Regulation of the NEDD8 Conjugation System by a Splicing Variant, NUB1L*

NEDD8 is a ubiquitin-like protein that controls vital biological events through its conjugation to target proteins. We previously identified a negative regulator of the NEDD8 conjugation system, NUB1, which works by recruiting NEDD8 and its conjugates to the proteasome for degradation. Recently, we found its splicing variant, NUB1L. It possesses an insertion of 14 amino acids that codes for a UBA domain. Structural study revealed that NUB1 has a NEDD8-binding site at the C terminus, whereas NUB1L has an additional site at the newly generated UBA domain. Interestingly, the sequence A(X4)L(X10)L(X3)L was conserved in these NEDD8-binding sites among human and other mammals. Mutational studies revealed that at least three Leu residues in the conserved sequence are required for binding with NEDD8. Functional study suggested that the NEDD8-binding ability at the C terminus of NUB1 and NUB1L is mainly involved in the down-regulation of NEDD8, but the NEDD8-binding ability at the UBA2 domain of NUB1L is minimally or not involved at all. The NEDD8-binding ability at the UBA2 domain might be required for an unknown function of NUB1L.

NEDD8 is a ubiquitin-like protein that controls vital biological events through its conjugation to target proteins. We previously identified a negative regulator of the NEDD8 conjugation system, NUB1, which works by recruiting NEDD8 and its conjugates to the proteasome for degradation. Recently, we found its splicing variant, NUB1L. It possesses an insertion of 14 amino acids that codes for a UBA domain. Structural study revealed that NUB1 has a NEDD8-binding site at the C terminus, whereas NUB1L has an additional site at the newly generated UBA domain. Interestingly, the sequence A(X 4 )L(X 10 )L(X 3 )L was conserved in these NEDD8-binding sites among human and other mammals. Mutational studies revealed that at least three Leu residues in the conserved sequence are required for binding with NEDD8. Functional study suggested that the NEDD8binding ability at the C terminus of NUB1 and NUB1L is mainly involved in the down-regulation of NEDD8, but the NEDD8-binding ability at the UBA2 domain of NUB1L is minimally or not involved at all. The NEDD8binding ability at the UBA2 domain might be required for an unknown function of NUB1L.
NEDD8 is a highly conserved 81-amino acid protein that shares 60% identity and 80% homology with ubiquitin. NEDD8 conjugates to a large number of target proteins (1), and this conjugation is thought to be catalyzed by four enzymes, NEDD8-carboxyl-terminal hydrolase (2), NEDD8-activating enzyme, NEDD8-conjugating enzyme, and NEDD8-ligating enzyme, in a manner analogous to ubiquitination and sentrinization (also known as SUMO-conjugation) (3). So far, all of the known NEDD8 targets are cullin family members, and these include Cul-1, -2, -3, -4A, -4B, and -5 (4,5). 1 Each cullin family member appears to be a component of the SCF complex, a ubiquitin E3 ligase composed of Skp1, Cullin, F-box protein, and ROC1 (also called RBX1 or HRT1) (3,6). For example, Cul-1 is a major component of an SCF complex that catalyzes the ubiquitination of IB␣, ␤-catenin, and p27 (Kip1) (7)(8)(9) and controls many biological events, such as cell-cycle transition, inflammation, and tumorigenesis. Recently, several groups reported that NEDD8 conjugation to Cul-1 is required for the ubiquitin-ligase activity of the Cul-1-containing SCF complex (10 -13). These observations suggest that the NEDD8 conjugation system is involved in many important biological functions. Indeed, the NEDD8 conjugation system was shown to be essential for cell cycle progression and morphogenesis in mice (14) and for eye development in Drosophila (15).
Recently, we identified a novel down-regulator of the NEDD8 conjugation system, NUB1 (NEDD8 ultimate buster-1), using a yeast two-hybrid system with NEDD8 as bait (16). NUB1 is a NEDD8-interacting protein composed of 601 amino acid residues with a calculated molecular mass of 69.1 kDa. It is an interferon-inducible protein and predominantly localizes in the nucleus. In a biochemical analysis, we found that NUB1 overexpression led to a severe reduction in the NEDD8 monomer and its conjugates in cells (16). Surprisingly, this reduction was completely blocked by proteasome inhibitors (17). Furthermore, we found that NUB1 interacted with the 19 S proteasome activator (PA700) (17). These results strongly suggested that NUB1 recruits NEDD8 and its conjugates to the proteasome for degradation, making NUB1 a down-regulator in the NEDD8 conjugation system.
The ubiquitin-associated (UBA) domain is a small domain of about 40 residues that was initially identified in ubiquitination enzymes, including E2s, E3s, and other proteins linked to ubiquitination (18). Human NUB1 has two UBA domains, whereas NUB1 homologues of other species such as mouse (accession number: AF534114), Drosophila (accession number: AE003752), and Arabidopsis (accession number: AC007295) have three UBA domains. This suggests that there is an isoform of human NUB1 with three UBA domains. To isolate isoforms of NUB1, we recently screened the human cDNA library by plaque hybridization and identified a splicing variant of NUB1 that encodes a longer protein, termed NUB1L (accession number: AF459743). In the studies described here, we characterized NUB1L and compared NUB1 and NUB1L.

EXPERIMENTAL PROCEDURES
Cell Culture-HeLa cells (American Type Culture Collection, Manassas, VA) and COS-M6 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics.
