Defective Acidification in Human Breast Tumor Cells and Implications for Chemotherapy

Bodipy-transferrin, LysoSensor Blue DND-167, FITC-transferrin, seminaphthorhodafluor (SNARF)-dextran, NBD-ceramide, and FITC-dextran were from Molecular Probes, Inc. (Eugene, OR). Adriamycin was from Calbiochem Corp. (San Diego, CA). Concanamycin A was from Fluka Chemical Corp. (Buchs, Switzerland). Bovine insulin and l-glutamine were from GIBCO BRL (Gaithersburg, MD), and FBS was from Gemini Bio-Products, Inc. (Calabasas, CA). The anti–lysosome-associated membrane protein (LAMP) 1 serum was from the Developmental Hybridoma Bank at Johns Hopkins University (Baltimore, MD), and goat anti–mouse secondary antibody Fab fragments conjugated to phycoerythrin were from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). All other reagents were from Sigma Chemical Co. (St. Louis, MO).

MCF-7 and MCF-7/ADR cells were obtained from Dr. William Wells of the Department of Biochemistry, Michigan State University. They were maintained in Modified Eagle's medium with phenol red, bovine insulin 10 μg/ml, l-glutamine, and 10% FBS in a humidified incubator at 37°C and 5% CO₂ (Forma Scientific, Inc., Marietta, OH). In addition, the MCF-7/ADR cells were maintained continuously in 0.8 μM Adriamycin. The MCF-10F cells were obtained from American Type Culture Collection (Rockville, MD).

All incubations were at 37°C with 5% CO₂. Unless otherwise stated, the cells were incubated in DME without phenol red or serum and with 20 mM Hepes, pH 7.3. All incubation and imaging was at 37°C in coverglass chambers (Lab-tek, Naperville, IL) with media preequilibrated with 5% CO₂, and the chamber was superfused with humified air containing 5% CO₂.

Unless specifically stated, all imaging measurements were performed on an Ultima confocal microscope (Meridian Instruments, Inc., Okemos, MI) equipped with an argon laser. Cells were visualized with a 60×/1.4NA oil objective (Olympus Optical Co., Ltd., Tokyo, Japan), and the data were collected with two photomultiplier tubes (R3896; Hamamatsu Photonics, Hamamatsu City, Japan).

A fluorescence microscope (Diaphot; Nikon, Inc., Melville, NY) was used for pH measurements within the lumens of recycling endosomes. The microscope was equipped with a 100 W Hg lamp and Uniblitz shutter (Vincent and Associates, Rochester, NY). The shuttering of the light source was controlled with a computer. The data were collected on an intensified charged coupled device (4910; Hamamatsu Photonics).

Adriamycin is a small heterocyclic amine (molecular mass 580 dalton) with a pK_a of 8.3 that can diffuse across membranes in the uncharged form. Adriamycin fluorescence is excited between 350 and 550 nm and emits between 400 and 700 nm. Cells were incubated with Adriamycin (10 μM) for 30 min at 37°C and then visualized using the confocal microscope with (λ_ex = 488 nm).

Cells were incubated with acridine orange (6 μM) for 15 min and then visualized on the confocal microscope (λ_ex = 488 nm). Emission light passed through a 560-nM short-pass dichroic mirror. The green emission passed through a 530/30-nm bandpass filter, and the red emission passed through a 610-nm longpass filter. The green and red emissions were collected using two photomultiplier tubes.

Cells were incubated with LysoSensor Blue DND-167 (2 μM, 1 mM stock in water) for 60 min, and then visualized on the confocal microscope (λ_ex = 353 nm). In some experiments, the cells were subsequently washed and incubated with Adriamycin (10 μM) for 30 min.

Cells were incubated in DME/20 mM Hepes, pH 7.3, with NBD-ceramide (5 μM) at 4°C for 10 min (30). They were then washed twice with DME/ 20 mM Hepes (pH 7.3)/10% FBS, incubated at 37°C for 30 min, and observed with the confocal microscope (λ_ex = 488 nm).

BODIPY-transferrin was used to label the recycling endosome compartment. Transferrin is endocytosed by specific receptors on the surface of the cell, then transported through the endosomes and recycled back to the surface. The transferrin receptor is not transported to the lysosomes, so probes that are conjugated to transferrin can be used to selectively monitor the recycling endocytic compartments (31). The endocytic pathway is known to undergo acidification (32). Thus, the fluorophore BODIPY was used, since its fluorescence is not very sensitive to pH. The cells were loaded with 50 μg/ml of BODIPY-transferrin in DME/20 mM Hepes, pH 7.3, for 25 min in a humidified incubator at 37°C and 5% CO₂ (31). The cells were washed rapidly three times with DME/Hepes, incubated for 2 min in a citric acid buffer, pH 4, at 37°C, and rinsed three times with HBSS containing 20 mM Hepes, pH 7.3. In some experiments, the cells were subsequently incubated at 37°C for 30 min to allow the transferrin to recycle to the surface, and then incubated with Adriamycin (10 μM) at 37°C for 30 min.

The pH-sensitive fluorophores, FITC and SNARF, were used to measure the pH within endosomes and the cytosol, respectively. LysoSensor Blue DND-167 was also used as an independent probe for calibration of the pH within the lumenal compartment of lysosomes. Both FITC and SNARF are ratiometric dyes. The emission intensity of FITC at 530 nm increases with increasing pH at λ_ex = 490 nm. However, it is unaffected by pH at λ_ex = 450 nm. Therefore, by taking the ratio of the emission intensities at the two excitation wavelengths, one can obtain a pH value independent of FITC concentration in a particular compartment. To convert the ratios to pH values, the cells were calibrated using monensin and nigericin with buffers of known pH (see below). FITC is most useful for measurement of pH values of 5.0–7.0.

