Functional Domain Analysis of the Yeast ABC Transporter Ycf1p by Site-directed Mutagenesis*

The yeast cadmium factor (Ycf1p) is a vacuolar protein involved in resistance to Cd2+ and to exogenous glutathione S-conjugate precursors in yeast. It belongs to the superfamily of ATP binding cassette transporters, which includes the human cystic fibrosis transmembrane conductance regulator and the multidrug resistance-associated protein. To examine the functional significance of conserved amino acid residues in Ycf1p, we performed an extensive mutational analysis. Twenty-two single amino acid substitutions or deletions were generated by site-directed mutagenesis in the nucleotide binding domains, the proposed regulatory domain, and the fourth cytoplasmic loop. Mutants were analyzed phenotypically by measuring their ability to grow in the presence of Cd2+. Expression and subcellular localization of the mutant proteins were examined by immunodetection in vacuolar membranes. For functional characterization of the Ycf1p variants, the kinetic parameters of glutathione S-conjugated leukotriene C4 transport were measured. Our analysis shows that residues Ile711, Leu712, Phe713, Glu927, and Gly1413 are essential for Ycf1p expression. Five other amino acids, Gly663, Gly756, Asp777, Gly1306, and Gly1311, are critical for Ycf1p function, and two residues, Glu709 and Asp821, are unnecessary for Ycf1p biogenesis and function. We also identify several regulatory domain mutants in which Cd2+ tolerance of the mutant strain and transport activity of the protein are dissociated.

Transporters designated ABC 1 proteins, or traffic ATPases, are present from micro-organisms to man (1,2). Eukaryotic ABC proteins are often composed of two homologous halves, each containing a transmembrane domain (TMD) with six pre-dicted membrane-spanning segments, and a NBF that contains the highly conserved Walker A (GXXGXGK(S/T)), Walker B (RX 6 -8 hyd 4 D) (3), and ABC "signature" (LSXGX(K/R)) (4) motifs. Walker A and Walker B are common to a wide variety of nucleotide-binding proteins, whereas the signature region, just upstream of the Walker B motif, is distinctive to the ABC family. Midway between the Walker A and B regions lies the "center" region, which does not form a true consensus sequence, although certain residues are conserved among ABC proteins subsets (5). Certain ABC transporters have an additional R domain that serves regulatory functions (6), and others have a large N-terminal hydrophobic region (7). Eukaryotic ABC transporters include the human CFTR, multidrug resistance and MRP1 proteins, and the yeast a-factor transporter (Ste6p), pleiotropic drug resistance (Pdr5p), and Ycf1p proteins (1, 8 -10).
Ycf1p is a vacuolar membrane protein that transports Cd 2ϩ ions and several drugs as glutathione conjugates (GS-conjugates) with a requirement for ATP hydrolysis, playing a critical role in Cd 2ϩ tolerance in the yeast Saccharomyces cerevisiae (11)(12)(13). Ycf1p has strong sequence similarity with CFTR and MRP1; in a comparison of all known yeast and human ABC transporters, these three proteins are classified in the same subgroup (5,14). Ycf1p and MRP1 share with CFTR a region comparable with the R domain in CFTR, which mediates cAMPdependent regulation of the chloride channel. The sequence identity of these proteins suggests that they probably have similar overall structures. Structure-function analysis of Ycf1p can thus provide valuable insights into the molecular mechanism of transport by ABC proteins.
