N-recognin/Ubc2 Interactions in the N-end Rule Pathway*

The N-end rule relates the in vivo half-life of a protein to the identity of its N-terminal residue. In the yeast Saccharomyces cereuisiae, substrates of the N- end rule pathway are targeted for degradation by a complex that includes the 225-kDa N-recognin, en- coded by UBRl, and the 20-kDa ubiquitin-conjugating enzyme encoded by UBC2. We report that both physi- cal stability and functional activity of the N-recognin Ubc2 complex require the presence of a highly acidic 23-residue region at the C terminus of Ubc2. Ubc2- C88A, an inactive variant of Ubc2 in which the active-site Cys-88 has been replaced by Ala, is shown to retain the affinity for N-recognin. Expression of Ubc2-CSSA inhibits the N-end rule pathway, apparently as a result of competition between Ubc2 and TJbc2-CSSA for bind- ing to N-recognin. The two-hybrid (interaction cloning) technique was used to identify a -170-residue C- terminal fragment of the 1,950-residue N-recognin as a Ubc2-interacting domain. We also show that the level of UBRl mRNA decreases upon overexpression of UBC2. This effect of UBC2 is observed with cells whose UBRl is expressed from an unrelated promoter but is not observed if UBRl contains a frameshift mutation, or if the Ubc2 protein lacks its C-terminal acidic region. The N-recognin Ubc2 complex

ll To whom correspondence should be addressed Div. of Biology, 147-75, California Institute of Technology, Pasadena, CA 91125. Tel.: 818-356-3785;Fax: 818-440-9821. The eukaryotic N-degron comprises at least two determinants: a destabilizing N-terminal residue and an internal Lys residue (or residues) (Bachmair and Varshavsky, 1989;Johnson et al., 1990;Dunten et al., 1991). The latter is the site of attachment of a multiubiquitin (multi-Ub) chain whose formation is required for the degradation of at least some N-end rule substrates. In a multi-Ub chain, several ubiquitin (Ub)' moieties are attached sequentially to an acceptor protein, forming a chain of Ub-Ub conjugates in which the C-terminal Gly-76 of one Ub is joined to Lys-48 of the adjacent Ub (Chau et al., 1989;Johnson et al., 1992). A substrate that bears a multi-Ub chain is degraded by the 26 S proteasome, a -1,500 kDa, ATP-dependent, multicatalytic protease that contains more than 20 distinct subunits. Different Ub-dependent pathways are mediated by distinct targeting complexes while apparently sharing at least a protease component (the 20 S or "core" proteasome) (reviewed by Finley and Chau, 1991;Rechsteiner, 1991;Hershko and Ciechanover, 1992;Goldberg, 1992;Hochstrasser, 1992;Varshavsky, 1992).
In the yeast Saccharomyces cerevisiae, the recognition component of the N-end rule pathway, called N-recognin (it is also known as E30 (Hershko and Ciechanover, 1992)), is encoded by the UBRl gene (Bartel et al., 1990). The 225-kDa N-recognin (Ubrl) selects potential N-end rule substrates by binding to their primary destabilizing N-terminal residues Phe, Leu, Trp, Tyr, Ile, Arg, Lys, and His. N-recognin has at least two substrate-binding sites. The type 1 site is specific for the basic N-terminal residues Arg, Lys, and His. The type 2 site is specific for the bulky hydrophobic N-terminal residues Phe, Leu, Trp, Tyr, and Ile (Reiss et al., 1988;Gonda et al., 1989;Baker and Varshavsky, 1991). These N-terminal residues are bound directly by N-recognin, whereas secondary destabilizing N-terminal residues (Asp and Glu in yeast) function through their conjugation, by Arg-tRNA-protein transferase (R-transferase), to Arg, one of the primary destabilizing residues. Tertiary destabilizing N-terminal residues, Asn and Gln, function through their conversion, by a specific amidase, into the secondary destabilizing residues Asp and Glu (Gonda et al., 1989;Balzi et al., 1990;reviewed by Varshavsky, 1992).
In S. cerevisiae, the ubiquitination of N-end rule substrates has been found to require the Ub-conjugating enzyme encoded by the UBC2 gene (its old name is RADG), and Ubc2 has been shown to be physically associated with N-recognin (Dohmen et al., 1991a). The yeast Ubc2 can also function as an N-end rule-mediating Ub-conjugating enzyme in a heterologous cellfree system such as an extract from rabbit reticulocytes (Sung et d . , 1991). Ubc2 is one of at least 10 distinct Ub-conjugating
Processes that are known to be perturbed by mutations in UBC2 include DNA repair, induced mutagenesis, cell cycle control, sporulation, regulation of retrotransposon transposition, and the N-end rule pathway of protein degradation (Reynolds et al., 1985;Borts et al., 1986;Friedberg, 1888Picologlou et al., 1990Madura et al., 1990;Ellison et al., 1991;Kang et al., 1991;Dohmen et al., 1991a, and references therein). The N-end rule pathway is inactive in either a ubc2 or a ubrl null mutant; however, the overall ubc2 phenotype is more severe than that of ubrl, indicating that the functions of the ubc2 enzyme are not confined to the N-end rule pathway (Dohmen et al., 1991a).
While it is likely that most of the E2 enzymes in yeast and other eukaryotes are guided to their in vivo substrates by E2associated "recognition" subunits (recognins) (Varshavsky, 1992), the yeast N-recognin . ubc2 complex remains the only example of such an assembly that has been defined genetically. A further biochemical and genetic dissection of this complex is described below.

EXPERIMENTAL PROCEDURES
Strains, Media, Genetic Techniques, and @-Galactosidase Assay-The S. cereuisiae strains used in this work are listed in Table I. 5' .