Antibodies-Mouse anti-RH antibody (specific for the amino acid sequences RGSHHHH and GGSHHHH) was purchased from Qiagen (Santa Clara, CA). Mouse anti-FLAG antibody (M5) was purchased from Sigma Chemical Co. (St. Louis, MO). GST-12, mouse monoclonal antibody specific for glutathione S-transferase (GST), was purchased from Santa Cruz Biotechnologies (Santa Cruz, CA). Rabbit anti-human NUB1 antiserum was generated by immunization with a GST fusion protein of NUB1 corresponding to amino acids 432-601 (16). Rabbit polyclonal anti-actin antibody (specific for the C-terminal actin fragment) was purchased from Sigma.
Plasmid Construction and Transfection-To express proteins tagged with epitope at the N terminus in mammalian cells, pcDNA3/RH-N and pcDNA3/FLAG-N were used as described previously (17,19). To express NUB1L, the cDNA insert was subcloned into pcDNA3/RH-N from pBK-CMV/NUB1L, which was isolated by cDNA library screening. The other human cDNAs used in this report were described previously. These include ubiquitin (20), NEDD8 (1), sentrin-3 (21), USP21 (22), HHR23B (23), glucocorticoid receptor (GR) (24), and NUB1 (16). These cDNAs were inserted into the aforementioned plasmid vectors, and the plasmids were transfected into COS-M6 cells or HeLa cells using FuGENE6 (Roche Applied Science, Indianapolis, IN). The transfected cells were harvested for immunostaining or Western blotting 20 h after transfection.
Screening of cDNA Library by Plaque Hybridization-A human testis cDNA library constructed in the ZAP Express vector (Stratagene, La Jolla, CA) was screened with the probe of NUB1 cDNA fragment, which was labeled with [␣-32 P]dCTP by a Ready-To-Go DNA labeling kit (Amersham Biosciences, Piscataway, NJ). Approximately 1 ϫ 10 6 plaques were transferred to Hybond-N ϩ replicated filters (Amersham Biosciences). The filters were hybridized with the probe in ExpressHyb hybridization solution (Clontech, Palo Alto, CA) and washed in 0.1ϫ SSC containing 0.1% SDS. Autoradiography was performed at Ϫ80°C.
Determination of Ubiquitin-like Domain and Ubiquitin-associated Domain-Ubiquitin-like (UBL) domains were determined by the Pro-fileScan program. Ubiquitin-associated (UBA) domains were determined by the SMART program. These programs are available via the World Wide Web.
RT-PCR-To detect messages of NUB1 and its isoform NUB1L in various human tissues, RT-PCR was performed using specific primers. For NUB1, a forward primer A (5Ј-CAGAGAGGAACTGGCCCAA-3Ј) and a reverse primer B (5Ј-GTACACCAATCGGTCAATG-3Ј) were used. For NUB1L, a forward primer A (5Ј-CAGAGAGGAACTGGCCCAA-3Ј) and a reverse primer C (5Ј-CTTCAGGGCCTCATCCAAGTTC-3Ј) were used. As a template, human multipanel cDNAs were purchased from Clontech (Palo Alto, CA). Two microliters of human multipanel cDNA was used for each PCR in a final volume of 15 l in the presence of 50 mM KCl, 25 mM Tris-HCl (pH 8.5), 1.5 mM MgCl 2 , 200 M dNTPs, 2% Me 2 SO, and 0.15 unit of Taq polymerase (Promega). The PCR condition was 30 s at 94°C, 30 s at 56°C, and 1 min at 72°C for 35 cycles. As a control, we amplified glyceraldehyde-3-phosphate dehydrogenase cDNA using a forward primer (5Ј-GTGGAGTCCACTGGCGTCTT-3Ј) and a reverse primer (5Ј-TGAGCTTCCCGTTCAGCTC-3Ј). The amplified products were loaded onto 1% agarose gel, followed by staining with ethidium bromide.
Treatment with Proteasome Inhibitors-N-Acetyl-L-leucinal-L-leucinal-L-norleucinal (LLnL) was purchased from Sigma. MG132 and lactacystin were purchased from Calbiochem (La Jolla, CA). 1 ϫ 10 6 COS-M6 cells were transfected by FuGENE6. After overnight culture, the culture medium was replaced with fresh medium containing proteasome inhibitor such as LLnL (50 M), MG132 (20 M), or lactacystin (10 M). The transfectants were cultured at 37°C with these proteasome inhibitors for 6 h. Then the cells were harvested, and the total cell lysates were prepared for Western blot analysis.
Western Blotting-Protein samples were treated at 45°C for 1 h in 150 l of 2% SDS-treating solution containing 5% ␤-mercaptoethanol. After SDS-PAGE, Western blotting was performed using the protocol provided with an ECL detection system (Amersham Biosciences). As a secondary antibody, horseradish peroxidase (HRP)-conjugated antimouse IgG antibody or anti-rabbit IgG antibody (Santa Cruz Biotechnologies) was used.
Immunostaining-Immunocytochemical staining was performed by the avidin-biotin-HRP complex (ABC-HRP) method, using the Vectastain ABC kit system (Vector, Burlingame, CA). Transfected HeLa cells on a coverslip were fixed in 3.7% paraformaldehyde solution for 20 min at 4°C and permeabilized in 0.1% Triton X-100 for 10 min at room temperature. After washing with PBS, the fixed cells were incubated with PBS containing 5% H 2 O 2 for 10 min to quench endogenous peroxidase activity and then washed with PBS. The cells were incubated for 10 min with PBS containing 5% horse serum for blocking, followed by additional incubation with anti-RH antibody for 30 min at 37°C. After rinsing with PBS, the cells were incubated with biotinylated anti-mouse IgG antibody for 30 min at 37°C, washed with PBS, and treated with the ABC reagent (avidin-biotin-HRP complex) for 30 min at 37°C. Finally, the enzymatic disclosing procedure was performed as reported previously (25).
