Phosphorylation of Candida glabrata ATP-binding Cassette Transporter Cdr1p Regulates Drug Efflux Activity and ATPase Stability* □

Fungal ATP-binding cassette transporter regulation was investigated using Candida glabrata Cdr1p and Pdh1p expressed in Saccharomyces cerevisiae . Rephosphorylation of Pdh1p and Cdr1p was protein kinase A inhibitor-sensitive but responded differentially to Tpk isoforms, stressors, and glucose concentration. Cdr1p Ser 307 , which borders the nucleotide binding domain 1 ABC signature motif, and Ser 484 , near the membrane, were dephosphorylated on glucose depletion and independently rephosphorylated during glucose exposure or under stress. The S484A enzyme retained half the wild type ATPase activity without affecting azole resistance, but the S307A enzyme was unstable to plasma membrane isolation. Studies of pump function suggested conformational interaction between Ser 484 and Ser 307 . An S307A/S484A double mutant, which failed to efflux the Cdr1p substrate rhodamine 6G, had a fluconazole susceptibility 4-fold greater than the Cdr1p expressing strain, twice that of the S307A mutant, but 64-fold less than the control null strain. Stable intragenic suppressors indicative of homodimer nucleotide binding domain 1-nucleotide binding domain 1 interactions partially restored rhodamine 6G pumping and increased fluconazole and rhodamine 6G resistance in the S307A/ S484A mutant. Nucleotide binding domain 1 of Cdr1p is a sensor of important physiological stimuli.

Infections caused by Candida sp. are most frequently seen in immunocompromised individuals, including AIDS and leukemia patients. Candida albicans remains the leading cause of candidiasis, but the incidence of drug-resistant non-albicans Candida infections has become an increasingly significant clin-ical problem. Candida glabrata is among the most common of these pathogens (1), with many clinical isolates showing a 16to 64-fold higher minimum inhibitory concentration (MIC) 1 of fluconazole (FLC) than C. albicans (2). Azole drugs such as FLC and itraconazole, which target lanosterol 14␣-demethylase and block the synthesis of ergosterol, are well tolerated and widely used in the treatment of fungal disease. They are, however, fungistatic substrates of pleiotropic drug resistance (PDR) family ATP-binding cassette (ABC) transporters, and resistant fungi that overexpress these pumps are frequently isolated in the clinic (3). The C. glabrata PDR family ABC transporters Cdr1p and Pdh1p, which efflux azole agents and structurally unrelated compounds, are among the primary causes of the intrinsic resistance of C. glabrata to azole drugs (4 -8). The two pumps have Ͼ70% amino acid sequence identity and transport a similar spectrum of substrates, but Cdr1p had greater drug efflux activity for most substrates. Structural information on fungal single subunit ABC transporters is rudimentary, and the molecular and regulatory features that determine their enzyme activity and substrate specificity are poorly understood. Such information is required for the rational design of pump inhibitors and antifungal drugs that are not pump substrates.
There are few reports of the effects of post-translational modification on the activity of fungal ABC transporters. Serine 420, a casein kinase-dependent phosphorylation site that determines Saccharomyces cerevisiae Pdr5p turnover (9), is the only experimentally proven fungal ABC transporter phosphorylation site. We have shown that the ATPase activity of Cdr1p and the drug efflux activity of Pdh1p are regulated by phosphorylation (7). Cdr1p cannot be phosphorylated at the position equivalent to Ser 420 , whereas Pdh1p phosphorylation was regulated by protein kinase A (PKA) at one or more sites not homologous to Pdr5p Ser 420 . More than one type of phosphorylation therefore occurs in fungal PDR family pumps.
This report describes the differential regulation of C. glabrata Cdr1p and Pdh1p expressed in S. cerevisiae. Immunological, physiological, and biochemical methods were applied to site-directed mutants in putative phosphorylation sites and to kinasedeletion mutants. PKA catalytic subunit isoforms differentially affected pump phosphorylation, and the effects of the phosphorylation of two putative novel sites in Cdr1p were determined. A phosphorylation site adjacent to the Cdr1p ABC signature motif in nucleotide binding domain 1 (NBD1) affected the extent of multidrug efflux and the in vitro stability of ATPase activity while mutation of another cytoplasmic site nearer the membrane diminished transport at low glucose concentrations. Mutation in both sites eliminated the pumping activity of Cdr1p, whereas intragenic suppressors obtained by exposing the double mutant to FLC partially restored pump function. The Cdr1p NBD1 is a functional sensor of cell physiology and stress that may regulate interactions between homodimers.

EXPERIMENTAL PROCEDURES
Yeast Strains and Growth Media-Growth and selection media, and the methods used to prepare and mutate S. cerevisiae strains expressing Pdh1p and Cdr1p, are described in the Supplemental Data section.
Drug Susceptibility Assays-Agar diffusion assays on YPD (Qbiogene, Irvine, CA) agar plates, microplate MIC assays in HEPES and MES-buffered CSM (Qbiogene, Inc.) were performed, and the drugs and chemical compounds used were obtained as previously described (7).
Analysis of Pump Protein Phosphorylation-Glucose-starved yeast were obtained by incubation in CSM minus glucose for 3.5 h, and crude plasma membrane fractions were prepared as previously described (7) using GTED-20 buffer (10 mM Tris-HCl, pH 7.0, 0.5 mM EDTA, and 20% (v/v) glycerol) instead of the previous homogenization buffer. The membrane fractions were analyzed by SDS-PAGE and immunoblotting as previously described (7) using 1/2000 dilutions of anti-phospho (serine/ threonine) protein kinase A substrate antibody, anti-phospho-(serine/ threonine) Akt substrate antibody, and anti-phospho threonine antibody from Cell Signaling Technology (Beverly, MA), together with antirabbit IgG conjugated with horseradish peroxidase secondary antibody (Amersham Biosciences).
