Address correspondence and reprint requests to Hiroyuki Nawa, Division of Molecular Neurobiology, Brain Research Institute, Niigata University, Asahimachi-dori 1–757, Niigata 951–8585, Japan. E-mail: email@example.com
Ubiquitin (Ub) ligation is implicated in active protein metabolism and subcellular trafficking and its impairment is involved in various neurologic diseases. In rat brain, we identified two novel Ub ligases, Momo and Sakura, carrying double zinc finger motif and RING finger domain. Momo expression is enriched in the brain gray matter and testis, and Sakura expression is more widely detected in the brain white matter as well as in many peripheral organs. Both proteins associate with the cell membranes of neuronal and/or glial cells. We examined their Ub ligase activity in vivo and in vitro using viral expression vectors carrying myc-tagged Momo and Sakura. Overexpression of either Momo or Sakura in mixed cortical cultures increased total polyubiquitination levels. In vitro ubiquitination assay revealed that the combination of Momo and UbcH4 and H5c, or of Sakura and UbcH4, H5c and H6 is required for the reaction. Deletion mutagenesis suggested that the E3 Ub ligase activity of Momo and Sakura depended on their C-terminal domains containing RING finger structure, while their N-terminal domains influenced their membrane association. In agreement, Sakura associating with the membrane was specifically palmitoylated. Although the molecular targets of their Ub ligation remain to be identified, these findings imply a novel function of the palmitoylated E3 Ub ligase(s).
The RING finger structure, named after a peculiar amino acid sequence found in ‘Really Interesting New Gene’, represents the zinc binding motif of Cys-X2-Cys-X9/39-Cys-X1/3-His-X2/3-Cys/His-X2-Cys-X4-48-Cys-X2-Cys (Lovering et al. 1993). At present, more than 100 members of this family contain this structure and most of them are suggested to function as a ubiquitin (Ub) ligase (E3) (Lorick et al. 1999). The RING finger proteins such as APC11 and Rbx1/ROC1 were initially isolated as oncogenes and probed to be E3 enzymes in the APC/C and SCF complexes (Tyers and Willems 1999). Although morphologic neuropathology indicates that brain neurons or glia contain Ub-containing inclusion bodies in a variety of neurologic diseases, biological implication of Ub accumulation in these diseases remains to be characterized (DiFiglia et al. 1997; Alves-Rodrigues et al. 1998; Hardy and Gwinn-Hardy 1998; Cummings et al. 1999). Recent progresses in molecular neurology have revealed that some neurodegenerative diseases involve impaired protein ubiquitination and abnormal proteolysis leading to neuronal death. Parkin, identified as a gene product responsible for familial Parkinsonism, contains RING finger domain and functions as an E3 Ub ligase in the nervous system (Kitada et al. 1998; Shimura et al. 2000; Chung et al. 2001b). These observations suggest that the Ub ligase activity of RING finger proteins may contribute to the pathogenesis of many neurodegenerative diseases.
The molecular roles of protein ubiquitination underlying oncogenesis have been intensively investigated and are correlated with catabolism of the regulatory proteins of cell growth and/or the cell cycle (Kamura et al. 1999; Joazeiro et al. 1999; Skowyra et al. 1999; Zachariae and Nasmyth 1999; Honda and Yasuda 2000; Ruffner et al. 2001). Although there are many unique membrane proteins in the nervous system, only a few E3 ligases corresponding to the substrates have been identified (i.e. Parkin, Siah and NEDD4). Siah is an enzyme that transfers Ub to the netrin-1 receptor, DDC (deleted in colorectal cancer), and is suggested to control degradation of the metabotropic glutamate receptor 1 (Hu and Fearon 1999; Ishikawa et al. 1999). NEDD4 is another member of the Ub ligase family that shares a common protein motif named HECT (homologous to E6AP C-terminus) and ubiquitinates amiloride-sensitive sodium channels (Staub et al. 1997). Although most of the targets for Ub ligases localize at special subcellular membrane components to exert their unique functions, the interaction of the Ub ligase with membranes is often indirect and mediated by other molecules (Kavsak et al. 2000; Shenoy et al. 2001; Soubeyran et al. 2002). Thus, it remains to be determined how membrane-bound neural proteins are ubiquitinated and catabolized in neurons and glia, and whether the dysfunction of these processes is associated with human neurodegenerative diseases.
In the present study, we have characterized the novel E3 Ub ligase and its homologue that both associate with neural cell membranes. Their distributions in the brain as well as their enzyme activities as a Ub ligase are examined and discussed.
