Characterization of phosphorylation-defective mutants of human P-glycoprotein expressed in mammalian cells.

To assess the role of phosphorylation of the human multidrug resistance MDR1 gene product P-glycoprotein for its drug transport activity, phosphorylation sites within its linker region were subjected to mutational analysis. We constructed a 5A mutant, in which serines at positions 661, 667, 671, 675, and 683 were replaced by nonphosphorylatable alanine residues, and a 5D mutant carrying aspartic acid residues at the respective positions to mimic permanently phosphorylated serine residues. Transfection studies revealed that both mutants were targeted properly to the cell surface and conferred multidrug resistance by diminishing drug accumulation. In contrast to wild-type P-glycoprotein, the overexpressed 5A and the 5D mutants exhibited no detectable levels of phosphorylation, either in vivo following metabolic labeling of cells with [P]orthophosphate or in vitro in phosphorylation assays with protein kinase C, cAMP-dependent protein kinase, or a P-glycoprotein-specific protein kinase purified from multidrug-resistant KB-V1 cells. These results reconfirm that the major P-glycoprotein phosphorylation sites are located within the linker region. Furthermore, the first direct evidence is provided that phosphorylation/dephosphorylation mechanisms do not play an essential role in the establishment of the multidrug resistance phenotype mediated by human P-glycoprotein.

To assess the role of phosphorylation of the human multidrug resistance MDR1 gene product P-glycoprotein for its drug transport activity, phosphorylation sites within its linker region were subjected to mutational analysis. We constructed a 5A mutant, in which serines at positions 661, 667, 671, 675, and 683 were replaced by nonphosphorylatable alanine residues, and a 5D mutant carrying aspartic acid residues at the respective positions to mimic permanently phosphorylated serine residues. Transfection studies revealed that both mutants were targeted properly to the cell surface and conferred multidrug resistance by diminishing drug accumulation. In contrast to wild-type P-glycoprotein, the overexpressed 5A and the 5D mutants exhibited no detectable levels of phosphorylation, either in vivo following metabolic labeling of cells with [ 32 P]orthophosphate or in vitro in phosphorylation assays with protein kinase C, cAMP-dependent protein kinase, or a P-glycoprotein-specific protein kinase purified from multidrugresistant KB-V1 cells. These results reconfirm that the major P-glycoprotein phosphorylation sites are located within the linker region. Furthermore, the first direct evidence is provided that phosphorylation/dephosphorylation mechanisms do not play an essential role in the establishment of the multidrug resistance phenotype mediated by human P-glycoprotein.
Multidrug resistance (MDR) 1 is a major impediment to effective cancer chemotherapy. In many human cancer cells, cross-resistance to a variety of natural product cytotoxic drugs is associated with the overexpression of the multidrug resistance MDR1 gene that encodes P-glycoprotein (reviewed in Refs. [1][2][3]. Highly homologous mdr or pgp genes have also been identified in rodents (reviewed in Refs. [1][2][3]. Based on the amino acid sequence deduced from the MDR1 cDNA sequence, P-glycoprotein is predicted to consist of two similar halves, each of which contains a transmembrane domain and a nucleotide binding fold (4,5). These structural elements identify the MDR1 gene product as a member of the superfamily of ATPbinding cassette transporters, which includes the cystic fibrosis transmembrane conductance regulator, and many other membrane-associated proteins from eukaryotic and prokaryotic origin (reviewed in Ref. 6). Gene transfer experiments involving MDR1 cDNA have corroborated that expression of P-glycoprotein is sufficient to endow drug-sensitive cells with multidrug resistance (reviewed in 2, 3). P-glycoprotein is an integral plasma membrane protein that functions as an energy-dependent drug efflux pump to reduce the intracellular accumulation of cytotoxic agents (reviewed in Refs. [1][2][3]. P-glycoprotein interacts directly with a variety of anticancer drugs and transports them across the plasma membrane lipid bilayer. P-glycoprotein exhibits a substrate-stimulated ATPase activity (7,8), suggesting that ATP hydrolysis may provide the energy required for the drug transport mechanism (reviewed in Refs. [1][2][3]. P-glycoprotein was described as a phosphoglycoprotein (9,10), and several studies have corroborated that both native and recombinant P-glycoproteins are phosphorylated in vivo (11)(12)(13)(14)(15)(16)(17)(18)(19)(20). Numerous studies have been conducted to address the importance of phosphorylation for the multidrug transporter activity of P-glycoprotein. Many multidrug-resistant cell lines were shown to express elevated levels of protein kinases, in particular protein kinase C (21)(22)(23)(24) and changes in levels and or activities of protein kinases, such as protein kinase C or cAMP-dependent protein kinase (protein kinase A) have been suggested to play a role in modulating levels of multidrug resistance mediated by P-glycoprotein. Attempts have been made to correlate the degree of phosphorylation of P-glycoprotein with its drug efflux activity (Refs. 11,13,16,22,25,and 26;reviewed in Ref. 27). Generally an increase in protein kinase activity and/or phosphorylation of P-glycoprotein has been associated with increased levels of multidrug resistance. Many of these studies involved the use of activators and/or inhibitors of protein kinases to modulate the state of phosphorylation of P-glycoprotein, but these regulatory molecules are not very specific and often cause multiple cellular effects. For example, several protein kinase inhibitors including staurosporine and derivatives thereof, calphostin C, or certain isoquinolinesulfonamide derivatives may directly interact with P-glycoprotein and affect its drug efflux activity independent of, or in addition to, their effects on P-glycoprotein phosphorylation (28 -31). Various protein kinase agonists (e.g. 12-O-tetradecanoylphorbol-13-acetate or diacylglycerol) and antagonists (e.g. stauro-sporine, H-87) may also affect MDR1 gene expression (32)(33)(34)(35). Thus, the role of phosphorylation of P-glycoprotein has not been clearly established.