ATPase Assay-Purified plasma membrane fractions were prepared from cells grown in YPD to early stationary phase (A 600 nm ϭ 7.0 -9.0) and oligomycin-sensitive ATPase activities of samples were measured as previously described (7).
Fluorometric Assay of Rhodamine 6G Efflux-Log phase (A 600 nm ϭ 1.5) cells grown in CSM-URA (Qbiogene, Inc., Irvine, CA) medium were stored overnight on ice. The cells were harvested by centrifugation, washed twice with distilled water, and then incubated in HEPES buffer (50 mM HEPES-NaOH, pH 7.0) containing 5 mM 2-deoxyglucose at 30°C for 30 min to deplete intracellular energy levels. The cells were preloaded with 15 M Rh6G for 30 min, washed twice, and resuspended in HEPES buffer at A 600 nm ϭ 15 (1.5 ϫ 10 8 cells ml Ϫ1 ). Cell samples (50 l) were incubated at 30°C for 5 min and 50 l of glucose at twice the final concentration added to start the reaction. After 8 min the cells in 80-l samples were removed by passage through a Multi-well Filter plate (Acro Prep, Pall Corp.) placed on a Multiscreen resist vacuum manifold (Millipore), with a 96-well black flat-bottom microtiter collection plate (BMG Labtechnologies GmbH, Offenburg, Germany) underneath. The Rh6G content of the eluate, combined with two 80-l washes with ice-cold HEPES buffer, was quantitated using a POLARstar OP-TIMA (BMG Labtechnologies) fluorometer (excitation and emission wavelengths of 485 and 520 nm, respectively) with Fluostar OPTIMA software and a standard curve of Rh6G in the HEPES buffer.

Antibodies Recognizing Cdr1p or Pdh1p
Phosphorylations-We have previously described the S. cerevisiae strains CDR1-AD and PDH1-AD, which contain a mutant Pdr1p transcriptional regulator and thus constitutively heterologously hyperexpress Cdr1p and Pdh1p, respectively, from the PDR5 promoter (7,10). Cdr1p and Pdh1p were hyperexpressed at comparable levels (ϳ10% of plasma membrane protein) and were readily distinguished as heterologous plasma membrane proteins on Coomassie Blue-stained SDS-polyacrylamide gels due to the deletion of seven similar-sized, endogenous pump proteins. Each strain also showed the expected high level resistance to azole drugs (7). The pumps heterologously expressed in S. cerevisiae were therefore correctly folded and should be post-translationally regulated as in C. glabrata, because these two closely related yeasts have similar intracellular molecular machinery.
The phosphorylation of both Cdr1p and Pdh1p was detected by 32 P labeling, but the signals were too weak to support more detailed experiments. Neither Cdr1p nor Pdh1p reacted with phospho-protein kinase C substrates antibody or phospho-tyrosine antibodies (7). An anti-phospho-(Ser/Thr) PKA substrate (p-PKAs) antibody detected phosphorylation of Pdh1p but not Cdr1p or background ABC transporters in Western blots of plasma membrane preparations (Fig. 1).
Antibodies with different specificities detected novel phosphorylation sites in Cdr1p and Pdh1p. The anti-phospho-Akt substrate (p-Akts) and anti-phospho-threonine (p-Thr) antibodies recognized the phosphorylation of Cdr1p and Pdh1p expressed in all growth phases of YPD culture (Fig. 1, black arrowheads). Cdr1p expression in CDR1-AD decreased slightly in stationary phase (Fig. 1, CBB staining), whereas the p-Akts antibody recognition signal increased. Conversely, the p-Thr antibody gave a stronger signal in the log phase preparation. Cdr1p is therefore multiply phosphorylated at discrete p-Akts and p-Thr antibody recognition sites. Pdh1p phosphorylation in the PDH1-AD strain was readily detected by the p-PKAs, p-Akts, and p-Thr antibodies, with constant signals obtained throughout all growth phases of YPD culture. The recognition of the 100-kDa protein, which co-migrates with the constitutively expressed plasma membrane ATPase Pma1p (Fig. 1, white arrowheads) by the three antibodies provided an internal control for both the CDR1-AD and PDH1-AD preparations.
Glucose-sensitive Phosphorylation of Pdh1p and Cdr1p-We previously found that the Cdr1p and Pdh1p pumps were extensively dephosphorylated after a few hours of glucose starvation and immediately rephosphorylated when 2% glucose was added (7). The glucose concentration dependence of each rephosphorylation was therefore determined ( Fig. 2A). Early stationary phase YPD cultures were glucose-starved for 3.5 h and then treated for 10 min with 10 M, 1 mM, or 100 mM glucose. Rephosphorylation of Cdr1p was detected with p-Akts and p-Thr antibodies, but only after exposure to 100 mM glucose. Pdh1p remained partially phosphorylated after glucose starvation, but all three antibodies detected glucose-dependent rephosphorylation at a glucose concentration (10 M) that failed to induce Pma1p phosphorylation. The different phosphorylation patterns detected with each antibody indicated that Pdh1p was also rephosphorylated at multiple sites. The Ͼ100-fold differential in the glucose-dependence of Cdr1p and Pdh1p phosphorylation suggested that the two pumps were modified by different kinases/phosphatases or had sites with different susceptibilities to phosphorylation.