We selected 22 amino acid residues of Ycf1p for extensive mutational analysis of the protein. Most of the selected residues are conserved in CFTR and MRP1 proteins, and the substitutions or deletions introduced are analogous to reported cystic fibrosis (CF)-associated mutations. Mutations were introduced by site-directed mutagenesis in NBF1, NBF2, the R domain, and ICL4 that joins TMD2 transmembrane segments X and XI. Mutants were analyzed phenotypically by assaying their Cd 2ϩ tolerance. The mutant protein expression level was quantitated by immunodetection in vacuolar membranes; those showing expression were functionally characterized by measuring ATP-dependent transport of LTC 4 into vacuolar membrane vesicles. The kinetic parameters for LTC 4 uptake were determined for the wild type and mutants. The results presented here show that the mutations introduced in the highly conserved NBF domains are extremely deleterious for Ycf1p function. In addition, the mutational analysis identified a group of mutants located in the putative R domain in which Cd 2ϩ tolerance was dissociated from the in vitro measured transport activity of the protein. We discuss a model for the regulatory role of the R-like domain in transporter function.  , ycf1⌬::URA3, ade2-1, his3-11, 15, leu2-3,  112, trp1-1, ura3-1), constructed as described below, was used. In all experiments, growth was at 30°C in SD medium (0.7% yeast nitrogen base without amino acids (U. S. Biological, Swampscott, MA), 2% glucose, pH 5.5) supplemented with required amino acids (100 g/ml). SD medium for Cd 2ϩ resistance assays was supplemented with drop-out mix (BIO 101, Joshua Way, Vista CA).
Disruption of YCF1 Gene-A 1.1-kb BamHI-SmaI fragment from pJJ242 (16) containing the URA3 gene, was used to replace a 4.7-kb BglII-SalI (blunt-ended) fragment of pIBIYCF1 (13), thus eliminating most of the promoter and coding sequence. A 3.5-kb SacI-MluI fragment containing the YCF1 disruption cassette was used to transform strain W303-1A to uracil prototrophy. Gene disruption was verified by Southern hybridization (17).
Plasmid Constructions-A 7.1-kb fragment from pIBIYCF1, excised by SphI (blunt-ended with Klenow fragment) and SacI digestion, containing YCF1 gene was subcloned into SmaI-SacI-digested pRS315 (18) to give pRS315YCF1. Digestion with XhoI-MluI, blunt-ending with Klenow fragment, and religating eliminated restriction sites between XhoI-SmaI sites in the pRS315 polylinker, giving plasmid pIB1157. Blunt-end ligation regenerated the MluI site. Plasmid pIB1197 was constructed by insertion of a 4.5-kb EcoRV-SalI fragment from pIBIYCF1 into pBluescript KS vector (Stratagene, La Jolla, CA), digestion with HindIII, and ligation to generate a 1.1-kb HindIII deletion internal to the EcoRV-SalI fragment.
Site-directed Mutagenesis-The following fragments from YCF1 were subcloned into M13mp19 (19) for mutagenesis: (i) 1.4-kb SalI-SphI fragment (epitope tagging); (ii) 2.1-kb StuI-SalI fragment (mutations D777N, R1143C, Q1148P, G1306E, G1311R, N1366K, G1413D); (iii) 1.1-kb HindIII fragment (remaining mutations). The StuI-SalI fragment was first subcloned into pSL301 (20) and recovered as SacI-SalI to be cloned into M13mp19. Mutagenesis was performed by the Eckstein method (21) with the Sculptor TM in vitro Mutagenesis System (Amersham Pharmacia Biotech). Epitope tagging of Ycf1p was performed by inserting the coding sequence for the 9-amino acid 12CA5 epitope (sequence: YPYDVPDYA) from human influenza hemagglutinin protein (HA) immediately before the YCF1 termination codon. The tagged YCF1 was designated YCF1-HA. All YCF1 fragments subjected to oligonucleotide-directed mutagenesis were sequenced. The YCF1 1.4-kb SalI-MluI fragment containing the HA epitope sequence was obtained from the replicative form of M13mp19 and exchanged with the corresponding wild type segments of pIB1157, generating pIB1201. The YCF1 1.1-kb HindIII fragments containing the mutations generated were obtained from the replicative form of M13mp19 and inserted into pIB1197. The 4.5-kb EcoRV-SalI fragments from the resultant plasmids were exchanged with the corresponding wild type segments of YCF1-HA in pIB1201. Mutagenized StuI-SalI fragments were liberated from the replicative M13mp19 form and used to replace the corresponding fragments in pIB1201. All replacements of wild type by mutant fragments were confirmed by sequencing. The 7.1-kb ApaI-SacI fragment from pIB1201 containing the promoter and open reading frame was also cloned into pRS425 (22) to render the YCF1-HA gene on multicopy plasmid. The resulting plasmids were used to transform the ⌬ycf1 strain using the lithium acetate procedure (23).
