NH₄⁺-induced down-regulation of the Saccharomyces cerevisiae Gap1p permease involves its ubiquitination with lysine-63-linked chains

Degradation of many eukaryotic proteins requires their prior modification by conjugation with ubiquitin (Ub). Ub molecules are transferred to lysine residues of target proteins via an E1-E2-E3 enzyme thioester cascade (Ub-activating enzyme/Ubconjugating enzyme/Ub-protein ligase) (Ciechanover, 1994). Ubiquitination is known mainly as a signal targeting substrate proteins for recognition and degradation by a multisubunit protease, the 26S proteasome (Pickart, 1997). This has been observed both for soluble proteins and for some integral membrane proteins that undergo endoplasmic-reticulumassociated degradation. Another function of ubiquitination has more recently been evidenced. The ubiquitination of some plasma-membrane proteins signals their internalization via the endocytic pathway and subsequent degradation in the lysosome/vacuole (Hicke, 1997). In the case of several such proteins it has been clearly demonstrated that this degradation is exclusively vacuolar and occurs normally in mutants with defective proteolytic or regulatory proteasome subunits (Galan et al., 1996; Hicke and Riezman, 1996; Egner et al., 1995; Riballo et al., 1995). This notably raises the question of why the proteasome does not target these ubiquitinated proteins. Recognition of ubiquitinated proteins by the proteasome requires formation of a polyubiquitin chain at least four subunits long. The chain is formed through linkage of the carboxy-terminal Gly of one Ub subunit to the ε-amino group of Lys48 of the previous Ub. This type of chain binds to subunit 5a of the proteasome (Pickart, 1997; Deveraux et al., 1994). In addition to Lys48, Ub carries six other lysine residues, and alternative types of polyubiquitin have been shown to form in vivo in yeast through Lys29 and Lys63 (Arnason and Ellison, 1994). The uracil permease, which undergoes basal and stressaccelerated Ub-dependent endocytosis (Galan et al., 1996), was found to be specifically modified by short Lys63-linked polyubiquitin chains (Galan and Haguenauer-Tsapis, 1997). Formation of these chains is required for maximum-rate internalization of the permease, but mono-ubiquitination at two lysine acceptor sites is sufficient to induce basal-rate permease endocytosis. Another interesting case is that of Ste2p, the G-protein-coupled receptor of α-factor pheromone. This receptor undergoes both basal and ligand-induced Ub-dependent downregulation (Hicke and Riezman, 1996). It was observed that mono-ubiquitination of a single lysine residue is sufficient to promote both basal and ligand-induced internalization of a C-terminally truncated form of Ste2p (Terrel et al., 1998). Although ligand binding to full-length Ste2p leads to formation of di- and tri-ubiquitinated Ste2p conjugates, there is evidence that this may mainly reflect addition of single Ub moieties onto multiple receptor lysines rather than formation of a polyubiquitin chain on one lysine (Terrel et al., 1998). In contrast to this situation, the mammalian growth hormone receptor was reported to undergo ligand-induced multiubiquitination (Govers et al., 1997) followed by lysosomal degradation (Strous et al., 1996).

This report focuses on the nitrogen-regulated ubiquitination of the general amino acid permease (Gap1p) of S. cerevisiae. Gap1p is tightly regulated according to the nitrogen source present in the culture medium (Grenson, 1992; Roberg et al., 1997; Springael and André, 1998). Its activity is maximal in cells using proline or urea as the sole nitrogen source. Upon addition of ammonium ions supporting optimal growth, its synthesis is repressed and pre-synthesized permease is inactivated (Grenson, 1983a,b) through progressive internalization and subsequent degradation in the vacuole (Springael and André, 1998). NH₄⁺-induced inactivation of Gap1p requires the product of the NPI1 and NPI2 genes (Grenson, 1983a,b). NPI1 codes for Rsp5p, a ubiquitin ligase essential to cell viability and highly conserved from yeast to man (Hein et al., 1995; Huibregtse et al., 1995). A recent analysis has shown that in proline-grown cells, only a small fraction of Gap1p molecules are ubiquitinated. After NH₄⁺ addition, the proportion of ubiquitin-conjugated forms increases, leading to rapid endocytosis and degradation of the permease. Ubiquitination and subsequent degradation of Gap1p are impaired in an npi1 strain displaying severely reduced amount of Npi1p Ub ligase (Hein et al., 1995; Springael and André, 1998). Npi1p is also required for stressinduced ubiquitination and degradation of the uracil permease (Hein et al., 1995; Galan et al., 1996) and for glucose-induced endocytosis and degradation of the maltose permease (Lucero and Lagunas, 1997). The carboxy-terminal tail of Gap1 plays a critical role in NH₄⁺-induced down-regulation of the permease (Hein and André, 1997; Springael and André, 1998). This region contains a di-leucine peptide and a glutamate residue located in a putative α-helix. Mutations within these sequences or in the last eleven amino acids of Gap1 protect the permease against NH₄⁺-induced down-regulation, but these mutant permeases still bind ubiquitin (Springael and André, 1998).