Pdh1p phosphorylation at p-PKAs sites is inhibited by PKA inhibitors (7). A panel of PKA catalytic subunit (Tpk) single and double deletion mutants was used to identify the TPK gene product that regulated Pdh1p phosphorylation in PDH1-AD. The three genes (TPK1, -2, and -3) encoding these PKA cata-lytic subunit isoforms can be deleted without lethality, either singly or up to two at a time (11). In YPD culture, identical Pdh1p phosphorylation signals were observed, even in the double deletion mutants PDH1-TPK1⌬2⌬, -TPK1⌬3⌬, and -TPK2⌬3⌬ (Fig. 2B). In contrast, rephosphorylation at the PKA and Akt sites was not induced by 10 M glucose in PDH1-TPK2⌬3⌬, and PKA site rephosphorylation was dramatically decreased in PDH1-TPK1⌬3⌬ (Fig. 2B). Thus Tpk3p was the main contributor to glucose-dependent rephosphorylation of Pdh1p PKA and Akts sites at low glucose concentrations, and only Tpk2p could augment this process.
Stress Sensitivity of Cdr1p and Pdh1p Phosphorylation-The glucose-dependent differences in Cdr1p and Pdh1p rephosphorylation patterns suggested that the two pumps might be phosphorylated in discrete physiological contexts. This hypothesis was initially tested for Cdr1p rephosphorylation in glucosestarved CDR1-AD by adding 1 mM glucose together with a stressor for 10 min (Fig. 3A). The 1 mM glucose supplied sufficient ATP as kinase substrate, because the Pma1p band was strongly phosphorylated even in the absence of Akt site phosphorylation of Cdr1p (Figs. 2A and 3A). The pump substrate FLC (45 g/ml) did not affect rephosphorylation (data not shown), but oxidative stress (2 mM H 2 O 2 ), osmotic stress (500 mM NaCl), and heat shock (42°C) for 10 min induced p-Akt site phosphorylation, albeit to a lesser extent than 100 mM glucose (Fig. 3A). Stress-dependent rephosphorylation was not detected with the p-Thr antibody. Pdh1p in PDH1-AD was highly phosphorylated in 1 mM glucose, and stress treatments caused no additional phosphorylation. In contrast, stress with 2 mM H 2 O 2 alone decreased Pdh1p phosphorylation and gave no rephosphorylation of Cdr1p. However, Cdr1p phosphorylation was not affected when the yeast in YPD culture were stressed (H 2 O 2 , NaCl, or 42°C) for 10 min (data not shown). Under these conditions, the effects of multiple kinases and phosphatases may have been complex and/or compensatory.
Rephosphorylation of Cdr1p Akt sites was inhibited by the PKA inhibitors H-89 and amide 14 -22 but not by the H-89 homologue H-8, which has a K i for PKA 30-fold higher than H-89 (12), or the protein kinase C inhibitor bisindolylmaleimide I (BIM) (Fig. 3B). A panel of TPK single or double deletion mutants constructed in CDR1-AD was tested for the effects of NaCl-and glucose-dependent rephosphorylation of Cdr1p (Fig.  3C). Apart from CDR1-TPK1⌬2⌬ and CDR1-TPK2⌬3⌬, which reduced Cdr1p rephosphorylation by about 50% during both treatments, p-Akts site re-phosphorylation was unaffected. Thus Tpk2p and to a lesser extent Tpk1p and Tpk3p may play a role with other kinases susceptible to PKA inhibitors in the rephosphorylation of Cdr1p Akts sites.
Identification of Putative Phosphorylation Sites in Cdr1p-Modest rephosphorylation signals suggested the presence of few p-Akts sites in Cdr1p. The p-Akts antibody recognizes phosphorylated Ser or Thr in the Ϫ5 (K/R)X Ϫ3 (K/R)XX 0 (S/T) motif and cross-reacts with the phosphorylated Ϫ3 (K/R) Ϫ2 (K/ R)X 0 (S/T) motif (manufacturer's information). The full size, single subunit, ABC transporter Cdr1p comprises two nucleotide binding domains (NBD1 and NBD2) that each contain the Walker A, Walker B, and ABC signature motifs, alternating with two pairs of six transmembrane segments, as illustrated in Fig. 4A. We constructed the yeast strains CDR1-M1 to CDR1-M9, which expressed equivalent amounts of Cdr1p (Fig.  4B), and each contained a point mutation (S/T 3 A) at each one of the nine putative p-Akts recognition sites. Of these sites, only M2, M4, and M7 may be recognized by the phospho-PKA substrates antibody (phosphorylated Ser or Thr in RXX(S/T)). Unlike the essentially normal phosphorylation of Cdr1p p-Akts sites in the CDR1-M3-CDR1-M9 mutants in 12-h YPD (early FIG. 2. Glucose-sensitive phosphorylation of Pdh1p. A, early stationary phase pSK-AD, CDR1-AD, or PDH1-AD cells from YPD culture were washed with distilled water and glucose-starved in CSM minus glucose (CSM-Gluc) for 3.5 h. Distilled water (0), 10 M, 1 mM, or 100 mM glucose was then added to samples of cells for 10 min, and crude membrane fractions were prepared. Membrane fractions of 5 g and 15 g of protein were analyzed by immunoblotting and Coomassie Brilliant Blue R250 (CBB) staining, respectively, as described in Fig. 1. Arrowheads indicate Cdr1p and Pdh1p (black) and Pma1p (white). Representative data from several experiments are shown. B, PDH1-AD and its derivative single TPK gene-defective strains, PDH1-TPK1⌬ (TPK1 deleted), PDH1-TPK2⌬, and PDH1-TPK3⌬, and the double TPK gene-defective strains, PDH1-TPK1⌬2⌬ (TPK1 and TPK2 deleted), PDH1-TPK1⌬3⌬, and PDH1-TPK2⌬3⌬ were grown in YPD to early stationary phase (YPD 12 h), some cells were then glucose-starved for 3.5 h and treated with 10 M glucose for 10 min. The phosphorylation and rephosphorylation pattern and expression level of Pdh1p in each strain was detected as described in Fig. 1. Representative data from several independently isolated clones of each strain are shown. stationary phase) cultures, Cdr1p phosphorylation was eliminated in CDR1-M1 and Ͼ90% inhibited in CDR1-M2 (Fig. 4B). Rephosphorylation of Cdr1p-M1 and Cdr1p-M2 was not detected in response to stressors plus 1 mM glucose (Fig. 4, C and D), and 100 mM glucose gave rephosphorylation to about 50% of the control level in each mutant (Fig. 4D). These results suggested that Ser 307 and Ser 484 are the dominant Akts sites in Cdr1p. This hypothesis was confirmed by finding, in the S307A/ S484A double mutant strain CDR1-M1,2, that 100 mM glucose gave no Cdr1p p-Akts site rephosphorylation (Fig. 4E). Thus, the phosphorylation of M1 and M2 sites responded comparably to 12-h YPD culture (low glucose) and stressors, whereas glu-cose-induced rephosphorylation of the two sites occurred independently. Interestingly, the S307A mutation of the M1 site mutation reduced Thr phosphorylation by one-third in 12-h YPD cultures (Fig. 4B), and the S509A mutation at the membrane-associated M3 site similarly affected Akts site rephosphorylation of Cdr1p in 100 mM glucose (Fig. 4E).