To express UBRl from the copper-inducible Pcupl promoter, we modified the high copy plasmid pJD1003, a derivative of pSOB35 that encoded Ubrl and contained a unique Sal1 site (produced by site-directed mutagenesis) 15 base pairs upstream of the start (ATG) codon of UBRZ? This plasmid was digested with PstI and EspI, treated with T4 DNApolymerase, and self-ligated, yieldingpKM1307, which lacks a 5'-proximal portion of the QCR9 gene that was present R. J. Dohmen, B. Bartel, and A. Varshavsky, unpublished data. in our earlier UBRI-expressing plasmids. QCR9, located next to UgRl on Chromosome VII, encodes subunit 9 of the mitochondrial ubiquinol-cytochrome c oxidoreductase complex (Trumpower, 1990;Phillips and Trumpower, 1993). Cells carrying UBRl -expressing, high copy plasmids that also expressed a truncated QCR9 had a mild petite-like phenotype (slow growth on inefficiently fermentable carbon sources such as raffinose or galactose, data not shown). The -0.35-kb, Pcup, promoter-containing region of the plasmid pJDC22-2 (see below) was excised by Sat1 and Sad, and ligated into SalI/ SacI-cut pKM1307, yielding pKM1315, which expressed UBRl from the Pcupl promoter. (pJDC22-2 was constructed by a multistep protocol (details available upon request) that included the insertion of a modified, PCup, promoter-containing fragment from the plasmid YEp46 (Ecker et al., 1987) (a gift from D. Finley, Harvard Medical School, Boston, MA) into the BamHI-cut polylinker of YCplac22 (Gietz and Sugino, 1988).) The plasmid pKM1315 was treated with SpeI and Klenow PoII, and self-ligated, producing a 4-base insertion/ frameshift mutation at the nucleotide position 510 in the UBRZ open reading frame (Bartel et al., 1990). The resulting plasmid, pKM1320, expressed a 170-residue N-terminal fragment of the 1,950-residue Ubrl protein (the mutation produced a stop codon immediately after the nucleotide position 510). Neither pKM1315 nor pKM1320 inhibited cell growth under conditions that induced their UBR1-linked Pcupl promoter (0.1 mM CUSOI, see Fig. 4).
Isolation of RNA and Northern Hybridization-A slight modification of the procedures described by Baker et al. (1992) was used. S.
cereuisiae strains KMY633,634 and 635 (Table I) were grown to A600 of -1 in SR medium containing 0.1 mM CuSO4 and lacking uracil and tryptophan. A 15-ml sample of each culture was centrifuged to pellet the cells, which were then washed with 1 ml of water. Galactose was added to the remainder of the cultures to a final concentration of 3%. Samples were withdrawn from the cultures 1,2, and 3 h after the addition of galactose. Washed cell pellets were resuspended in 0.25 ml of ice-cold TEN buffer (0.5 M NaC1, 10 mM Na-EDTA, 0.2 M Tris-HC1, pH 7.5). An equal volume of glass beads and 0.25 ml of phenol-chloroform (l:l, equilibrated at pH 8) were added, followed by vortexing at maximum speed for 2 min, with intermittent cooling on ice (Silverman, 1987). The samples were centrifuged at 12,000 X g for 1 min, and the aqueous phase was reextracted twice with phenolchloroform, followed by the addition of 2 volumes of 95% ethanol and incubation at -20 "C to precipitate nucleic acids, which were recovered by centrifugation at 12,000 x g for 5 min. The nucleic acid pellets were dissolved in 2 mM dithiothreitol containing 500 units/ml of RNAsin (Promega) The concentration of RNA was determined by measuring A2w (these preparations contained negligible amounts of DNA, data not shown). For Northern hybridization, a slightly modified procedure of McMaster and Carmichael (1977) was used. 25 pg of total RNA was treated with 1.5 M deionized glyoxal, 50% dimethyl sulfoxide, 10 mM sodium phosphate, pH 6.5, at 50 "C for 1 h, cooled, and electrophoresed at 4 V/cm, with recirculation of electrophoretic buffer, in an 0.8% agarose gel containing 10 mM sodium phosphate, pH 6.5. The gel was incubated in 25 mM sodium phosphate, pH 6.5, for 15 min, and the RNA was transferred by blotting in the same buffer to a Genescreen membrane (Du Pont). The gel was stained with ethidium bromide and photographed at 300 nm before and after the transfer to verify equimolarity of RNA sample loads and the uniformity of transfer. The membrane was heated at 85 "C for 3 h, placed in 7% SDS, 1% bovine serum albumin, 0.5 M sodium phosphate, pH 7.0, sealed in a plastic bag, and prehybridized for 30 min at 65 'C. 32P-Labeled DNA probes were prepared by the method of Feinberg and Vogelstein (1984), using 0.1 pg of gel-purified DNA fragments. The heat-denatured DNA probes were added to the plastic bag and incubated with the membrane-immobilized RNA for -15 h at 65 "C. The membrane was then washed twice for 30 min at 65 T in 0.2% SDS, 0.3 M NaC1, 30 mM sodium citrate, pH 7.0.
The UBRZ-specific probe was a -5.3-kb SpeI-KpnI fragment of pSOB44 (Bartel et al., 1990) that is contained within the UBRl open reading frame. The UBCZ-specific probe was a -0.6-kb EcoRI fragment of pDG309 (see above) that included the entire 516-base pair open reading frame of UBC2. The probes were labeled with 32P and mixed together before hybridization. After completion of the Northern analysis, a control hybridization was carried out with the same membrane (without removing the hybridized UBRl and UBC2 probes), using a 32P-labeled ACT1 (yeast actin gene) probe (a -2-kb BamHI-PstI fragment of pTB392; a gift of J. Jones, McArdle Laboratory, University of Wisconsin, Madison, WI). Hybridization patterns were detected either by standard autoradiography with x-ray films or using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) (Johnston et al., 1990;Tobias et al., 1991) and quantified as described in the legend to Fig. 40.