We here report the molecular characterization of NPI2, a second gene required for NH₄⁺ inactivation of Gap1 (Grenson, 1983a,b). We show that this gene codes for the ubiquitin hydrolase Doa4p/Ubp4p/Ssv7p and is required for normal ubiquitination and NH₄⁺-induced down-regulation of the Gap1 permease. In npi2 cells, the amount of free monomeric ubiquitin is about 4 times lower than in wild-type cells. Effective ubiquitination and down-regulation can be restored in an npi2 strain by overproduction of Ub. We provide evidence, based on experiments with npi2 cells overproducing several mutant Ubs unable to form specific kinds of Ub chains, that Gap1p is ubiquitinated at two acceptor sites and that NH₄⁺ triggers formation of Lys63-linked polyubiquitin chains. We also show that formation of such chains accelerates permease internalization.

Strains 23346c (MATa, ura3); 27061b (MATa, ura3, trp1); 27002d (MATa, ura3, npi2), and 27071b (MATa, ura3, trp1, npi2) used in this study are all isogenic with the wild-type Σ1278b except for the mutations mentioned (Béchet et al., 1970). MHY501 (MATahis3- Δ200 leu2-3, 112 ura3-52 lys2-801 trp1-1) and its isogenic doa4Δ derivative (MHY693) are described by Papa and Hochstrasser (1993). Cells were grown in minimal buffered medium (pH 6.1) with 3% glucose as the carbon source (Jacobs et al., 1980). Since this medium apparently chelates copper, experiments requiring copper induction of ubiquitin were carried out on cells grown in YNB medium (Difco) without ammonium. Nitrogen sources were added as indicated at the following final concentrations: (NH4)2SO4 (10 mM), proline (0.1%), serine (0.1%), threonine (0.1%). The 2μ-based multi-copy plasmid YEp96 contains a synthetic yeast ubiquitin gene under the control of the copper-inducible CUP1 promoter (Hochtrasser et al., 1991). Plasmids encoding mutant ubiquitins in which Lys29 (UbK29R), Lys48 (UbK48R), Lys63 (UbK63R), or all three of these (UbRRR) were replaced by arginine, are derivatives of YEp96 (Arnason and Ellison, 1994). HA-UbK63R is identical to UbK63R except that it contains an haemagglutinin (HA) tag on the N terminus of Ub (Terrel et al., 1998). Overproduction of Ubs was induced with 0.1 mM CuSO₄ for two hours. Yeast cells treated with lithium acetate (Ito et al., 1983) were transformed according to the method of Sherman et al. (1986). The E.coli strain used was JM109. All procedures for manipulating DNA were standard ones (Ausubel et al., 1995; Sambrook et al., 1997).

S. cerevisiae strain 27002d (MATa, ura3, npi2) was transformed with a low-copy-number library representing the whole genome of strain Σ1278b (Marini et al., 1994). One of the 77000 Ura3⁺ transformants exhibited an Npi2⁺ phenotype on selection media (see text). DNA was isolated from this clone and used to transform E. coli. The recovered plasmid contained a DNA fragment about 6.5 kb long (YCpJYS-10). A 4 kb subclone constructed in the centromere-based vector pFL38 (YCpJYS-22) suppressed the phenotype linked to the npi2 mutation. The npi2 mutant allele was cloned by screening bacterial clones of a library representing the genome of the npi2 strain with an NPI2 probe. The NPI2 and npi2 genes were sequenced using oligonucleotide primers derived from the sequence of the YDR069c gene established by systematic sequencing of chromosome IV (Jacq et al., 1997).