Recent approaches have focused on the identification of the phosphorylation sites within the primary structure of P-glycoprotein. In human P-glycoprotein, phosphoserine, but not phosphothreonine or phosphotyrosine, has been detected by phosphoamino acid analysis (15,16,19). Chambers and co-workers (36,37) demonstrated that the major phosphorylation sites of the human MDR1 gene product are confined to a central cytosolic segment of approximately 60 amino acids that connects the two homologous halves of P-glycoprotein. This region, commonly referred to as the linker region, is characterized by a high content of charged amino acids (approximately 30 -40%) and contains several consensus sequences for phosphorylation by protein kinases requiring basic amino acid residues near the phosphoacceptor group (e.g. protein kinase C, protein kinase A). A cluster of four serine residues was shown to be phosphorylated in vitro by protein kinase C (Ser-661, Ser-667, Ser-671) and/or protein kinase A (Ser-667, Ser-671, and Ser-683) (36,37). Three of these four serine residues appear to be phosphorylated in vivo, namely Ser-661, Ser-667, and Ser-671 (26,36,37). Similarly, the linker region of the mouse mdr1 P-glycoprotein has been demonstrated to be phosphorylated in vitro at analogous serine residues, namely Ser-669 by protein kinase C and at Ser-681 by protein kinase A (38). Several protein kinase C and/or protein kinase A consensus phosphorylation sites are also present in the linker region of the mouse mdr3 and the hamster pgp1 P-glycoproteins, but the actual sites of phosphorylation have not yet been described. In analogy with the R domain between the two halves of the cystic fibrosis transmembrane conductance regulator, a target for multisite phosphorylation by protein kinase A believed to regulate the cAMP-dependent cystic fibrosis transmembrane conductance regulator chloride channel activity, it has been hypothesized that the phosphorylatable linker region of P-glycoprotein may be a regulatory domain that controls its drug transport function (36,38).
The identification of the major sites of phosphorylation provides an opportunity to use site-directed mutagenesis to address the role of phosphorylation of P-glycoprotein for its drug efflux activity. Our approach was to substitute five consensus sites for phosphorylation by protein kinase C (Ser-661, Ser-667, Ser-671, Ser-675, Ser-683) within the linker region of P-glycoprotein by nonphosphorylatable alanine residues (5A mutant), or by aspartic acid residues to mimic permanently phosphorylated serine-like residues (5D mutant). The 5A and 5D mutants of P-glycoprotein were tested for their ability to confer multidrug resistance to drug-sensitive cells and were characterized for drug-binding capacity and state of phosphorylation.

MATERIALS AND METHODS
Genetic Engineering of Expression Vectors for Mutants of P-glycoprotein-To facilitate the construction of P-glycoprotein phosphorylation mutants, two unique restriction sites, ClaI and XbaI, were introduced into the nucleotide sequence of the human MDR1 cDNA open reading frame at positions 1941 and 2060, flanking the sequences encoding the linker region. The ClaI and XbaI sites were created by a T 3 C transition at position 1943 and a G 3 A transition at position 2063. These changes in the nucleotide did not affect the encoded amino acid sequence. Three MDR1 cDNA fragments, an ApaI-ClaI fragment encompassing nucleotides 1586 -1946, a ClaI-XbaI fragment encompassing nucleotides 1941-2065, and an XbaI-Asp718 fragment encompassing nucleotides 2060 -2780 were generated by the polymerase chain reaction using pMDR2000XS (39) as a template. Each of these three fragments was subcloned individually into the appropriate restriction sites of a pGem2-derived vector pEACXA, 2 and the correctness of their sequence was confirmed. Then the three subcloned PCR fragments were ligated together and inserted to replace the native ApaI-Asp718 MDR1 cDNA fragment in pSX-MDR1/A-wt, a pGem2-derived plasmid carrying the full-length MDR1 cDNA isolated from the multidrugresistant KB-V1 cell line (40) as an SstII-XhoI fragment. 2 This control plasmid was named pSX-MDR1/A-wt-CX and contains novel ClaI and XbaI restriction sites.