Coimmunoprecipitation Assay-Cells from 20-ml exponential yeast cultures ( A m < 1) of the strain BBY68 (ubc2A ubrlA, Table I) transformed with various plasmids (see "Results") were collected by centrifugation at 2,500 X g for 5 min, resuspended in 0.5 ml of SD medium lacking methionine, and labeled with 0.2 mCi of [%]Translabel ( E N ) for 30 rnin at 30 "C. The labeled cells were pelleted by centrifugation, resuspended in 0.5 ml of DB buffer (50 mM NaCl, 1 mM Na-EDTA, 50 mM Na-HEPES, pH 7.5) containing protease inhibitors leupeptin, pepstatin A, antipain, chymostatin, and aprotinin (each at 20 pg/ml, Sigma), and lysed by vortexing with glass beads (Bartel et al., 1990). The extracts were centrifuged at 12,000 x g for 10 min. Immunoprecipitation assays were carried out with the supernatants by adding ascitic fluid (1 p1/0.3 ml of a supernatant) containing the anti-ha monoclonal antibody 12CA5 (Field et al., 1988) (Babco Inc., Richmond, CA) to the single or mixed extracts described under "Results" and in the legend to Fig. 2. All samples also contained a -10-fold excess of an identically prepared yeast extract from untransformed, unlabeled BBY68 (ubc2A ubrlA) cells. The samples were incubated on ice for 2 h; 20 pl of Protein A-Sepharose (Repligen, Cambridge, MA) was then added, and the suspensions were incubated, with end-over-end rotation, for 1 h at 4 "C, followed by a 5-8 centrifugation in a microcentrifuge. Pellets were washed four times with 0.8 ml of DB buffer containing 5% glycerol, resuspended in electrophoretic sample buffer, heated at 100 "C for 3 min, and subjected to SDS-PAGE (13%), followed by fluorography. '%-Labeled SDS-PAGE protein standards were from Amersham Corp. Interaction Cloning of Ubc2-binding Protein Domains-The S. cerevisiae UBC2 gene containing an EcoRI site immediately 5' to its start (ATG) codon (a gift from R. D. Gietz) was isolated as a 525base pair EcoRI-SalI fragment and ligated into EcoRIISalI-cut pMA424 (Ma and Ptashne, 1987), yielding the plasmid pKM1210-1, which expressed a fusion containing the DNA-binding domain of Gal4 upstream of the complete sequence of Ubc2. The S. cereuisiae strain GGY1::171 (Gill and Ptashne, 1987) was transformed (Schiestl and Gietz, 1989) with pKM12lO-1. His+ transformants were selected on DOBA-his plates (BIOlOl), yielding KMY528 (Table I). This strain was transformed with DNA pools from the three libraries pGAD1,2, and 3, in which yeast genomic DNA fragments (produced by a partial Sau3A digestion) had been fused, in three different frames, to an upstream fragment that encoded the activation domain of Gal4 (Chien et al., 1991; a gift from P. Bartel, SUNY, Stony Brook, NY). Transformants were selected on DOBA-his-leu plates (BIOlOl), and -1.5 X lo6 colonies were replica-plated onto the same medium containing the chromogenic 8-galactosidase substrate XGal (Chien et al., 1991).
This screen yielded 14 blue transformants that remained blue upon restreaking. The HZS3-based pKM1210-1 was evicted from these transformants by growing them on histidine-containing medium. The library-derived plasmids were isolated from the resulting His-transformants as described by Hoffman and Winston (1987) and transformed into the Escherichia coli strain DH5a The amplified plasmids were purified from E. coli and transformed into KMY528 (Table I), confirming the results of the initial XGal screen. Nucleotide sequencing, using the chain termination method (Ausubel et al., 1989), of the yeast DNA inserts in the plasmids showed that all but one insert encoded Ga14. This insert was found to encode a -170-residue Cterminal fragment of the 1,950-residue Ubrl protein (plasmid pKM1312, the exact number of Ubrl residues encoded by the pKM1312 insert is not given because the corresponding sequencing gel was unreadable close to the sequencing primer site; data not shown). In a negative control, pKM1312 was transformed into the GGY1::171 strain carrying pEE5, a plasmid that expressed a fusion of the DNA-binding domain of Gal4 to Snfl, a protein presumably unrelated to Ubc2 (Yang et al., 1992). As expected, these transformants yielded white colonies on XGal plates.

RESULTS
The Acidic C-terminal Region of Ubc2 Is Required for Degradation of N-end Rule Substrates-Degradation of proteins by the N-end rule pathway was assayed in vivo using derivatives of E. coli @-galactosidase (@gal) as model substrates. In eukaryotes, Ub fusions such as Ub-X-@gal are precisely deubiquitinated either in vivo or in cell-free extracts by Ub-specific processing proteases to yield X-@gal test proteins bearing the desired residue X at the N-terminus (Bachmair et al., 1986;Gonda et al., 1989). In contrast to the function of Ub in protein degradation, the role of Ub in these engineered Ub fusions is to allow the in uiuo generation of otherwise identical proteins bearing different N-terminal residues. Depending on the identity of X, the X-@gal proteins are either long-lived or metabolically unstable, with destabilizing Nterminal residues conferring short half-lives on the corresponding X-@gals (Bachmair et al., 1986;Tobias et al,, 1991). This N-terminal degradation signal (the N-degron) is manifested as the N-end rule.
Previous work (Dohmen et ai., 1991a) has shown that the normally short-lived N-end rule substrates such as Arg-@gal are long-lived in S. cereuisiae strains lacking the UBC2 gene. One feature of the 172-residue Ubc2 Ub-conjugating enzyme is its highly acidic 23-residue C-terminal region which contains 20 acidic residues (Asp or Glu) (Fig. lA, Reynolds et al., 1985). The acidic "tail" of Ubc2 is required for some but apparently not for all of the Ubc2 functions. In particular, strains whose mutant Ubc2 protein lacks the acidic tail are unable to sporulate (as homozygous ubc2 diploids), but they are nearly wild-type in their ability to carry out DNA repair and induced mutagenesis, in marked contrast to congenic ubc2A strains (Morrison et al., 1988).

