Rma1, a novel type of RING finger protein conserved from Arabidopsis to human, is a membrane-bound ubiquitin ligase

In a screen to identify molecules that function in the plant secretory pathway, we previously isolated the Arabidopsis thaliana cDNA RMA1 as a clone that complements a yeast secretory mutant, sec15. This cDNA encodes a protein with a typical RING finger motif and a C-terminal membrane-anchoring domain, and was thus designated RMA1 for a RING finger protein with a membrane anchor (Matsuda and Nakano, 1998). The RING finger motif is defined as Cys-X2-Cys-Xn-Cys-X-His-X2-Cys-X2-Cys-Xn-Cys-X2-Cys, where X is an arbitrary amino acid residue (Saurin et al., 1996). This motif has been found in a variety of eukaryotic proteins, but not in any prokaryotic proteins. The RING finger motif was first discovered in a putative transcription factor and was claimed to be a common motif for DNA binding proteins (Freemont et al., 1991). However, this motif and its variants are now found in more than 200 proteins with diverse functions. They are implicated in tumor suppression (Cbl, BRCA1 and Mdm2), signal transduction (TRAFs and Ste5), apoptosis (c-IAPs and XIAP), peroxisomal biogenesis (Pex2 and Pex10), cell cycle (Apc11 and Rbx1), DNA repair (Rad5, Rad16 and Rad18) and vesicular transport (Vps8, Pep5 and Vps18). In the case of Arabidopsis Rma1, its role in vesicular transport remains obscure.

Until recently, the diversity of RING finger proteins was puzzling (Saurin et al., 1996), but now emerging evidence directs the role of the RING finger motif to the ubiquitin-ligase (E3) activity (for reviews, see Deshaies, 1999; Tyers and Jorgensen, 2000; Freemont, 2000; Jackson et al., 2000; Joazeiro and Weissman, 2000; and references therein). Ubiquitination is performed by sequential actions of three enzymes. Ubiquitin is first activated by E1 through the formation of a high-energy thioester bond, and is subsequently transferred to E2 via the thioester bond. Finally E2 transfers ubiquitin to its substrate with the help of a ubiquitin ligase, E3 (Yamao, 1999; Ciechanover et al., 2000). Ubiquitination and subsequent degradation by the proteasome at correct timing are essential for many cellular processes. The variety of functions fulfilled by different RING finger proteins will be explained if they are acting as ubiquitin ligases on different proteins in diverse cellular reactions. This situation tempted us to examine whether Rma1 has an E3 activity, and if so which E2 is its partner.

Human cDNA clones that show similarity to AtRMA1 were found in the EST database and obtained from Incyte Genomics Inc. (St Louis, Missouri, USA). These clones, AA687756, A1929339 and AI146471, are reported to derive from human normal prostate, frontal lobe of brain and heart, respectively. Their complete nucleotide sequences were determined to obtain the contiguous sequence of HsRMA1.

Plasmids and oligonucleotides used in this study are listed in Table 1. An underline indicates mutated or inserted nucleotides. To express fusion proteins with the maltose binding protein (MBP), pMAL-p2 (New England BioLabs, Beverly, Massachusetts, USA) was utilized. AtRMA1ΔC, in which a stop codon was inserted just before the membrane-anchoring domain, was made by polymerase chain reaction (PCR) using primers NM68 and NM70. The PCR product was sequenced, digested with BglII and XhoI, and inserted into the BamHI-SalI sites of pMAL-p2 to express the MBP-AtRma1ΔC fusion protein (MBP-AtRma1). The resulting plasmid was designated pNMR41. To yield MBP-AtRma1C63S, which harbors the C63S point mutation in the RING finger motif, the AtRMA1C63S mutant (Matsuda and Nakano, 1998) was used as a PCR template. The obtained plasmid was designated pNMR7. To obtain recombinant MBP-3HA-AtRma1ΔC, the PCR fragment was amplified from 3HA-AtRMA1 (Matsuda and Nakano, 1998) using primers NM68 and NM79, digested with BglII and XhoI, and inserted into the BamHI-SalI sites of pMAL-p2. The resulting plasmid was designated pNMR13. A plasmid to express MBP-3HA-AtRma1C63SΔC was similarly constructed using 3HA-AtRMA1C63S and was named pNMR14. HsRMA1ΔC, which lacks the membrane-anchoring domain, was made by PCR using primers NM122 and NM123. A plasmid to express MBP-HsRma1 was similarly constructed using HsRMA1ΔC and pMAL-p2, and was named pNMHR2. The C42S point mutation of HsRma1, which is equivalent to the C63S mutation of AtRma1, was introduced by PCR using primers NM127 and NM128. The plasmid to express MBP-HsRma1C42SΔC was constructed by the same way as pNMR7 and was designated pNMHR5. The K68R, K75R, K86R and K93R mutations of HsRma1 were introduced by PCR using primers NM137 and NM138, NM139 and NM140, NM141 and NM142, and NM143 and NM144, respectively. The plasmids to express MBP-HsRma1K68RΔC and MBP-HsRma1K68R, K75R, K86R, K93RΔC were designated pNMHR6 and pNMHR10, respectively. All these recombinant proteins were purified from E. coli lysate according to the instructions of the supplier (New England BioLabs).