Materials and methods
cDNA library screening
A Sal I-EcoR I genomic fragment spanning 75 bp (849–923 nt; AATCCTGGCTCGGAATTTTGTCAACTATTCTGGCTGCTGTGAAAAGTGGGAGCTGGTGGAGAAAGTCAACCGGCT) was isolated as a part of the down-stream gene by monitoring the binding activity of the neuronal transcription factor HIT-4 (Hirano et al. 2001). Subsequently, 300 000 plaques of the lambda Zap library carrying rat hippocampal cDNA (Stratagene, La Jolla, CA, USA) were screened against the genomic sequence. Eleven clones were isolated and subjected to automatic DNA sequencing (ABI 7700; Applied Bio Instruments, Foster City, CA, USA). All positive clones contained the same sequence and one clone contained a 1952-bp cDNA corresponding to a full open reading frame of Momo. To obtain cDNA homologues to the Momo gene, the same library was screened again at a lower stringency with the entire cDNA fragment for Momo. Accordingly, we obtained its homologue covering the entire translation frame and named the clone Sakura. The nucleotide sequence of Sakura cDNA was 2808 bp long.
PolyA + RNA was extracted and purified from adult male Sprague–Dawley rats (SLC, Shizuoka, Japan) using the acid guanidium–phenol–chloroform method (Chomczynski and Sacchi 1987). RNA samples were denatured in the presence of 50% formamide and 6% formaldehyde, separated on a 1.5% formaldehyde-agarose gel, and transferred onto a nylon membrane (Pall Scientific, East Hills, NY, USA). A 32P-labeled cDNA probe was generated using the Random primed DNA labeling kit (Boehringer Mannheim). The probe (2 × 106 cpm/mL) was hybridized to filters for 20 h at 42°C in 50% formamide, 5 × sodium chloride/sodium citrate buffer (SSC), 5 × Denhardt's solution and 1% sodium dodecyl sulfate (SDS) followed by washing with 0.1 × SSC, 0.1% SDS at 60°C, and exposure to films.
In situ hybridization
Sense oligoDNAs (AGCACTGGTCCATTCAGGTTTACACCGAGTTCTGACTTTCCTACCTAC for Momo, ACAAGTCACCTCTGTGTTAGCCCAGGATCAGGAAACTCAGCAGGCCATT for Sakura) and antisense oligoDNAs (GTAGGTAGGAAAGTCAGAACTCGGTGTAAACCTGAATGGACCAGTGCT for Momo, AATGGCCTGCTGAGTTTCCTGATCCTGGGCTAACACAGAGGTGACTTGT for Sakura) were selected not to have any significant homology to other genes and labeled with terminal deoxynucleotide transferase (Toyobo Biotech, Osaka, Japan) and [α-35S]dATP (New England Nuclear, Japan, ∼3000 Ci/mmol) (Katagiri et al. 1993). Brain sections were incubated with the hybridization solution [50% formamide, 5 × SSC, 5 × Denhardt's solution, 1 mm EDTA, 0.1 m dithiothreitol (DTT) and 0.5 mg/mL denatured salmon sperm DNA] followed by the solution containing [35S]-labeled oligoDNA (4 × 106 cpm/mL). After overnight hybridization at 42°C, sections were washed with 2 × SSC/4 mm DTT followed by 0.1 × SSC/4 mm DTT and dehydrated with ethanol. Sections were first processed in a BAS 2000 phosphoimager (Fuji Film, Tokyo, Japan) and then exposed to BioMax™ MS film (Eastman Kodak, Rochester, NY, USA).
Recombinant fusion proteins of the N-terminal portion of Momo (1–93 amino acids) and Sakura (1–103 amino acids) were produced in an Escherichia coli BL21 strain from a prokaryotic expression vector, pGEX–2T (Pharmacia, Uppsala, Sweden). Cell lysate from E. coli was purified with a gluthathione-conjugated affinity column (Pharmacia) and subjected to SDS–polyacrylamide gel elecrophoresis (PAGE). An induced protein band (36 kDa for Momo or 37 kDa for Sakura) was recovered and emulsified with Freund's complete adjuvants and used to immunize rabbits. Serum batches showing higher titers were subjected to antigen-affinity chromatography. The affinity column had a bed of Affi-Gel 10 (1 mL; Bio-Rad Laboratories, Hercules, CA, USA) that was coupled to synthetic peptides (1 mg; H-GAVRGQQSAFAGATGPFRFTPN-OH for Momo or H-LDGQPEEVPPPQGARMQAYSNPG-OH for Sakura), or the recombinant proteins used as the antigen. The peptide sequences were derived from their human orthologues (GenBank #AAM29180 for Momo and #AAM29181 for Sakura). As the antigen-affinity purified antibodies cross-reacted with many unidentified proteins (data not shown), the peptide affinity-purified rabbit antibodies for Momo and Sakura were used for immunoblotting.
