Evidence that natural killer cells express mini P-glycoproteins but not classic 170 kDa P-glycoprotein


Dr Greg Woods, Discipline of Pathology, University of Tasmania, GPO Box 252–29, Hobart, Tasmania, Australia. E-mail: G.M.Woods@utas.edu.au


Several lines of evidence including reverse transcription polymerase chain reaction, immunoreactivity and their ability to efflux rhodamine 123 have implied the existence of P-glycoprotein in natural killer (NK) cells. It has been a natural tendency to assume that NK-cell P-glycoprotein is identical to the P-glycoprotein of multidrug resistant (MDR) cell lines, however, the present study uncovered major differences. Functionally, NK cells demonstrated a restricted substrate profile, being unable to transport daunorubicin and calcein acetoxymethylester while efficiently transporting other P-glycoprotein substrates. Furthermore, physical differences in NK-cell P-glycoprotein were established by differential reactivity with P-glycoprotein antibodies. NK cells demonstrated strong reactivity with C494 and JSB-1, but did not react appreciably with C219. In addition, NK cells were unable to bind to the antibody MM4·17 unless they had been fixed and permeabilized, yet this antibody normally recognizes an extracellular epitope of P-glycoprotein. These differences culminated in the demonstration using Western analysis that NK cells did not express detectable levels of 170 kDa P-glycoprotein. Instead, NK cells expressed small-molecular-weight ‘mini P-glycoprotein’ products, of approximately 70 and 80 kDa. Collectively, these data indicate that the predominant P-glycoprotein species of NK cells are novel mini P-glycoproteins and not the classic P-glycoprotein of MDR models.

Multidrug resistant (MDR) cancers are refractory to a wide range of anti-neoplastic drugs and the difficulty in treating such tumours has inspired extensive efforts to clarify the underlying causes. The best studied mediator of this phenomenon is the product of the MDR1 gene, P-glycoprotein, which is a 170 kDa transporter situated in the plasma membrane of MDR cells. It uses energy from ATP hydrolysis to export a diverse range of substrates (Gottesman & Pastan, 1993) including anti-cancer drugs such as vinca alkaloids, anthracyclines and taxoids (Gottesman & Pastan, 1993) and fluorescent dyes such as rhodamine 123, calcein acetoxymethylester (AM) and fluo-3 AM (Neyfakh, 1988; Homolya et al, 1993).

The roles of P-glycoprotein have also received attention outside cancer research, as P-glycoprotein is expressed in organs of detoxification such as the intestine and blood–brain barrier where it mediates the extrusion of natural toxins and protects the organism from toxicity (Schinkel et al, 1994; Sparreboom et al, 1997; Schinkel, 1998). However, other cell types that express P-glycoprotein, e.g. the haematopoietic lineage, do not fulfil detoxification functions and the role of P-glycoprotein in these cells is not yet clear (Chaudhary & Roninson, 1991; Drach et al, 1992).

The natural killer (NK) cell features the strongest P-glycoprotein expression within the haematopoietic compartment. NK-cell P-glycoprotein has been detected at the protein and mRNA levels (Chaudhary et al, 1992; Drach et al, 1992; Klimecki et al, 1994, 1995), and its transport competence is suggested by the ability of NK cells to export rhodamine 123 (Chong et al, 1993; Kobayashi et al, 1994). Several putative functions have been attributed to P-glycoprotein in NK cells, with roles being hypothesized in cell-mediated cytotoxicity and cytokine secretion (Chong et al, 1993; Markham et al, 1993; Kobayashi et al, 1994; Klimecki et al, 1995). Little is known of the specific properties of P-glycoprotein and its regulation in NK cells, rather NK-cell P-glycoprotein is assumed to be identical to its better known counterpart expressed in MDR cell lines (Robey et al, 1999). We describe deviations in the behaviour of P-glycoprotein in NK cells compared with the classic P-glycoprotein of MDR models.