Yeast Two-hybrid Analysis-Using PCR, we prepared cDNAs of the mutated NUB1 and NUB1L shown in Fig. 8. To examine the in vivo interaction of NEDD8 with these mutants, the yeast MATCHMAKER two-hybrid system 3 (Clontech) was used as described previously (26). The cDNA of NEDD8-G was subcloned into pGADT7 (Gal4 DNA-activating domain vector for Gal4-AD fusion) from pcDNA3/HA-NEDD8-G (1), whereas the cDNA of each mutant of NUB1 and NUB1L was subcloned into pGBKT7 (Gal4 DNA-binding domain vector for Gal4-BD fusion). The plasmids for the two fusion constructs were cotransfected into AH109 yeast cells using the lithium acetate method (27). Transformed yeast cells were grown on a His Ϫ /Trp Ϫ /Leu Ϫ synthetic agar plate for 3 days at 30°C. The specific protein-protein interaction was determined by the growth of the cells on the selection plate. Liquid ␤-galactosidase assays were performed according to manual of the manufacturer (Clontech). The ␤-galactosidase expression levels in liquid yeast cultures were measured in six independent transformants. Site-directed Mutagenesis-Leu-to-Ala substitutions were generated in NUB1 at Leu-573 and Leu-577 and in NUB1L at Leu-464, Leu-468, Leu-587, and Leu-591. The cDNAs of wild type NUB1 and NUB1L were mutated by PCR-based site-directed mutagenesis as described previously (28). The mutated cDNAs were subcloned into pTrcHis (Invitrogen) or pcDNA3/RH-N (1).
GST Pull-down Assay-RH-tagged proteins and GST fusion proteins were expressed in Escherichia coli BL21 using pTrcHis plasmid (Invitrogen, Carlsbad, CA) and pGEX-2TK plasmid (Amersham Biosciences), respectively. Cells were resuspended in lysis buffer (25 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.1% Nonidet P-40) containing protease inhibitor mixture (Roche Applied Science) and then lysed by brief sonication. The GST fusion proteins were purified as described previously (28). The crude bacterial lysate containing RH-tagged proteins was centrifuged at 14,000 ϫ g for 5 min, and the supernatant was incubated for 3 h at 4°C with GST fusion proteins immobilized on glutathione-Sepharose beads (Amersham Biosciences). The beads were then washed four times with lysis buffer. The precipitated proteins on the beads were solubilized in sample treating buffer containing 2% SDS and 5% ␤-mercaptoethanol, followed by Western blot analysis using anti-RH antibody.

Domain Structure of Human NUB1 and Its Homologues in
Other Species-To examine NUB1 homologues in other species, we searched the protein data base of GenBank TM and found the homologous sequences of Drosophila and Arabidopsis. These homologues had three UBA domains. In addition, we found the mouse homologue BS4 (accession number: U27462) in the data base. On the basis of the cDNA sequence, BS4 is composed of 677 amino acids. Unlike NUB1 and the other homologues we examined, BS4 had a longer C-terminal region next to the UBA repeat. In the C-terminal region, BS4 shared no homology to other homologues, implying sequencing errors in the BS4 cDNA. To confirm this, we isolated and sequenced the cDNA from the testis cDNA library of BALB/c mouse. It revealed that the cDNA isolated in our laboratory was not identical to that of BS4. In particular, we found six differences in the coding region between the sequences. All the differences appeared to result from the sequencing errors of BS4, which in turn caused a frameshift. The cDNA and protein sequences of our clone were submitted to GenBank TM (accession number: AF534114). We termed the gene product "mouse NUB1." It is composed of 614 amino acids and possesses three UBA domains, as do the Drosophila homologue and the Arabidopsis homologue ( Fig. 1). In the data base, we could not find any NUB1 homologues containing two UBA domains (Fig. 1).
cDNA Cloning of NUB1 Isoform-Whereas human NUB1 has two UBA domains, a search for NUB1 homologues in the protein data base of GenBank TM revealed that the homologues in other species have three UBA domains, as described above (Fig. 1). This observation suggested that there is an isoform of human NUB1 possessing three UBA domains. To isolate this possible isoform of NUB1, we recently screened the human testis cDNA library by plaque hybridization and obtained seven positive clones. Subsequent analysis by DNA sequencing showed that three of the seven clones had a 3113-bp insert containing a single open reading frame from nucleotides 78 to 1883 (Fig. 2, Clone 1). The cDNA sequence of Clone 1 was identical to that of NUB1, which was previously reported by our group (accession number: AF300717) (16). Another two of the seven clones had a 2687-bp insert containing a single open reading frame from nucleotides 37 to 1842 (Fig. 2, Clone 2 (accession number: AY129295)). Although the non-coding regions at both the 5Ј-and 3Ј-ends were shorter than those of Clone 1, its coding region (nucleotides 37-1842) was identical to that of Clone 1 (nucleotides 78 -1883). The remaining two of the seven clones had a 3114-bp insert containing a single open reading frame from nucleotides 37 to 1884 (Fig. 2, Clone 3 (accession number: AF459743)). The non-coding region at the 5Ј-end (nucleotides 1-36) was identical to that of Clone 2 (nucleotides 1-36), whereas the non-coding region at the 3Ј end (nucleotides 1885-3114) was identical to that of Clone 1 (nucleotides 1884 -3113). The coding region of Clone 3 was 42 bp longer than that of Clone 1 and Clone 2 because of a single insertion from nucleotides 1390 to 1431. As a result, Clone 3 coded for a longer form of NUB1 that had an insertion of 14 amino acids. We termed this isoform NUB1L.