N-recogninlUbc2 Interactions 12049
Arg-Bgal) or Met-@gal (Ub-Met-Bgal) (see the legend to Fig.  1). Arg is a type 1 primary destabilizing residue in the N-end rule, whereas Met is a stabilizing residue (see Introduction). Metabolic stabilities of Arg-(?gal and Met-(?gal were compared by determining their intracellular concentrations (Fig. 1B). Previous work (Bartel et al., 1990;Balzi et al., 1990;Baker and Varshavsky, 1991;Dohmen et al., 1991a) has shown that the steady state level of an X-(?gal protein in yeast cells is a sensitive indicator of its metabolic stability; compare, for example, the levels of Met-@gal (till > 20 h) and Arg-(?gal ( t 1 / 2 of -3 min) (Bachmair et al., 1986) expressed from identical vectors in cells that also expressed the wild-type UBC2 gene (Fig. lB, Ubc2-172). However, in the congenic ubc2A cells, the level of Arg-Bgal was much higher, close to that of Met-Bgal (Fig. lB, &PA), in agreement with earlier findings (Dohmen et al., 1991a). As can be seen from the data in Fig. lB, the shortening of the 23-residve acidic tail of Ubc2 by 8 residues (yielding Ubc2-164) did not impair the ability of Ubc2 to support the degradation of Arg-(?gal, whereas the Ubc2-153 derivative which lacked all but 4 residues of the acidic tail was severely impaired in this function (Fig. lB,. The residual activity of Ubc2-153 in the N-end rule pathway could be eliminated by deleting the 4 remaining residues of the acidic tail (Fig. lB,. The level of the normally short-lived Arg-(?gal in cells expressing the entirely tailless Ubc2-149 was indistinguishable from the level of the normally long-lived Met-(?gal (Fig. lB,. In contrast, the level of Arg-Bgal was significantly (and reproducibly) lower than that of Met-(?gal in ubc2A cells (Fig. lB, Ubc2A). Cells lacking Ubc2 are known to contain elevated levels of other Ub-conjugating (E2) enzymes, in particular Ubc4 (data not shown). This apparently "reciprocal" cross-regulation among E2 enzymes suggests the following explanation for a paradoxically higher residual activity of the N-end rule pathway in the total absence of Ubc2 than in the presence of the tailless Ubc2. First, it is assumed that at least some of the E2 enzymes other than Ubc2 have a weak but significant affinity for the UBRl-encoded N-recognin. Second, it is assumed that expression of either the wildtype Ubc2 or its tailless derivative such as Ubc2-149 downregulates the expression of the postulated "cross-complementing" E2 enzymes. This model could account for the above results (Fig. lB, Ubc2-149 versus UbcPA), because a (hypothetical) cross-complementing E2 enzyme would be induced in a ubc2A strain but not in a ubc2-149 strain. Examples of at least partial functional complementation between different E2 enzymes in yeast have been reported previously. For instance, overexpression of the yeast UBC5 gene partially complements proteolytic defects of the h 4 A mutant (Seufert and Jentsch, 1990).

The Acidic C-terminal Region of Ubc2 Is Required for High
Affinity Binding of Ubc2 to N-recognin-Earlier coimmunoprecipitation experiments have shown that the UBRl-encoded N-recognin is physically associated with Ubc2 (Dohmen et al., 1991a). We asked whether a Ubrl.Ubc2 complex that is stable enough to be detectable in a coimmunoprecipitation assay requires the presence of the acidic C-terminal region in Ubc2. A modified Ubrl protein used in these experiments had its C-terminus extended with a 9-residue sequence derived from hemagglutinin (ha) of influenza virus and containing an epitope (the ha tag) recognizable by the monoclonal antibody 12CA5 (Bartel et al., 1990;Dohmen et al., 1991a). The epitopetagged Ubrl (Ubrl-ha) was functionally active, in that the UBRl -ha gene and the unmodified UBRl gene were indistinguishable in their ability to complement the ubrlA mutant (Bartel et al., 1990). When overexpressed from the ADHl promoter on a high copy plasmid, the -226-kDa Ubrl-ha protein could be immunoprecipitated from ["S]methioninelabeled cell extracts with the anti-ha monoclonal antibody ( Fig. 2A, lane b; compare with lone a ) (dots in Fig. 2A mark the two Ubrl-unrelated yeast proteins that cross-reacted with the anti-ha monoclonal antibody; these proteins were also present in cells lacking the UBRl and URRl-ha genes, and in addition almost comigrated with Ubrl-ha upon SDS-PAGE; see also Bartel et al. (1990) and Dohmen et af. (1991a)).