Rhodamine 6G Efflux from Pdh1p, Cdr1p, and Point Mutant Yeast-The effect of the M1 and M2 mutations on energy-dependent drug efflux by Cdr1p was quantitated by comparably pre-loading CDR1-AD, CDR1-M1, CDR1-M2, CDR1-M1,2 and AD1-8u Ϫ cells with the pump substrate rhodamine 6G (Rh6G) after 2-deoxyglucose treatment and then stimulating efflux by  (14 -22), bisindolylmaleimide I (BIM), or a solvent control (1% Me 2 SO plus 1% water). After 10 min, 500 mM NaCl plus 1 mM glucose, or 100 mM glucose was added, the cells cultured for 10 min, and crude membrane fractions were prepared for analysis. C, single and double TPK gene deletion mutants were constructed from CDR1-AD as described for PDH1-AD (Fig. 2). Glucosestarved CDR1-AD and the derivative TPK deletion mutants were treated with 500 mM NaCl, and 1 mM glucose, or 100 mM glucose for 10 min. The phosphorylation patterns and protein expression levels of the crude membrane fractions were analyzed.
adding glucose (Fig. 5). Fluorometric measurements showed that the CDR1-M2 strain (half-maximal rate of Rh6G pumping, 2.5 mM glucose; maximal pumping rate, 5-10 mM glucose) pumped Rh6G at rates up to 80% of the CDR1-AD strain (half-maximal rate of Rh6G pumping, 1 mM glucose; maximal rate, 5-10 mM glucose). In contrast, the CDR1-M1 strain required at least 5 mM glucose for a significant rate of Rh6G efflux and reached a maximal rate at 20 mM glucose, which was only 30% that of the CDR1-AD strain. The CDR1-M1,2 strain showed no glucose-dependent Rh6G efflux, even at 100 mM glucose. All strains that effluxed Rh6G showed a 30% decrease in the pumping rate at glucose concentrations between 20 and 100 mM glucose.
The pumping activities of Cdr1p and its point mutants were confirmed by flow cytometric measurement of cellular Rh6G content, which monitors combined dye uptake and efflux (see Supplemental Data, Fig. S1). These data also showed that strain CDR1-M1,2 did not efflux Rh6G even in the presence of 100 mM glucose and that phosphorylation of the Cdr1p-M1 site may be required for Rh6G efflux at low glucose concentrations.