Materials and methods

Cell culture CEMVLB (Dr Ross Davey, Royal North Shore Hospital, Australia), MM170 (Dr Rob Whitehead Ludwig, Institute for Cancer Research, Australia) and OKT3 (Dr Hilary Warren, The Canberra Hospital, Australia) cell lines were used. These cell lines and K562 (ATCC) were grown in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% fetal calf serum (FCS), 5 × 10−5 mol/l 2-mercaptoethanol (2-ME), 30 U/ml gentamicin and 2 mmol/l l-glutamine. Medium for CEMVLB was also supplemented with 400 ng/ml vinblastine (David Bull Laboratories). Cells were maintained at 37°C in a fully humidified atmosphere of 5% CO2 in air.

Purification and culture of human NK cells NK cells were cultured using minor modifications of a protocol described previously (Warren & Skipsey, 1991). Peripheral blood was obtained by venepuncture of healthy, consenting adults (University of Tasmania Ethics Committee, Human experimentation permit number H4417) and collected into EDTA vacutainers (Greiner). Mononuclear cells were separated using Histopaque 1077 (Sigma), monocytes were depleted using a 1 h, 37°C adherence step in tissue culture flasks, and T cells were depleted by two rounds of complement lysis, using OKT3 supernatant and rabbit complement (Cedar Lane). After thorough washing, the remaining cells were co-cultured in 24-well plates together with 40 Gy X-irradiated MM-170 melanoma cells in Dulbecco's modified Eagles medium (DMEM) supplemented with 15% FCS, 4 ng/ml recombinant human interleukin 2 (IL-2; Pharmingen), 5 × 10−5 mol/l 2-ME, 30 U/ml gentamicin and 2 mmol/l l-glutamine. The starting density of NK cells and MM170 was 105 and 1·25 × 105 cells/well, respectively, with a total of 2 ml/well. Cultures were incubated at 37°C in a fully humidified atmosphere of 10% CO2 in air.

The initial date of plating the NK cells with MM170 was regarded as d 0. During d 4 to d 6 of the culture period, cells were fed by aspirating half the medium and replacing with fresh medium. The cells were split each day from d 7 (when confluent) until approximately d 18. When proliferation had ceased, the cells were fed every 2–3 d by aspirating half the medium and replacing with fresh medium. For the purposes of the present study, the cells were used at d 11–14 of culture.

Flow cytometric analysis of NK-cell antigens and P-glycoprotein The purity and phenotype of NK cells was evaluated by staining 5 × 105 NK cells with 10 μl of Simultest CD3/16 + 56 antibody, or the IgG2a/IgG1 isotype control (both from Becton Dickinson), for 20 min at room temperature, washed three times with phosphate-buffered saline (PBS) containing 2% bovine serum albumin (BSA) and 0·1% sodium azide (PBS/BSA/Azide), and then fluorescence monitored using an EPICS Elite ESP flow cytometer (Coulter).

For flow cytometric detection of P-glycoprotein expression, different fixation procedures were employed for different antibodies. For JSB-1 (Signet), C494 and C219 (Dako), each used at 1 μg per 5 × 105 cells, the cells were fixed with 70% methanol for 1 min at room temperature, then washed twice with PBS/BSA/Azide prior to staining. For MM4·17 (20 μl of culture supernatant) cells were either unfixed or fixed with 4% paraformaldehyde for 10 min at room temperature and permeabilized by saponin (0·1%), which was present in all subsequent incubations and washes. Primary antibodies were incubated for 30 min at room temperature, then detected using fluorescein isothiocyanate (FITC)-conjugated sheep anti-mouse Ig (Silenus). Isotype controls (IgG1 and IgG2a; Sigma) were used at identical concentrations to the P-glycoprotein antibodies.

Cytotoxicity Cell-mediated cytotoxicity was assessed using the standard chromium release assay, conducted in complete DMEM without IL-2. Briefly, NK cells were mixed with 51Cr-labelled K562 in V-bottomed 96-well plates. Spontaneous (spon) and maximum (max) release values were assessed by incubating K562 in medium alone or in 1% Triton X-100 respectively. The plates were centrifuged at 400 g for 4 min then incubated for 4 h at 37°C, in a fully humidified atmosphere of 5% CO2 in air, then 100 μl of supernatant was harvested from each well and gamma radiation evaluated using a Wallac KB Gammamaster.

The percentage specific cytotoxicity = (experimental cpm − spon cpm)/(max cpm − spont cpm) × 100.