For constructing P-glycoprotein mutants in which five serine residues at positions Ser-661, Ser-667, Ser-671, Ser-675, and Ser-683 were substituted with either alanine or aspartic acid, modified ClaI-XbaI fragments were generated by chemical synthesis of a series of eight complementary and overlapping oligodeoxynucleotides. , and UAG-121 (5Ј-CTA-GAGCCTCTTTGGTATCAAGCTTTCT-3Ј) for the 5D mutant. All oligodeoxynucleotides were gel-purified and (except for UAG-114, UAG-121, and UAG-123) phosphorylated at the 5Ј end using T4 polynucleotide kinase according to standard procedures (41). Equimolar amounts of eight oligodeoxynucleotides were annealed in the presence of 10 mM MgCl 2 by heating to 80°C and slow cooling to room temperature. Subsequently, the annealed oligodeoxynucleotides were introduced into ClaI and XbaI double-digested pSX-MDR1/A-wt-CX, and their DNA sequences were confirmed. Finally, the wild type and two mutant MDR1 cDNAs were isolated as SstII-XhoI fragments and placed under control of Harvey murine sarcoma virus long terminal repeats in the pCO1 retroviral vector (42), to give the expression vectors pHa-MDR1/A-wt-CX (wild-type control), pHaMDR1/A-wt-5A (encoding mutant carrying five alanine residues at positions 661, 667, 671, 675, and 683), and pHaMDR1/A-wt-5D (encoding mutant carrying five aspartic acid residues at positions 661, 667, 671, 675, and 683).
Mass populations of highly drug-resistant transfectants were selected in stepwise increasing concentrations of vincristine as follows. Approximately 200,000 cells were seeded per 10-cm dish, and increasing amounts of vincristine were added. Cells were grown for 7-14 days until colonies were visible to the eye. The highest vincristine concentration survived by all three different transfectants was chosen to adapt cell populations during two passages. Then the next step of selection was initiated as described above. Retrospectively, adaptation concentrations of vincristine for drug selection were 180 ng/ml, 600 ng/ml, and 2400 ng/ml for NIH 3T3 sublines, and 18 ng/ml, 48 ng/ml, and 300 ng/ml for KB-3-1 sublines.
Drug resistance profiles of the NIH 3T3 and KB-3-1 parental cell lines and vincristine-selected transfectants were determined by measuring cell survival in colony formation assays as described (44). Average cloning efficiencies for these assays were 10 -20% for NIH 3T3 sublines and approximately 50% for KB sublines. Drug accumulation assays were performed as described previously (26).
Fluorescence-activated Cell Sorting (FACS) Analysis-Subconfluent cells from a 10-cm dish were harvested by trypsinization into phos-phate-buffered saline (PBS), washed twice with PBS containing 1% bovine serum albumin (PBS-BSA), and resuspended in PBS-BSA at a concentration of approximately 10 6 cells/ml. 2.5 ϫ 10 5 cells were incubated for 30 min at room temperature with 5 g of human P-glycoprotein-specific MRK16 monoclonal antibody (a gift of Hoechst Japan Ltd.) or 5 g IgG2a isotype control antibody (Pharmingen), washed twice with PBS-BSA, and reacted with fluorescein-conjugated (FITC-labeled) goat-anti-mouse IgG2a antibody (1:10 diluted; Jackson Immunoresearch Laboratories, West Grove, PA) for 30 min at room temperature. The cells were again washed twice as described above, and the levels of FL1 fluorescence were analyzed using a FACSort flow cytometer with LYSIS II software (Beckton-Dickinson FACS System, San Jose, CA). Each sample was individually compared with the respective isotype control.
For preparation of crude membranes, cells were scraped as described above and washed once with PBSAp and once with lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1 mM MgCl 2 , 1% (v/v) aprotinin). Cells were resuspended in lysis buffer, incubated on ice for 45 min, and Dounce-homogenized (30 times with pestles A and B). An equal volume of TSNa (10 mM Tris-HCl, pH 7.5, 250 mM sucrose, 50 mM NaCl, 1% (v/v) aprotinin) was added to the lysate, followed by centrifugation at 500 ϫ g for 10 min. The low speed supernatant was centrifuged at 100,000 ϫ g for 1 h. The high speed pellet was washed once with TSNa, resuspended in TSNa, and stored at Ϫ80°C until use.