Using the rabbit reticulocyte lysate as a source of E1 and E2, the ubiquitination assay of MBP-HsRma1 was performed as follows. The reaction mixture (150 μl) containing 40 mM Tris-HCl, pH 7.5, 5 mM MgCl2, 2 mM ATP, 2 mM dithiothreitol (DTT), 300 ng/μl ubiquitin (Sigma, St Louis, Missouri, USA), 25 μM MG132 (Peptide Inc., Osaka, Japan), 5 μl rabbit reticulocyte lysate (Promega, Madison, Wisconsin, USA) and 400 ng MBP-HsRma1 was incubated at 30°C for appropriate times and subjected to immunoblotting with an anti-MBP antibody (New England BioLabs). In vitro ubiquitination assay of MBP-AtRma1, MBP-3HA-AtRma1 and MBP-3HA-AtRma1C63S was performed using the reaction mixture (150 μl) containing 400 ng recombinant MBP-AtRma1 protein or its derivative and 5 μl wheat germ lysate (Promega) instead of MBP-HsRma1 and the rabbit reticulocyte lysate. They were subjected to immunoblotting with the anti-influenza haemagglutin epitope (HA) monoclonal antibody 16B12 (Berkeley Antibody, Richmond, California, USA) or an anti-MBP antibody. To deplete ATP from the reaction mixture, 1 μg/μl hexokinase (Sigma) and 20 mM 2-deoxyglucose (Nacalai Tesque Inc., Kyoto, Japan) were added in each assay.

The reaction mixture (750 μl) containing 40 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 2 mM ATP, 2 mM DTT, 300 ng/μl ubiquitin, 25 μM MG132, 40 μl rabbit reticulocyte lysate and 15 μl amylose resin (New England BioLabs), with conjugated MBP-HsRma1, was incubated at 30°C for appropriate times with gentle agitation. The resin was washed three times with the column buffer containing 20 mM Tris-HCl, pH 7.5, 200 mM NaCl, and 1 mM EDTA. MBP-HsRma1 and its putative modifier were eluted from the amylose resin with 150 μl column buffer containing 10 mM maltose, and subjected to immunoblotting with the anti-MBP antibody (New England BioLabs) or an anti-ubiquitin monoclonal antibody (Medical and Biological Laboratories, Nagoya, Japan). The pull-down assay of MBP-AtRma1 was also performed using 15 μl amylose resin with conjugated MBP-AtRma1 and 40 μl wheat germ lysate instead of the rabbit reticulocyte lysate.