Extracts were prepared from ["S]methionine-labeled ubrlA ubc2A cells transformed with either a control plasmid (vector alone), a plasmid expressing Ubrl-ha, or a plasmid expressing the wild-type Ubc2 or its mutated derivatives. The control  (Rartel et al., 1990)); pAI)HURC2, pJD647, pJD617, pJD616. and pJD615 (the YEplacl95ADH-had plasmids express-in& respectively, Ubc2-172 (wt), UhcZ-CMA, Ubc2-164, Uhc2-1.53 and Ubc2-149 from the ADHI promoter) (see "Experimental Pmedures"). Extracts from %labeled cells (-1 X 10' acid-insoluble counts/minute/extract) carrying either the control plasmid (vector alone, labeledatop the lanes a, c, e, R, i, and k) or the plasmid expressing Ubrl-ha (labeled + atop the lanes 6. d, f. h, j , and I ) were mixed either with a labeled extract from "S-labeled control cells (carrying vector alone; lanes a and 6) or with extracts from ".%-labeled congenic cells carrying plasmids that expressed Uhc2 and its derivatives (-0.75 X lo' acid-insoluble counts/minute/extract): Ubc2-172 (wild type) (lanes c and d ) ; Ubc2-CfMA (kanes e and f);   and h); Ubc2-153 (lanes i and j ) ; and Ubc2-149 (lanes k and I ) . After removing a portion of each sample for the analysis of total protein, the mixed extracts were subjected to immunoprecipitation with a monoclonal antibody to the ha epitope, followed by SDS-PAGE (13%) and fluorography (see "Experimental Procedures"). A, A 3-h fluorographic exposure of an upper portion of the gel that shows the immunoprecipitated -226-kDa Ubrl-ha (lanes 6, d, f. h. j ,  and 1). DoLv mark two unrelated yeast proteins (present in all lanes) that cross-react with the anti-ha antibody (Dohmen et 01,. 1991a). R. see also main text). C, the same molecular mass range as in R, in an identically run gel that was loaded with mmples (1% of total "S counts/minute) from each of the mixed extracts before immunoprecipitation, confirming the approximate equimolarity of Uhc2 variante in the initial extracts. Positions and lengths of Uhc2-172 and iLq C terminally truncated derivatives are indicated by numbers on the right. extract (lacking both Ubrl-ha and Ubc2 or its derivatives) was processed for immunoprecipitation with the anti-ha monoclonal antibody either alone (Fig. 2, A and B, lane a ) or after having been mixed either with the extract containing Ubrl-ha (Fig. 2, A and B, lune b ) or with the extract containing Ubc2 (or its derivatives) (Fig. 2, A and B, lanes c, e, g, i, and  k). Each of the Ubc2-containing extracts was also mixed with the extract containing Ubrl-ha, and the mixed extracts were incubated with the anti-ha monoclonal antibody (Fig. 2, A  and B, lunes d, f, h, j , and I). The antibody precipitated the Ubrl-ha protein from all samples that contained Ubrl-ha ( Fig. 2A, lanes b, d, f, h, j , and 1). Crucially, the antibody did not precipitate proteins the size of Ubc2 (19.7 kDa) or its derivatives from samples containing Ubc2 but lacking Ubrlha (Fig. 2, A and B, lanes c, e, g, i, and k). However, immunoprecipitation from the sample containing both Ubrl-ha and the wild-type Ubc2 protein yielded not only Ubrl-ha but also a protein the size of Ubc2 (Fig. 2, A and B, lanes d; compare with lane c), confirming the earlier finding that the wild-type Ubc2 is physically associated with Ubrl (Dohmen et al., 1991a).
Similar assays with Ubrl-ha and the C-terminally truncated Ubc2 derivative Ubc2-164, which lacks a portion of the acidic tail but is fully active in the N-end rule pathway (Fig. lB,, also resulted in the coimmunoprecipitation of Ubc2-164 and Ubrl-ha (Fig. 2, A and B, lane h; compare with lane g). The same experiment was carried out with more extensively truncated Ubc2 derivatives, Ubc2-153 and Ubc2-149. The former lacks all but 4 residues of the acidic tail and is nearly inactive in the N-end rule pathway, while Ubc2-149 lacks the entire acidic tail and has no detectable activity in the pathway (Fig. lB, Ubc2-153 and Ubc2-149). Neither Ubc2-153 nor Ubc2-149 were coimmunoprecipitated with Ubrl-ha, in contrast to either the wild-type Ubc2 or Ubc2-164 (Fig. 2, A and B, lanes j and I; compare with, for instance,

lane h ) .
We conclude that a portion of the 23-residue acidic tail of UbcP that is longer than 4 residues but shorter than 16 residues is sufficient for both a high affinity (coimmunnoprecipitation assay-detectable) binding of Ubc2 to Ubrl and for the function of Ubc2 in the N-end rule pathway. Regions of the Ubc2 enzyme other than the acidic tail may also be involved in functionally relevant interactions with N-recognin, or in other N-end rule-related functions, because neither the Ubc3 (Silver et al., Kolman et al., 1992) nor the Ubc4 Ub-conjugating enzyme that have been modified by the addition of the Ubc2-specific acidic tail could complement the N-end rule defect of a ubc2A train.^

N-recognin Is Bound to a -22-kDa Protein-The anti-ha
antibody precipitated not only the -226-kDa Ubrl-ha and the -20-kDa UbcP (see above), but also a protein with an apparent molecular mass of -22 kDa, termed p22 (Fig. 2B,  lanes b, d, f, h, j , and 1). The presence of p22 in an anti-ha immunoprecipitate required the presence of Ubrl-ha in the initial extract but did not depend on the presence of Ubc2 (data not shown and Fig. 2B, lanes b and d; compare with  lanes a and c). Thus, the apparent association of p22 with Nrecognin does not require an intact N-recognin . Ubc2 complex. The possibility that p22 is a C-terminal proteolytic fragment of Ubrl-ha that bears the ha tag and is therefore immunoprecipitable with the anti-ha antibody was tested by constructing a Ubrl variant that bore two consecutive C-R. J. Dohmen, K. Madura, and A. Varshavsky, unpublished data (a plasmid expressing a derivative of Ubc3 that bore the Ubc2-specific acidic tail was kindly provided by M. Ellison, University of Alberta, Edmonton, Canada). Levels of @gal activity in the S. cerevisiae stain YPH500 (UBRl UBCZ) carrying the LEU2-based, high copy plasmid pRL2 that expressed Ub-Arg-@gal (Bartel et al., 1990, and data not shown), and also either the plasmid YEplacl95ADH (vector alone; control) or its derivatives (pADHUBC2, pJD615, and pJD647) that expressed, respectively, Ubc2-172 (wild-type), Ubc2-149, and Ubc2-C88A from the ADHl promoter (see also "Experimental Procedures" and Table   I). Values shown are the means of four measurements carried out with four independent transformants. Standard deviations are shown above the burs.
terminal ha tags: Immunoprecipitation from an extract containing Ubrl-ha-ha should have resulted in a detectable (-1.5 kDa) increase in the size of the p22 protein if it were a Cterminal (ha-bearing) fragment of Ubrl-ha-ha. No such increase was observed (data not shown), suggesting that p22 is a distinct Ubrl-associated protein.