Gap1p permease activity was determined by measuring incorporation of ¹⁴C-labelled citrulline (20 μM) as described by Grenson et al. (1966). All permease activities were measured in cells having reached the state of balanced growth (Wiame et al., 1985). The permease was inactivated by adding prewarmed (NH₄) ₂SO₄ to the culture at the final concentration of 10 mM.

Crude extracts were prepared as previously described (Hein et al., 1995). Membrane-enriched preparations were obtained as described elsewhere (Springael and André, 1998). For western blot analysis of Gap1p, solubilized proteins were loaded on a 12% SDS-polyacrylamide gel (11% in some experiments) in a Tricine system (Schägger and von Jagow, 1987). After transfer to nitrocellulose the proteins were probed with rabbit antiserum raised against the N-terminal region of Gap1p (1:20000) or of the plasma-membrane H+-ATPase (Pma1) (1:10000) (J. O. De Craene and B. André, unpublished). For anti-ubiquitin immunoblot analysis, cell extracts were prepared as previously described, except that they were boiled after resuspension in sample buffer. Proteins were transferred to Immobilon-P membranes (Millipore), after which the blots were boiled in water for 15 minutes, before saturation and incubation with polyclonal anti-ubiquitin antibodies (Sigma 1:3000). Primary antibodies were detected with horseradish peroxidase-conjugated (HRP-conjugated) anti-rabbit IgG secondary antibody followed by detection of chemoluminescence (ECL, Amersham).

The product of the NPI2 gene is required for NH₄⁺-induced loss of Gap1p permease activity (Grenson, 1983a,b). Recently we have shown that NH₄⁺ added to proline-grown cells triggers internalization and degradation of the permease and that this down-regulation requires conversion of Gap1p to Ub-conjugated forms (Springael and André, 1998). To investigate the role of NPI2 in down-regulation of Gap1p, we monitored the fate of Gap1p after addition of NH₄⁺ to cultures of wild-type and npi2 cells growing on proline (Fig. 1A and B). NH₄⁺ induced a loss of permease activity in wild-type cells and a parallel drop in permease immunodetected in total protein extracts. In contrast, npi2 cells were protected against both loss of permease activity and Gap1p degradation. This resistance of Gap1p to NH₄⁺-induced down-regulation might be due either to impaired ubiquitination of the permease, as in the npi1 mutant, or to a defect in a subsequent step such as endocytosis, as in the act1-1 mutant altered in actin (Springael and André, 1998). To further investigate the role of NPI2 in NH₄⁺-induced down-regulation of Gap1p, we tested the efficiency of Gap1p ubiquitination in npi2 cells. Immunoblotting was performed with membraneenriched fractions to facilitate detection of ubiquitin-permease conjugates (Fig. 1C). The Gap1p signal detected in prolinegrown wild-type cells consists of a doublet around 60 kDa and two minor bands of higher molecular mass. Previous work has shown that the minor upper bands correspond to Ub-conjugated forms of the permease (Springael and André, 1998). In the early minutes after addition of NH₄⁺, the amount of Ub-conjugated Gap1p increased markedly in wild-type-cells. After 30 minutes, while the control signal corresponding to plasma-membrane H⁺-ATPase remained stable, the intensity of the Gap1p signal was reduced and ubiquitin-permease conjugates were barely detectable (Fig. 1C). Proline-grown npi2 cells displayed a more intense Gap1p doublet than did proline-grown wild-type cells (Fig. 1C, lower panel), but this did not lead to higher permease activity. This sugests that the extra-amount of Gap1p could be intracellular or that some elements are limitant for Gap1p activity. The mutant cells, furthermore, displayed a lesser proportion of Ub-conjugated Gap1p forms. After addition of NH₄⁺, the amount of these forms increased but very slightly and the main Gap1p immunodetected signal remained nearly stable. Taken together, these results show that permease ubiquitination is strongly reduced in npi2 cells. This ubiquitination defect appears to stabilize Gap1p in proline-grown cells and to protect the permease against NH₄⁺-induced down-regulation.