Alternative Splicing of NUB1 Gene and the Domain Structure of Its Product-To determine how the cDNA of NUB1L was generated, we searched the genomic data base of Gen-Bank TM and found the genomic sequence of human NUB1 in chromosome 7q36 (accession number: AC005486). Sequence analysis revealed that the NUB1 gene was composed of 15 exons and that the 42-bp insertion in the NUB1L cDNA was generated by alternative splicing of an extended exon 12 of the NUB1 gene. As described above, the 42-bp insertion of the cDNA resulted in a 14-amino acid insertion between residue 451 and 452 in the protein sequence of NUB1 (Fig. 1, see also Fig. 8C). This short insert in NUB1L led to the generation of an additional UBA domain (termed UBA2) between the two original UBA domains (termed UBA1 and UBA3). Thus, we found a novel isoform, NUB1L, containing three UBA domains as initially expected (see Figs. 1 and 8C).
Subcellular Localization of NUB1L-The subcellular localization of NUB1L was determined in HeLa cells transfected with a plasmid containing an insert of RH-NUB1L cDNA. As controls, we transfected HeLa cells with either an expression plasmid alone or a plasmid containing an insert of RH-USP21 or RH-NUB1 cDNA. The cells were then fixed, permeabilized, and stained with anti-RH antibody. As shown in Fig. 3, RH-USP21 was detected equally in both the cytosol and the nucleus, whereas RH-NUB1 and RH-NUB1L were mainly found in the nucleus. The nuclear localization of NUB1 and NUB1L is consonant with the fact that both molecules have an NLS located from Arg-414 to Arg-431 (16). However, NUB1 and NUB1L were also detected weakly in the cytosol.
Tissue Distribution of NUB1 and NUB1L-To identify the expression of NUB1 and NUB1L, we employed RT-PCR. As shown in Fig. 4B, the NUB1 message was detected almost equally in all examined tissues other than the pancreas (lane 13). In contrast, the NUB1L message could not be detected equally. The NUB1L message in the prostate (lane 3), leukocyte (lane 7), liver (lane 10), and skeletal muscle (lane 11) was detected much less than that in the other tissues.
Activity of NUB1L in the Down-regulation of NEDD8 Expression-The COS cell coexpression assay was used to examine the functions of NUB1 and NUB1L. In brief, FLAG-tagged NEDD8 was coexpressed in COS cells with empty vector, RHtagged glucocorticoid receptor (GR), NUB1, or NUB1L. As shown in Fig. 5A, when FLAG-NEDD8 was expressed with empty vector (lane 2) or RH-tagged GR (lane 3), we clearly detected a 7.0-kDa band of unconjugated FLAG-NEDD8 and high molecular mass bands of NEDD8-conjugated proteins. This NEDD8 expression pattern was identical to the one we observed previously (1). When FLAG-NEDD8 was coexpressed with RH-NUB1 (lane 4) or RH-NUB1L (lane 5), both the unconjugated and conjugated forms of NEDD8 were dramatically reduced. This result suggested that NUB1 and NUB1L possess the ability to down-regulate the expression of NEDD8 monomer and its conjugates. As controls, we also coexpressed FLAGtagged ubiquitin or sentrin-3 with empty vector, RH-GR, RH-NUB1, or RH-NUB1L. As shown in Fig. 5 (B and C), the overexpression of NUB1 and NUB1L did not reduce the expression of ubiquitin or sentrin-3 (lanes 4 and 5). Thus, the downregulation induced by NUB1 and NUB1L was specific to NEDD8.
We previously reported that the reduction of NEDD8 expres- sion by NUB1 was completely blocked by proteasome inhibitors (17), indicating the involvement of the proteasome in NUB1 function. In this article, we examined the same possible involvement in NUB1L function. RH-NUB1L was coexpressed with FLAG-NEDD8 in COS cells in the absence or presence of various proteasome inhibitors, including N-acetyl-L-leucinal-Lleucinal-L-norleucinal (LLnL), MG132, and lactacystin. As shown in the upper panel of Fig. 6, the overexpression of NUB1L led to a dramatic reduction of NEDD8 expression (lane 2 versus lane 3). As expected, all of the proteasome inhibitors prevented the reduction in NEDD8 conjugates caused by NUB1L overexpression (lanes 4 -6). However, they did not restore the levels of unconjugated NEDD8. This is probably because the unconjugated NEDD8 was quickly consumed for the conjugation of its targets during the treatment with proteasome inhibitors.
