found that human NEDD8 (yeast Rub1 homolog) was covalently ligated to Cul-4A by a novel Ub-related conjugation pathway consisting of an E1-like APP-BP1-Uba3 heterodimer and E2-like Ubc12 (41). Very recently, we also found that NEDD8 was conjugated to six Cul-family proteins in vitro (42). Therefore, we assume that the larger form of Cul-1 is the result of modification by NEDD8.
␤TrCP1 Forms a Homodimer but not a Trimer Complex-Next we examined whether ␤TrCP1 forms a dimer or multimer complex. When three FLAG-, HA-, or Myc-tagged ␤TrCP1 were coexpressed in HEK293 cells, immunoprecipitates prepared by anti-FLAG antibody contained all FLAG-, HA-, and Myctagged ␤TrCP1 (Fig. 5, far left lane). Similar results were observed by immunoprecipitation with anti-HA or anti-Myc antibody (data not shown). Subsequently, immunoprecipitates prepared by anti-FLAG antibody were treated with FLAG peptide. The eluates were split and subjected to a second immunoprecipitation with anti-HA or anti-Myc antibodies. The resulting immunoprecipitates were used for Western analysis using anti-FLAG, -HA, or -Myc antibodies. As shown in Fig. 5  (right two lanes), the anti-HA antibody coimmunoprecipitated FLAG-and HA-␤TrCP1 but not Myc-␤TrCP1. Similarly, the anti-Myc antibody coimmunoprecipitated FLAG-and Myc-␤TrCP1, but not HA-␤TrCP1. Thus it was clear that expression of three independent tagged ␤TrCP1 resulted in two-tagged ␤TrCP1 interaction, but not a three-tagged ␤TrCP1 complex, strongly indicating that ␤TrCP1 did not form a trimer or multimer complex. Thus it was concluded that ␤TrCP1 is assembled into the dimeric complex.
Dimerization of ␤TrCP1 and ␤TrCP2 Occurs at Their NH 2terminal Conserved Regions Termed the "D-domain"-We next examined how both ␤TrCP1 and ␤TrCP2 interact with each other to form homo-and heterocomplexes using various deletion mutants of ␤TrCP1 and ␤TrCP2. In this experiment, the S-protein purification method was used. As shown in Fig. 6A, S-tagged ␤TrCP1 and ␤TrCP2 copurified wild-type FLAG-␤TrCP1 or deletion mutants lacking F-box or all WD40 repeats (named ⌬F or ⌬W 1-7 , respectively; see Fig. 1, upper panel), but not the NH 2 -terminal deletion mutant named FLAG-␤TrCP1⌬NF, indicating that the F-box alone is dispensable, whereas the NH 2 -terminal domain together with the F-box is required for dimer formation.
To verify further that the F-box is not required and that the NH 2 -terminal region is sufficient for dimer formation, S-tagged ␤TrCP1⌬F and ␤TrCP2⌬F were used to test their interactions with FLAG-tagged ␤TrCP constructs. As shown in Fig. 6B,  4. Formation of homo-and heterocomplexes of ␤TrCP1 and ␤TrCP2 which are assembled to the SCF complex. HEK293 cells were cotransfected with the indicated combinations of pEF-Bos-␤TrCP1-FLAG or pEF-Bos-␤TrCP2-FLAG (10 g each) and pcDNA3.1-T7-␤TrCP1 or pcDNA3.1-T7-␤TrCP2 (10 g each). 36 h after transfection, TNF-␣ treatment, preparation of cell lysates, and immunoprecipitation with anti-FLAG were carried out as described under "Experimental Procedures." Cell lysates and immunoprecipitates were analyzed by immunoblotting with the indicated antibodies. Asterisks indicate truncated ␤TrCP1-FLAG and ␤TrCP2-FLAG, which may be produced by proteolytic digestion or synthesized from a different translation initiation site. It is noteworthy that these faster migrating forms of ␤TrCP1-FLAG or ␤TrCP2-FLAG were not coimmunoprecipitated with T7-tagged ␤TrCPs (Footnote 2), supporting the importance of the NH 2 -terminal region for complex formation (for details, see "Results" and Figs. 6 and 7). either S-␤TrCP1⌬F or S-␤TrCP2⌬F was complexed with FLAG-tagged wild-type or deletion mutants lacking the F-box, but not with FLAG-tagged ␤TrCP1⌬NF, confirming again that the F-box is not required for dimer formation, but the NH 2terminal region beyond the the F-box domain is sufficient for efficient dimer formation.