The npi2 mutant was found to grow very slowly on minimal medium containing L-serine (MSer) or L-threonine (MThr) as the sole nitrogen source. Although this phenotype is still unexplained, it was used to clone the NPI2 gene by functional complementation. The recipient strain (ura3, npi2) was transformed with a low-copy-number plasmid library representing the genome of strain Σ1278b. Transformants were screened for growth on MSer medium. A single plasmid carrying a 6.5 kb insert complementing the npi2 mutation was recovered. Subcloning experiments and phenotypic analysis on MSer and MThr revealed that the complementing gene corresponds to the YDR069c/DOA4/SSV7 open reading frame encoding the ubiquitin hydrolase Ubp4p (Papa and Hochtrasser, 1993). A plasmid bearing the cloned NPI2 gene was introduced into the npi2 strain and tested for its ability to restore NH₄⁺-induced down-regulation of the Gap1p permease. As expected, addition of NH₄⁺ ions to this strain triggered down-regulation of Gap1p similar to that observed in the wildtype strain (Fig. 1A). To confirm that NPI2 is allelic with DOA4, we cloned a DNA fragment containing the YDR069c ORF of an npi2 strain and showed that it was unable to complement the npi2 mutation (data not shown). DOA4 belongs to a large family of genes encoding de-ubiquitinating enzymes, or Ubps, involved in removing Ub from other proteins (Wilkinson, 1997). Loss of DOA4 produces several phenotypes including thermosensitivity, a deficient general response to stress, and a deficient DNA replication (Hochtrasser, 1996). In doa4 cells, proteolysis of all tested substrates of the ubiquitin/proteasome pathway, such as the Matα2 repressor (Chen et al., 1993) and the Matα2-β-galactosidase fusion protein, is impaired (Papa and Hochtrasser, 1993). More recently it was shown that ubiquitination of the Fur4p uracil permease (Galan and Haguenauer-Tsapis, 1997) and Ste2p α-factor receptor (Terrel et al., 1998) and internalization and degradation of the maltose permease (Lucero and Lagunas, 1997) are impaired in cells with a defective Doa4p. Our data showing that NPI2 is identical to DOA4/UBP4 thus further demonstrate that lack of Doa4p Ub hydrolase affects ubiquitination of plasmamembrane proteins.

Ub-specific processing proteases (Ubps) (16 in S. cerevisiae) vary greatly in length and structural complexity, suggesting functional diversity (Wilkinson, 1997). While the deduced amino-acid sequences of these proteins display little similarity, sequence comparisons reveal two conserved domains, the Cys and His boxes. The Cys domain contains a Cys residue that serves as the active nucleophile of the enzyme. The His domain contains a His residue that contributes to the enzyme active site (Baker et al., 1992). Yeast Ubp-family members further display additional conserved boxes in the region between the Cys and His domains (Wilkinson, 1997). Sequencing of the npi2 mutant allele revealed a single A→T substitution resulting in replacement of tryptophan 782 by an arginine. This replacement occurs within one of the conserved blocks located between the Cys and His boxes. Although this 14-residue block (EVLGGDNXWYCPKC) is reported to occur in 13 of the 16 yeast Ubp enzymes (Wilkinson, 1997), we found it to be significantly conserved in only 8. We found, moreover, that this block is a part of a larger conserved sequence 33 amino acids long (Fig. 2). This 33-residue motif is also found in the human Tre2 proto-oncogene (Onno et al., 1993), the Drosophila Faf protein involved in eye development (Huang et al., 1995), and in several other Ubps present in other organisms. Alignment of these sequences reveals that a glutamate, a leucine, and two cysteine residues are invariant in all of them. We therefore propose to name this sequence ‘ELC box’. The aromatic residue replaced by arginine in the npi2 mutant is also highly conserved. Although we cannot exclude an effect of the npi2 mutation on the stability of the protein, our results are also consistent with an important role of the ELC box in the normal function of Npi2p/Doa4p. In keeping with both hypotheses, the uracil permease was recently found to be identically protected against stress-induced degradation in doa4 null-mutant cells and in cells with the above-mentioned npi2 point mutation (data not shown). The role of the ELC box remains to be determined, but its presence in a few members of the Ubp family suggests that these hydrolases might be functionally distinct from the other proteins of the same family.