In Vivo Phosphorylation-Cultures were metabolically labeled with 0.1 mCi/ml [ 32 P]orthophosphate for 4.5 h at 37°C. Cells were washed twice and harvested into ice-cold PBS. PBS and all solutions for immunoprecipitations were supplemented with 1 mM phenylmethylsulfonyl fluoride, 0.2 mg/ml aprotinin, 5 mM sodium fluoride, 0.2 mM sodium orthovanadate, and 20 nM okadaic acid. After centrifugation at 500 ϫ g for 5 min the cell pellets were suspended in 0.5 ml of radioimmune precipitation buffer, incubated on ice for 30 min, and centrifuged at 100,000 ϫ g for 30 min. Protein concentrations of the supernatants were determined by the method of Bradford (48). Equivalent amounts of extracts (0.5 mg of protein) were subjected to immunoprecipitation with 13 l of P-glycoprotein-specific antibody PEPG13 (46) as described previously (47). Proteins were precipitated with acetone (47) and redissolved in 60 l SDS-polyacrylamide gel electrophoresis sample buffer without boiling. Samples were divided in two, and each half was electrophoresed on an 8% SDS-polyacrylamide gel. One gel was stained with Coomassie Blue, destained, dried, and subjected to autoradiogra-3 S. V. Ambudkar, unpublished data.

FIG. 1. Design of mutants of P-glycoprotein with substitutions in linker region.
A, the P-glycoprotein polypeptide chain consisting of 1280 amino acids is schematically represented as a line. Bars 1-12 refer to predicted transmembrane regions, and nucleotide binding folds (NB) are circled. The wild-type (WT) P-glycoprotein encoded by pHaMDR1/A-wt-CX harbors a cluster of five putative phosphorylation sites including Ser-661, Ser-667, Ser-671, Ser-675, and Ser-683 within the linker region (enlarged insert). In the 5A mutant encoded by pHaMDR1/A-wt-5A these serine residues are replaced by nonphosphorylatable alanine residues, whereas the 5D mutant encoded bÿ pHaMDR1/A-wt-5D contains aspartic acid residues to mimic permanently phosphorylated serine-like residues. B, amino acid sequence of the linker region of human P-glycoprotein between amino acids 633 and 692. Arrows point to serine residues targeted by site-directed mutagenesis.
phy with Kodak XAR5 film at Ϫ70°C. The duplicate gel was transferred to nitrocellulose, and P-glycoprotein was detected by immunostaining with C219 monoclonal antibody as described above.

Construction of 5A and 5D Mutant MDR1 cDNA Expression
Vectors-Oligodeoxynucleotide-mediated mutagenesis was used to alter the major putative phosphorylation sites within the linker region of the human MDR1 gene product (Fig. 1). Five clustered serine residues at positions 661, 667, 671, 675, and 683, all representing protein kinase C and/or protein kinase A consensus sequences, were targeted for substitution with either nonphosphorylatable alanine residues or aspartic residues to mimic permanently phosphorylated serine-like residues. Four of these five targeted serine residues (Ser-661, Ser-667, Ser-671, and Ser-683) have previously been demonstrated to be phosphorylated in vitro by protein kinase C and/or protein kinase A (36,37).
To facilitate the construction of these mutations within the MDR1 cDNA, two unique ClaI and XbaI restriction sites, flanking the linker region that encodes the phosphorylation site cluster, were introduced without affecting the encoded amino acid sequence. Wild-type and 5A and 5D mutant ClaI-XbaI linker subfragments were generated by gene synthesis, reintroduced into full-length MDR1 cDNA, and cloned into a pCO1derived retroviral expression vector under the control of long terminal repeats (LTRs) of the Harvey murine sarcoma virus. The three different retroviral expression constructs, termed pHaMDR1/A-wt-CX, pHaMDR1/A-wt-5A, and pHaMDR1/Awt-5D, encode the wild-type human MDR1 gene product with five serine residues at positions 661, 667, 671, 675, and 683 or two mutant forms of P-glycoprotein with five alanine (5A mutant) or aspartic acid residues (5D mutant) at the respective positions.
Biological Activity of 5A and 5D Mutants of P-glycoprotein-The pHaMDR1/A-wt-CX, pHaMDR1/A-wt-5A, and pHaMDR1/ A-wt-5D expression vectors were first tested for their ability to confer drug resistance to drug-sensitive murine NIH 3T3 or human KB-3-1 host cells upon transfection. Initial selection of transfected cell populations in the presence of cytotoxic concentrations of vincristine or colchicine indicated that all three constructs encode a functional multidrug transporter (Table I). Similar numbers of vincristine-resistant colonies were obtained for all wild-type and mutant constructs, but the numbers of colchicine-resistant colonies were somewhat lower for pHaMDR1/A-wt-5D and even more so for pHaMDR1/A-wt-5A in NIH 3T3 host cells, when compared with the pHaMDR1/Awt-CX control.