Increased Interaction of NUB1L with NEDD8 -In our initial study, we isolated the cDNA of the C-terminal fragment (Glu-372 to Asn-601) of NUB1 by a yeast two-hybrid screening using NEDD8 as bait (16). This indicated that the NEDD8-binding site is located on this C-terminal region of NUB1. We therefore hypothesized that the inserted sequence in the C-terminal region of NUB1L may exert an influence on its interaction with NEDD8. To test this hypothesis, we examined the interaction of NEDD8 with the C-terminal region of NUB1 (termed NUB1c) or NUB1L (termed NUB1Lc) by a liquid culture assay of the yeast two-hybrid system in which ␤-galactosidase activity was measured to quantify the interaction. Before the assay, we prepared cDNAs of NUB1c and NUB1Lc. NUB1c is the C-terminal fragment (Glu-372 to Asn-601) of NUB1, and NUB1Lc is the C-terminal fragment (Glu-372 to Asn-615) of NUB1L. These cDNAs were inserted into a plasmid, pGBKT7 (Gal4 DNA-binding domain vector for Gal4-BD fusion). We also used a plasmid, pGADT7 (Gal4 DNA-activating domain vector for Gal4-AD fusion), in which a cDNA of NEDD8-G had been inserted (pGADT7/NEDD8-G). Because NEDD8-G has a Cterminal deletion from Gly-76 to Gln-81 and is unable to form conjugates with target proteins, its use allowed us to examine the interaction of NUB1c and NUB1Lc with NEDD8 itself, not with the target proteins of NEDD8 (1,17). For the yeast twohybrid assay, yeast cells were cotransformed with pGADT7/ NEDD8-G and pGBKT7 plasmid described above. As shown in the lower panel of Fig. 7A, yeast cells transformed with pGADT7/NEDD8-G alone (negative control) produced low levels of ␤-galactosidase. Yeast cells transformed with pGADT7/ NEDD8-G and pGBKT7/HHR23B (negative control) also produced low levels of ␤-galactosidase due to the lack of interaction between NEDD8 and HHR23B. In contrast, yeast cells transformed with pGADT7/NEDD8-G and pGBKT7/ NUB1c displayed high levels of ␤-galactosidase because of the interaction between NEDD8 and NUB1c. Yeast cells transformed with pGADT7/NEDD8-G and pGBKT7/NUB1Lc displayed even higher levels of ␤-galactosidase indicating a stronger interaction between NEDD8 and NUB1Lc. Importantly, NUB1Lc produced levels of ␤-galactosidase almost double those produced by NUB1c. Although the result of the ␤-galactosidase assay was clear, it was possible that the observed difference was due to the differing expression levels between NUB1c and NUB1Lc. To rule out this possibility, we examined the expression levels of NUB1c and NUB1Lc in yeast cells by Western blotting using rabbit anti-NUB1 antibody. As expected, the expression levels of these two proteins were the same (Fig. 7A, upper panel), suggesting that the binding ability of the C-terminal NUB1L (Glu-372 to Asn-615) was almost twice that of the C-terminal NUB1 (Glu-372 to Asn-601) in yeast cells. To confirm this finding, a GST pull-down assay was also performed. Poly-His-tagged full-length NUB1 and NUB1L were expressed in E. coli and precipitated with GST-NEDD8 immobilized on beads. As shown in Fig. 7B, GST-NEDD8 could precipitate NUB1L more effectively than NUB1 (lane 3 versus lane 6), supporting the result of liquid ␤-galactosidase assay.
Identification of NEDD8-binding Sites on NUB1 and NUB1L-We formulated two hypotheses to explain the increased ability of NUB1L to bind to NEDD8. The first was that

FIG. 4. Distribution of NUB1 and NUB1L in human tissues. In
A: primers for RT-PCR of NUB1/NUB1L. A is a forward prime, whose sequence is located between nucleotides 1242 and 1260 in the coding region of NUB1/NUB1L cDNA. B is a reverse primer, whose sequence is located between nucleotides 1440 and 1458 in NUB1 cDNA and between nucleotides 1482 and 1500 in NUB1L cDNA. C is a reverse primer, whose sequence is located only in NUB1L between nucleotides 1374 and 1395. The coding regions of NUB1 and NUB1L are represented by open boxes. A 42-bp insertion in NUB1L cDNA is indicated by a gray box. In B: detection of cDNA fragments of NUB1/NUB1L. The fragment of NUB1 cDNA (217 bp) was amplified by PCR using primers A and B. The fragment of NUB1L cDNA (154 bp) was amplified by PCR using primers A and C. In the PCR, cDNAs derived from various human tissues were used as template. After RT-PCR, the products were resolved on 1% agarose gel and stained with ethidium bromide. PCRs of NUB1 and NUB1L were performed by 35 cycles, whereas glyceraldehyde-3-phosphate dehydrogenase, which expressed much more than NUB1/NUB1L, was amplified by 30 cycles. The amplified fragment of glyceraldehyde-3-phosphate dehydrogenase cDNA is 400 bp in size.
an additional NEDD8-binding site is generated in NUB1L as a result of the insertion encoded by the alternatively spliced sequence. If this were true, NUB1L would have more binding sites with NEDD8 than NUB1 does, which would allow NUB1L to interact with more molecules of NEDD8. Our second hypothesis was that NUB1 and NUB1L have the same binding site for NEDD8, but the NEDD8-binding ability of the site in NUB1L is qualitatively increased by the insertion. To examine these hypotheses, we precisely identified the NEDD8-binding sites in both NUB1 and NUB1L. We narrowed down the NEDD8-binding areas using deletion mutants of NUB1 and NUB1L in a yeast two-hybrid interaction assay.