The plasmids to express mouse E1, E2-20k (Kaiser et al., 1994), E2-25k (Kalchman et al., 1996), Ubc3 (Plon et al., 1993) and UbcH5a (Scheffner et al., 1994) were kind gifts from Kazuhiro Iwai of Kyoto University. Ubc4 (Rolfe et al., 1995), whose cDNA was kindly provided by Mitsuyoshi Nakao of Kumamoto University, and Ubc8 (Kumar et al., 1997) isolated by PCR, were subcloned into the pT7-7 bacterial expression vector. Recombinant mouse E1 and E. coli (BL21-DE3) lysates expressing E2-20k, E2-25k, Ubc3, Ubc4, UbcH5a and Ubc8 were prepared as described previously (Iwai et al., 1999). Sufficient expression and solubilization of each E2 was confirmed by Coomassie Brilliant Blue staining. The reaction mixture (75 μl) containing 40 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 2 mM ATP, 2 mM DTT, 25 μM MG132, 300 ng/μl ubiquitin, 225 ng mouse E1, 15 μg E2-expressing E. coli lysate, and 200 ng MBP-HsRma1 or MBP-AtRma1 was incubated at 35°C for 180 minutes, and subjected to immunoblotting with the anti-MBP antibody (New England BioLabs). To examine whether the RING finger motif is essential for the ubiquitin-ligase activity, MBP-HsRma1C42S and MBP-AtRma1C63S were used instead of MBP-HsRma1 and MBP-AtRma1, respectively. To determine whether the MBP moiety is ubiquitinated, MBP-HsRma1K68R and MBP-(KallR)HsRma1 (HsRma1K68R, K75R, K86R, K93R) were used instead of MBP-HsRma1.

800 ng recombinant MBP-3HA-AtRma1 was incubated with 400 ng Factor Xa protease (New England BioLabs) on ice for 180 minutes in the digestion buffer (300 μl) containing 40 mM Tris-HCl, pH 7.5 and 2 mM CaCl₂. Factor Xa was removed using 50 μl equilibrated Benzamidine Sepharose 6B (Amersham Pharmacia Biotech, Little Chalfont, Buckinghamshire, England) and then in vitro ubiquitination was performed by adding 5 mM MgCl2, 2 mM ATP, 2 mM DTT, 300 ng/μl ubiquitin, 25 μM MG132, 300 μM phenylmethane sulfonyl fluoride (PMSF) and 5 μl wheat germ lysate. The reaction mixture was incubated at 30°C for appropriate times and subjected to immunoblotting with the anti-HA monoclonal antibody or an anti-MBP antibody.

We have already reported that human, mouse and nematode possess close homologs of Arabidopsis RMA1 (Matsuda and Nakano, 1998). We wished to isolate the human RMA1 homolog because we planned to perform in vitro ubiquitination assays using mammalian E1 and E2 enzymes. We searched a human EST database and found several cDNA clones encoding the putative human homolog of Rma1. Three human EST clones were obtained and sequenced. The longest ORF within the contiguous sequence encoded a protein of 180 amino acid residues with a typical RING finger motif and a putative C-terminal membrane-anchoring domain (Fig. 1). We designated this ORF HsRMA1. The amino acid sequence of HsRma1 is completely identical to human hypothetical protein HSMHC3W5A and is 97% identical to mouse hypothetical protein T09063. HSMHC3W5A and T09063 are registered in the database as proteins residing in the major histocompatibility complex (MHC) class III region, but their functions are unknown at present. The amino acid sequence identity is 26% between the original Arabidopsis Rma1 (AtRma1) and HsRma1.

To obtain recombinant AtRma1, we first constructed plasmids to express GST-AtRma1ΔC, His6-AtRma1ΔC and AtRma1ΔC-His6, but the purification of these fusion proteins was unsuccessful. Next we attempted to purify a maltose binding protein (MBP)-AtRma1 fusion protein. To express MBP-AtRma1, a stop codon was inserted just before the membrane-anchoring domain and the resulting AtRMA1ΔC was subcloned into pMAL-p2. A sufficient quantity of the recombinant MBP-AtRma1 fusion was expressed and purified by this method. A good amount of MBP-HsRma1 was obtained in a similar way. Because E. coli does not have the ubiquitination system, these recombinant proteins are free of contaminating E3 enzymes.