Mass populations of stably transfected, drug-selected cells were grown in the presence of stepwise increasing amounts of vincristine to achieve high levels of expression of the wild-type and mutant forms of P-glycoprotein, to compare relative resist-ance to different drugs of populations of transfectants, and for biochemical analysis. Generally, selection for cells exhibiting enhanced drug resistance was achieved easily for all three different types of transfectants. The pHaMDR1/A-wt-5A transfectants were rate-limiting for all steps of selection of the NIH 3T3 sublines but not for all steps of selection of the KB-3-1 sublines. The concentration of vincristine in the growth medium was raised three times for adaptation of NIH 3T3 transfectants from 30 to 180 ng/ml, and then to 600 ng/ml, and finally to 2400 ng/ml. Similarly, highly drug-resistant KB-3-1 transfectants were selected in four steps by adaptation at 3 ng/ml, 18 ng/ml, 48 ng/ml, and 300 ng/ml vincristine.
Enhanced vincristine resistance of subpopulations is usually paralleled by an increase in the levels of expression of transfected MDR1 gene products. Western blot analysis of whole cell extracts was used to determine the overall P-glycoprotein content in various transfected sublines. As shown in Fig. 2, both the 5A and 5D mutant P-glycoproteins were found to exhibit electrophoretic mobility comparable with that of the wild type, and both 5A and 5D mutants were recognized by the C219 monoclonal antibody, which binds to an epitope near the nucleotide binding regions of P-glycoprotein (49). Fig. 2A reveals that the increased resistance of NIH 3T3 transfectants from 30 ng/ml to 2400 ng/ml vincristine was accompanied by a dramatic and comparable increase in levels of expression of wild-type or mutant P-glycoproteins. The increase was estimated to be  3 and 6), and pHaMDR1/A-wt-5D-transfectants (lanes 4 and 7) that were selected in the presence of 30 ng/ml (N3V30, N4V30, N5V30) or 2400 ng/ml (N3V2400, N4V2400, N5V2400) vincristine. A, total cellular proteins were size-fractionated by SDS-polyacrylamide gel electrophoresis and analyzed by Western blot using P-glycoproteinspecific C219 monoclonal antibody as described under "Materials and Methods." B, crude membranes were prepared from the same cell lines and stained with Coomassie Blue after separation by SDS-polyacrylamide gel electrophoresis. Arrows indicate wild-type and 5A and 5D mutant P-glycoproteins, and sizes of molecular mass markers are given in kDa.

. Expression of 5A and 5D mutants in stably transfected, vincristine-selected NIH 3T3 cells. Cellular extracts or crude membranes were prepared from parental NIH 3T3 cells (lane 1), pHaMDR1/ A-wt-CX-transfectants (lanes 2 and 5), pHaMDR1/A-wt-5A transfectants (lanes
Ͼ100-fold based on densitometry scanning analysis. The NIH 3T3 populations maintained at 2400 ng/ml vincristine harboring pHaMDR1/A-wt-CX (N3V2400 cells), pHaMDR1/A-wt-5A (N4V2400 cells), and pHaMDR1/A-wt-5D (N5V2400 cells) sequences appeared to contain similar amounts of transfected MDR1 gene product based on the relative intensities of the immunoreactive bands. The levels of the transfected wild-type and mutant P-glycoproteins were high enough to allow detection by Coomassie Blue staining after SDS-polyacrylamide gel electrophoresis of crude membrane preparations of N3V2400, N4V2400, and N5V2400 cells (Fig. 2B). Cell populations grown at intermediate concentrations of vincristine contained intermediate levels of transfected MDR1 gene products according to their relative vincristine resistance (data not shown). Similar data were also obtained for the KB-3-1 transfectants, but overall levels of wild-type and mutant P-glycoprotein expression in vincristine-selected sublines were lower than in NIH 3T3 transfectants (data not shown). Generally, recombinant wildtype and mutant human P-glycoproteins expressed in murine NIH 3T3 transfectants had increased electrophoretic mobility FIG. 3. Cell surface expression of 5A and 5D mutants of P-glycoprotein. Parental NIH 3T3 cells, control KB-3-1 and KB-V1 cells, and NIH 3T3 transfectants expressing wild-type P-glycoprotein (N3V30, N3V2400), 5A mutant (N4V30, N4V2400), or 5D mutant (N5V30, N5V2400) were subjected to FACS analysis after staining with monoclonal human P-glycoprotein-specific antibody MRK-16 or IgG2a isotype control and FITC-labeled anti-mouse IgG2a antibody as described under "Materials and Methods." Transfectants were maintained at 30 ng/ml (N3V30, N4V30, N5V30) or 2400 ng/ml vincristine (N3V2400, N4V2400, N5V2400). compared with P-glycoproteins expressed in human KB-3-1 transfectants, due to altered carbohydrate content. 3 A total of four different P-glycoprotein-specific detection reagents were used for Western blot analyses of N3V30, N4V30, N5V30, N3V2400, N4V2400, and N5V2400 cells, including C219 monoclonal antibody (49), 4007 antiserum (45), PEPG2 antiserum (46), and PEPG13 antiserum (46), and they all gave comparable results (Fig. 2, data not shown). The PEPG13 antiserum is specific for the human MDR1 gene product and does not cross-react with human MDR2 or mouse P-glycoproteins. 4 Thus, these results indicate that the levels of the transfected wild-type or mutant human P-glycoproteins, but not the endogenous mouse P-glycoproteins, were increased by the vincristine selection of the various NIH 3T3 transfectants. Southern blot hybridizations of genomic DNAs isolated from NIH 3T3, N3V30, N4V30, N5V30, N3V2400, N4V2400, and N5V2400 cells with the pHDR.5 probe (50) indicated that drug selection of the transfectants was accompanied by amplification of the transfected human MDR1 cDNAs (data not shown). Washes at low stringency did not indicate amplification of endogenous mouse mdr genes (data not shown).