Subcellular fractionation and immunoblotting
Cerebral cortex and medulla, cerebellar cortex and medulla, and their whole tissues of postnatal day 9 rats were dissected under a macroscope (Olympus, Tokyo, Japan) and homogenized with 10 volumes of 0.25 m sucrose/phosphate-buffered saline (PBS) with a Potter-type homogenizer. Cultured cells were similarly homogenized in 0.25 m sucrose/PBS. Unbroken cells and nuclei were removed by centrifugation at 600 × g for 10 min at 4°C. To obtain a crude mitochondrial fraction (P2), the supernatant was centrifuged at 5000 × g for 10 min at 4°C. Supernatants were further separated into a microsomal fraction (P3) and a cytoplasmic fraction (S3). Both S3 and P3 fractions were denatured and separated by SDS–PAGE using 4–20% gradient gel (NEN, Tokyo, Japan), and then transferred to a nitrocellulose membrane (Schleicher and Schuell, Dassel, Germany). The membrane was incubated with the affinity-purified anti-Momo antibody (10 µg/mL), anti-Sakura antibody (2.5 µg/mL), anti-GM130 (5 µg/mL; BD Transduction Laboratories, Lexington, KY, USA), or anti-catalase (10 µg/mL of Cathepsin C; Nordic Immunological Laboratories, Tilburg, the Netherlands). Immunoreactivity was detected with a goat anti-rabbit immunoglobulin conjugated to peroxidase (1 : 10 000) followed by a chemiluminescence reaction combined with film exposure (Western Lighting Chemiluminescence Reagent Plus; Perkin Elmer, Boston, MA, USA).
Momo (1–381 amino acids) and Sakura (1–362 amino acids) cDNAs containing their entire coding regions were amplified by polymerase chain reaction (PCR) using KOD plus polymerase (Toyobo Biotech). The PCR products were inserted into the pCi vector (Promega Corp., Madison, WI, USA) carrying a myc-tag sequence at their amino-terminal end; pCi + myc-Momo and pCi + myc-Sakura. The N- and C-terminal portions of Momo and Sakura were deleted using the PCR technique and subcloned again into the pCi + myc vector; pCi + myc-ΔN-Momo and pCi + myc-ΔN-Sakura, and pCi + myc-ΔC-Momo and pCi + myc-ΔC-Sakura, respectively. The expression vector carrying FLAG-tagged Ub was also cotransfected into HEK293 cells; pcDNA3.1(+) FLAG-Ub (Shimura et al. 2000).
Modified Sindbis virus vector plasmid (pSinEGdsp#9) originated from pSinRep5 (Invitrogen, Carlsbad, CA, USA) as described previously (Kawamura et al. 2003). The pSinEGdsp#9 plasmid carried duplicated subgenomic promoters (Psg). The myc-tagged Momo and Sakura fragments were inserted into multiple cloning sites (Xba I-Mlu I) of the pSinEGdsp#9 vector to generate pSin + myc-Momo/EG and pSin + myc-Sakura/EG plasmids, respectively. Additionally the third Psg was inserted into pSinEGdsp#9 and used for the coexpression of FLAG-tagged Ub; pSin + myc-Momo/FLAG-Ub/EG and pSin + myc-Sakura/FLAG-Ub/EG. Viral genome RNA was transcribed using an Invitroscript™ Cap Kit (Invitrogen), and cotransfected with a helper mRNA (DH-26S) (Invitrogen) in BHK cells by electroporation (Gene Pulser, Bio-Rad). The titer of Sindbis virus vector was determined by monitoring EGFP-positive cells. Typically, the viral titer was greater than 108 infectious unit/mL.
Ubiquitination analysis in cells
Primary cortical cells were prepared from embryos of Sprague–Dawley rats (embryonic day 18) as previously described (Narisawa-Saito et al. 1999). The primary culture contained 5–10% glial population and were infected at a multiplicity of infection (MOI) of 1–2 for 1 h at 37°C. The cells were harvested at 22–24-h postinfection and lysed in solubilizing buffer [50 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1% NP-40, and protease inhibitors (2 µg/mL aprotinin, 0.5 µg/mL leupeptin, and 1 µg/mL pepstatin A)]. Alternatively, pCi vectors carrying Momo, Sakura, and their mutants were transfected into HEK293 cells using an alternative calcium phosphate method (Chen and Okayama 1988). The cells were harvested at 42–44-h post-transfection, and lysed with the above solubilizing buffer. During infection or transfection, cells were often treated with 50 µm MG132 (Peptide Inc. Osaka, Japan) for 6 h unless otherwise noted.