P-glycoprotein functional assays Flow cytometric dye efflux assays were employed to assess P-glycoprotein activity. Rhodamine 123 (150 ng/ml), BODIPY FL verapamil (1 μmol/l), ST BODIPY dihydropyridine (0·1 μmol/l), fluo-3 AM (1 μmol/l) and calcein AM (0·5 μmol/l) were purchased from Molecular Probes, and daunorubicin (1 μmol/l) was from David Bull Laboratories. Cells at 106/ml were incubated at 37°C with the indicated concentration of dye in the presence or absence of 20 μg/ml verapamil (Knoll) or 20 μg/ml vinblastine (David Bull Laboratories). For all dyes except rhodamine 123, the incubation period was 30 min. The cells were then washed twice with ice-cold PBS/BSA/Azide and maintained on ice prior to flow cytometric assessment of fluorescence. For rhodamine efflux, cells were incubated for 15 min at 37°C with rhodamine 123 ± verapamil or vinblastine, washed twice with ice-cold complete medium, then resuspended in dye-free medium ± verapamil or vinblastine and incubated for a further 30 min at 37°C. The cells were washed and analysed as described above.

Confocal microscopy Cells at 106/ml in PBS were adhered onto poly lysine-coated coverslips, then fixed with 4% paraformaldehyde for 10 min. The cells were washed twice with saponin buffer consisting of PBS containing 0·1% saponin, 0·1% azide and 1 mmol/l EGTA. All subsequent washes and incubations were conducted in saponin buffer. The cells were incubated with C494 (1 μg/ml) for 45 min at room temperature, washed three times and then incubated with Alexa 488-conjugated goat anti-mouse IgG (Molecular Probes) for 30 min at room temperature in the dark. After three washes, the cells were air-dried briefly, then mounted onto glass slides using Permafluor (Coulter) and examined using an OptiScan scanning confocal microscope equipped with a krypton/argon laser and attached to an Olympus microscope.

Western blot analysis Cells were washed twice in PBS then lysed at 4°C for 30 min in ice-cold lysis buffer (pH 7·5) comprising 50 mmol/l Tris-HCl, 150 mmol/l NaCl, 1% Nonidet P-40, 20 μg/ml aprotinin, 2 mmol/l phenylmethylsulphonyl fluoride and 10 μg/ml leupeptin. Cellular debris was removed by centrifugation at 12 000 g for 20 min at 4°C and the supernatants were frozen until required. Proteins were separated on sodium dodecyl sulphate (SDS)−7·5% polyacrylamide gels and electroblotted onto polyvinylidene difluoride (PVDF) membranes. Antibodies C494, C219 and JSB-1 were used at 1 μg/ml and horseradish peroxidase-conjugated anti-mouse Ig (Santa Cruz) was used at 0·2 μg/ml. Membranes were visualized by enhanced chemiluminescence (NEN Life Science Products).


Characterization of cultured NK cells

NK-cell cultures routinely comprised at least 98% CD3-negative, CD16- and/or CD56-positive cells, which mediated potent cytotoxicity against K562, as shown in Fig 1.

Figure 1.

 Phenotype and activity of cultured NK cells. (A) Dot plot shows two-colour staining of NK cells with CD3 (fluorescein isothiocyanate; FITC) and CD16/CD56 (phycoerythrin; PE). The NK cells derived from the culture procedure were routinely 98% CD3 negative and CD16/CD56 positive and (B) mediated potent cytotoxicity against the classic NK target K562.

Restricted P-glycoprotein activity in NK cells

The ability of NK cells to extrude rhodamine 123 has been interpreted as evidence of classic P-glycoprotein activity in these cells. While NK cells were competent in the transport of rhodamine 123, fluo-3 AM, ST BODIPY dihydripyridine and BODIPY FL verapamil, no calcein AM or daunorubicin transport was observed (Fig 2). This contrasts with the behaviour of CEMVLB, which was capable of transporting all dyes examined and suggests a restricted P-glycoprotein substrate profile in NK cells.

Figure 2.