DNA Sequencing-DNA was sequenced by the dideoxy chain termination method (24) modified for single-stranded and double-stranded template with Sequenase, as described by the enzyme supplier (U. S. Biochemical Corp.).
Cadmium Resistance Assays-Qualitative and quantitative determinations of Cd 2ϩ resistance were performed. Cells were cultured for 2 days and suspended in water to an A 660 of 0.4 (2.4 ϫ 10 7 cells/ml) to be used as inoculum. For qualitative assays, 5 l were dropped on plates alone or with CdCl 2 at the indicated concentrations. Growth was scored after a 3-day incubation. For quantitative determination of the minimal inhibitory concentration (MIC), flat-bottom 96-well microtiter plates containing medium with CdCl 2 concentrations ranging from 0 to1 mM were inoculated to a final cell density of 6 ϫ 10 5 cells/ml. Inoculum-free wells were also included. The optical density (A 595 ) of each well was determined after a 2-day incubation. Data were fitted to a sigmoidal dose-response equation using Prism 2.0 GraphPad Software. The MIC value is defined as the lowest Cd 2ϩ concentration at which prominent inhibition of cell growth (90 -95%) is observed.
Isolation of Vacuolar Membrane Vesicles-Vacuolar membrane vesicles were prepared from cultures grown to an A 660 of 1.0. After washing cells with H 2 O, spheroplasts were obtained with Zymolyase 20T (U. S. Biological, Swampscott, MA) (25), and intact vacuoles were isolated by flotation centrifugation of spheroplast lysates on Ficoll 400 step gradients as described (26), using a shortened procedure with only one cycle of flotation centrifugation. The resulting vacuole fraction was vesiculated in 5 mM MgCl 2 , 25 mM KCl, 10 mM Tris-MES, pH 6.9, pelleted by centrifugation (37,000 ϫ g, 25 min), and resuspended in buffer (1.1 M glycerol, 2 mM dithiothreitol, 1 mM EGTA, 5 mM Tris-MES, pH 7.6) (11). Dithiothreitol and EGTA were omitted in vacuolar preparations used for transport assays. All buffers used contained a protease inhibitor mixture (1 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, and 1 mM phenylmethylsulfonyl fluoride). 4 Uptake-Standard uptake experiments were performed at 30°C in TS buffer (250 mM sucrose, 25 mM Tris-MES, pH 8.0) containing 4 mM ATP, 10 mM MgCl 2 , 10 mM creatine phosphate, 20 units/ml creatine kinase, and 50 nM [ 3 H]LTC 4 (13 nCi/ pmol) in a final volume of 55 l. Uptake was initiated by the addition of vesicles (10 g of protein). LTC 4 uptake into vacuolar vesicles increased linearly with the amount of vacuolar membrane protein, at least to 30 g. Aliquots (10 l) were removed at the times indicated, diluted in 1 ml of ice-cold TS buffer, immediately filtered through nitrocellulose filters (pore size 0.45 m, Millipore, Bedford, MA) presoaked in TS buffer and washed twice with 5 ml of ice-cold TS buffer. The retained radioactivity was counted using liquid scintillation fluid. Initial rates were calculated from the first 1.5 min of uptake.

Measurement of [ 3 H]LTC
Protein Analysis-Protein concentration was measured by the Bradford method (27), using the Bio-Rad protein assay reagent and bovine IgG as standard. SDS-polyacrylamide gel electrophoresis on 7% gels was performed as described (28). Sample solubilization and Western blot analysis were performed as described (29,30). Reversible protein staining with Ponceau S (31) and immunodetection of Ycf1p-HA using mouse anti-HA monoclonal antibody and a second antibody coupled to alkaline phosphatase (Bio-Rad) was as described (32). Western blots were scanned with a StudioscanII (Agfa-Gevaert, Leverkusen, Germany), and Adobe Photoshop and NIH Image 1.60/fat software were used to quantitate the Ycf1p amounts. Standard default settings were used for all measurements, and all mutant enzymes were compared with an internal wild type control on the same gel. Chemicals-[ 3 H]LTC 4 (165 Ci/mmol) was obtained from NEN Life Science Products. Unlabeled LTC 4 was from Sigma. All other reagents were of analytical grade and purchased from Sigma, Roche Molecular Biochemicals, Amersham Pharmacia Biotech, or U. S. Biological.