E3 accelerates the ubiquitination of its substrate and sometimes E3 is itself ubiquitinated. Therefore, the in vitro demonstration of substrate-independent auto-ubiquitination is a good indication of the E3 activity. We decided to examine whether MBP-Rma1 catalyzes auto-ubiquitination in the presence of ATP, E1, E2 and ubiquitin. First, we used the rabbit reticulocyte lysate as the source of E1 and E2, because it is known to contain high ubiquitin-dependent proteolysis activity (Etlinger and Goldberg, 1977; Wilkinson et al., 1980; Hershko et al., 1980). MBP-HsRma1 was incubated at 30°C with ATP, ubiquitin and the rabbit reticulocyte lysate, and subjected to immunoblotting with the anti-MBP antibody. As shown in Fig. 2A, incubation of MBP-HsRma1 with the reticulocyte lysate led to the formation of slower-migrating ladders, which were recognized by the anti-MBP antibody. For examination of Arabidopsis Rma1, the wheat germ lysate was used instead of the reticulocyte lysate as the source of plant E1 and E2, because it also possesses a high ubiquitination activity (Hatfield and Vierstra, 1989). When MBP-AtRma1 was incubated with ATP, ubiquitin and the wheat germ lysate, high molecular-mass ladders appeared again (Fig. 2B). MBP alone or the omission of cell lysates failed to give such modification. Depletion of ATP by hexokinase and 2-deoxyglucose inhibited this reaction as well (Fig. 2). These results indicate that MBP-HsRma1 and MBP-AtRma1 receive an ATP-dependent modification when incubated with the cell lysates.

To test whether the modification acquired by MBP-Rma1 is indeed the polyubiquitination, a pull-down assay was performed. MBP-HsRma1 was immobilized on the amylose resin and incubated with ATP, ubiquitin and the reticulocyte lysate as described above. MBP-AtRma1 was also conjugated on the amylose resin and similarly incubated with the wheat germ lysate instead of the reticulocyte lysate. After incubation for appropriate times, the amylose resin was washed thoroughly by a buffer to remove free ubiquitin and other ubiquitinated proteins. MBP-HsRma1 and MBP-AtRma1 with the modification were then eluted from the resin with 10 mM maltose and then subjected to immunoblotting with anti-MBP and anti-ubiquitin antibodies. As shown in Fig. 3, only the slower-migrating ladders of MBP-HsRma1 (A) and MBP-AtRma1 (B) reacted with the anti-ubiquitin antibody, indicating that the modification on MBP-Rma1 was in fact polyubiquitination.

We next tried to reconstitute autoubiquitination of MBP-Rma1 by purified components only, i.e. ubiquitin, E1, E2 and MBP-Rma1. E3 requires specific E2 to mediate ubiquitination. E. coli lysates expressing E2-20k, E2-25k, Ubc3, Ubc4, UbcH5a and Ubc8 were used as a source of E2 in this assay. MBP-HsRma1 was incubated with ATP, ubiquitin, recombinant mouse E1 and one of the several E2 enzymes described above, and subjected to immunoblotting with the anti-MBP antibody. As shown in Fig. 4A, high molecular-mass ladders were clearly observed when MBP-HsRma1 was incubated with Ubc4 and UbcH5a, indicating that MBP-HsRma1 possesses an intrinsic autoubiquitination activity. Other E2 enzymes tested, E2-20k, E2-25k, Ubc3 and Ubc8, did not render this modification. A similar result was obtained with AtRma1. Slower-migrating ladders were observed only when MBP-AtRma1 was incubated with Ubc4 or UbcH5a, and not with any other E2 enzymes examined (Fig. 4B). These results clearly demonstrate that HsRma1 and AtRma1 catalyze ubiquitination with the help of Ubc4 and UbcH5a. In other words, Rma1 proteins are ubiquitin ligase that functions in cooperation with the Ubc4/5 subfamily.

It has been shown that the RING finger motif is required for the ubiquitin ligase activity for almost all the RING finger-containing E3 enzymes. To test whether this also holds true for HsRma1 and AtRma1, the conserved third cysteine in the RING finger motif was changed to serine (Fig. 5A). The mutants, AtRMA1C63S and HsRMA1C42S, were subcloned into pMAL-p2 and the resulting MBP-AtRma1C63S and MBP-HsRma1C42S were purified. We have previously reported that the C63S mutation in AtRma1 impaired the suppression activity toward sec15 (Matsuda and Nakano, 1998). When incubated with ATP, ubiquitin, E1 and UbcH5a, neither MBP-AtRma1C63S nor MBP-HsRma1C42S produced any high molecular-mass ladders (Fig. 5B, lanes 4-6 and 10-12). As controls, MBP-AtRma1 and MBP-HsRma1 in this assay formed slower-migrating bands (Fig. 5B, lanes 1-3 and 7-9). Thus the RING finger motif is essential for the ubiquitin-ligase activity for the case of the Rma1 proteins as well.