 Restricted P-glycoprotein activity in NK cells. The accumulation of various dyes was monitored alone (shaded) and in the presence of verapamil (solid line, unfilled) or vinblastine (dotted line, unfilled). (A) CEMVLB demonstrated P-glycoprotein-mediated export of all dyes examined, characterized by an increase in dye fluorescence in the presence of either verapamil or vinblastine. (B) In contrast, NK cells demonstrated a restricted P-glycoprotein substrate profile. For the dyes ST BODIPY dihydropyridine, BODIPY FL verapamil, fluo-3 AM and rhodamine 123, verapamil and vinblastine increased dye accumulation several fold, indicating these dyes are substrates for NK-cell P-glycoprotein. However, no changes in accumulation of calcein AM or daunorubicin were detected in the presence of these inhibitors, indicating no P-glycoprotein-mediated export of these molecules by NK cells.

Differential reactivity of NK cells with P-glycoprotein antibodies

Flow cytometric evaluation of P-glycoprotein expression in NK cells and CEMVLB revealed a marked deviation in antibody reactivity between the two cell types. This was observed using antibodies that bind to intracellular P-glycoprotein epitopes, which were used to stain methanol-fixed cells. CEMVLB demonstrated strong reactivity with the antibodies C494, JSB-1 and C219, consistent with the high P-glycoprotein expression of this cell line. However, while NK cells demonstrated strong C494 and JSB-1 reactivity, they did not show significant reactivity with C219 (Fig 3).

Figure 3.

 Reactivity of CEMVLB cells and NK cells to P-glycoprotein antibodies. Shaded histograms are the isotype controls and unfilled, solid lines are the appropriate antibody histograms. Both NK cells and CEMVLB showed strong immunoreactivity with C494 (A) and JSB-1 (B). However, while CEMVLB reacted with C219, very little staining of NK cells was observed with this antibody (C).

NK cells also differed from CEMVLB in terms of binding to the antibody MM4·17, which recognizes a cell surface epitope of P-glycoprotein (Cianfriglia et al, 1994). Because of the external nature of its epitope, MM4·17 is normally used to stain unfixed cells, and unfixed CEMVLB demonstrated strong reactivity with this antibody. In contrast, we observed negligible staining of unfixed NK cells with MM4·17 (Fig 4). Further analysis revealed that NK cells do possess the MM4·17 epitope, however, reactivity was only observed after fixation and permeabilization of the cells.

Figure 4.

 Inaccessibility of cell-surface epitope of P-glycoprotein in NK cells. Unfixed CEMVLB demonstrated strong reactivity with the MM4·17 antibody (A), while unfixed NK cells did not stain with this antibody (B). Fixation and permeabilization of the NK cells, however, potentiated strong binding with MM4·17 (C).

Distribution of P-glycoprotein in NK cells

The inaccessibility of the MM4·17 epitope in unfixed NK cells suggested either an intracellular location of P-glycoprotein or a novel conformation within the plasma membrane, and confocal microscopy was employed to explore these possibilities. Figure 5 shows a representative confocal image taken through the approximate centre of an NK cell at d 14 of culture. There was evidence for P-glycoprotein expression at the cell surface and also throughout the cytoplasm.

Figure 5.

 Distribution of P-glycoprotein in NK cells. NK cells at d 14 of culture were adhered onto poly lysine-coated coverslips and fixed with 4% paraformaldehyde. Staining was carried out in the presence of saponin using C494 at 10 μg/ml and Alexa 488-conjugated goat anti-mouse IgG at 1 μg/ml. C494 staining was apparent at the plasma membrane of NK cells and also throughout the cytoplasm.

Absence of 170 kDa P-glycoprotein and expression of small-molecular-weight P-glycoproteins in NK cells

The unusual properties of NK-cell P-glycoprotein prompted further analysis via Western blot analysis. In CEMVLB lysates, classic P-glycoprotein was detected as a broad band migrating at approximately 170 kDa and above, and reactive with C219, JSB-1 and C494 (Fig 6). This high-molecular-weight band was not detected in NK-cell lysates, which instead featured smaller P-glycoprotein products of approximately 70 and 80 kDa. Both products showed reactivity with C494, while only the 70 kDa product reacted with JSB-1. Neither product reacted with C219, confirming the absence of C219-reactive proteins in NK cells. The 70 and 80 kDa P-glycoprotein molecules were also detected in CEMVLB lysates, however, in this MDR cell line the most abundant product was classic, high-molecular-weight P-glycoprotein. Although C494 and JSB-1 cross-react with the mitochondrial enzyme pyruvate carboxylase (Rao et al, 1994, 1995), it should be emphasized that C494 and JSB-1 did not detect any proteins of a corresponding molecular weight (120–130 kDa) in NK-cell lysates, establishing the fact that pyruvate carboxylase cross-reactivity is negligible in this study.