To analyze further whether the NH 2 -terminal region is required for dimer formation irrespective of the F-box, two additional deletion mutants, ␤TrCP⌬N and ␤TrCP⌬D, which lack the NH 2 -terminal entire region beyond the F-box or just the conserved region flanking the F-box, respectively, were constructed (see Fig. 1, upper panel). As shown in Fig. 7A, none of the constructs that lack the NH 2 -terminal region beyond the F-box (␤TrCP1⌬N, ␤TrCP2⌬N, ␤TrCP2⌬D, and ␤TrCP2⌬D) was complexed with S-tagged wild-type ␤TrCP1 or ␤TrCP2. Reciprocal immunoprecipitation of FLAG-tagged constructs also revealed that the S-tagged ␤TrCP1 or ␤TrCP2 cannot bind ␤TrCP⌬N and ⌬D constructs, indicating that the NH 2 -terminal region is the site for dimer formation (Fig. 7B).
To ascertain that the F-box or the F-box-binding protein Skp1 does not play a role in dimer formation, we investigated Skp1 interaction with these FLAG-tagged ␤TrCP mutants. As shown in Fig. 7B, wild-type ␤TrCPs and mutants lacking the NH 2 -terminal region but not the F-box (␤TrCPs⌬N and ␤TrCPs⌬D) bind to Skp1. On the other hand, ␤TrCPs⌬NF and ␤TrCPs⌬F that lack the F-box failed to bind Skp1. The facts that the deletion of the F-box does not affect the dimerization property and that dimerization-defective mutants could still bind to Skp1 strongly indicate that Skp1 binding is not necessary and sufficient for dimer formation. Furthermore, it was indicated that the dimer formation of ␤TrCPs is not a prerequisite of Skp1 binding. Thus, we concluded that the NH 2terminal region is necessary and sufficient for dimer formation and provisionally named these specific regions needed for dimerization the D-domain. Although the dimer formation of ␤TrCPs can occur independently of TNF stimulation or substrate binding, substrates may help the dimerization of ␤TrCPs. Indeed, S-␤TrCP2⌬F could interact with FLAG-␤TrCP2⌬NF, although the extent was not very strong (Fig. 6B,  right panel). They may have indirectly interacted with each other through the WD40 repeat domain probably via the substrate. Further study is required for the clarification of this point.
␤TrCP1 and ␤TrCP2 Form Homo-and Heterodimer Complexes, but Only the Homodimer Is Recruited to pIB␣-As shown in Fig. 4, FLAG-tagged ␤TrCP1 coimmunoprecipitated not only T7-␤TrCP1 and T7-␤TrCP2 but also pIB␣ under the stimulation of TNF-␣. This result indicates that ␤TrCP1 and ␤TrCP2 form a heterocomplex but does not mean that this heterodimer complex is associated with pIB␣ because there is a possibility that ectopically expressed ␤TrCP1-FLAG or T7-␤TrCP2 forms a homodimer or heterodimer with endogenous ␤TrCP1, ␤TrCP2, or other F-box proteins. To analyze whether both homodimers and the heterodimer complex are able to associate with pIB␣, two-step immunoprecipitation analysis was carried out. As shown in Fig. 8, FLAG-and HA-tagged ␤TrCP1 or ␤TrCP2 plasmids were transfected and treated with TNF-␣. Immunoprecipitation was first carried out using anti-FLAG antibody, and the resulting immunoprecipitates were eluted with the FLAG peptide. The resulting eluates were split and subsequently immunoprecipitated individually with anti-HA or anti-IB␣ antibodies, and the resultant immunoprecipitates were analyzed by immunoblots with anti-FLAG, anti-HA, and anti-IB␣ antibodies.