We next examined whether the autoubiquitination of MBP-Rma1 is catalyzed by an intermolecular or intramolecular reaction. To discern these two possibilities, we prepared 3HA (influenza hemagglutinin epitope)-tagged versions of MBP-AtRma1 and MBP-AtRma1C63S. When recombinant MBP-3HA-AtRma1 was incubated with ATP, ubiquitin and the wheat germ lysate, high molecular-mass ladders derived from ubiquitination were observed (Fig. 6, lanes 1-3). On the contrary, MBP-3HA-AtRma1C63S gave no such slower-migrating ladders (Fig. 6, lanes 4-6). We then mixed MBP-3HA-AtRma1C63S and wild-type MBP-AtRma1 in the reaction. If auto-ubiquitination of MBP-Rma1 is a trans-molecular reaction, MBP-3HA-AtRma1C63S should be ubiquitinated. As shown in Fig. 6, however, MBP-AtRma1 was ubiquitinated (lanes 10-12) but MBP-3HA-AtRma1C63S was not at all (lanes 7-9). This result clearly indicates that the auto-ubiquitination of MBP-Rma1 is a cis-molecular reaction.

AtRma1 and HsRma1 possess nine and four lysine residues, respectively. One of them is located near the RING finger motif (K97 of AtRma1 and K68 of HsRma1) and is well conserved among all Rma1 homologs. To test whether this lysine residue is the ubiquitination site, K68 of HsRma1 was changed to arginine. When incubated with ATP, ubiquitin, E1 and UbcH5a, however, MBP-HsRma1K68R underwent ubiquitination and gave high molecular-mass ladders as well as wild-type HsRma1 (Fig. 7). Moreover, even when all four lysine residues of HsRma1 were changed to arginine, MBP-HsRma1 (KallR) was still ubiquitinated (Fig. 7). This result reveals that ubiquitin is conjugated to the lysine residue(s) in the MBP moiety of MBP-HsRma1. Ubiquitination of the original lysine residues of HsRma1 is not yet ruled out, but note that they are not essential for the ubiquitin ligase activity.

We further asked whether Rma1 itself is ubiquitinated. Recombinant MBP-3HA-AtRma1, in which a Factor Xa protease site is inserted to separate MBP and 3HA-AtRma1, was constructed and purified (Fig. 8A). The 3HA tag possesses no lysine residue and is free from ubiquitination. After complete cleavage of MBP-3HA-AtRma1, Factor Xa was removed by benzamidine-sepharose and then the in vitro ubiquitination assay was performed. As shown in Fig. 8B, no ubiquitination of 3HA-Rma1 itself was detected by immunoblotting with the anti-HA antibody (lanes 3, 4). In the control, ubiquitination of MBP-3HA-AtRma1 was observed when the same experiment was performed without Factor Xa (lanes 1, 2). This result demonstrates that Rma1 mediates virtually no ubiquitination on its own lysine residues.

Is MBP a general good substrate for Rma1 then? We next tested whether Rma1 can ubiquitinate free MBP in solution. Soluble MBP and 3HA-AtRma1 were similarly prepared from MBP-3HA-AtRma1 by the Factor Xa cleavage. After removal of Factor Xa, they were subjected to the in vitro ubiquitination assay and immunoblotting with anti-MBP antibody. Interestingly, soluble MBP was not ubiquitinated by 3HA-AtRma1 at all (Fig. 8B, lanes 7, 8). Ubiquitination was again observed in the control experiment done in parallel (Fig. 8B, lanes 5, 6). Since the in-frame-fused MBP was efficiently ubiquitinated by Rma1 (Fig. 7), these results indicate that MBP can be a substrate of the Rma1 ubiquitin ligase only when it is in the physical vicinity of Rma1.

RMA1 proteins are found in various species of higher eukaryote, including nematode (C. elegans), fruit fly (D. melanogaster), human (H. sapiens) and higher plants (A. thaliana), but not in lower eukaryotes like yeast. It suggests that this family of proteins fulfils a very important function, which was acquired during evolution. In this paper, we have demonstrated that Arabidopsis and human Rma1 are a membrane-bound ubiquitin ligase, E3, which requires Ubc4/5 as E2.