Lability of ATPase Activity in CDR1-M1-The Cdr1p drug efflux pumps were functional in CDR1-M1, CDR1-M2, and CDR1-M8 cells and therefore expected to retain significant ATPase activity on cell fractionation. Plasma membrane fractions from these strains contained equivalent amounts of Cdr1p proteins, CDR1-M8 had a normal oligomycin-sensitive ATPase activity with a broad pH profile, but CDR1-M2 showed a decrease of ϳ50% in the ATPase activity compared with CDR1-AD (Fig. 6). CDR1-M8 was chosen as a control, because the T1007A mutation borders the NBD2 ABC signature motif FIG. 4. Phosphorylation sites of Cdr1p detected by the phospho-(Ser/Thr) Akt substrate antibody. A, the nine putative phospho-Akt substrate antibody recognition sites (p-Akts sites) Ϫ5 (K/R)X Ϫ3 (K/R)XX 0 (S/T) or Ϫ3 (K/R) Ϫ2 (K/R)X 0 (S/T) of Cdr1p are shown (M1-M9), with Ser in black and Thr in gray. The conserved motifs of the nucleotide binding cassette, Walker A, B, and ABC signature, and twelve trans-membrane domains (TMD) are depicted with gray bars, whereas regions external to the plasma membrane (PM) are shown in black. CDR1-M1 through CDR1-M9 yeast strains, whose Ser or Thr residues were replaced by Ala at each M1-M9 site, were constructed. B, the phosphorylation patterns and expression levels of Cdr1p in crude membrane fractions from CDR1-AD and the point mutated derivatives CDR1-M1 through CDR1-M9 grown to early stationary phase in YPD medium were measured as described in previous figures. The ratios of intensities relative to CDR1-AD (1.0), given at the bottom of each panel, were measured using Scion Image. C, CDR1-AD and its derivative mutant strains were glucose-starved for 3.5 h and treated with 1 mM glucose and 500 mM NaCl for 10 min. The phosphorylation status of the Akt sites in each mutant Cdr1p was analyzed as above. D, glucose-starved CDR1-AD, -M1, and -M2 yeasts were treated with stressors and 1 mM glucose, or 100 mM glucose for 10 min as indicated in the top panel. The patterns of Cdr1p phosphorylation were analyzed as above. The relative band intensities are shown for each stress condition. E, CDR1-AD, the point mutants CDR1-M1 through CDR1-M9, and CDR1-M1,2 were glucose-starved for 3.5 h and then treated with 100 mM glucose for 10 min. The phosphorylation of Cdr1p was analyzed as above. In B-E, Cdr1p and the equivalent bands from the point mutant derivatives are indicated with the black arrowhead, whereas the Pma1p band is indicated with the white arrowhead. Representative data, from multiple experiments using several independently isolated clones of each strain, are shown. residue and aligns with Ser 307 of NBD1. The membranes from CDR1-M1, however, lacked detectable oligomycin-sensitive ATPase activity. The M1 site may therefore be critical for the in vitro ATPase activity. Experiments that included ATP during membrane isolation allowed the recovery of small amounts (10%) of vanadate and oligomycin-sensitive ATPase activity compared with the control strain (see Supplemental Data Table  SIII). A statistically significant proportion of this activity (p Ͻ 0.01) can be attributed to S307A Cdr1p that survived membrane isolation due to the protective ATP.
Drug Susceptibilities of CDR1-AD and PDH1-AD Derivative Yeasts-The MIC 80 values for antifungal agents were determined for PDH1-AD and CDR1-AD derivative strains to assess whether the phosphorylation of Cdr1p and Pdh1p affected drug efflux activity (Table I). As expected, the Pdh1p-and Cdr1pexpressing strains were more resistant to azole agents than the control pSK-AD strain but were equally susceptible to flucytosine and amphotericin B, which are not pump substrates. The flucytosine and amphotericin B susceptibilities for all the strains were within two dilutions of the values for pSK-AD. Among the CDR1-AD derivatives, only CDR1-M1 and CDR1-M1,2 consistently showed lower FLC susceptibilities than CDR1-AD, giving values that were reproducibly 2-fold and 4-fold lower, respectively. The TPK2 deletion mutants CDR1-TPK2⌬, -TPK1⌬2⌬, and -TPK2⌬3⌬ gave detectably greater azole resistance than the control Cdr1p-expressing strain. Repeated testing of Cdr1p phosphorylation in TPK2 deletion mutants found no relationship between the phosphorylation status of Cdr1p and the apparent azole resistance (data not shown). In contrast, the azole resistance of PDH1-AD TPK mutants was unchanged compared with the control parental strain. A SCH9 kinase deletion mutant and a HOG1 kinase deletion mutant of CDR1-AD both gave slightly increased resistance to azole agents. The former showed increased phosphorylation of Cdr1p Akts sites, and the latter had an unchanged phosphorylation pattern compared with the parental strain (data not shown).
Agar diffusion tests of susceptibilities of CDR1-AD, CDR1-M1, CDR1-M2, and CDR1-M1,2 to structurally unrelated xenobiotics extended the data on antifungal susceptibility reported in Table I and suggested that the M1 and M2 sites are synergistic effectors of Cdr1p drug efflux activity (see Supplemental Data Fig. S2). Of the 17 compounds to which CDR1-AD was strongly resistant (7), CDR1-M1 showed increased susceptibility to eight compounds, whereas CDR1-M1,2 had even greater susceptibility to these and two other compounds. In contrast, the susceptibilities of CDR1-AD and CDR1-M2 were indistinguishable.
The relatively high FLC MIC value for the CDR1-M1 strain (Table I) can be explained by partially functional mutant Cdr1p operating at glucose concentrations Ͼ5 mM. The lack of Rh6G pumping by CDR1-M1,2 (Fig. 5), however, seemed inconsistent with its relatively high MIC for FLC and other azoles (Table I), although this was less pronounced in disk diffusion assays. CDR1-M1,2 was also more resistant to FLC than either PDH1-AD or the AD1-8u Ϫ host strain (Table I). Even though identical FLC and Rh6G susceptibilities were observed with three separately isolated clones for each construct under study, we excluded the possibility that secondary mutations in the genetic background of the strains might affect susceptibilities to azoles. CgCDR1-URA3 DNA, obtained by PCR of genomic DNA of CDR1-AD, CDR1-M1, CDR1-M2, and CDR1-M1,2, was used to transform the hypersensitive AD1-8u Ϫ strain. All tested Ura ϩ transformants that grew normally on 5 g/ml FLC (to ensure the incorporation of the CDR1 gene) showed FLC and Rh6G susceptibilities identical to those for the strains providing the transforming DNA (Table II). The drug resistance of each donor strain was thus conferred by the expression of Cdr1p and not an extragenic determinant. The unexpectedly high FLC and Rh6G liquid MIC 80 values for CDR1-M1,2 were therefore due to Cdr1-M1,2p overexpression. It was possible, however, that suppressor mutations were selected during growth in liquid MIC assays.