FACS analysis of NIH 3T3 transfectants (Fig. 3, data not shown) and KB-3-1 transfectants (data not shown) using the human P-glycoprotein-specific monoclonal antibody MRK16 or an IgG2a isotype control, and an FITC-labeled anti-mouse IgG2a secondary antibody, confirmed the Western blot findings described above and indicated proper targeting of the 5A and 5D mutants to the cell surface. A comparable increase in cell surface fluorescence intensity of MRK16 signals was observed when increasingly resistant cell populations were investigated. Thus, similarly elevated levels of expression of the wild-type human P-glycoprotein and the 5A and 5D mutants at the cell surface of murine NIH 3T3 and human KB-3-1 transfectants correlated with their increased resistance to vincristine.
Colony formation assays in the presence of vinblastine, colchicine, and adriamycin were performed to demonstrate that the vincristine-selected pHaMDR1/A-wt-CX-, pHaMDR1/Awt-5A and pHaMDR1/A-wt-5D-transfectants were cross-resistant to various cytotoxic drugs. As demonstrated by the killing curves in Fig. 4, both 5A and 5D mutants conferred multidrug resistance to the transfected cells, similar to wild-type P-glycoprotein. For all vincristine-selected transfectants, the rela-   (47). Arrows indicate wild-type and 5A and 5D mutant P-glycoproteins, and sizes of molecular mass markers are given in kDa.
tive resistance to the Vinca alkaloid tested (vinblastine) was higher than to colchicine or the anthracycline adriamycin. The relative resistance to vinblastine of the initially selected transfectants (N3V30, N4V30, N5V30 cells) was less than 10-fold when compared with the parental NIH 3T3 cell line, whereas cells maintained at the highest concentration of 2400 ng/ml vincristine (N3V2400, N4V2400, N5V2400 cells) were approximately 100 -150-fold resistant. Similar data were obtained for KB-3-1 transfectants (data not shown). Thus, like wild-type P-glycoprotein, both 5A and 5D mutants conferred simultaneous resistance to a variety of hydrophobic cytotoxic agents.
Drug-binding Activity of 5A and 5D Mutants of P-glycoprotein-Photoaffinity labeling experiments were carried out to characterize the drug-binding capacity of the 5A and 5D mutant P-glycoproteins in more detail. These studies were performed with N3V2400, N4V2400, and N5V2400 sublines because of their high and similar expression of the transfected MDR1 gene products. As shown in Fig. 5, both 5A and 5D mutants were labeled specifically with [ 3 H]azidopine, with ef-ficiencies similar to wild-type P-glycoprotein. A 250-fold excess of nonradioactive vinblastine displaced the [ 3 H]azidopine label completely from all three different forms of P-glycoprotein analyzed, indicating that both 5A and 5D mutants were capable of specific interactions with cytotoxic drug substrates.