Cell lysates were sonicated using a Handy Sonic (Tomy Seiko Co., Ltd, Tokyo, Japan) and centrifuged at 12 000 g for 20 min at 4°C. Protein concentrations of the supernatants were determined using a micro-BCA Protein Assay Reagent (Pierce, Rockford, IL, USA). Each supernatant including 200 µg of protein was incubated with anti-myc rabbit polyclonal antibody (1 µg) and Protein G Sepharose beads (10 µL; Amersham Pharmacia, Buckinghamshire, UK). The immunoprecipitates were eluted from the beads by boiling in 2 × SDS sample buffer and subjected to immunoblotting. The membranes were probed with anti-FLAG (M2), biotin-conjugated anti-FLAG (M2) antibodies (both from Sigma), or an anti-myc monoclonal antibody (9E10; ATCC, Manassas, VA, USA). After washing the membranes, the immunoreaction was detected with the secondary antibodies or streptoavidin conjugated to horseradish peroxidase. Bands were visualized using Western Lighting Chemiluminescence Reagent Plus (Perkin Elmer).
In vitro ubiquitination assay
Semi-purified myc-Momo and myc-Sakura were obtained from cultured cortical neurons infected with Sindbis viral vectors carrying Momo and Sakura, respectively. Culture lysate (200 µg protein) was subjected to immunoprecipitation with 1 µg of anti-myc rabbit IgG-linked protein G Sepharose beads. The immunoprecipitate served as a putative E3 Ub ligase in an in vitro ubiquitination assay with an ATP regenerating system (Suzuki et al. 1999). In brief, 50 µL of reaction mixture contained 10 µL protein G Sepharose bead slurries of the immunoprecipitate, 50 mm Tris-HCl (pH 7.5), 10 mm MgCl2, 20 mm ATP, 0.5 mm DTT, 0.6 U/mL inorganic pyrophosphatase (Sigma), 5 µm MG132, 100 ng of purified recombinant mouse E1 (Ub activating enzyme) (Suzuki et al. 1999), 0.5–1 µm (His)6-tagged E2 enzyme (Murata et al. 2001), 1 mg/mL bovine Ub (Sigma) and 125I-labeled Ub (4 × 106 cpm). The mixture was incubated at 30°C for 2 h. After terminating the reaction by adding 3 × SDS sample buffer, 25 µL of the reaction mixture was separated by SDS–PAGE on a 4–20% gradient gel and visualized by autoradiography.
Producing bacterial recombinant Momo and Sakura, we also tested their E3 activity in vitro (Matsuda et al. 2001; Imai et al. 2003). cDNA for Momo and Sakura was subcloned to a prokaryotic expression vector, pMAL-p2 (New England BioLabs, Beverly, MA). Recombinant Momo and Sakura were produced in E. coli as a fusion protein of maltose binding protein, and purified with amylose resin (New England BioLabs). In vitro ubiquitination assay was carried out similarly in the presence of E1 and UbcH4 (E2).
The HEK293 cells were metabolically labeled with [3H]palmitic acid as described previously (Topinka and Bredt 1998). In brief, 40 h after transfection, cells were labeled with 1 mCi/mL [3H]palmitic acid (50 Ci/mmol; New England Nuclear) for 5 h in the presence of cerulenin (2 µg/mL; Sigma) and fatty acid-free bovine serum albumin (10 µg/mL; Sigma). Proteins in the cell suspension (5%) were precipitated with a 50% v/v concentration of acetone to monitor whole [3H]palmitic acid incorporation. The remaining cell suspension (95%) was homogenized in 0.25 m sucrose/PBS by passing through a 25-gauge needle. Unbroken cells and nuclei were removed by centrifugation at 600 × g for 10 min at 4°C. To obtain a soluble fraction (S) and a crude membrane fraction (P), the supernatant was centrifuged at 100 000 × g for 60 min at 4°C. The P fraction was solubilized with RIPA buffer [50 mm Tris-HCl (pH 7.5), 150 mm NaCl, 1% Triton X-100, 1% sodium deoxycholate and protease inhibitors]. Both S and P fractions (200 µg) were subjected to an immunoprecipitation reaction with 1 µg of anti-myc rabbit polyclonal antibody followed by western blotting or autoradiography. For autoradiography, an SDS–PAGE gel was treated with Amplify (Amersham Pharmacia) and exposed to Hyperfilm™ MP (Amersham Pharmacia) at − 80°C.