Recent experiments have shown that overproduction of Ub in doa4Δ cells can reverse some of the doa4 phenotypes, notably sensitivity to canavanine or cadmium (Hicke, 1997) and defective ubiquitination of the Fur4p uracil permease (Galan and Haguenauer-Tsapis, 1997) and Ste2p α-factor receptor (Terrel et al., 1998). It was thus interesting to see whether NH₄⁺-induced ubiquitination and down-regulation of Gap1p are also restored. Wild-type and npi2 cells were transformed with a plasmid encoding a synthetic ubiquitin gene under the control of the inducible CUP1 promoter. Overproduction of ubiquitin did not affect permease down-regulation in wild-type cells, but in npi2 cells it totally restored NH₄⁺-induced Gap1p ubiquitination (as detailed below) and down-regulation (Fig. 3). This supports the view that the amount of Ub in npi2/doa4 cells might be limiting for some cellular processes (Galan and Haguenauer-Tsapis, 1997; Terrel et al., 1998). To investigate this point under our experimental conditions, we used antibodies raised against Ub to probe crude extracts prepared from wild-type, npi2, and doa4Δ cells growing exponentially on proline medium. No significant difference was evidenced in the high molecular mass Ub-conjugates present in all cells (not shown). In the low molecular mass ranges, wild-type cells displayed an intense signal corresponding to monomeric Ub and a faint signal likely corresponding to di-Ub (Fig. 4, lane 1). Both the doa4Δ and the npi2 strains displayed several bands in the region of di-Ub, probably corresponding to Ubs dimers linked to small oligopeptides previously reported to accumulate in doa4 cells (Papa and Hochtrasser, 1993), and more strikingly a reduced signal for free monomeric Ub (lanes 2 and 4). Serial dilution of the extracts prepared from wild-type cells analyzed on the same gel (lanes 5-7) allowed to estimate that the mutants contain about four times less monomeric Ub than the wild type. Overexpression of Ub from the inducible CUP1 promoter resulted in at least 20 fold increase in the steady state level of free Ub relative to that in the corresponding untransformed cells (compare lanes 3-4 to lanes 8-9). The level of free Ub in induced transformed npi2 cells thus exceeds that in untransformed wildtype cells. These data show that defective Gap1p downregulation in npi2/doa4 cells is due mainly to the limiting intracellular pool of free monomeric Ub.

The possibility of restoring normal ubiquitination of at least some proteins in doa4 cells by overproduction of Ub was recently used to study the type of ubiquitination undergone by two cell-surface proteins (Galan and Haguenauer-Tsapis, 1997; Terrel et al., 1998). In these studies, various mutant Ubs unable to form certain types of Ub chain were overproduced in the doa4 cells. We have used here a similar approach to analyze Gap1 ubiquitination before and after addition of NH₄⁺. Overproduction of ubiquitin in wild-type and npi2 cells grown on proline medium led to the presence of two intense bands above the major Gap1 signal (Fig. 5, lanes 1 and 5). These upper bands, previously detected as much weaker signals in wild-type cells, correspond to ubiquitinated forms of the protein (see Springael and André, 1998, and Fig. 1C). Accumulation of these ubiquitinated forms did not lead to reduced Gap1p activity (not shown). Addition of NH₄⁺ triggered the very rapid appearance of 2 to 3 additional bands, corresponding to ubiquitin-permease conjugates of higher molecular mass (Fig. 5, lanes 1-2 and 5-6). This modification was accompanied by a decrease in Gap1p activity and a decrease in the amount of Gap1p immunodetected in crude extracts (Fig. 6B). Thus, NH₄⁺ triggers binding of additional Ub molecules to Gap1p. This induced ubiquitination could involve Ub binding to additional lysine residues of the target protein or the formation of Ub chains in which the C-terminal glycine residue of a Ub is linked to the internal lysine residue of a previously bound Ub. We thus overproduced in npi2 cells a mutant Ub (UbRRR) in which Lys29, Lys48, and Lys63 are replaced by arginine. This Ub variant cannot form multi-Ub chains in vivo (Arnason and Ellison, 1994). When NH₄⁺ was added to npi2 cells overproducing UbRRR, no bands were formed above the monoand di-ubiquitinated forms of Gap1p (Fig. 5, lane 10). Although no ubiquitination sites have yet been identified within the Gap1p permease, this result suggests that two residues of Gap1 can act as Ub acceptors. In proline-grown cells, a fraction of the Gap1p molecules would thus be mainly mono-ubiquitinated at one or both acceptor sites; NH₄⁺ addition would induce formation of Ub chains on at least one of these sites. Formation of these polyubiquitin chains is not absolutely essential to NH₄⁺-induced down-regulation of Gap1, but it does appear necessary to allow its occurrence at an optimal rate: upon addition of NH₄⁺, npi2 cells producing UbRRR lost Gap1p activity only half as quickly as cells producing the wild-type Ub (Fig. 6A). Furthermore, immunoblots obtained with crude extracts showed that Gap1p is degraded in npi2 cells producing the mutant Ub, but more slowly than in the same cells producing wild-type Ub.