As shown in Fig. 8A, we generated eight mutants of NUB1, M1 to M8, to examine the interaction with NEDD8. Each mutant has a C-terminal deletion and/or an N-terminal deletion. For example, M1 has a C-terminal deletion from Lys-371 to Asn-601, resulting in the loss of two UBA domains (UBA1 and UBA3) and a PEST domain. M2 has an N-terminal deletion from Met-1 to Phe-370, resulting in the loss of a UBL domain. Using these mutants and a wild type of NUB1, we then examined the interaction with NEDD8-G in yeast cells. As noted earlier, NEDD8-G allowed us to examine the interaction of NUB1 mutants with NEDD8-G itself but not with target proteins of NEDD8 (1,17). In the yeast two-hybrid assay, NEDD8-G fused to the Gal4 activation domain was used for the interaction with a panel of NUB1 mutants fused to the Gal4 DNA-binding domain. As shown in Fig. 8A, NEDD8-G interacted with NUB1 wild-type (WT), NUB1 371-601 (M2), NUB1 461-601 (M6), and NUB1 515-601 (M8), but not with NUB1 1-370 (M1), NUB1 1-418 (M3), NUB1 371-418 (M4), NUB1 427-460 (M5), and NUB1 472-514 (M7). These results indicated that a NEDD8-binding site was located at the C terminus of NUB1 between amino acid residues 515 and 601. To pinpoint the NEDD8-binding site at the C terminus of NUB1, we generated five more mutants, M9 to M13, and examined their interaction with NEDD8-G using the yeast two-hybrid assay. As shown in Fig. 8B, NEDD8-G interacted with NUB1 515-584 (M9), NUB1 536 -601 (M11), and NUB1 536 -584 (M13) but not with NUB1 515-568 (M10) and NUB1 596 -601 (M12). These results indicated that the NEDD8-binding site was located at the C terminus of NUB1 between amino acid residues 536 and 584. Interestingly, this region contains the latter half of the PEST domain.
We also investigated the NEDD8-binding sites on NUB1L. Because the difference between NUB1L and NUB1 is the 14amino acid insertion that generates an additional UBA domain (UBA2), we focused on the interaction between NEDD8 and the newly generated UBA domain. As shown in Fig. 8A, we made two mutants of NUB1L with deletions. NEDD8-G interacted with NUB1L wild type (LWT), NUB1 371-615 (LM1), and NUB1L 427-474 (LM2), indicating that NUB1L had an additional NEDD8-binding site between amino acid residues 427 and 474.
In summary, NUB1 possesses a NEDD8-binding site at the C terminus, but NUB1L has an additional site for binding with NEDD8. As expected, the additional site in NUB1L was located at the UBA2 domain that includes the inserted sequence (Fig.  8C). Thus, we concluded that the greater ability of NUB1L to bind to NEDD8 originates with the additional binding site.
NEDD8-binding Motif in NUB1 and NUB1L-To predict amino acid residues that are responsible for NEDD8 binding, we aligned amino acid sequences of NEDD8-binding sites at the UBA2 domain and the C-terminal region obtained from human NUB1, human NUB1L, mouse NUB1 homologue, and cow NUB1 homologue (29). As shown in Fig. 9A, this alignment revealed that four amino acid residues, one Ala and three Leu, were conserved and found in the sequence A(X 4 )L(X 10 )L(X 3 )L.
To determine whether these conserved residues are important for binding with NEDD8, a yeast two-hybrid interaction assay was performed. We generated M13 mutants of C-terminal NUB1 that had either an Ala-to-Val substitution at Ala-557 or a Leu-to-Ala substitution at Leu-562, Leu-573, or Leu-577 (see Figs. 8B, 8C, and 9A). Furthermore, we generated LM2 mutants of the UBA2 domain of NUB1L that had an Ala-to-Val substitution at Ala-448 or a Leu-to-Ala substitution at Leu-453, Leu-464, or Leu-468 (see Figs. 8A, 8C, and 9A). The yeast two-hybrid assay revealed that the interaction of M13 and LM2 with NEDD8-G was strongly, but not completely, inhibited by all of these Leu-to-Ala substitutions (data not shown). In contrast, Ala-to-Val substitutions did not affect the interaction at all (data not shown). These results suggest that at least three Leu residues in the sequence A(X 4 )L(X 10 )L(X 3 )L are important for the interaction with NEDD8.
NEDD8-binding Ability of Full-length NUB1 and NUB1L with Leu-to-Ala Substitutions-To confirm the effect of these Leu-to-Ala substitutions on the interaction of NUB1 with NEDD8, we next performed a GST pull-down assay. In this assay, RH-tagged full-length NUB1 was expressed in bacteria and precipitated with GST or GST-NEDD8 immobilized on beads. The precipitates were then analyzed by Western blotting using anti-RH antibody to detect RH-NUB1. As in the yeast two-hybrid assay, we detected a weak interaction between NEDD8 and full-length NUB1 molecules with a single substitution of Leu-to-Ala (data not shown). To completely abolish the ability of NUB1 to bind to NEDD8, we substituted both Leu-573 and Leu-577 with Ala at the C-terminal region in full-length NUB1 (see Figs. 8C and 9A). As shown in the upper panel of Fig. 9B, the wild-type NUB1 could be precipitated with GST-NEDD8 (lane 3), but the mutant NUB1 with both L573A and L577A could not (lane 6), indicating that the mutant could not interact with NEDD8.  7. Quantitative study of interaction between NEDD8 and NUB1/NUB1L. A, yeast two-hybrid interaction of NEDD8 with the C-terminal NUB1 (NUB1c) and the C-terminal NUB1L (NUB1Lc). The yeast strain AH109 was transformed with pGADT7/NEDD8-G alone or transformed with pGADT7/NEDD8-G and pGBKT7 construct expressing HHR23B, NUB1c, or NUB1Lc and grown in liquid yeast culture. The strength of the interaction with NEDD8-G was determined by measuring ␤-galactosidase activity in six independent transformants (lower panel). In addition, the expression levels of Gal4-BD-fused NUB1c and NUB1Lc in yeast cells were examined by Western blotting using rabbit anti-NUB1 antibody (upper panel). NUB1c is the C-terminal fragment (Glu-372 to Asn-601) of NUB1. NUB1Lc is the C-terminal fragment (Glu-372 to Asn-615) of NUB1L. Note: NEDD8-G is a mutant of NEDD8, which has a C-terminal deletion from Gly-76 to Gln-81 and is unable to form conjugates with target proteins. Its use allowed us to examine the interaction with NEDD8 itself, not with the target proteins of NEDD8. B, in vitro interaction of NEDD8 with full-length NUB1 and NUB1L. GST and GST-NEDD8 were expressed in bacteria and purified by glutathione-Sepharose beads. Using these beads, recombinant RHtagged NUB1L and NUB1 were precipitated. The precipitates were analyzed by Western blotting with anti-RH antibody (upper panel) or anti-GST antibody (lower panel).