Novel primary structures of Momo and Sakura and their mRNA expression
We isolated a novel RING finger protein, Momo, and its structural homologue named Sakura from the rat cDNA library (Fig. 1a). Momo cDNA (2224 bp) contained the largest open reading frame, which encodes 381 amino acid residues (42695 Da), while Sakura cDNA (3107 bp) encoded a protein of 362 amino acid residues (40407 Da) (GenBank accession #AY157968 for rat Momo and #AY157969 for rat Sakura). The overall nucleotide homology between Momo and Sakura was 62% and their amino acid identity was 46%. The Pfam domain search program revealed that Momo and Sakura both contain two distinct zinc finger domains; consecutive two C2C2 zinc finger motifs (Klug and Schwabe 1995) and a RING finger motif (Freemont 2000). The duplicated C2C2 zinc finger motifs are located at the amino terminal region and the RING finger motif at the carboxyl terminal region. In addition, there is a putative site(s) for protein palmitoylation at most N-terminals of both Momo and Sakura. The double cysteine residues (Cys-Cys, Cys-X-Cys, or Cys-X-X-Cys) flanked by hydrophobic amino acids represent a potential palmitoylation signal (Mumby 1997). A GenBank search revealed that the Drosophila CG17019-PA gene has a 48% amino acid homology to Sakura over the entire coding region (Adams et al. 2000). Thus, Sakura is thought to be a mammalian orthologue for the Drosophila CG17019-PA gene, although a Drosophila mutant for the CG17019-PA gene has been not isolated. Their structural similarity to the RING finger proteins suggests that Momo and Sakura might function as an E3 Ub ligase.
To study mRNA distribution of Momo and Sakura, we performed northern blotting among brain regions and peripheral tissues examined in rats, a signal for Momo mRNA was detected at the 2.6-kb position and weakly at 5-kb position in the brain and testis (Fig. 1b). In contrast, Sakura mRNA appeared as multiple sizes of 1.6, 3.8 and 7 kb on northern blots. As we employed stringent washing conditions for blotting, these bands presumably reflected mRNA species produced from a single gene through alternative splicing or differential poly A addition. Relatively higher signals for Sakura mRNA were detected in the brain, liver, kidney, lung, and spleen. Only in the testis, the major signal for Sakura mRNA appeared as a 1.6-kb band. In contrast to the uniform distribution of Momo mRNA in the brain, Sakura mRNA was most enriched in the midbrain, including the pituitary and in the brain stem regions. In situ hybridization analysis revealed a more precise distribution in the brain: a stronger signal for Momo mRNA was detected in the pituitary as well as in the molecular layer of the cerebellar cortex (Fig. 1c). In contrast, Sakura mRNA was detected in the corpus callosum, brain stem, spinal cord and cerebellar white matter.
Cellular distributions of Momo and Sakura proteins
To confirm the distributions of Momo and Sakura in neurons or glial cells, protein lysates from cerebellar gray matter and white matter were prepared and subjected to western blotting (Fig. 2). When Momo and Sakura genes were overexpressed in cortical cultures containing neurons and glial cells, Momo immunoreactivity appeared as four bands around 38–45 kDa, and the Sakura immunoreactivity as two bands around 46 kDa and 40 kDa (Fig. 2a,b; left). Some bands exhibited higher or lower mobility in a gel in comparison with their calculated molecular sizes. The mobility shift might reflect their post-translational modifications including palmitoylation, myristylation, and proteolytic processing (Kahns et al. 2002). Momo immunoreactivity was most enriched in the gray matter fractions of the cerebral cortex and cerebellar cortex while Sakura immunoreactivity was recovered mainly in the white matter fractions. The sizes of the endogenous Momo and Sakura immunoreactivities matched those in positive controls. Momo and Sakura lack a transmembrane region and a typical signal peptide motif for membrane anchoring, but they both carry putative palmitoylation site(s) in the N-terminal region to interact with the membrane via a fatty acid moiety (Resh 1999). The membrane (P3) and cytoplasmic (S3) fractions were prepared from the cerebellum. Western blotting of each fraction revealed that the immunoreactivity for both Momo and Sakura was enriched in the membrane fraction. Almost none of the Sakura immunoreactivity was recovered in the cytosolic fraction.