When HA-tagged ␤TrCP plasmids were cotransfected to generate the homodimer in combination with FLAG-tagged plasmids (i.e. FLAG-␤TrCP1-HA-␤TrCP1 or FLAG-␤TrCP2-HA-␤TrCP2), the second immunoprecipitates obtained by anti-HA antibody contained pIB␣, and in reciprocal anti-IB␣ immunoprecipitates, HA-tagged ␤TrCPs were observed (Fig. 8, left   FIG. 6. The F-box of ␤TrCP1 and ␤TrCP2 is not required, but the NH 2terminal region is sufficient for each homo-and heterodimer formation. Stagged wild-type (panel A) or the F-box deletion mutant (panel B) of ␤TrCP1 and ␤TrCP2 was transfected in HEK293 cells in combination with various FLAGtagged wild-type or deletion mutants of ␤TrCP (␤TrCP⌬F, ␤TrCP⌬NF, and ␤TrCP⌬W 1-7 ), as indicated. 36 h posttransfection, cell lysates were prepared and subjected to the affinity purification of the S-protein complex (S-protein ppt) and then analyzed by immunoblotting with anti-FLAG and anti-S (K- 14) antibodies. and right panels), strongly indicating that the homodimer of ␤TrCP1 or ␤TrCP2 is associated with pIB␣. The reason that the extent of pIB␣ obtained by immunoprecipitation with the anti-HA antibody was considerably less than that obtained by the anti-IB␣ antibody may be in part because FLAG-tagged ␤TrCP1 or ␤TrCP2 can homodimerize with itself or with endogenous ␤TrCPs. On the other hand, when plasmids were cotransfected to produce heterodimers (i.e. FLAG-␤TrCP1-HA-␤TrCP2 or FLAG-␤TrCP2-HA-␤TrCP1), the second immunoprecipitates obtained by anti-HA antibody contained no pIB␣, and HA-tagged ␤TrCPs were not seen in reciprocal anti-IB␣ immunoprecipitates (Fig. 8, middle two panels), strongly indicating that the heterodimers of ␤TrCP1 and ␤TrCP2 do not bind to pIB␣. Therefore, it is clear that the heterodimeric complex of ␤TrCP1 and ␤TrCP2 cannot be recruited to associate with pIB␣, and in turn, the homodimers of ␤TrCP1 and ␤TrCP2 complexed with Skp1 and the Cul-1 function as an SCF Ub-ligase complex responsible for the ubiquitination of pIB␣. DISCUSSION Accumulating evidence shows that SCF ␤TrCP1 is responsible for the ubiquitination of pIB␣ (30 -34). In addition, we also reported that not only SCF ␤TrCP1 but also SCF ␤TrCP2 functions for ubiquitination, based on the findings that ␤TrCP2, like ␤TrCP1, was associated with pIB␣ and that overexpression of ␤TrCP1 and ␤TrCP2 individually augmented the ubiquitination of pIB␣ (39). Here we extended the original finding for the involvement of ␤TrCP2 in addition to ␤TrCP1 for ubiquitination of pIB␣ and revealed that ␤TrCP1 and ␤TrCP2 are complexed with Skp1 and Cul-1 to form SCF ␤TrCP1 and SCF ␤TrCP2 , respectively, which are recruited to bind to pIB␣ phosphorylated in response to TNF-␣ stimulation. Indeed, their F-box deletion mutants prevented the ubiquitination and the subsequent degradation of pIB␣. Moreover, these two mutant ␤TrCP expressions caused significant prevention of signal-dependent transcriptional activation of NF-B. However, it is unknown why two homologous F-box/WD40-repeat proteins display apparently similar roles for pIB␣ destruction. Perhaps these related proteins have redundant functions to maintain the important NF-B signaling pathway.
There are many reports that SCF ␤TrCP1 is involved in the ubiquitination not only of pIB␣ but also of phosphorylated ␤-catenin (30,(35)(36)(37). This is reasonable considering the similarity of the two phosphorylation sites of IB␣ and ␤-catenin (43), although obviously different kinases are involved, respectively i.e. IB␣ kinase and glycogen synthase kinase 3␤ (18). Moreover, Fuchs et al. (44) reported that ␤TrCP2 is also responsible for the ubiquitination-dependent breakdown of ␤-catenin as well as IB␣, which is in contrast to the finding that ␤TrCP2, unlike ␤TrCP1, seems not to interact directly with ␤-catenin based on yeast two-hybrid analysis (36). Whether ␤TrCP2 is involved in the destruction of ␤-catenin requires further study. Fuchs et al. (44) did not show the effect of ␤TrCP1 and ␤TrCP2 for the degradation of IB␣ and ␤-catenin simultaneously; therefore, it is unknown whether the role of ␤TrCP2 for ubiquitination of ␤-catenin occurs physiologically. The similarity between ␤TrCP1 and ␤TrCP2 is very close, displaying an overall identity of 76%, with almost an identical WD40 repeat domain that is capable of binding phosphorylated IB␣ and ␤-catenin. Hence, overproduction of the related family of the ␤TrCP protein may bind directly to the substrates or indirectly through dimerization via the D-domain under nonphysiological conditions. This assumption is also supported by the results in which excess production of wild-type ␤TrCP proteins leads to inhibition of the NF-B-dependent transcriptional activation (data not shown). Finally, we found that overexpression of not only ␤TrCP1 but also ␤TrCP2 caused coimmunoprecipitation of ␤-catenin; however, the homodimer of ␤TrCP2 or the heterodimer of ␤TrCP1 and ␤TrCP2 does not associate with ␤-catenin, although the homodimer of ␤TrCP1 is firmly bound, 2 which clearly differs from their effect for pIB␣ (see below). Therefore, we would favor a hypothesis that these two F-box proteins do not simply have redundant functions, rather they confer substrate specificity; i.e. ␤TrCP1, compared with ␤TrCP2, has a dominant effect on the ubiquitination of ␤-catenin. However, it remains unknown whether a similar substrate specificity is present for the ubiquitination of IB␣ because ␤TrCP1 and ␤TrCP2 appear to have equivalent activity as shown by the present study. Further investigation is needed for the demonstration of this interesting hypothesis.