Next, NUB1L with Leu-to-Ala substitutions were examined. In this assay, RH-tagged full-length NUB1L was expressed in bacteria and precipitated with GST or GST-NEDD8 immobilized on beads. The precipitates were analyzed by Western blotting using anti-RH antibody to detect RH-NUB1L. Because NUB1L possesses two NEDD8-binding sites, we generated three mutants of full-length NUB1L. The first mutant had two substitutions at the UBA2 domain (L464A and L468A). The second mutant had two substitutions at the C-terminal region (L587A and L591A). The third mutant had four substitutions, two at the C-terminal region (L587A and L591A) and two at the UBA2 domain (L464A and L468A) (see Figs. 8C and 9A). Using the wild type and these mutants of the full-length NUB1L, a GST pull-down assay was performed. As shown in the upper panel of Fig. 9C, GST-NEDD8 precipitated the wild type (lane 3), the first mutant with two substitutions at the UBA2 domain (lane 6), and the second mutant with C-terminal substitutions (lane 9), but it did not precipitate the third mutant with four substitutions at both the UBA2 domain and the C-terminal region (lane 12). This result indicated that the first and second mutants could still interact with NEDD8 through the intact C-terminal region or the intact UBA2, but this ability was totally abolished in the third mutant as a result of the four substitutions.
NEDD8-buster Function of NUB1 and NUB1L Lacking NEDD8-binding Ability-In the previous section, we demonstrated that the interaction of NUB1 and NUB1L with NEDD8 was abolished by Leu-to-Ala substitutions in their C-terminal region and UBA2 domain in vitro. In this section, we examined whether these substitutions affected the function of NUB1 and NUB1L in vivo. First, we focused on the C-terminal substitutions of L573A and L577A in NUB1. An RH-epitope was tagged to wild-type NUB1 and its mutant with substitutions of L573A and L577A. These RH-tagged proteins and empty vector (control) were coexpressed with FLAG-tagged NEDD8 in COS cells. Total cell lysate was then prepared from the COS transfectants and analyzed by Western blotting using anti-FLAG antibody to detect the expression of FLAG-NEDD8. As shown in Fig. 10A, the overexpression of RH-tagged wild-type NUB1 caused a dramatic reduction in the FLAG-NEDD8 monomer and its conjugates (lane 2 versus lane 3). In contrast, the overexpression of the RH-tagged NUB1 mutant did not cause any reduction in the FLAG-NEDD8 monomer and its conjugates (lane 2 versus lane 4), suggesting that the C-terminal interaction of NUB1 with NEDD8 was required for its NEDD8-buster function to occur.
In a second experiment, we examined how the function of NUB1L was affected by substitutions in the UBA2 domain (L464A and L468A) and/or substitutions in the C terminus (L587A and L591A). We first generated three mutants of NUB1L. To abolish the interaction of its UBA2 domain with NEDD8, we generated a mutant with substitutions of L464A and L468A. To abolish the interaction of its C-terminal region with NEDD8, we generated a mutant with substitutions of L587A and L591A. To abolish the interaction of both sites with NEDD8, we generated a mutant with substitutions of L464A, L468A, L587A, and L591A (see Figs. 8C and 9A). Using the wild-type NUB1L and these NUB1L mutants, their function was then examined. The proteins were tagged with RH-epitope to allow detection. These RH-tagged proteins and empty vector (control) were then coexpressed with FLAG-tagged NEDD8 in COS cells. Total cell lysate was prepared from the COS trans- A, interaction of NEDD8-G with mutant NUB1 and NUB1L in yeast two-hybrid system. The yeast strain AH109 was transformed with pGADT7/NEDD8-G and pGBKT7 construct to express wild-type NUB1 (WT), mutant NUB1 (M1-8), wild-type NUB1L (LWT), or mutant NUB1L (LM1 and LM2). Transformed yeast cells were grown on His Ϫ / Trp Ϫ /Leu Ϫ synthetic agar plate for 3 days at 30°C. The specific proteinprotein interaction was determined by the growth of the cells on the selection plate. B, interaction of NEDD8-G with C-terminal NUB1 mutants in yeast two-hybrid system. The yeast strain AH109 was transformed with pGADT7/NEDD8-G and pGBKT7 construct to express mutant NUB1 (M8 -13). Transformed yeast cells were grown on His Ϫ /Trp Ϫ /Leu Ϫ synthetic agar plate for 3 days at 30°C. The specific protein-protein interaction was determined by the growth of the cells on the selection plate. C, location of NEDD8-binding sites on NUB1 and NUB1L, indicated by arrows. fectants and analyzed by Western blotting using anti-FLAG antibody to detect the expression of FLAG-NEDD8. As shown in the upper panel of Fig. 10B, the mutant NUB1L with two substitutions (L464A and L468A) in the UBA2 domain caused a severe reduction in the FLAG-NEDD8 monomer and its conjugates (lane 2 versus lane 4) as well as the wild-type NUB1L FIG. 9. In vitro interaction of NEDD8 with full-length NUB1 and NUB1L possessing Leu-to-Ala substitutions. A, amino acid sequence alignment