State of Phosphorylation of 5A and 5D Mutants of P-glycoprotein-In the 5A and 5D mutants, five consensus sites of phosphorylation by protein kinase C and/or protein kinase A including the previously identified major phosphorylation sites of P-glycoprotein were substituted with the nonphosphorylatable amino acid residue alanine or aspartic acid. To exclude the presence of cryptic phosphorylation sites within the mutant P-glycoprotein polypeptide chains, it was important to evaluate the state of phosphorylation of the 5A and 5D mutants by comparing it with that of wild-type P-glycoprotein. The N3V2400, N4V2400, and N5V2400 sublines were chosen for both in vivo and in vitro phosphorylation experiments, because they were the cell lines that expressed the highest and similar levels of transfected MDR1 gene products (Fig. 2). Phosphorylation was studied in intact cells by measuring incorporation of [ 32 P]orthophosphate into the proteins. Immunoprecipitates were prepared from extracts of metabolically 32 P-labeled cells using PEPG13 antiserum. This polyclonal antiserum directed against amino acids 592-636 of the MDR1 gene product is human P-glycoprotein-specific and does not cross-react with mouse P-glycoproteins. 3 The autoradiogram presented in Fig.  6A, even upon prolonged exposure (data not shown) indicated no significant levels of phosphorylation of the 5A and 5D mutants of P-glycoprotein expressed in N4V2400 and N5V2400 cells, respectively. The background signals were comparable with those obtained with drug-sensitive NIH 3T3 cells or KB-3-1 cells, that do not produce human P-glycoprotein. In contrast, the wild-type P-glycoprotein in N3V2400 cells and KB-V1 control cells was highly phospholabeled. The immunoblots presented in Fig. 6B confirmed that the immunoprecipitates from N3V2400, N4V2400, and N5V2400 extracts contained approx-  6) were metabolically labeled with [ 32 P]orthophosphate (0.1 mCi/ml) for 4.5 h, and P-glycoprotein was immunoprecipitated with PEPG13 antiserum as described under "Materials and Methods." A, 32 P labeling of P-glycoprotein was visualized by autoradiography; exposure was for 24 h with an intensifying screen. A longer exposure of 5 days confirmed essentially no radioactive labeling of P-glycoprotein 5A and 5D mutants in lanes 3 and 4. B, P-glycoprotein was detected following SDS-polyacrylamide gel electrophoresis by immunoblotting with P-glycoprotein-specific C219 monoclonal antibody as described under "Materials and Methods." Arrows indicate wild-type and 5A and 5D mutant P-glycoproteins, and sizes of molecular mass markers are given in kDa.
imately equal amounts of the various forms of human P-glycoprotein. These data suggest that the five serine residues that were substituted with alanine or aspartic acid in the 5A and 5D mutants include all major phosphorylation sites of human Pglycoprotein. No major cryptic phosphorylation site(s) was detectable in either the 5A or the 5D mutant of P-glycoprotein.
Additional in vitro phosphorylation studies were conducted using crude membrane preparations from N3V2400, N4V2400, and N5V2400 cells. As shown in Fig. 7, wild-type P-glycoprotein, but not the 5A or the 5D mutant, was phosphorylated by protein kinase C as well as by a novel, membrane-bound protein kinase (V-1 kinase) 3 isolated from KB-V1 cells. Similar data were obtained using protein kinase A (data not shown). We also did not detect phosphorylation of the 5A and 5D mutant P-glycoproteins after immunoprecipitation with the monoclonal antibody C219 known to cross-react with all human and rodent P-glycoproteins (49). These results confirm earlier findings (see above and Fig. 3) indicating that only human Pglycoproteins (wild type or mutants) are overexpressed in the transfectants. No significant incorporation of radioactive phosphate was detected for the 5A and 5D mutants of P-glycoprotein, even in the presence of vanadate, a general inhibitor of phosphatases and ATPases (data not shown), confirming the results from the in vivo phosphorylation experiments described above.
Taken together, these data suggest that the major sites for in vitro and in vivo phosphorylation of P-glycoprotein are absent in the 5A and 5D mutants due to substitution by nonphosphorylatable alanine or aspartic acid residues. Clearly, the analyzed phosphorylation-and dephosphorylation-defective mutants of P-glycoprotein confer multidrug resistance to drugsensitive cells with efficiency comparable with the wild-type P-glycoprotein and have similar drug-binding capacity. Although we cannot exclude completely the possibility that the 5A and 5D mutants are phosphorylated at very low levels below detection limits, our data infer that phosphorylation of the major sites in the linker region is not essential for the drug efflux activity of human P-glycoprotein. Multidrug resistance can be mediated by P-glycoprotein in the absence of phosphorylation, as well as in a state mimicking stable phosphorylation, compensated by negatively charged residues. DISCUSSION We have employed site-directed mutagenesis of predicted phosphorylation sites within the linker region of human Pglycoprotein to assess the importance of phosphorylation for its multidrug transport function and found that the activity of P-glycoprotein is not affected by its state of phosphorylation. Previously, analyzing P-glycoprotein in human KB-V1 cells, a total of three serine residues were found to be phosphorylated in vivo (26). Subsequent in vitro analyses identified these as Ser-661, Ser-667, and Ser-671 (36,37). Our mutational analysis targeted these three predicted phosphorylation sites within the P-glycoprotein linker region as well as Ser-675 and Ser-683. The latter serine residue was included because it was found to be phosphorylated by protein kinase A in vitro (37). Ser-675 was mutagenized because it represents a nearby minimal consensus site for phosphorylation by protein kinase C that may become modified in the absence of the major phosphorylation sites. Two types of mutants were constructed: a 5A mutant, in which the five serine residues were replaced by nonphosphorylatable alanine residues, and a 5D mutant with five aspartic acid residues at the respective positions to mimic permanently phosphorylated serine-like residues. Stable transfection experiments with murine NIH 3T3 and human KB-3-1 cells demonstrated that both the 5A and 5D mutants of P-glycoprotein are expressed at the cell surface and endow the drug-sensitive host cells with resistance to a variety of cytotoxic drugs including Vinca alkaloids, anthracyclines, and colchicine. Subpopulations overexpressing high levels of transfected wild-type or mutant P-glycoproteins were selected to facilitate biochemical analyses. Highly drug-resistant murine transfectants were found to specifically overexpress the transfected MDR1 gene products, and there was no evidence for increased expression of endogenous mouse P-glycoproteins that could account for increased levels of multidrug resistance. In contrast to wild-type P-glycoprotein, no significant levels of phosphorylation were detected for the 5A and 5D mutants by phosphorylation studies performed in vitro and in vivo, despite their high levels of expression. Thus, in agreement with the biochemical data reported by Chambers et al. (36,37) our mutational analysis confirmed that all the detectable phosphorylation sites are located in the linker region of human P-glycoprotein.