To test which lysine residue of Ub is involved in forming polyubiquitin chains on Gap1, we conducted experiments in which a variant Ub singly mutated at either Lys29, Lys48, or Lys63 was produced in npi2 cells. Overproduction of UbK29R and UbK48R restored the formation of five bands above the major Gap1p signal in response to NH₄⁺ addition. In the presence of UbK63R, only the monoand di-ubiquitinated forms of Gap1p were evidenced after addition of NH₄⁺ (Fig. 5, lane 9). Both bands exhibited a retarded gel mobility when mutant UbK63R was extended by the addition of HA tag (Fig. 5, lane 11). This clearly indicates that Lys63 is the acceptor residue required for NH₄⁺-induced ubiquitination of Gap1p. In npi2 cells overproducing UbK63R, moreover, the loss of permease activity induced by NH₄⁺ was only partial, as in the case of overproduction of UbRRR. We conclude that NH₄⁺ induces formation of Lys63-linked polyubiquitin chains on Gap1p, required for rapid and complete NH₄⁺-induced downregulation of the permease.

After addition of NH₄⁺ to yeast cells using proline as the sole nitrogen source, Gap1 permease pre-accumulated in the plasma membrane undergoes rapid ubiquitination followed by endocytosis and vacuolar degradation (Springael and André, 1998). Gap1p ubiquitination requires Npi1p/Rsp5p, a ubiquitin protein ligase required for cell viability (Hein et al., 1995; Springael and André, 1998). We here report cloning of NPI2, a second gene required for NH₄⁺-induced down-regulation of the Gap1p permease, and show that it encodes the ubiquitin hydrolase Doa4p. Ubiquitin hydrolases, which cleave Ub from linear or branched conjugates, constitute an unexpectedly large enzyme family, of which 16 members have been identified in S. cerevisiae (Wilkinson, 1997; Hochtrasser, 1996). These enzymes share high sequence similarity in two proposed functional domains known as the Cys and His boxes (Baker et al., 1992). These boxes appear to delimit a catalytic core of about 300 residues. Additional blocks of similarity were found between the Cys and His boxes of many isoforms and are assumed to confer specificity or control localization of the enzyme (Wilkinson et al., 1995; Wilkinson, 1997). We show here that the npi2 mutation impairing Gap1p permease downregulation results in replacement of a conserved aromatic residue within one of these blocks, which we propose to name the ELC box (Fig. 3). For some of the phenotypes analyzed, the npi2 mutation appears equivalent to disruption of DOA4. We have not determined whether the npi2 mutation merely reduces the stability of Doa4p or affects a functionally important site, but the critical role of the aromatic residue and the presence of the ELC box in a subset of Ubps (a few in yeast and some in higher eukaryotes, such as the human Tre2 oncogen or the drosophila fat-facets) suggest that these enzymes might share specific functional properties. There is considerable evidence of specific functions among Ub isopeptidases, even though their precise physiological substrates have not yet been identified. For instance, cell fate determination in the Drosophila eye is controlled by the fatfacets isopeptidase (Huang et al., 1995), another isopeptidase (Uch-D) is specifically involved in Drosophila oogenesis (Zhang et al., 1993), and the yeast Ubp3p appears as an inhibitor of silencing (Moazed and Johnson, 1996).