Ycf1p-dependent Transport Activity in Vacuolar Membrane
Vesicles-YCF1 has been identified as a gene necessary for the detoxification of Cd 2ϩ (13) and shown to be a pump for GSconjugates (GS-X pump) responsible for the vacuolar sequestration of organic GS-conjugates (11) and GS-cadmium complexes (Cd⅐GS 2 ) (12). We chose LTC 4 as substrate of Ycf1p since human MRP1, a GS-X pump that transports LTC 4 and structurally related conjugates in mammalian cells (33), complements Ycf1p function by restoring Cd 2ϩ resistance in a ⌬ycf1 deletion mutant (34). We measured [ 3 H]LTC 4 uptake into vacuolar membrane vesicles from a ⌬ycf1 yeast strain harboring the wild type gene carried on either single or multiple copy plasmid. Incubation of the vacuolar vesicles with 50 nM [ 3 H]LTC 4 in the presence of ATP resulted in rapid LTC 4 uptake (Fig. 1A). In vesicles from wild type cells expressing YCF1 from a single or a multicopy plasmid, the initial rates in the presence of ATP were 45 and 182 pmol⅐min Ϫ1 ⅐mg of protein Ϫ1 , respectively, and 0.1 and 15 pmol⅐min Ϫ1 ⅐mg protein Ϫ1 in the absence of ATP. LTC 4 uptake in the presence of a nonhydrolyzable ATP analog, adenosine 5Ј-O-(3-thiotriphosphate), or into vacuolar vesicles prepared from the ⌬ycf1 deletion mutant was negligible under these conditions (not shown). Ycf1p-dependent transport activity correlated roughly with Cd 2ϩ tolerance (Fig. 1B) and the amount of Ycf1p (Fig. 1C) of these strains. This indicates that LTC 4 uptake is Ycf1p-dependent, allowing the use of quantitative transport assays as an indicator of Ycf1p function.
The effect of LTC 4 or ATP concentration on the rate of LTC 4 uptake was measured in vacuolar vesicles prepared from a ⌬ycf1 strain expressing the epitope-tagged YCF1-HA from a single copy plasmid. The uptake was saturated with respect to LTC 4 and ATP concentrations with an apparent K m of 0.82 Ϯ 0.14 M for LTC 4 and 55 Ϯ 9.7 M for ATP. The V max was 1.85 Ϯ 0.12 nmol⅐min Ϫ1 ⅐mg of protein Ϫ1 (Table III). The results revealed that Ycf1p affinity is higher for LTC 4 than for other substrates previously assayed, such as S- (2,4- (12) and with the same order of magnitude as the LTC 4 affinity for its physiological transporter, MRP1 (K m ϭ 0.11 M) (33).
A large variety of GS-conjugates bearing aliphatic or aromatic S-substitutes inhibit transport activity by ATP-dependent GS-conjugate carriers in membrane vesicles (11,35,36). We examined inhibition of LTC 4 uptake by different glutathione-related compounds. Glutathione (1 mM GSH) had a weak inhibitory effect, concurring with the low affinity of Ycf1p for this compound (37), whereas 100 M GS-conjugates such as S-decyl glutathione or S-azidophenacyl glutathione inhibited LTC 4 transport by 60 -63% (Table I). LTC 4 uptake was also inhibited by 45% with 500 M Cd 2ϩ in the presence of 1 mM GSH. Agents that dissipate the H ϩ electrochemical gradient established by the vacuolar H ϩ -ATPase did not decrease LTC 4 transport significantly. In contrast, DIDS and probenecid, potent inhibitors of anion transporters (36), completely abolished LTC 4 uptake at the concentrations used. These results show that the energy-dependent LTC 4 uptake is not linked to the vacuolar electrochemical H ϩ gradient and has properties similar to those of GS-conjugate transport by other GS-X pumps.