Progeny that survived FLC exposure during MIC determinations in buffered FLC-containing CSM-URA medium, using either glucose or the non-fermentable substrate glycerol as energy source, were isolated. Sequencing of several independent isolates showed that the S307A and S484A mutations were maintained in these progeny independent of the energy source. In addition, three independently isolated intragenic suppressor mutations were obtained that did not change the Ser 307 and Ser 484 background: a ⌬A349 mutation after growth on glycerol and D32H and V353L mutations after growth on glucose. In contrast to the inactive parental CDR1-M1,2 strain, all three suppressor strains showed glucose-dependent Rh6G pumping comparable to the CDR1-M1 mutant. An Rh6G efflux experiment conducted with the representative suppressor strain CDR1-M1,2 48a is shown in Fig. 5. The CgCDR1-URA3 cassette was obtained by PCR of genomic DNA from each of the three suppressor strains and used to transform strain AD1-8u Ϫ . The Ura ϩ transformants that also grew on 5 g/ml FLC showed FLC and Rh6G resistance comparable to the CDR1-M1 strain (Table II). The suppressor mutation phenotypes therefore result from intragenic modification of Cdr1p in CDR1-M1,2. The selection of the suppressor mutants occurred at a low frequency (suppressor strains were detectable in Ͻ10% of MIC determinations). This frequency was not high enough to compromise the MICs for azole drugs and other xenobiotic substrates of CDR1-M1,2.

DISCUSSION
Phosphorylation of ABC Transporters-Phosphorylation mediated by PKA and PKC affects the function of numerous human ABC transporters, including ABCA1, MDR1, and CFTR (13)(14)(15). We previously demonstrated that PKA-dependent phosphorylation of C. glabrata Pdh1p was important for drug efflux and Cdr1p ATPase specific activity was glucose-dependent and possibly regulated by phosphorylation (7). C. glabrata Cdr1p and Pdh1p have about 70% amino acid sequence identity with S. cerevisiae Pdr5p and thus belong to the PDR family, the  Table II. largest ABC transporter family in S. cerevisiae (16,17). The S. cerevisiae drug efflux transporters Pdr5p, Snq2p, and Yor1p are all phosphorylated, and casein kinase I phosphorylation of Ser 420 in Pdr5p is important for enzyme turnover (9), but the equivalent site is absent in Cdr1p. Phosphorylation of Thr 613 and Ser 623 in the D-box linking the two homologous halves of Ste6p has been implicated in enzyme turnover (18). The Cdr1p p-Akts antibody recognition sites at Ser 307 and Ser 484 are therefore distinct from the sites affecting Pdr5p and Ste6p turnover. The Ser 307 and Ser 484 sites appear physiologically important. They affect the glucose dependence of pump activity, and glucose-dependent Rh6G efflux was eliminated by the S307A/S484A double mutation. The expression of S307A, S484A, and the S307A/S484A double mutant Cdr1ps at levels comparable with the wild type protein implies that misfolding does not target the mutant proteins for early degradation. The PKA-dependent phosphorylation of human ABCA1 (13) provides a precedent for phosphorylation at Ser 307 .
Kinases Responsible For the Phosphorylation of Pdh1p and of Cdr1p at M1 and M2 Sites-The glucose-dependent rephosphorylation of Pdh1p in starved cells was blocked by PKA inhibitors (7) and was primarily affected by Tpk3p and to a lesser extent by Tpk2p. Antibody recognition of phosphorylated PKA sites implies that these phosphorylations were direct. Of the nine putative phosphorylation sites tested in glucose-starved Cdr1p, only Ser 307 and Ser 484 strongly affected the rephosphorylation pattern. Unlike Pdh1p phosphorylation, neither site was recognized by the p-PKA substrates antibody after stress or at glucose concentrations 100-fold higher than those required for Pdh1p p-Akts site rephosphorylation. The blockage FIG. 6. Loss of ATPase activity in mutated Cdr1p. Plasma membrane fractions were prepared from pSK-AD, CDR1-AD, CDR1-M1, CDR1-M2, CDR1-M1,2, and CDR1-M8 yeast at early stationary phase of YPD culture. Oligomycin-sensitive ATPase activities of each fraction were measured at each pH indicated. Experiments were performed several times using the membrane fractions prepared from independently isolated clones of the yeast. Representative data for ATPase activities are shown on the right (mean Ϯ S.E., n ϭ 3), and the SDS-PAGE Coomassie Brilliant Blue R250 (CBB) staining patterns for 5 g of protein of each membrane fraction used in the experiment are shown on the left. The arrowhead indicates Cdr1p.

Phosphorylation of C. glabrata Cdr1p
of Cdr1p rephosphorylation by the PKA inhibitors H89 and 14 -22 and the 50% reduction in Cdr1p phosphorylation in CDR1-TPK1⌬2⌬ and CDR1-TPK2⌬3⌬ only, but not CDR1-TPK2⌬, suggested a role for Tpk2p in phosphorylation at either Ser 307 or Ser 484 that may be complemented by Tpk1p plus Tpk3p. Alternatively, other kinases affected by PKA inhibitors could directly or indirectly cause the Ser 307 and Ser 484 rephosphorylation signals.