Strikingly different genetic approaches have led to identification of the DOA4/SSV7/UBP4 gene: searches for mutants defective in Matα2 degradation (doa for degradation of alpha) (Hochtrasser and Varshavsky, 1990), isolation of saltsensitive vacuolar (ssv) mutants (Latterich and Watson, 1991), of cells with uncoordinated DNA replication (Singer et al., 1996). Our finding that the npi2 (nitrogen permease inactivator) mutation is a point mutation in DOA4 further extends this list. Overproduction of Ub reverses several of the numerous phenotypes conferred by lack of Doa4p, notably sensitivity to cadmium and canavanine (Hicke, 1997) and the inability to ubiquitinate and subsequently degrade certain proteins such as the Ste2p receptor (Terrel et al., 1998), the uracil permease (Galan and Haguenauer-Tsapis, 1997), and as shown in this paper, Gap1p. These observations suggest that some doa4 phenotypes are due to decreased Ub levels in doa4 cells. It has been reported that Ub pools drop dramatically in doa4Δ cells entering the stationary phase (Chen and Hochtrasser, 1995) and that free monomeric Ub is no longer detectable in doa4Δ cells subjected to prolonged heat stress (Singer et al., 1996). We here extend these observations, showing that the amount of free monomeric Ub is reduced at least 4-fold in npi2/doa4Δ cells growing exponentially in minimal medium. Obviously, however, not all Ub-dependent processes are equally affected by lowering of the Ub pools in doa4 cells. Various proteins of the Ub-proteasome pathway are stabilized to strikingly different degrees (Hochtrasser, 1996). As an extreme example, the Ub pool of doa4 cells appears sufficient for normal turnover of a membrane-bound protein subject to endoplasmic-reticulum-associated degradation (Galan et al., 1998).

The exact physiological role of Doa4p, still a matter of debate, obviously extends beyond its effect, likely indirect, on the intracellular Ub pool. Doa4p-deficient mutants have been shown to accumulate small Ub-linked peptide fragments (mostly Ub dimers) (Papa and Hochtrasser, 1993, Amerik et al., 1997). A model has been proposed in which Npi2p/Doa4p works in conjunction with the proteasome, cleaving Ub chains still attached to peptide remnants of proteolysis, creating unanchored Ub chain substrates (Amerik et al., 1997). Some doa4 phenotypes not complemented by overproduction of Ub (Singer et al., 1996), such as uncoordinated replication, might be more directly related to the true function of Doa4p, whatever it may be.

NH₄⁺-induced ubiquitination and down-regulation of Gap1p are thus inhibited in npi2 cells and both restored by overproduction of Ub. This fact has enabled us to investigate the type of Ub modification involved in these processes. In proline-grown wild-type and npi2 cells overproducing Ub, a larger fraction of the Gap1p molecules appears to undergo ubiquitination, resulting in accumulation of monoand diubiquitinated species. This modification does not lead to reduced Gap1p activity, suggesting that the permease remains stable at the plasma membrane. Addition of NH₄⁺ to the medium leads within minutes to formation of Ub-permease conjugates bearing up to four or five Ub moieties. The same species are observed, after NH₄⁺ addition, in npi2 cells overproducing a mutant Ub in which Lys29 or Lys48 have been mutated to arginine. In contrast, npi2 cells overproducing the triple Ub mutant UbRRR, unable to form Ub chains, or the Ub mutant with an arginine residue instead of Lys63, displayed only di-ubiquitinated Ub-permease conjugates. This strongly suggests that two lysine residues in Gap1p serve as Ubacceptor sites, and that NH₄⁺ promotes the formation of Lys63-linked polyubiquitin chains. Furthermore, formation of these chains promotes more rapid down-regulation of Gap1p. The uracil permease displays a very similar ubiquitination pattern (Galan and Haguenauer-Tsapis, 1997). Ubiquitination of both transporters appears to involve formation of short Lys63-linked polyubiquitin chains, and in both cases the rate of internalization appears clearly linked to the ability to form these chains. In the case of Gap1p, interestingly, the specific shift from monoto poly-ubiquitination is triggered by addition of NH₄⁺ to the medium, indicating that polyubiquitination is subject to nitrogen-source regulation. Addition of NH₄⁺ also generally increases conversion of Gap1p to Ub-conjugated forms (Springael and André, 1998). The mechanisms underlying nitrogen control of Gap1p ubiquitination are unknown. Addition of NH₄⁺ might alter Gap1p phosphorylation, which in turn could render the permease more accessible to monoand poly-ubiquitination. Regulatory factors responding to NH₄⁺ are likely involved. One such factor could be Npr1p, a protein kinase homologue (Vandenbol et al., 1990) which appears to protect Gap1p and other NH₄⁺-sensitive permeases from down-regulation in cells growing on proline or urea as the sole nitrogen source (Grenson, 1983b).