Selection of Amino Acids for Site-directed Mutagenesis-Individual Ycf1p amino acids were selected for site-directed mutagenesis with the following criteria. In the NBF domains and ICL4 loop, residues highly conserved in ABC proteins were selected, and for the less well conserved R-like domain, the selected amino acids were conserved at least in Ycf1p and CFTR. In all cases, the amino acid changes introduced are those described as CF-associated mutations (38). Alignment of the conserved regions of the N-and C-terminal NBF domains and the ICL4 sequence of CFTR, MRP1, and Ycf1p are shown in Fig. 2A; Fig. 2B shows all substitutions introduced in Ycf1p and their localization within the protein. We generated 11 mutations in Ycf1p NBF1 and NBF2 domains. Two other mutations, R1143C and Q1148P, are localized in a protein region homologous to the ICL4 in MRP1 and CFTR. ICL4 is suggested to couple activity of the NBF domains to channel gating in or the wild type YCF1-HA (see below) gene in centromeric (CEN) or multicopy (2 m) plasmids were suspended in water to an A 660 of 0.4, and 5 l were spotted onto SD drop-out plates containing the indicated CdCl 2 concentrations. Growth was scored after a 3-day incubation at 30°C for Cd 2ϩ -containing plates and a 1-day incubation for the control plate. MICs were determined as described under "Experimental Procedures." Values are the average of two independent experiments. C, immunodetection and quantitation of Ycf1p in vacuolar membrane vesicles from the same strains as in B was performed as described under "Experimental Procedures." Only the relevant part of the membrane of a representative experiment is shown. All YCF1 alleles constructed contained a C-terminal HA epitope tag to localize the protein (see "Experimental Procedures"). The HA tag did not influence Ycf1p function (data not shown).  4 , as described under "Experimental Procedures." Additions were made 5 min before initiation of the uptake reaction. Before adding glutathione (GSH) and Cd 2ϩ to the uptake medium, a mixture of 20 mM GSH and 10 mM CdSO 4 was incubated in 0.2 M phosphate buffer, pH 8.0, at 45°C for 24 h to allow Cd ⅐ GS 2 to form (12). Azidophenacyl-GS was added from a Me 2 SO stock solution. Gramicidin D and carbonyl cyanide m-chlorophenyl hydrazone (CCCP) were added from ethanol stock solutions. Me 2 SO or ethanol alone (Ͻ2%) had no effect on transport activity. The rate for the wild-type strain (100%) was 50 pmol ⅐ min Ϫ1 ⅐ mg of protein Ϫ1 . Values are the means of two independent experiments differing less than 10%. DIDS, 4,4Ј-diisothiocyanatostilbene-2,2Ј-disulfonate. The R domain of CFTR contains several consensus phosphorylation sites for cyclic AMP-dependent protein kinase A and for protein kinase C, thought to be important for regulation of its activity (40). The N-terminal part of this domain has been highly conserved during evolution, whereas the C-terminal two-thirds is poorly conserved (41). As mutagenesis of the putative protein kinase A phosphorylation site in the R-like domain of Ycf1p abolishes Cd 2ϩ resistance, Ycf1p is proposed to require a similar domain for normal function, but the manner in which this region participates in Ycf1p activity has not been established (13). To determine the role of the R domain, we introduced six mutations in the highly conserved N-terminal part of the domain (L817S, D821G, L825T, L826S, G835R, and I840P) and three in the less conserved C-terminal part (Y855L, A910G, and E927K). The activity of the Ycf1p variants was evaluated by measuring Cd 2ϩ tolerance and Ycf1p-mediated LTC 4 transport in vacuolar membrane vesicles.
Mutant ycf1 Strains Growth on Cadmium Medium-Yeast transformants carrying the ycf1 mutant alleles on a centromeric plasmid were tested for growth on plates containing 50 or 100 M CdCl 2 (Fig. 3). Cd 2ϩ MIC was determined for each strain, and mutants were classified in three groups according to their Cd 2ϩ tolerance. Substitutions G663V, I711S, ⌬L712, ⌬F713, G756D, D777N, E927K, G1306E, G1311R, N1366K and G1413D affected essential residues, since strains with these mutant alleles exhibited a Cd 2ϩ hypersensitive phenotype ( Fig.  3 and Table II). These mutations are all located in the NBF domains except for E927K, in the R domain. Strains expressing ycf1 mutations L817S, L825T, L826S, G835R, I840P, Y855L, A910G, R1143P, and Q1148P were more sensitive to Cd 2ϩ than the wild type strain (Fig. 3), with MIC ranging from 111 to 172 M, with 200 M the MIC of the wild type strain (Table II). These mutations are all located in the R-like domain and ICL4. Only two substitutions, E709Q and D821G in NBF1 and R domains, respectively, did not modify Ycf1p in vivo activity, since a wild type growth pattern was observed in yeast expressing these mutant enzymes.