Figure 6.

 Absence of 170 kDa P-glycoprotein and presence of small-molecular-weight P-glycoproteins in NK-cell lysates. Proteins were separated using sodium dodecyl sulphate polyacrylamide gel electrophoresis and then blotted with C494, JSB-1 and C219 antibodies. The left lane shows NK lysates and the right lane shows lysates from CEMVLB. Classic P-glycoprotein in CEMVLB was observed as a broad band migrating at approximately 170 kDa. NK lysates did not display an equivalent high-molecular-weight P-glycoprotein band, but demonstrated 70 and 80 kDa bands (arrow heads) reactive with C494 and JSB-1, but not C219. Similar 70 and 80 kDa products were observed in CEMVLB lysates.


Complementary evidence from reverse transcription polymerase chain reaction (RT-PCR), antibody reactivity and rhodamine efflux studies has led to the acceptance that NK cells possess the classic MDR1 gene product (Chaudhary et al, 1992; Drach et al, 1992; Chong et al, 1993; Klimecki et al, 1994, <1995; Kobayashi et al, 1994). We now provide evidence that the P-glycoprotein expressed in human NK cells deviates markedly from classic P-glycoprotein expressed in MDR models.

Functional assays provided the first indication that NK-cell P-glycoprotein was distinct from classic P-glycoprotein. Despite efficiently transporting rhodamine 123, fluo-3 AM, BODIPY FL verapamil and ST BODIPY dihydropyridine, NK cells did not export daunorubicin and calcein AM, each of which was transported by CEMVLB. Furthermore, differential antibody reactivity was established using three antibodies that bind to internal regions of P-glycoprotein. CEMVLB reacted with each of these antibodies, however, NK cells demonstrated reactivity with C494 and JSB-1 but did not react appreciably with C219. The use of MM4·17, an antibody recognizing an external epitope, also highlighted differences between the two cell types. Unfixed CEMVLB demonstrated strong reactivity with MM4·17, however, unfixed NK cells did not react with MM4·17 and reactivity was only evident after fixation and permeabilization of the NK cells. Confocal microscopy demonstrated substantial P-glycoprotein expression within the plasma membrane of NK cells, indicating that this fixation requirement could not be entirely explained by a cytoplasmic location of P-glycoprotein in NK cells. Collectively, these data suggest that the P-glycoprotein product of NK cells is distinct from classic P-glycoprotein.

Western blot analysis provided the greatest insight into the unique nature of NK-cell P-glycoprotein. In CEMVLB lysates, classic P-glycoprotein migrated as a broad band at 170 kDa and above, reactive with the antibodies C494, JSB-1 and C219. Notably, this high-molecular-weight band was not detected in NK-cell lysates, suggesting negligible expression of classic, 170 kDa P-glycoprotein. Instead, NK cells featured expression of 70 and 80 kDa proteins, herein referred to as ‘mini P-glycoproteins’. In addition to differences in molecular weight, these products were distinguished by their differential antibody reactivity. The 70 kDa form was reactive with both C494 and JSB-1, whereas the 80 kDa form reacted only with C494. Neither product reacted with C219. Corresponding mini P-glycoproteins were also detected in CEMVLB, however, at significantly lower levels than classic P-glycoprotein. It is highly improbable that these small-molecular-weight proteins are breakdown products resulting from the proteolytic cleavage of full-length P-glycoprotein, as our evidence for the absence of 170 kDa P-glycoprotein in NK cells is not confined to Western analysis, but is supported by our flow cytometric data. Thus, the negligible C219 reactivity in NK cells, the requirement for fixation in order to liberate MM4·17 reactivity despite the apparent plasma membrane expression of the protein and the atypical dye efflux profile all argue against the expression of classic P-glycoprotein in NK cells.