Although E3 plays an imperative role in ubiquitination, our knowledge of plant E3 is quite limited (Vierstra and Callis, 1999; Gray and Estelle, 2000). In higher plants, the E3 activity has only been demonstrated for UPL1, a protein which contains the HECT (homologous to E6-AP carboxyl terminus) domain but not the RING finger motif (Bates and Vierstra, 1999). On the other hand, dozens of genes encoding RING finger proteins have been found in Arabidopsis thaliana, but their biochemical functions remain to be tested. They include COP1 (Deng et al., 1992), CIP8 (Torii et al., 1999), PRT1 (Potuschak et al., 1998), ATL2-6 (Salinas-Mondragon et al., 1999), RHN (Jensen et al., 1998), RMR (Jiang et al., 2000), A-RZF (Zou and Taylor, 1997) and RMA1 (Matsuda and Nakano, 1998). PRT1, COP1 and CIP8 are interesting in the context of ubiquitination. Bachmair and coworkers isolated prt1 mutant in which a normally short-lived reporter protein was stabilized (Bachmair et al., 1993). In a recent report they showed that PRT1 encodes a novel protein with two RING finger motifs and speculated that it may be a component of E3 (Potuschak et al., 1998). cop1 was isolated as a mutant that mimicked the light-grown morphology even in darkness, indicating that COP1 represses photomorphogenesis in the dark. COP1 encodes a protein with a RING finger motif (Deng et al., 1992). Another RING finger protein, CIP8, was isolated by two-hybrid screening as a putative partner of COP1 (Torii et al., 1999). COP1 also interacts with HY5, a bZIP transcription factor that advances photomorphogenesis (Oyama et al., 1997; Ang et al., 1998). Very recently, Osterlund et al. (Osterlund et al., 2000) showed that HY5 is degraded in the dark and that a cop1 mutation or treatment with a proteasome inhibitor prevents its degradation. These results suggest that COP1/CIP8 complex functions as a ubiquitin ligase for HY5 (Deshaies and Meyerowitz, 2000; Hardtke and Deng, 2000). All these studies appear to point to a common role of Arabidopsis RING finger proteins as E3, but unfortunately, direct biochemical evidence for the ubiquitin ligase activities has yet to be demonstrated except the case of AtRma1.

Although RMA1 was originally isolated as a cDNA that complemented the yeast sec15 mutation, RMA1 and SEC15 are not structurally related. The yeast genome does not contain any RMA1 homologs at all. How can RMA1 suppress the growth defect of the sec15 mutant then? We previously hypothesized that RMA1 suppresses sec15 via some component(s) conserved between yeast and plants (Matsuda and Nakano, 1998). Now we know that Rma1 is a ubiquitin ligase depending on Ubc4/5. Ubiquitination may link RMA1 and SEC15. Ubiquitin, E1 and E2 enzymes are well conserved between yeast and higher plants (Vierstra and Callis, 1999). Indeed, yeast Ubc4/5 and their Arabidopsis homologs AtUbc8-12 show 80% identity and 90% similarity at the amino acid sequence level (Girod et al., 1993). Yeast Ubc4/5 is known to be important for degradation of abnormal proteins and for turnover of various short-lived proteins (Seufert and Jentsch, 1990). To examine whether AtRma1 stabilizes Sec15-1p in the sec15 mutant, we performed an immunoblotting analysis using the anti-Sec15 antibody (a gift from P. Novick). However, the amount of Sec15-1p was not increased by overproduction of AtRma1 (data not shown). On the other hand, the growth defect of sec15 is suppressed by the overexpression of various genes, for instance SEC1, SEC4, SSO1/2, SEC9, SEB1/SBH1, RHO3 and SEM1 (Salminen and Novick, 1987; Aalto et al., 1993; Brenwald et al., 1994; Toikkanen et al., 1996; Adamo et al., 1999; Jäntti et al., 1999). It is thus likely that heterologous overproduction of AtRma1 perturbs the function of yeast Ubc4/5, which affects the life span of various proteins including the suppressors of sec15 described above, and consequently leads to the suppression of the sec15 mutation. Consistent with this hypothesis, AtRma1C63S, which lacks the ubiquitin-ligase activity (Fig. 5) cannot suppress the sec15 mutation (Matsuda and Nakano, 1998). Physical interaction between yeast Ubc5 and AtRma1 has also been demonstrated by a two-hybrid assay (N. Matsuda and A. Nakano, unpublished result). Interestingly, a few recent studies report genetic interaction between secretory genes and ubiquitin-proteasome genes (Amerik et al., 2000; Kimata et al., 2000; Wiederkehr et al., 2000). It would again suggest that some secretory proteins are regulated by ubiquitination.