Effect of ycf1 Mutations on Subcellular Protein Localization-Ycf1p is an integral protein of the vacuolar membrane (11). Intracellular sorting of the protein or its turnover may be affected by mutations, contributing to the changes observed in Cd 2ϩ resistance of the mutants. We estimated the relative amount of mutant Ycf1p in the vacuolar membrane by quantitative immunoassay (Table II). Five of the mutant proteins, I711S, ⌬L712, ⌬F713, E927K, and G1413 D, were undetectable either in the vacuole or in total membrane pellets, indicating that the Cd 2ϩ toxicity exhibited by mutants expressing those ycf1 alleles may therefore be explained by defective mutant protein biogenesis; they were not studied further. Four mutants, I840P, R1143C, Q1148P, and N1366K, gave significantly lower yields of protein expression compared with the wild type control, which may explain their decreased capacity for Cd 2ϩ detoxification, although an alteration in protein activity cannot be excluded (see below). The remaining mutant proteins, G663V, E709Q, G756D, D777N, L817S, D821G, L825T, L826S, G835R, Y855L, A910G, G1306E, and G1311R, reached the vacuolar membrane at levels close to those of the wild type control or even higher, as was the case for G1306E (Table II). Within this group, only E709Q and D821G mutants exhibited wild type phenotype for Cd 2ϩ resistance, in agreement with their wild type Ycf1p amount in the vacuole. In the other cases, an alteration in protein activity may account for Cd 2ϩ toxicity.
LTC 4 Uptake in Vacuolar Membrane Vesicles from ycf1 Mutants-To measure the magnitude of change introduced by the mutations on the transport activity of the protein, we measured LTC 4 uptake into vacuolar membrane vesicles by the mutant proteins. A group of mutants was found to have significantly altered LTC 4 uptake (Table II). Of these, four mutations (G663V, G756D, G1306E, and G1311R) completely abolished transport activity, whereas six of theYcf1p variants (D777N, L826S, G835R, I840P, R1143C, and N1366K) showed a partial reduction in LTC 4 uptake with values ranging from 30% (N1366K) to 72% (R1143C) of the wild type control. Two mutant proteins, Y855L and A910G, showed values above wild type, whereas the transport rate of the five remaining mutant enzymes was found to be essentially wild type. In all cases in which LTC 4 uptake was detectable, the kinetic transport parameters were determined (Table III). The apparent K m for LTC 4 or ATP of all mutant enzymes was not significantly changed, with the exception of the D777N variant, which had a 25-fold increase in the K m for ATP; thus, transport alterations were due to V max modifications. In mutants exhibiting impaired transport activity, the changes in LTC 4 transport paralleled the observed defective Cd 2ϩ detoxification. Nevertheless, mutant enzymes showing either wild type (E709Q, L817S, D821G, and L825T) or above wild type V max values (Y855L and A910G) all had diminished ability to detoxify Cd 2ϩ ions, except for E709Q and D821G (Fig. 3 and Table II). These results indicate that Leu 817 , Leu 825 , Tyr 855 , and Ala 910 are nonessential residues for correct intracellular sorting and LTC 4 transport activity of Ycf1p but are important for Cd 2ϩ resistance. When transport activity was corrected for the relative amount of Ycf1p in the vacuolar membrane, it became evident that not only mutants expressing the Y855L and A910G mutations, but also those carrying I840P and R1143C changes, showed a gainof-function phenotype with respect to LTC 4 uptake when Cd 2ϩ b MICs for Cd 2ϩ were determined as described under "Experimental Procedures." Values are the average of independent duplicate experiments.
c The relative amount of Ycf1p in vacuolar membrane vesicles from wild-type and mutant strains was determined by quantitative immunoblotting (see "Experimental Procedures"). Values are the mean of two to three independent experiments with an average S.E. Ͻ15%. d The initial rate of LTC 4 uptake in vacuolar membrane vesicles was assayed with 50 nM LTC 4 , as described under "Experimental Procedures." Values are the average obtained from two to six determinations in different vacuolar membrane preparations isolated independently. e ND, not determined.