The M1 site in Cdr1p (Fig. 7A) contributes to putative yeast PKA (KRVS), and CKII (SIAE) sites (see www.cbs.dtu.dk/databases/PhosphoBase/predict/predict.php) and was, as expected, detected by cross-reactivity with the p-Akts antibody and not by the p-PKAs antibody. Ser 484 of M2 (Fig. 7D) contributes to putative PKA, Akt, and CaMII kinase sites. Although not close homologues of Akt, deletion of SCH9, one of the four yeast kinases categorized in the Akt family (19), gave slightly higher levels of multidrug drug resistance and an increased, rather than decreased, Akt site phosphorylation that cannot directly involve Ser 307 . Conservation of a Ϫ2 PX 0 (S/ T) ϩ1 P motif at the M2 site of Cdr1p and its homologs suggests proline-directed kinases, such as mitogen-activated protein kinases and cyclin-dependent kinases, phosphorylate this site. Deletion of the yeast HOG1 kinase, an osmotic stress-activated homologue of the stress-activated human mitogen-activated protein kinase protein p38 (20), did not alter the stress-induced phosphorylation of Cdr1p Akt sites and only gave a slight increase in resistance to some drugs (Table I). Both direct and indirect stress-responsive phosphorylation by PKC (21,22) was excluded, because phosphoprotein kinase C antibodies did not detect Cdr1p phosphorylation (7), and the phosphorylation of Akt sites was unaffected by the PKC inhibitor BIM (Fig. 3B). The kinase(s) causing the Ser 484 p-Akts site signal have yet to be identified.
Because the context-independent p-PKAs antibody did not detect rephosphorylated Cdr1p, it is possible, assuming absolute antibody specificity, that Ser 484 is not phosphorylated and that the S484A mutation causes a conformational change that affects a nearby phosphorylation site. The membrane-associated M3 (Ser 509 ) site, which may also undergo significant rephosphorylation in the presence of 100 mM glucose (Fig. 4E), is a possible candidate. The identification of phosphopeptides obtained from rephosphorylated Cdr1p will be needed to clarify this uncertainty and to demonstrate directly phosphorylation of Ser 307 or Ser 484 .
The kinase-deleted strains CDR1-TPK2⌬, CDR1-TPK1⌬2⌬, and CDR1-TPK2⌬3⌬ all showed slightly (2-fold) higher FLC MIC 80 values than the parental CDR1-AD strain. Tpk2p is a unique PKA isoform that is differentially active during pseudohyphal development, iron uptake, and respiratory function (23,24). A response to respiratory function is consistent with enhanced Cdr1p Akts phosphorylation when the cells enter diauxic phase. The absence of p-PKAs antibody recognition in Cdr1p and the partial inhibition of Akts site phosphorylation in the TPK2 deletion mutants (Fig. 3C) are consistent with Tpk2p phosphorylating Ser 307 but not Ser 484 . In contrast to the TPK2 deletion mutants of CDR1-AD, the comparable PDH1-AD mutants did not show increased azole resistance. The enzyme-specific resistance suggests that Tpk2p can negatively regulate Cdr1p, and this may be reflected in decreased glucose-dependent Rh6G pumping at glucose concentrations Ͼ10 mM. Furthermore, Pdh1p Akts site rephosphorylation was primarily regulated by the Tpk3p isoform. This specificity might explain why glucose at Ͼ1 mM minimally increased Rh6G pumping by the PDH1-AD strain.
Structural and Functional Analysis of the M1 Phosphorylation Site-Conservation of the amino acid sequences around the M1 and M2 sites in many fungal ABC transporters (Fig. 7) would suggest regulation comparable with Cdr1p and Pdh1p might occur widely in fungi, including fungal pathogens like C. albicans. Ser 307 at the Cdr1p M1 site is five amino acids C-terminal to the NBD1 ABC signature "LSGGQ" motif, which is not perfectly conserved in fungi. The sequence around Ser 307 is highly conserved in the PDR family transporters, SNQ2, and known PDR homologues in other pathogenic fungi, including C. albicans, Aspergillus fumigatus, and Filobasidiella neoformans (Fig. 7A). In addition, an M1 Ser or Thr is found in four S. cerevisiae non-PDR family ABC transporters.
The LSGGQ consensus motifs are normally highly conserved between NBDs of ABC transporters. The NBD1 Walker ABC signature motif VSGGE of fungal PDR family proteins corresponds with a LNVEQ consensus in the NBD2 of these proteins (Fig. 7B). A similar sequence to the M1 region is found in mammalian (hABCG1, hABCG2, and hABCG8) and insect (White and Brown) ABCG proteins (Fig. 7C) (17). ABC transporters with sequences similar to that around the M1 site of human MDR1, a multidrug pump functional homologue of Cdr1p, are widely distributed from Protista to mammals, and include the yeast mitochondrial peptide transporter MDL1 (Fig. 7C). Most of the human ABCA and ABCC (CFTR/MRP) family transporters have Ser or Thr at M1 equivalent sites. Alanine is the main amino acid in the M1 and M8 sites of MDR1 type multidrug transporters, and all ABCB (MDR1) family transporters have alanine at the M1 site, consistent with the substantial pump function of the CDR1-M1 strain. PKA-dependent phosphorylation was recently reported at sites equivalent to M1 in both NBDs of the full-size human ABCA1 (13). Phosphorylation of the NBD2 site was important for phospholipid efflux, but the effect of phosphorylation on ATPase activity was not reported.
The multiple defects caused by S307A mutation in NBD1, together with the phosphorylation of M1 but not the aligned M8 site in Cdr1p, indicate that the two NBDs in Cdr1p are functionally non-equivalent. Furthermore, the Rh6G efflux activity, drug resistance, p-Akts site phosphorylation, and plasma membrane nucleoside triphosphatase activity of CDR1-M8 and the CDR1-AD strain were identical. Thus the M8 site in NBD2 is not critical for Cdr1p function. The C193A mutation, in a catalytically active GST fusion of the N-terminal cytoplasmic domain of C. albicans Cdr1p, causes a loss of 95% of enzyme activity (25). We recently found that the equivalent C189A mutation in the NBD1 Walker A motif of Cdr1p gave 60% of normal ATPase activity and susceptibilities to FLC and  Table I.