Ubiquitination induced by overexpression of Momo and Sakura in cortical culture
Using rat cortical cultures containing both neurons and glial cells, we assessed their biological activity as a Ub ligase. Myc-tagged Momo and/or FLAG-tagged Ub were overexpressed in cultured cortical cells using the Sindbis virus expression system (Kawamura et al. 2003) (Fig. 3a). Incorporation of FLAG-tagged Ub into total protein was monitored by western blotting using the anti-FLAG antibody (Fig. 3b). When cortical cultures were transfected with the viral vector carrying FLAG-tagged Ub alone, basal ligation of the tagged-Ub was only detected in the presence of a proteasome inhibitor, MG132, as observed by broad bands. There was no apparent influence of the Momo expression on total ubiquitination levels. We also examined Momo-dependent ubiquitination in the immune complex of an anti-myc antibody (Fig. 3c). The overexpression of myc-tagged Momo induced polyubiquitination of target protein(s) in the Momo-carrying immune complex. The ubiquitination was less pronounced in the absence of the proteasome inhibitor, MG132. Although cortical neurons endogenously expressed Momo, basal reaction of ubiqutination was almost undetectable because the strength of basal Momo expression was less than one hundredth of that by Sindbis virus-mediated overexpression (data not shown). Thus, these results suggest that Momo is involved in polyubiquitination.
When myc-tagged Sakura was similarly expressed by a Sindbis virus expression vector together with FLAG-tagged Ub, a similar ubiquitination reaction in the absence of the MG132 was observed in total ubiquitination as well as in the immune complex of Sakura (Fig. 4). In contrast to the reaction with Momo, however, marked amounts of polyubiquitinated protein(s) induced by the Sakura expression were not influenced by the presence or absence of the inhibitor, MG132. The results suggest that the ubiquitination by Sakura is relatively stable and might not be recognized by the conventional MG132-sensitive 26S proteasome (Hershko and Ciechanover 1998).
In vitro ubiquitination of Momo and Sakura and their interaction with various E2 components
We attempted to reconstitute the ubiquitination in vitro with a purified recombinant Ub activating enzyme (E1) and various Ub conjugating enzymes (E2; Ubc) as well as Momo or Sakura as a Ub ligase (E3). Each UbcH3, H4, H5c, H6, and H8 was tested as an E2 enzyme and ubiquitination reaction was monitored in vitro in the presence of 125I-labeled Ub. The combination of immunoprecipitated myc-tagged Momo with UbcH4 and H5c, but not with UbcH3, H6, nor H8, triggered ubiquitination (Fig. 5a). The immune complex of myc-tagged Momo alone failed to induce ubiquitination (data not shown). Western blotting for the anti-myc antibody suggested that the polyubiquitination might include self-ubiquitination of Momo as the amounts of authentic Momo were reduced and its size shifted to higher molecular ranges, although the shift was not apparent in the ubiquitination assay using cultured neural cells (see Fig. 3c). In contrast, myc-tagged Sakura recruited UbcH4, H5c and H6 to induce polyubiquitination (Fig. 5c). In contrast to the potential self-ubiquitination of Momo, the Sakura-induced self-ubiquitination was not apparent in the western blots with the anti-myc antibody. In the absence of Momo or Sakura, however, there was no endogenous ubiquitination (Fig. 5b).
To avoid the possibility that the immune complexes carried not only Momo or Sakura but also other E3 enzyme(s) and thus caused the in vitro ubiqutination, we attempted to reconstitute the reaction using purified recombinant Momo or Sakura (Fig. 5d). Momo and Sakura proteins were produced in E. coli as a fusion protein of maltose binding protein. Incubation of the recombinant Momo and Sakura with the purified E1 enzyme and UbcH4 similarly caused their self-ubiquitination as monitored by immunoblotting using the anti-maltose binding protein antibody. However, we cannot rule out the possibility that the self-ubiquitination might result from their overexpression or the lack of proper substrates, as we failed to detect the self-ubiquitination in the brain or neural cells.
Dependency of ubiquitination on C-terminal RING finger
The structural dependency of Momo- and Sakura-mediated ubiquitination was examined in a human embryonic kidney cell line, HEK293. We produced deletion mutants of myc-tagged Momo and Sakura, omitting their N-terminal portion containing the palmitoylation sites/double zinc finger domain (ΔN) or their C-terminal region containing the RING finger domain (ΔC) (Fig. 6a). The transfection of cDNA for wild-type Momo and Sakura as well as that for mutants resulted in a significant level of protein expression in HEK293 cells, although the N-terminal deletion appeared to reduce their stability (Fig. 6b). The transfection of wild-type Momo cDNA together with the FLAG-tagged Ub increased Ub immunoreactivity as revealed with the anti-FLAG antibody (Fig. 6c). The deletions in the N-terminal as well as in the C-terminal of Momo significantly diminished the Ub immunoreactivity in the ubiquitination reaction in culture. In contrast, while the C-terminal deletion mutant of Sakura almost completely abolished the ubiquitination reaction, its N-terminal deletion only influenced the sensitivity of the Sakura-induced ubiquitination to the proteasome inhibitor, MG132, but not the ubiquitination itself (Fig. 6d). These results confirm that the RING finger domain of Momo and Sakura has a primary role in ubiquitination.