FIG. 3. Cadmium resistance of wild type and mutant ycf1 strains. Cells of a ⌬ycf1 yeast strain harboring the centromeric plasmid pRS315 alone (⌬ycf1) or carrying the wild type YCF1-HA gene (wt) or the ycf1 mutant alleles created by sitedirected mutagenesis (indicated by the position of the mutated residue) were suspended in water to an A 660 of 0.4, and 5 l were spotted onto SD drop-out plates containing the indicated CdCl 2 concentrations. Growth was scored after a 3-day incubation at 30°C for Cd 2ϩ containing plates and a 1-day incubation for the control plate. The amino acid changes introduced in the positions indicated are shown in Fig. 2. resistance was affected. The decreased Cd 2ϩ resistance of these two mutants could thus be due to their defect in protein biogenesis. DISCUSSION We performed a mutational analysis of Ycf1p, a yeast homolog of both human CFTR and MRP1 transporters, generating 22 mutations by site-directed mutagenesis. Mutant phenotypes were characterized by measuring growth in the presence of Cd 2ϩ ions, the relative amount of mutant protein in the vacuolar membrane, and Ycf1p-dependent transport in vacuolar membrane vesicles .
For the sake of discussion, the mutants described here can be divided into four groups. Group 1 includes mutants E709Q and D821G, which showed essentially wild type phenotype for the three parameters tested. These residues can be assumed to be relatively unimportant for Ycf1p biogenesis and function. Interestingly, the Glu 709 residue lies in the position that corresponds to Glu 504 of CFTR, predicted to be the NBF1 catalytic carboxylate residue involved in ATP hydrolysis by comparison to the ␤ subunit of F 1 -ATPase (42,43). Although CFTR mutation E504Q is associated to CF, to our knowledge the effect of this particular change on transporter activity has not been tested in any ABC protein. Moreover, recent crystal structure resolution of the Salmonella typhimurium histidine permease (HisP) NBF domain points to other residues as the putative activating base (44).
Group 2 consists of mutant enzymes defective in biogenesis. Mutants of particular interest in this group are those located in the center region of NBF1 (I711S, ⌬L712, ⌬F713) and NBF2 (N1366K). The equivalent CFTR changes, ⌬F508 and N1303K, are frequent CF-associated mutations that have been described as processing mutants (6,45). In fact, mutant protein ⌬F508 is retained in the endoplasmic reticulum membrane and degraded by the ubiquitin-proteasome system (46,47). The introduction into YCF1 of the analogous mutation, ⌬F713, or the proximal mutations I711S and ⌬L712 led to mutant proteins undetectable even in a pre1pre2 mutant strain defective in the chymotrypsin-like activity of the proteasome (data not shown). These results suggest that an alternative pathway is used in yeast to eliminate aberrant Ycf1 proteins. Other groups (13,48) found that mutant ⌬F713 protein expressed in a multicopy plasmid was mislocalized, although a mislocalization due to overexpression, as shown for other membrane proteins (49), cannot be discarded. The inability of mutations I711S, ⌬L712, and ⌬F713 to confer Cd 2ϩ resistance is clearly caused by a biogenesis defect of the corresponding proteins, as is also the case for mutations E927K and G1413D. Two other mutants showing reduced protein expression, I840P and R1143C, were also affected in LTC 4 transport and have been included in group 4. In the case of mutants R1148P and N1366K, protein biogenesis was altered, but they retained fully active LTC 4 transport activity.