Interestingly, our phosphorylation studies performed with the 5A and 5D mutants did not indicate any other major cryptic phosphorylation site(s) present within or outside of the Pglycoprotein linker region, although the possibility of nondetectable minor phosphorylation sites cannot be completely excluded. Studies with additional P-glycoprotein mutants containing four or fewer alanine or aspartic acid substitutions at positions 661, 667, 671, 675, and 683 are ongoing to determine the exact in vivo phosphorylation sites by mutational analysis. As there is no evidence for any phosphorylatable residue(s) within the 5A and 5D mutants, they should serve as useful tools to reevaluate the effects of protein kinase activators and inhibitors on levels of multidrug resistance. They may also help to clarify the mechanisms of action of protein kinase modulators.
The 5A and 5D mutants mimic a phosphorylation-deficient and permanently phosphorylated P-glycoprotein, respectively. We found that both the 5A and 5D mutants of P-glycoprotein confer multidrug resistance to drug-sensitive cells. Our data imply that phosphorylation and dephosphorylation of P-glycoprotein is not essential for its drug efflux activity. A similar conclusion was implied in earlier work by Buschman and Gros (51), who showed that a chimeric mouse mdr1 P-glycoprotein containing the mouse mdr2 linker region was functional as a drug transporter. Since the mouse mdr2 linker region lacks the classic RRXS recognition sequence that is present in the mdr1 gene product, the argument was made that phosphorylation of this site was dispensable, at least for the multidrug transport activity of mouse P-glycoprotein (51).
Recently, a study involving site-directed mutagenesis of eight serine/threonine consensus sites for protein kinase C in the linker region of human P-glycoprotein has suggested that protein kinase C-mediated phosphorylation of P-glycoprotein may function to regulate the activity of an endogenous chloride channel (52). Despite the fact that confirmatory phosphorylation studies of the P-glycoprotein mutants were not performed, the eight sites mutated included the five sites targeted by us. Therefore, the P-glycoprotein mutants studied by Hardy et al. (52) were most likely phosphorylation-defective. Thus, one possible role for P-glycoprotein phosphorylation may be to affect indirectly the activity of another protein or proteins.
Our data suggest that phosphorylation and dephosphorylation of P-glycoprotein may not be essential for its overall ability to interact with different cytotoxic agents. It is possible, however, that phosphorylation of P-glycoprotein may contribute to regulation of its drug substrate specificity, as has been suggested by Bates et al. (11). Although there is no obvious pattern that correlates the state of phosphorylation mimicked by the 5A and 5D mutants with levels of resistance to a certain drug, it is interesting to note that both mutants in comparison with wild-type P-glycoprotein appear to have a somewhat reduced capacity to confer resistance to colchicine, and adriamycin, but not Vinca alkaloids. Our results from both the initial drug selection of transfected cell populations and the drug resistance profiles of the highly drug-resistant transfectants indicated this tendency. Additional studies with cloned sublines and P-glycoprotein mutants in which not all phosphorylation sites are mutagenized should address a possible role of phosphorylation in regulating substrate specificity.
Phosphorylation of P-glycoprotein has also been suggested to influence the kinetics (velocity) of drug transport (25) or to modulate multidrug resistance by affecting the half-life of Pglycoprotein (34). The 5A and 5D mutants of P-glycoprotein represent good models to address these issues. Ultimately, experiments involving functional reconstitution of purified wild-type and phosphorylation-or dephosphorylation-defective mutants of P-glycoprotein in phospholipid vesicles may establish the degree to which the drug efflux function of the multidrug transporter may be regulated by a phosphorylation/dephosphorylation mechanism. One intriguing idea, however, is that phosphorylation of P-glycoprotein may be less important for drug efflux than for the regulation of the transport of yet to be determined physiologic substrate(s).