For the first time, we also present direct evidence that ectopically expressed ␤TrCP1 and ␤TrCP2 form both a homodimer and a heterodimer, irrespective of the treatment of TNF-␣, without displaying the multimer complex. This finding is consistent with recent observations that fission yeast SCF, which regulates the level of the cyclin-dependent kinase inhibitor Rum1 and the S-phase regulator Cdc18 (45,46), is composed of two related F-box-WD40 repeat proteins termed Pop1-Ste16 and Pop2-Sud1, forming three distinct complexes, SCF Pop1-Pop1 , SCF Pop1-Pop2 , and SCF Pop2-Pop2 (47).
We demonstrated here that dimer formation of ␤TrCP1 and/or ␤TrCP2 takes place at the D-domain in their NH 2terminal regions. This NH 2 -terminal domain beyond the F-box is necessary and sufficient for their dimerization. Unexpectedly, Skp1 binding to the F-box did not require the dimer formation of F-box proteins, and vice versa. Initially, we thought that two F-box interfaces are required for the Skp1 binding because Skp1 was not associated with the heterodimer consisting of ␤TrCP⌬F and wild-type ␤TrCP, raising the possibility that dimer formation is a prerequisite of Skp1 binding. However, based on the finding that both ␤TrCP⌬D and ␤TrCP⌬N could bind to Skp1, we concluded that dimer formation is not essential for Skp1 binding, and hence SCF complex formation occurs (data not shown). Indeed, ␤TrCP⌬N, in contrast to ␤TrCP⌬F, did not show significant effect of pIB␣ destabilization compared with that of ␤TrCP⌬F (data not shown), probably because of its functional SCF complex assembly or in part because of the lack of dimerization property with endogenous F-box proteins. Because F-box proteins bind the substrate, the combination of the F-box would regulate the affinity of the SCF complexes for the substrates. From this point of view, it is tempting to determine whether dimerization via the D-domain is regulated by other factors.
Consistent with our findings, Pop1-Ste16 and Pop2-Sud1 are associated with each other at their NH 2 -terminal regions (48). Moreover, based on genetic evidence, Wolf et al. (48) reported that only the heterocomplex of Pop1 and Pop2 mediated Cdc18 and Rum1 proteolysis, even though Pop1 and Pop2 formed both homo-and heterocomplexes (47). They also showed that the Pop genes failed to complement the deleted phenotypes of each other, indicating that the functions of Pop1 and Pop2 are not redundant. On the other hand, we demonstrated that the homodimers, but not the heterodimers, of ␤TrCP1 and ␤TrCP2 were selectively recruited to pIB␣ induced by TNF-␣, revealing that not only ␤TrCP1 but also ␤TrCP2 participates in the ubiquitination-dependent destruction of pIB␣ by forming their homodimeric SCF ␤TrCP1-␤TrCP1 and SCF ␤TrCP2-␤TrCP2 complexes. Considering these differences, we cannot exclude the possibility that SCF ␤TrCP1-␤TrCP2 is responsible for other function(s) so far unknown because it is possible that ␤TrCP1 or ␤TrCP2 may have interacted with other factors that would form the functional complexes to contribute as protein-Ub ligases at physiological levels. It is notable that the F-box/WD40repeat proteins consist of a large gene family in various eukaryotic cells (49), and it is known that another ␤TrCP protein with a close similarity to ␤TrCP1, displaying 97% identity, exists in Xenopus laevis (12). If the homolog exists in mammalian cells, the protein may interact with ␤TrCP1 or ␤TrCP2 to form different SCF complexes that function for the ubiquitination of other targets.