The antibodies reactive with NK-cell P-glycoprotein all bind to the C-terminal half of P-glycoprotein. The C494 epitope has been mapped to aa 1027–1032 (Georges et al, 1990), the JSB-1 epitope is believed to overlap this site (Rao et al, 1995) and the predominant MM4·17 epitope occurs at aa 740–747 (Cianfriglia et al, 1994). In contrast, C219 binds to two sites within the molecule, at aa 567–572 in the N-terminal half and aa 1213–1218 in the C-terminal half (Georges et al, 1990). Therefore, an attractive explanation accounting for NK mini P-glycoproteins is that they are truncated MDR1 products comprising the C-terminal region between the C219 epitopes. Such a truncation would yield an approximately 640 aa protein corresponding to an area between aa 573 and 1213 of the MDR1 sequence, consistent with the size of the proteins reported herein and the pattern of antibody reactivity observed.

Alternative splicing of the MDR1 gene could account for smaller-molecular-weight P-glycoproteins and examples of alternative splicing of mdr genes have been reported previously (Barg et al, 1999). Alternatively, NK mini P-glycoproteins may be encoded by distinct genes with high structural homology to the C-terminal half of MDR1. Regardless of their origins, the fact that mini P-glycoproteins are constitutively expressed in primary human NK cells warrants an exploration of their physiological function in this cell type. The discovery of small-molecular-weight P-glycoproteins in NK cells adds to a growing list of mini P-glycoproteins reported in different species (Kawai et al, 1994; Thévenod et al, 1996; Barg et al, 1999; Ma et al, 1999).

The existence of smaller-molecular-weight P-glycoproteins in NK cells warrants reappraisal of the previous studies that guided the general acceptance of classic P-glycoprotein expression in this cell type. With regard to the numerous reports of PCR-based MDR1 detection, it is prudent to note that the primers chosen for NK-cell studies amplify regions located exclusively in the C-terminal half of the molecule (Drach et al, 1992; Gupta et al, 1992; Markham et al, 1993; Wilisch et al, 1993; Kobayashi et al, 1994; Yamashiro et al, 1998; Egashira et al, 1999). We suggest that these studies may have detected mini P-glycoproteins rather than full-length MDR1. With regard to immunological detection of P-glycoprotein via flow cytometry, the majority of studies in NK cells have been performed using the monoclonal antibody MRK16. The epitope recognized by MRK16 is a conformational one, encompassing regions of both the N- and C-termini, within the first and fourth external loops respectively (Georges et al, 1993). It is interesting to note, however, that a synthetic fourth loop epitope demonstrates strong and independent reactivity with MRK16 (Georges et al, 1993). On the basis of this precedent, the possibility that MRK16 reacts with hypothetical C-terminal mini P-glycoproteins is therefore feasible.

The dye efflux observed in NK cells is not compatible with the transport properties of multidrug resistance-associated protein (Davey et al, 1996; Hollo et al, 1996; Germann et al, 1997), sister of P-glycoprotein (Lecureur et al, 2000) or the mitoxantrone resistance protein (Litman et al, 2000). In contrast, C-terminal mini P-glycoproteins are predicted to contain drug-binding sites in transmembrane domain 12 (Ambudkar et al, 1999) and, therefore, deserve further scrutiny as possible mediators of the P-glycoprotein-like dye transport executed by NK cells.

The present study raises important issues concerning the diagnostic detection of P-glycoprotein and the chemotherapeutic strategies employed to treat P-glycoprotein-positive malignancy. Evidence of P-glycoprotein expression and activity cannot be automatically extrapolated to assume expression of the fully functional P-glycoprotein observed in MDR cell lines. The differential handling of P-glycoprotein substrates by NK cells suggests the potential for a restricted MDR phenotype, which has ramifications for the spectrum of drugs and inhibitors employed in its treatment. Consequently, maximizing therapeutic utility will require discrimination between these different phenotypes.


The authors wish to thank Dr Hilary Warren for technical expertise and reagents for establishing the NK cultures, Mark Cozens for flow cytometry, Drs Bob Chappell and Mike Groth and the WP Holman Clinic for irradiation of cell lines. This work was supported by grants from the David Collins Leukaemia Foundation and the Australian Research Council. Z. Wang is supported by the Dick Buttfield Memorial Fellowship. M. Cianfriglia is supported in part by grants from Ministero della Sanità Italiana, Progetto Ricerche AIDS.