A Dominant Negative Allele of UBC2-Cys-88, the only Cys residue in the Ubc2 Ub-conjugating enzyme, has been shown to be the site of Ub thioester formation; replacement of Cys-88 in Ubc2 with either Ala or Val eliminates the Ub-conjugating activity of Ubc2 and renders it biologically inactive (Sung et al., 1990). As could be expected from these earlier findings, the Ubc2-C88A (Cys-Ala) derivative was found to be inactive in the N-end rule pathway (Fig. lB, Ubc2-C88A). However, in contrast to the effect of overexpressing the tailless Ubc2-149 (which does not bind to N-recognin and is also inactive in the N-end rule pathway; see above), overexpression of Ubc2-C88A in wild-type (UBC2) cells resulted in strong inhibition of the N-end rule pathway (Fig. 3). No such inhibition was observed upon overexpression of the wild-type UBC2 (Fig. 3). A likely reason for the difference between the effects of overexpressing Ubc2-149 and Ubc2-C88A is indicated by the finding that Ubc2-C88A, unlike Ubc2-149, retains the ability of the wild-type Ubc2 to form a high affinity complex with the UBRl-encoded N-recognin (Fig. 2, A  Thus, ubcZ-C88A is a dominant negative allele of the UBC2 gene, whose interference with the N-end rule pathway is likely to result from competition between the wild-type Ubc2 and Ubc2-C88A for binding to N-recognin. Although the Ubc2 regions in contact with N-recognin are unlikely to be confined to the acidic tail of Ubc2 (see above), the overall affinity of "non-tail" Ubc2 regions for N-recognin is at most low: Ubc2-149 is not coimmunoprecipitated with N-recognin (Fig. 2), and the inhibition of the N-end rule pathway by Ubc2-149 is much weaker than the inhibition by Ubc2-C88A (Fig. 3, compare Ubc2-149 with UbcZ-C88A).
Ubc2-C88A, which lacks the thioester-forming Cys residue of the wild-type Ubc2, cannot accept Ub from the Ub-activating (El) enzyme (Sung et al., 1990). The apparently undiminished affinity of Ubc2-C88A for N-recognin (see above) suggests that interactions between N-recognin and Ubc2 do not involve the thioester-bound Ub moiety of the wild-type Ubc2 (at the in vivo concentrations of U b , Ubc2, and the UBAl -encoded El enzyme, the bulk of Ubc2 is likely to bear the thioester-bound Ub moiety (Pickart, 1988)). Ubc2-interacting Domain of N-recognin-To identify the domains of N-recognin that interact with the Ubc2 Ub-conjugating enzyme in vivo, we have used the "two-hybrid" interaction cloning technique of Fields and Song (1989). In this method, the expression of the E. coli lac2 reporter gene from the GAIA-dependent P G A L l promoter in S. cerevisiae is designed to require the reconstitution of Gal4 activity from two unlinked but interacting protein domains. Specifically, the DNA-binding domain of the Gal4 protein is expressed in yeast as a fusion whose C-terminal portion is a polypeptide of interest. These cells are transformed with a library in which quasi-randomly generated DNA fragments (from yeast or other species) are fused to a fragment that encodes the Gal4 transcription activation domain. Transformants expressing a polypeptide from the library that has a significant affinity to the polypeptide of interest can be identified as those that induce their resident PGALl-hZ reporter gene because the DNA-binding and activation domains of Gal4 are brought into spatial proximity through the interaction of polypeptides fused to each domain (Fields and Song, 1989;Chien et al., 1991;Chevray and Nathans, 1992;Yang et al., 1992). The S. cerevisiae strain GGY1::171, which lacks both GAL4 and GAD0 genes and contains an integrated PGALl-lacZ reporter construct (Chien et al., 1991), was transformed with the plasmid pKM1210-1 which expressed the Gal4 DNAbinding domain fused to the N-terminus of the full-length Ubc2. The resulting strain, KMY528, was transformed with PGAD-based DNA libraries that expressed the DNA-binding domain of Gal4 fused (in all three frames) to partial Sau3A fragments of the yeast genomic DNA (Chien et al., 1991; see "Experimental Procedures"). 14 lacZ-expressing transformants were identified among -1.5 X lo6 colonies screened. While most of the pGAD plasmids from the positive transformants carried the intact GAL4 gene (the pGAD libraries were derived from a GAIA yeast strain), one transformant was found to express a -170-residue C-terminal fragment of N-recognin (Ubrl) fused to the Gal4 activation domain. The corresponding pGAD-based plasmid, termed pKM131.2, reproducibly activated the expression of the PGALl-hZ reporter gene upon retransformation of the KMY528 strain (GGY::171 carrying pKM1210-1). In a control experiment (data not shown), the same plasmid did not activate the expression of P G A L I -~~ in GGY1::171 carrying the plasmid pEE5, which expressed the DNA-binding domain of Gal4 fused to a region of a presumably unrelated protein Snfl (Yang et al., 1992). We conclude that a C-terminal region of N-recognin that encompasses approximately 10% of this 1,950-residue protein contains a Ubc2-interacting domain.