Membrane association via the N-terminal domain of Momo and Sakura
Using the above deletion constructs, we also monitored subcellular localization of the Momo and Sakura mutant proteins. Wild-type Momo and Sakura as well as the mutant proteins were overexpressed in HEK293 cells, and subsequently the cells were fractionated into crude membrane and cytoplasmic fraction (Fig. 7a). Subcellular fractionation was confirmed by immunoblotting for a Golgi-associated protein, GM130, and a cytoplasmic enzyme, catalase. The cytoplasmic fraction contained larger amounts of catalase while the membrane fraction carried predominantly higher levels of GM130 in all preparations. Synthesized wild-type myc-tagged Momo and Sakura had similar molecular sizes to those in neural culture (see Fig. 2a,b) and were recovered in both the membrane and cytoplasmic fractions under the fractionation condition. In contrast, when cells expressing the Momo and Sakura mutants lacking the N-terminal domain were subjected to the fractionation, these mutant proteins were recovered predominantly in the cytoplasmic fraction. The C-terminal deletion mutants of Momo and Sakura exhibited the same subcellular distributions as their wild types (data not shown). These data indicate that the N-terminal domain of Momo and Sakura was responsible for their membrane association. The cytoplasmic recovery of endogenous Momo and Sakura was quite low in the tissue preparation, however (see Fig. 2). The higher cytoplasmic recovery in HEK293 cells might result from their overexpression in the heterologous non-neural cells.
Palmitoylation of Sakura
Both Momo and Sakura carry the putative signal sites for palmitoylation. To confirm their palmitoylation, we transfected the eukaryotic expression vectors carrying myc-tagged Momo and Sakura cDNAs into HEK293 cells and then allowed cells to incorporate [3H]palmitic acid. HEK293 cells were subjected to the above subcellular fractionation and the crude membrane fraction was solubilized with a mild detergent of Triton X-100 (1%). Myc-tagged Sakura in the cytoplasmic and membrane fractions was immunoprecipitated with the anti-myc antibody, and the immunoprecipitates were subjected to SDS–PAGE followed by autoradiography or immunoblotting for the anti-myc antibody. [3H]palmitic acid incorporation into the protein fraction was confirmed in the presence of an inhibitor for endogenous fatty acid synthesis (Fig. 8a). Film autoradiography revealed a protein band in the membrane fraction that was linked to [3H]palmitic acid and matched the size of Sakura (Fig. 8b). Although immunoblotting confirmed the presence of myc-tagged Sakura both in the membrane and cytoplasmic fractions, only membrane-bound Sakura appeared to be palmitoylated (Fig. 8c). In contrast, we failed to obtain any significant palmitoylation signal for myc-tagged Momo (data not shown). As the expression of Momo was quite restricted in neurons, palmitoylation of Momo might not be completed in the non-neuronal cells. Alternatively, the myc tagging might cause steric hindrance against palmitoylation of Momo. The palmitoylation of Momo remains to be examined with a different approach.
We characterized novel membrane-associating E3 Ub ligases with a RING finger domain, Momo and Sakura. Both E3 Ub ligases are structurally homologous to each other but had distinct distributions in the brain as well as in HEK293 cells. Momo mRNA and protein expression was pronounced in the cerebellar gray matter and Sakura mRNA and protein expression was more limited to the cerebellar white matter and lower brain stem. The mRNA and protein distribution patterns suggest that neurons and oligodendrocytes are likely to express Momo and Sakura, respectively. Although molecular targets of these Ub ligases remain to be identified, these enzymes appear to have a primal role in the nervous system. Structural analysis revealed several unique functional domains; putative palmitoylation sites, double zinc finger structure and RING finger domain. Consistent with possession of the RING finger domain, Momo and Sakura exhibited Ub ligase activity in vitro. In particular, self-ubiquitination of Momo and Sakura was detected in vitro assay as reported in other E3 ligases (Fang et al. 2000; Yang et al. 2000).