Phosphorylation of C. glabrata Cdr1p
Rh6G intermediate between CDR1-M1 and CDR1-M2. 2 In contrast, a K899A mutant in the aligned residue in the Walker A motif of NBD2 showed no ATPase activity and had FLC and Rh6G susceptibilities comparable with strain AD1-8u Ϫ . These results confirm the functional non-equivalence of the two NBDs of Cdr1p, show that NBD2 is required for basal pump activity, and suggest that NBD1 regulates enzyme activity. If NBD1-TM6-NBD2-TM6 transporters like Cdr1p were functional monomers, drug pumping should involve the binding of ATP molecules using the Walker A and B motifs from one NBD and the ABC signature motif from the other NBD (26). The residue homologous to Ser 307 in the human TAP1 NBD (Protein Data Base file: 1JJ7) can be approached via a narrow groove. Phosphorylation of Ser 307 in Cdr1p, which confers enzyme stability in vitro and would probably need to occur in an open conformation of the enzyme, could significantly modify the disposition of the helix supporting the signature motif and affect interactions with the ␥-phosphate of the ATP bound by a second NBD. Selective phosphorylation of the Ser 307 is therefore proposed as a sensing device that allows cells to respond to stress in millimolar glucose environments by forming a productive complex between ATP and two NBDs. Reconstructed single molecule images show that the yeast Pdr5p transporter is a homodimer with rotational symmetry that allows NBDs to present equivalent faces across the center of the dimeric complex like the Bacillus subtilis half transporter YvcC (27). ATPmediated interactions between NBD1 pairs or NBD2 pairs in the homodimer would give productive complexes and, unless a 90°change in NBD orientation occurs, interactions between NBDs within the same polypeptide chain are unlikely. Differential phosphorylation of the NBD1s of this structure could give the enzyme activation seen at glucose concentrations be-low 20 mM and significant enzyme deactivation at glucose concentrations above 20 mM by regulating the frequency and/or turnover of productive contacts between like NBDs.
Functional Role of M2 Phosphorylation-The primary sequence around the M2 site, which is C-terminal of NBD1 and closer to the first transmembrane domain, is strongly conserved in Cdr1p, Pdh1p, and PDR homologues and differs from Snq2p, which has basic residues at both the Ϫ3 and Ϫ2 positions (Fig.  7D). The S484A mutation blocked M1 and M2 site phosphorylation in a 12-h YPD culture, decreased plasma membrane ATPase activity by 50%, but only modestly affected transport activity at glucose concentrations Ͻ10 mM. Partial suppression of the effects of the S307A mutation on Rh6G pumping by Ser 484 and the non-additive drug resistance properties of the S307A/S484A double mutation suggest conformational interactions between these two sites. Lack of suitable structural data for homology modeling precludes further interpretation.
Drug Resistance of Mutant Strains-Although CDR1-M1 and CDR1-M1,2 showed major reductions in Rh6G efflux, both strains were 2-to 5-fold more susceptible than CDR1-AD to FLC and still showed higher drug resistance than the control null strain or even PDH1-AD, in both liquid MIC and agar diffusion assays. These observations can be reconciled by assuming that the S307A mutation affected the formation of productive ATPase complexes at low glucose concentrations and that this defect was partially suppressed by the phosphorylation/conformational change at S484A at higher glucose concentrations. Thus, hyperexpression of partially functional Cdr1p in the 2% (110 mM) glucose in the nutrient-rich media used for drug susceptibility assays should support sufficient pump function to detoxify the M1 mutant at glucose concentration Ͼ5 mM. Because ATPase activity is required for ABC transporter pump function, the CDR1-M1 enzyme must be sufficiently stable for phosphorylation/conformational change at Ser 484 to support partial pump function, even though only ϳ5% of the activity of the enzyme survives plasma membrane isolation in the presence of protective concentrations of ATP. The CDR1-M1,2 strain, at glucose concentrations that activate the S307A or S484A pumps, failed to efflux Rh6G. The identical FLC and Rh6G resistances, obtained for independent CDR1-M1,2 isolates and upon recloning of this CDR1-URA3 construct in the FLC-and Rh6G-hypersensitive AD1-8u Ϫ strain, show that overexpression of an apparently non-functional Cdr1p gives significant drug resistance. Although the mechanism responsible has yet to be determined, enhanced ergosterol-rich lipid raft formation, an undetected enzyme-stabilizing modification, or residual activity of the hyperexpressed Cdr1p may be important.
Drug resistance assays selected a low frequency of intragenic suppressors (e.g. D32H, V353L, and ⌬A349) that assembled productive S307A/S484A Cdr1p with enhanced Rh6G pumping and diminished susceptibility to FLC and Rh6G. We speculate that the ⌬A349 or V353L mutations may modify the structure of the H-loop in NBD1. The Cdr1p H-loop lacks the "invariant" H thought to contribute to NBD interactions as a result of ATP binding. The Blastp algorithm aligns Ala 349 with this amino acid, and its deletion may alter the configuration of neighboring amino acids. The D32H mutation produces an His 3 motif just three residues C-terminal to another His residue. This might introduce a structure-deforming divalent metal ion binding pocket or facilitate complementary interaction between homodimer N termini. The effects of these suppressor mutations on enzyme function and their possible location at homodimer inter-subunit boundaries are consistent with NBD1 mediating multiple responses to growth and stress by modulating the efficiency of productive inter-subunit interactions. A complex pattern of phosphorylation events, primarily involving the Ser 307 site, the compensatory Ser 484 site, and possibly involving other sites, regulate the overall ATPase and pumping activity of homodimeric Cdr1p. A similar, but differentially glucose-sensitive mechanism, probably regulates Pdh1p.