Regulatory Interactions between UBCZ and UBRI-In the course of earlier experiments (Dohmen et al., 1991a), it was noticed that Ubrl and Ubc2 could not be overproduced simultaneously within the same yeast cells. It was subsequently found that the levels of UBRl mRNA (expressed from the strong P A D H I promoter on a high copy plasmid) were high in the absence of UBCZ expression (in ubc2A cells) but much lower if cells carried a high copy plasmid overexpressing UBC2 from the PADHI promoter. In addition, the copy number of a UBRl -expressing plasmid was decreased in the presence of a plasmid expressing UBCZ but not in the presence of an otherwise identical plasmid lacking UBC2, suggesting that simultaneous overexpression of UBRl and UBC2 is cytotoxic.' To verify and analyze this effect, we constructed a high copy plasmid that expressed UBRl from the copper-inducible Pcupl promoter, as well as a low copy (CEN-based) plasmid that expressed UBCZ from the galactose-inducible PGAL~ promoter. Both plasmids were transformed into the ubrl A ubc2A strain KMY618, yielding the strain KMY633 (Table I), which was propagated in the raffinose-containing SR medium in the absence of added copper ions, conditions which do not induce the P G A L l and Pcupl promoters. (Induction of the P G A L~ promoter by galactose in cells that have been growing in the presence of raffinose is much faster than in cells that have been growing in the presence of dextrose (Griggs and Johnston, 1991;Johnston, 1987).) The strain KMY633 was grown in SR medium containing 0.1 mM CuS04 (conditions that induce PCUpl-UBRl but not PGALI-UBCZ) to a midexponential phase (Asoo 5 l ) , followed by the addition of galactose to a final concentration of 3% (conditions that induce PGALl-UBCS as well). RNA was isolated from cells either immediately before or (at hourly intervals) after the addition of galactose, followed by Northern analysis with UBRl and UBCZ hybridization probes (Fig. 4). While no expression of UBC2 was detectable in SR medium, UBC2 mRNA accumulated to a high level within 1 h after the addition of galactose (Fig. 4 A , UBC2; compare lanes a and b).
Conversely, the level of UBRl mRNA was high in SR medium but decreased by approximately 2.5-fold within 1 h after the addition of galactose, and remained low afterward (Fig. 4A, UBRI; compare lanes a and b-d; see also Fig. 40). The rapid decrease of UBRl mRNA level upon induction of UBC2 is likely to be caused by metabolic destabilization of UBRl mRNA, inasmuch as both UBCZ and UBRl were expressed from heterologous promoters that are active in a coppersupplemented, galactose-containing medium. (Similar results were obtained when UBRl was expressed from the P A D H l promoter or from its natural (PUBR1) promoter (data not shown).) Moreover, since the doubling time of the KMY633 cells in this medium is approximately 3 h at 30 "C (data not shown), the rapidity of the observed decrease in the level of UBRl mRNA (Fig. 4, A and D ) is unlikely to result from dilution of this mRNA due to cellular growth. A UBCZinduced destabilization of UBRl mRNA remains to be verified directly.
The inhibitory effect of UBC2 expression on the level of UBRl mRNA was found to require the region that encodes the acidic C-terminal tail of Ubc2 (Fig. 1A): no significant decrease in the level of UBRl mRNA was observed upon induction of ubc2-148 which encoded a tailless Ubc2 (Fig. 4  to galactose-containing medium did not, by itself, cause the observed decrease in UBRl mRNA. To determine whether the UBRl-suppressing effect of Ubc2 also required a functional Ubrl protein, we asked whether the induction of UBC2 decreased the level of an mRNA produced from an insertion/ frameshift ubrl allele that encoded a 170-residue N-terminal region of the 1,950-residue Ubrl protein. No significant effect of UBCZ on this ubrl allele was observed (Fig. 4, C and D ) , strongly suggesting that a functionally active Ubrl protein is required for down-regulation of UBRl mRNA by UBC2.
A likely explanation of these findings (Fig. 4) is that the N-recognin . Ubc2 complex participates in the regulation of the UBRl-encoded N-recognin. For reasons considered above this regulation appears to involve an enhanced degradation of UBRl mRNA upon an increase in UBC2 expression.

DISCUSSION
Our previous work has shown that the yeast N-recognin is physically associated with Ubc2, one of at least 10 Ub-conjugating enzymes in S. cereuisiae and an essential component of the yeast N-end rule pathway (Dohmen et al., 1991a). Further analysis resulted in the following main findings.
(i) Both physical stability and functional activity of the Nrecognin . Ubc2 complex require the presence of a 23-residue C-terminal region in Ubc2 that contains 20 acidic residues (Asp or Glu).
(ii) A Ubc2-interacting domain of the UBRl-encoded Nrecognin is located within the 170-residue C-terminal region of the 1,950-residue N-recognin. Whether this is the only Ubc2-interacting region of N-recognin and whether this region interacts with the C-terminal acidic region of Ubc2 remains to be determined.
(iii) The ubc2-CB8A gene, in which the codon encoding the active-site Cys of Ubc2 has been replaced by an Ala codon, acts as a dominant negative allele of UBC2, most likely as a result of competition between the wild-type Ubc2 and its enzymatically inactive variant for binding to N-recognin.
(iv) Expression of UBC2 decreases the level of UBRl mRNA which encodes N-recognin. This effect requires the intact N-recognin-Ubc2 complex and is likely to result from enhanced degradation of UBRl mRNA upon an increase in UBC2 expression.
Targeting Complex of the N-end Rule Pathway-The known components of the S. cereuisiae N-end rule pathway that mediate steps prior to or distinguishable from the actual proteolysis of a substrate are N-terminal amidase (Nt-amidase), encoded by the NTAl gene: Arg-tRNA-protein transferase, encoded by the ATE1 gene (Balzi et al., 1990), Nrecognin, encoded by the UBRl gene (Bartel et ul., 1990), the Ub-conjugating (E2) enzyme, encoded by the UBC2 gene (Reynolds et al., 1985;Jentsch et al., 1987;Dohmen et al., 1991a), and the Ub-activating (El) enzyme, encoded by the UBAl gene (McGrath et al., 1991). In addition to direct (immunoprecipitation-based) evidence for the physical association between N-recognin and Ubc2 (see "Results"), there is also circumstantial (overexpression-based) evidence for the existence of a ternary complex between N-recognin, Arg-tRNA-protein transferase, and Nt-amidase' (these latter components of the pathway modify N-termini of certain N-end rule substrates prior to their binding by N-recognin; see Introduction).