Group 3 includes mutant enzymes that, although correctly located, exhibit impaired Ycf1p Cd 2ϩ tolerance and LTC 4 transport function. Mutations of this group localized in the NBF domains completely abolished (G663V, G756D, G1306E, and G1311R) or severely damaged (D777N) Ycf1p function, whereas mutations in the R-like domain (L826S and G835R) had a milder effect. Such detrimental mutations include substitutions in the conserved Walker A, Walker B, and signature motifs of NBF1 and the Walker A motif of NBF2, corroborating their importance in the ABC transporter function. Mutant D777N shows a much lower affinity for ATP than the wild type Ycf1p. This residue has been proposed, based on HisP crystal structure (44), to interact with magnesium during ATP hydrolysis. Identical changes at corresponding positions in the signature region of NBF1 and NBF2, G756D, and G1413D produced quite distinct phenotypes. Mutant protein G756D was present in the vacuole at wild type levels but inactive, whereas G1413D was undetectable. The equivalent substitutions to several of group 3 NBF mutations in other ABC proteins, CFTR (50), multidrug resistance protein (51,52), Ste6p (53), or HisP (54), also give rise to normally expressed although nonfunctional mutant proteins. In contrast, the analogous changes to L826S and G835R in the N-terminal part of the R-like domain have only been characterized in CFTR, resulting in aberrant processing (55).
Group 4 is of particular interest and includes mutants in which Cd 2ϩ resistance and LTC 4 transport activity were dissociated. Cd 2ϩ tolerance of some R domain or ICL4 mutants was reduced as compared with the wild type strain, but their transport activity was near wild type, with variations lower than 30% (L817S and L825T) or even increased above wild type (Y855L and A910G). The gain-of-function effect also became  4 . The initial rate of LTC 4 uptake in vacuolar membrane vesicles was assayed with ATP concentrations from 0.35 to 6 mM, as described under "Experimental Procedures." d The uptake values were normalized on the basis of the amount of mutant protein detected in the vacuolar membrane fractions relative to the wild-type control. evident for I840P and R1143C mutants when transport activity was corrected for the relative amount of Ycf1p. A similar gainof-function phenotype has been observed in CFTR R domain mutants H620Q and A800G (55).
Several hypotheses can be considered to explain this striking phenotype. A simple explanation would be that the limiting step in Cd 2ϩ detoxification is not transport into the vacuole but an earlier event such as complexation with GSH. In this case, an increase in GS-X pump activity would not lead to an increase in Cd 2ϩ tolerance. This hypothesis is discarded based on the finding that Ycf1p overexpression led to an increase in LTC 4 transport and to a parallel increase in Cd 2ϩ tolerance of the yeast strain (Fig. 1). Affinity or specificity changes of the transporter do not seem probable, since the K m for LTC 4 or ATP were not modified in these mutants, and LTC 4 uptake in mutant vacuolar vesicles was inhibited by Cd 2ϩ plus GSH or azidophenacyl-GS to the same extent as in wild type vesicles (data not shown). In all these mutants, only an increase in the V max of LTC 4 transport was detected. The observed gain-offunction phenotype for LTC 4 transport of some R domain and ICL4 loop mutants may be due to the release from an inhibitory interaction between these two protein regions, leading to the observed increase in the V max . In CFTR and other ABC proteins, it has been speculated that the intracellular loops may link the NBFs to the TMDs to couple ATP binding and hydrolysis to the transporter activity (39,56,57). This coupling may be modulated by the R-like domain in Ycf1p. Intragenic suppressor analysis of R domain and/or ICL4 loop mutations will shed light on the proposed domain interaction. This hypothesis would account for the augmented LTC 4 uptake in vacuolar membrane vesicles but would not explain the reduced Cd 2ϩ resistance observed in vivo in group 4 mutants. The necessary functional coupling between Cd⅐GS 2 synthesis and their efficient transport into the vacuole may have been lost in these mutants. It is tempting to speculate that the R domain may interact in vivo both with ICL4 and with a protein that would couple synthesis of GS-conjugates to transport. Interaction of the R domain with such a protein would lead to a disruption in its interaction with ICL4, with an increase of the transport activity of the protein. The similarity in properties of many of the mutants studied to those of their counterparts in mammalian ABC proteins leads to the expectation that the mechanistic information gathered for yeast Ycf1p may be extrapolated to their human homologs.