The RING finger sequence of Momo exhibited a 42% amino acid identity and 64% similarity to that of IAP (inhibitor of apoptosis) (Strausberg 2002), and that of Sakura 51% identity and 70% similarity to that of IAP. Both Momo and Sakura, however, lack the BIR domain of IAP, which interacts with caspases (Deveraux et al. 1997; Roy et al. 1997). A human orthologue of Momo was recently isolated and termed as hRFI (human RING Finger homologue to IAP type) and its anti-apoptotic activity was examined based on its structural similarity to IAP (Sasaki et al. 2002). In contrast, we failed to detect an influence of Momo or Sakura overexpression on the cell survival of neurons (data not shown). Momo and Sakura also carry C2C2 zinc finger motifs at their N-terminal portion, which contain common cysteine repeats among B-box, RING finger, and other zinc fingers. According to the amino acid alignment therefore the zinc finger region of Momo shares a 32% identity and that of Sakura shares a 30% identity to the cysteine finger, so-called FYVE, of human EEA1 (Mu et al. 1995). Both zinc finger regions of Momo and Sakura, however, lack essential RRHH amino acid residues in the core of the FYVE domain (Stenmark and Aasland 1999). Accordingly, we classified this domain as a conventional C2C2 zinc finger structure. Mutation analysis revealed that the N-terminal domain of Momo and Sakura influenced the ubiquitination reaction. In addition, the lack of a C-terminal RING finger in Momo did not fully inhibit the reaction. In this context, we cannot rule out the possibility that the N-terminal zinc finger structure functions as an E3 Ub ligase. Alternatively, this zinc finger domain might be involved in protein–protein interactions to associate with other components (Kuroda et al. 1996; Rodgers et al. 1996; Lyngso et al. 2000).
Although the indirect interaction of E3 Ub ligases with cell membranes via substrates or other associating proteins is often reported (Kavsak et al. 2000; Shenoy et al. 2001; Soubeyran et al. 2002; Plant et al. 2000), information on the direct membrane anchor of E3 Ub ligases is quite limited. Among various RING finger-type E3 Ub ligases, Sakura is the first reported E3 molecule that is covalently linked to palmitic acid. The palmitoylation is likely to contribute to the membrane binding capability of Sakura. An interesting regulation of the ubiquitination reaction was reported on another type of E3 Ub ligase with respect to membrane association. The E3 Ub ligase, NEDD4, carries the HECT domain in place of the RING finger and exerts Ub ligase activity on amiloride-sensitive epithelial sodium channels (Staub et al. 1997). This HECT-type Ub ligase contains C2 domains that interact with phosphatidyl serine in a calcium-dependent manner (Plant et al. 1997). Thus, ubiquitination and subsequent proteasomal degradation of amiloride-sensitive sodium channels is regulated by the membrane translocation of NEDD4, which is induced by an elevated intracellular Ca2+ level (Staub et al. 2000). Although Sakura and Momo immunoreactivities were enriched in the membrane fraction of the brain preparations (Fig. 2), they were recovered in both cytoplasmic and membrane fractions when it was overexpressed in HEK293 cells. This observation suggests that the palmitoylation of Sakura, and potentially that of Momo, might be regulated to control their subcellular distributions and substrate recognition (Plant et al. 2000). The regulatory mechanism of the palmitoylation remains to be characterized, however.
In the nervous system, an impaired Ub-proteasome pathway is often implicated in neurodegenerative diseases in which there is accumulation of abnormal protein deposits in the nervous system (Paulson 1999; Saigoh et al. 1999; Bence et al. 2001; Chung et al. 2001a). Recent studies indicate that protein ubiquitination serves not only as a signal for proteasomal degradation but also as a routing code for intracellular molecular trafficking (Shih et al. 2000; Dupre et al. 2001; Hicke 2001; Katzmann et al. 2001; Polo et al. 2002). The degradation of Sakura ubiquitination products was not sensitive to an inhibitor of the conventional 26S proteasome, potentially suggesting the use of unconventional lysine sites for Ub ligation. Given the distributions and activities of these Ub ligases, future studies should elucidate their contribution to molecular trafficking as well as to protein degradation.
We thank Ms. Yuriko Iwakura and Mr Tadasato Nagano for technical assistance and Ms. Hiromi Kato for typing. We also thank Dr Mako Narisawa-Saito for helpful discussion. The nucleotide sequences of human Momo and Sakura cDNAs were similarly determined and deposited in GenBank (GenBank accession #AAM29180 for human Momo and #AAM29181 for human Sakura). This work was supported by the Japanese Society for the Promotion of Science (RFTF-96L00203) and a Grant-in-Aid for Creative Scientific Research.