Proteins that participate in a complex metabolic transformation are often physically associated. The examples of a multienzyme assembly range from pyruvate dehydrogenase (synthesis of acetyl-coA) to replisome (DNA replication), spliceosome (RNA splicing), ribosome (protein synthesis), and proteasome complexes (protein degradation). Mechanistic advantages of such assemblies stem from their increased fidelity due to often present editing capabilities, and also from the properties of processivity and channeling that are central to the functioning of multienzyme machines. For example, the association between N-recognin and the Ubc2 Ub-conjugating enzyme not only endows the latter with the ability to recognize a substrate that bears a destabilizing N-terminal residue but may also contribute to processivity of the Ubc2mediated synthesis of a substrate-linked multi-Ub chain.
The targeting complex of the S. cereuisiae N-end rule pathway is likely to contain at least five components, Ntal, Atel, Ubrl, and Ubc2 (whose subunit molecular masses are, respectively, 52, 58, 225, and 20 kDa), and also the at least transiently associated 114-kDa Ubal (El) enzyme, which must be bound to Ubc2 during the E1+E2 transfer of an activated Ub moiety. That this list is incomplete is suggested by the R. Baker and A. Varshavsky, manuscript in preparation.  finding of a -22-kDa protein (distinct from the 20-kDa Ubc2) in association with N-recognin (see "Results") and by the earlier data indicating that a partially purified Arg-tRNAprotein transferase from rabbit reticulocytes is a -360-kDa complex of several subunits of Arg-tRNA-protein transferase associated with several subunits of Arg-tRNA synthetase (Ciechanover et al., 1988). If the yeast Arg-tRNA-protein transferase (encoded by A T E l ) is organized similarly to its mammalian counterpart, the mass of the entire targeting complex in S. cereuisiae could be about 1,500 kDa. In uiuo, this complex should be bound to the rest of proteasome at least some of the time. The complete proteolytic machine that implements the N-end rule is thus a strikingly diverse assembly of enzymes and binding factors whose combined size is comparable to those of ribosomal subunits.
Evolution of the Targeting Complex-Homologs of the S. cereuisiae Ubc2 enzyme are present in all eukaryotes examined, from yeasts to mammals. However, while a slightly shorter counterpart of the acidic C-terminal tail of the S. cereuisiae Ubc2 is also present in Ubc2 from the related budding yeast Kluyueromyces l~c t i s ,~ this region is absent from the otherwise close homologs of Ubc2 such as the Rhp6 Ubconjugating enzyme of the fission yeast Schizosacchuromyces pombe (Reynolds et al., 1990), and from the Drosophila, rabbit, and human homologs of UbcP as well (Koken et al., 1991a;Wing et al., 1992;Schneider et al., 1990;Koken et al., 1991b).
In S. cereuisiae, the tailless derivative of Ubc2 is inactive in sporulation and in the N-end rule pathway but is able to support other Ubc2 functions such as DNA repair and induced mutagenesis at nearly wild-type levels (see " Results" and Morrison et al., 1988). The sporulation defect in cells containing exclusively the tailless UbcP is unrelated to their lack of the N-recognin-Ubc2 complex because S. cereuisiae is capable of sporulation in the absence of N-recognin and hence in the absence of the N-end rule pathway (Bartel et al., 1990). Thus, it is likely that the acidic tail of Ubc2 is also required for an interaction with an unknown Ubc2-specific recognin whose function is essential for sporulation in S. cereuisiae.
The naturally tailless Ubc2 homolog Rhp6 of S. pombe was shown to complement the known phenotypes of ubc2A S. cereuisiae except for its defect in sporulation (complementation of the N-end rule defect has not been tested) (Reynolds et al., 1990). Conversely, the tailless derivative of the S. cereuisiae Ubc2 enzyme could complement all of the tested defects of an rhp6A S. pombe mutant, including its sporulation defect (Reynolds et al., 1990). Either the wild-type or tailless Ubc2 of S. cereuisiae could function as an N-end rule-mediating enzyme in an extract from rabbit reticulocytes (Sung et al., 1992).
The tailless Ubc2 might have preceded its acidic tail-containing variant in a lineage of organisms that yielded S. cereuisiae and R lactis. In this model, N-recognin and UbcP of the S. cereuisiae lineage (but not the homologous protein pairs in predecessors of fission yeast and larger eukaryotes) coevolved in a way that resulted in a positively charged surface of a domain in N-recognin accommodating the multiple Asp and Glu residues of the acidic tail in Ubc2. Throughout this coevolution, UbcP would be expected to retain its affinities for other recognins (which mediate other functions of Ubc2) in ways that did not depend on the presence of an acidic tail in Ubc2. However, the converse interpretation, in which a tail-bearing Ubc2 and a correspondingly adapted N-recognin is the ancestral configuration, is also plausible, in part because nothing is known about the nature of selective pressures that could underlie a tail-related coevolution of Ubc2 and Nrecognin in some but not in all species.

Regulatory Interactions in the N-end Rule Pathway-The
intact N-recognin . Ubc2 complex was shown to be required for an observed decrease in the level of UBRI mRNA (which encodes N-recognin) upon an increase in UBC2 expression. A variety of evidence suggests that a decrease in the level of UBRl mRNA is brought about by a UBC2-mediated decrease in its metabolic stability (see "Results"). It remains to be seen whether this new effect of Ubc2 is mediated by the presently known function of the N-recognin Ubc2 complex, i.e. by an N-end rule-mediated degradation of a short-lived protein that directly or indirectly stabilizes UBRl mRNA, or whether a proteolysis-independent mechanism is involved. Expression of genes that encode interacting proteins is ' P. R. H. Waller and A. Varshavsky, unpublished data. often coregulated to maintain an appropriate concentration of a multiprotein complex as well as correct stoichiometries of its components. The effect of UBC2 expression on the level of UBRl mRNA (see "Results") is a likely example of such a regulation. The expression of S. cerevisiae genes encoding Ntamidase and Arg-tRNA-protein transferase appears to be coregulated as well, presumably to maintain correct stoichiometry of these interacting enzymes.' The evidence for coregulation includes the presence of common nucleotide sequence motifs in upstream regions of genes that encode components of the N-end rule pathway.6