There is evidence that a RING finger protein functions as E3 by itself or as an essential subunit of the E3 complex in the cases of Ubr1 (Bartel et al., 1990; Kwon et al., 1998), MDM2 (Honda et al., 1997; Fang et al., 2000), Apc11 (Zachariae et al., 1998; Gmachl et al., 2000; Leverson et al., 2000), AO7 (Lorick et al., 1999), Cbl (Joazeiro et al., 1999; Yokouchi et al., 1999; Levkowitz et al., 1999), Rbx1/Hrt1/Roc1 (Kamura et al., 1999; Skowyra et al., 1999; Seol et al., 1999; Ohta et al., 1999; Tan et al., 1999), Parkin (Shimura et al., 2000), TRAF6 (Deng et al., 2000) and cIAPs (Huang et al., 2000; Yang et al., 2000). However, these are only a small fraction of numerous RING finger proteins and it is still unclear whether all RING finger proteins possess E3 activity. Among E3 enzymes identified hitherto, Rma1 is unique because it has a membrane-anchoring tail at the C terminus. No such protein has been identified as E3 before. Hrd1/Der3 is an integral membrane protein with an N-terminal membrane spanning region and a C-terminal RING finger motif and is involved in ER-associated degradation (Hampton et al., 1996; Bordallo et al., 1998). During the revision of this manuscript, two groups reported biochemical evidence for the E3 activity of Hrd1/Der3 (Bays et al., 2001; Deak and Wolf, 2001). Although the membrane anchor of Rma1 is dispensable for the ubiquitin ligase activity in vitro, it must play a critical role for correct subcellular localization and substrate recognition in vivo. In fact, AtRma1ΔC is unable to suppress the temperature-sensitive growth of the yeast sec15-1 mutant (data not shown). Identification of the physiological substrate will be essential to reveal the precise role of Rma1 in vivo and is now under way.

E3 enzymes are thought to be classified into two groups, namely those with the HECT domains and those containing RING finger motifs (Deshaies, 1999; Tyers and Jorgensen, 2000; Freemont, 2000; Jackson et al., 2000; Joazeiro and Weissman, 2000). E3 with the HECT domain accepts ubiquitin from E2 again through a high-energy thioester bond and then transfers it to the substrate. On the other hand, E3 containing a RING finger motif does not form a thioester bond with ubiquitin but mediates ubiquitination by facilitating the direct transfer of ubiquitin from E2 to the substrate. However, how RING finger proteins facilitate ubiquitination remains to be established. We have shown in this paper that Rma1 ubiquitinates in-frame-fused MBP (Fig. 7) but not free MBP in solution (Fig. 8). Rma1 cannot ubiquitinate the MBP moiety of MBP-Rma1C63S in a trans-molecular fashion either (Fig. 6). These observations indicate that Rma1 recognizes MBP as a substrate not because of its amino acid sequence but because of its physical vicinity to Rma1 (Fig. 9). We propose that the primary function of RING finger E3 would be to bring its substrate into physical contact with the E2-ubiquitin conjugate and thus catalyze the transfer of ubiquitin.

We are grateful to Kazuhiro Iwai of Kyoto University and Mitsuyoshi Nakao of Kumamoto University for plasmids and to Peter Novick of Yale University for an anti-Sec15 antibody. We also thank Ken Sato for technical advice on protein purification and the members of the Nakano laboratory for helpful discussions. This work was supported by a Grant-in-Aid for Exploratory Research from the Ministry of Education, Science, Sports and Culture of Japan and by a Special Coordination Fund for Promotion of Science and Technology from the Science and Technology Agency of Japan. N.M. is a recipient of fellowships for Junior Research Associate and Special Postdoctoral Researcher of RIKEN.