of NEDD8-binding sites in NUB1/NUB1L of human, mouse, and cow. The amino acid sequences were deduced from the cDNA sequences, which were isolated in our laboratories. The accession numbers in GenBank TM are: human NUB1, AF300717; human NUB1L, AF459743; mouse NUB1 homologue, AF534114; and cow (Bos taurus) NUB1 homologue, AF514279. Arrowheads indicate the amino acid residues conserved in the sites for NEDD8 binding. The conserved residues are found in the sequence A(X 4 )L(X 10 )L(X 3 )L. B, GST pull-down assay for the interaction between NEDD8 and NUB1 possessing Leu-to-Ala substitutions. Wildtype NUB1 (lanes 1-3) and mutant NUB1 with L573A and L577A substitutions (lanes 4 -6) were tagged with RH-epitope and expressed in bacteria. The bacterial lysates were precipitated by GST (lanes 2 and 5) or GST-NEDD8 (lanes 3 and 6). The precipitates were analyzed by Western blotting using anti-RH antibody to detect RH-NUB1 (upper panel) and anti-GST antibody to detect GST or GST-NEDD8 (lower panel). C, GST pull-down assay for the interaction between NEDD8 and NUB1L possessing Leu-to-Ala substitutions. The RH-epitope was tagged to wild-type NUB1L (lanes 1-3), mutant NUB1L with L464A and L468A substitutions (lanes 4 -6), mutant NUB1L with L587A and L591A substitutions (lanes 7-9), and mutant NUB1L with L464A, L468A, L587, and L591A substitutions (lanes 10 -12). These RH-tagged proteins were expressed in bacteria and precipitated by GST (lanes 2, 5, 8, and 11) or GST-NEDD8 (lanes 3, 6, 9, and 12) from the bacterial lysates. The precipitates were analyzed by Western blotting using anti-RH antibody to detect RH-NUB1L (upper panel) and anti-GST antibody to detect GST or GST-NEDD8 (lower panel).  6). These results suggested that the C terminus is mainly involved in NEDD8-buster function through its interaction with NEDD8. DISCUSSION We previously reported our findings of a novel NEDD8-binding protein, NUB1, that recruits NEDD8 monomer and its conjugates to the proteasome for degradation (16,17). Recently, we identified a splicing variant of human NUB1, termed NUB1L. This isoform had a short insertion of 14 amino acids coded by an alternatively spliced mRNA of exon 12. Interestingly, this insertion generated an additional UBA domain between the two original UBA domains in NUB1L. Thus, NUB1 has two UBA domains, whereas NUB1L has three UBA domains, which are conserved in other species. So far, no NUB1 homologue possessing two UBA domains has been reported in other species.
Structural study revealed that NUB1 has a NEDD8-binding site at the C terminus, whereas NUB1L has an additional NEDD8-binding site at the newly generated UBA domain, UBA2. Comparing the amino acid sequences of the C terminus and the UBA domain in human, mouse, and cow, a NEDD8binding motif could be formulated. It appears that the NEDD8binding site contains the sequence, A(X 4 )L(X 10 )L(X 3 )L. Based on the results of yeast two-hybrid assay and GST pull-down assay using substitution mutants of NUB1 and NUB1L, three Leu residues in A(X 4 )L(X 10 )L(X 3 )L are important for the NEDD8 binding. The requirement of an Ala residue in this motif is unclear. Although an Ala to Val substitution had no effect on the NEDD8 binding, it might result from the high similarity between Ala and Val residues. Taken together, at least three Leu residues are required for the NEDD8 binding in NUB1/NUB1L.
Functional study established that NUB1L has the activity to down-regulate NEDD8 expression as well as NUB1. Repeated experiments showed that the down-regulation of NEDD8 by NUB1L is slightly greater than that by NUB1. The mutational study using a wild-type NUB1L and its three mutants revealed that the wild-type NUB1L and the first mutant, which lacked the ability to interact with NEDD8 at the UBA2 domain, showed a high NEDD8-buster activity. The second mutant, which lacked the ability to interact with NEDD8 at the C terminus, and the third mutant, which lacked the ability to interact with NEDD8 at both sites, showed almost no activity. These results suggested that the down-regulation of NEDD8 is mainly contributed by the NEDD8-binding site at the C terminus in NUB1L. Although the UBA2 domain also has an affinity to NEDD8, it minimally or does not contribute to the downregulation of NEDD8. The NEDD8-binding ability at the UBA2 domain might be required for an unknown function of NUB1L.
We previously detected two major messages of 3.1-3.5 kb and 2.3-2.7 kb in the Northern blot analysis for NUB1 isoforms (16). Although we did not identify these mRNAs, the longer message might be the mRNAs of Clone 1 and Clone 3 that encode NUB1 and NUB1L, respectively, and the shorter message might be the mRNA of Clone 2 that encodes NUB1 (Fig. 2).
In this study, we demonstrated that the tissue distribution of NUB1L message is different from that of NUB1 message. The alternative splicing might control the balance between NUB1 and NUB1L and hence regulate the NEDD8 conjugation system in each tissue.