Since the Fur4p and Gap1p permeases seem to undergo the same type of ubiquitination, might Lys63-linked polyubiquitin chains be a general feature of plasma-membrane proteins in yeast? A truncated form of the receptor Ste2p, another plasmamembrane protein, is reported to undergo only monoubiquitination (Terrel et al., 1998). Perhaps Fur4p and Gap1p ubiquitination on the one hand and Ste2p ubiquitination on the other involve distinct E2/E3 pairs. Ubiquitination of Ste2p is reported to be Ubc4p/Ubc5p-dependent (Hicke and Riezman, 1996), whereas ubiquitination of Fur4p appears normal in ubc4Δubc5Δ cells (Galan and Haguenauer-Tsapis, 1997). On the other hand, the Npi1p ubiquitin ligase is required for ubiquitination of both Fur4p and Gap1p (Galan and Haguenauer-Tsapis, 1997; Springael and André, 1998). In yeast, this Ub ligase appears to play a central role in endocytosis of numerous plasma-membrane proteins, or at least of several transporters. It is involved in catabolite inactivation of the maltose transporter (Lucero and Lagunas, 1997) and in NH₄⁺-induced inactivation of numerous nitrogencompound permeases in addition to Gap1p (Grenson, 1992). It is tempting to imagine an even broader role for this type of enzyme in ubiquitination of plasma-membrane proteins of eukaryotic cells, since rNedd4, the rat homolog of Npi1p, is involved in the ubiquitination of some epithelial Na channel (ENaC) subunits (Staub et al., 1996, 1997). In addition, it was recently demonstrated that Nedd4 mediates Ub-dependent down-regulation of ENaC in response to increased intracellular Na⁺ (Dinudom et al., 1998).

Gap1p is thus the second example, after the uracil permease, of a yeast plasma-membrane transporter whose rapid internalization requires formation of polyubiquitin chains with linkage through Lys63. It is clear that neither monoubiquitination nor the formation of such chains targets proteins for proteasome recognition. The role of these Ub modifications as internalization signals thus remains to be determined. It has been suggested that Ub (chains) might be recognized by a component of the endocytic machinery or might alter target proteins so as to facilitate their binding to such a component (Hicke, 1997). Alternatively, ubiquitination might promote movement of ubiquitinated proteins into membrane regions that actively endocytose, as suggested by the polarized cellsurface staining of Ste6p in mutants displaying reduced ubiquitination (Loayza and Michaelis, 1998). Ubiquitination or lysine-63-mediated polyubiquitination might also contribute to determining lysosomal degradation rather than recycling of internalized plasma-membrane proteins.

We are grateful to D. Wolf for anti-ubiquitin antibodies and to M. Hochstrasser and M. Ellison for strains and plasmids. We thank A.-M. Marini for discussions as the work was in progress and for critical comments on the manuscript. We are also grateful to C. Jauniaux for her help with cloning and sequencing. This work was supported by the Fund for Medical Scientific Research (Belgium F.R.S.M. 3.4602.94), the Communauté Française de Belgique, Direction de la Recherche Scientifique, Action de Recherche Concertée (98/03-223), and by a special grant of the CNRS (program ‘Biologie Cellulaire’, project no. 96105). During part of this work, J.Y.S. was the recipient of a predoctoral fellowship from the Fonds pour la Formation à la Recherche dans l’Industrie et dans l’Agriculture (F.R.I.A.). He is presently a fellow of the Fondation Universitaire David et Alice Van Buuren. J.-M. Galan was supported by a grant from the Fondation de la Recherche Médicale.