SNARF at λ_ex = 514 nm emits at two wavelengths, 570 and 630 nm. The protonated fluorophore emits at 570 nm, and the neutral fluorophore emits at 630 nm. Again, the ratio of the two emissions corresponds to a pH value that is theoretically independent of the concentration of the dye in that compartment. SNARF can be calibrated reliably over the pH range 6.2–9.0. LysoSensor blue DND-167 was used to semiquantitatively investigate the pH in lysosomes. The fluorescence of LysoSensor Blue is dependent on pH. LysoSensor Blue has a functional group that, when deprotonated, leads to a loss of fluorescence of the molecule. The pK of this group is 5.1. Therefore, at pH <5.1, a greater percentage of the dye will be protonated and fluorescent. There is little fluorescence above pH 5.8.

At the end of each experiment, the fluorescence emission of each dye was calibrated with solutions of known pH. For pH calibration of endosomes, the cells were incubated in solutions of 150 mM NaCl, 20 mM Hepes or Mes (depending on the pH of the calibration solution), 5 mM KCl, 1 mM MgSO₄ buffered at pH 5, 6, 6.5, and 7, containing monensin (20 μM) and nigericin (10 μM) for 5 min before recording the fluorescence (33). For the pH calibrations of cytosol and nucleoplasm, the cells were incubated in solutions of 140 mM KCl, 10 mM Mops, 5 mM MgSO₄, 1 mM CaCl₂ buffered at pH 6, 7, and 7.5 containing nigericin (20 μM).

FITC bound to transferrin was used to selectively probe the pH of the endocytic compartment (31, 32). The cells were loaded with 150 μg/ml FITC-transferrin for 30 min. Based on competition binding studies with unlabeled transferrin, this amount of FITC-transferrin specifically labels the recycling endocytic compartment in MCF-7 and MCF-7/ADR cells. The cells were then subsequently washed six times with HBSS with 20 mM Hepes, pH 7.3, and visualized with epifluorescence microscopy (λ_ex = 490 and 450 nm, and λ_em = 515 nm). The pH was calibrated from the FITC fluorescence as described above.

To measure the pH within the lysosomes, the cells were incubated with 5 mg/ml of FITC-dextran 10 kD for 30 min (28), then washed four times in DME with 20 mM Hepes, pH 7.3, and incubated in this medium for 90 min. They were then visualized with epifluorescence with FITC excitation filters (see above). The pH was calibrated from the FITC fluorescence as described above. Alternatively, the cells were incubated with LysoSensor Blue as described above.

The pH within the cytoplasm and nucleoplasm was selectively probed by loading these compartments with SNARF conjugated to dextrans using a procedure referred to as “scrape loading” (34). Briefly, the cells were plated on polystyrene plates at 50% confluence 24–36 h before loading with dextrans. The medium was aspirated, and the cells were rapidly (<10 s) scraped off in the presence of 10 mg/ml of 70- or 10-kD SNARF-dextran and washed in ice-cold media to minimize endocytosis before plating on coverslips as described previously (34). Confocal fluorescence microscopy was used to prepare optical sections through the cell. The fluorescence intensity of the nucleoplasm and cytoplasm could then be quantified. The fluorescence from the SNARF-conjugated dextrans was recorded 24–36 h after scrape loading. The pH was calibrated from the fluorescence as described above.

For immunolocalization of lysosomes, anti–LAMP-1 serum was used as described previously (35). Cells were fixed with 2% paraformaldehyde in 50 mM phosphate buffer, pH 7.8, containing lysine (9 mg/ml) for 2 h. They were then permeabilized with 0.01% saponin for 5 min. Anti–LAMP-1 sera was used undiluted for 30 min at room temperature. Cells were washed extensively with PBS and then incubated for 15 min with goat anti–mouse secondary antibody Fab fragments conjugated to phycoerythrin at 1:150 dilution at room temperature. Cells were washed in PBS and visualized with the confocal microscope using λ_ex = 488 nm.

The PSS hypothesis predicts that weak base chemotherapeutic drugs should accumulate in the acidic secretory organelles of drug-resistant cells. Adriamycin was chosen as the model chemotherapeutic drug to characterize the subcellular distribution of these agents in drug-sensitive and drug-resistant tumor cells because its natural fluorescence allows it to be tracked visually, and because it is widely administered in the treatment of many different types of cancers. MCF-7 and MCF-7/ADR cells were used as a pair of drug-sensitive and drug-resistant cell lines, respectively. They are human breast carcinoma cells that are used as an in vitro model system for breast cancer. The drug-resistant MCF-7/ADR cell line is derived from the MCF-7 cell line by selection in the chemotherapeutic Adriamycin (36). MCF-7/ADR cells are also cross-resistant to a number of other chemotherapeutic drugs, including vincristine, vinblastine, and colchicine (36).

In the drug-sensitive MCF-7 cells, Adriamycin fluorescence was seen throughout the cytoplasm and nucleoplasm (Fig. 1,b), with some localized increased fluorescence in both the compartments. The slightly reticular distribution of Adriamycin is likely the consequence of its binding to cytoskeleton (37). One of the primary targets for Adriamycin is in the nucleus, where it binds to DNA and inhibits the DNA metabolic enzyme topoisomerase II, thereby blocking DNA replication and transcription (38, 39). In contrast, in the drug-resistant MCF-7/ADR cells, there was little Adriamycin fluorescence in the nucleoplasm and an apparent reduction of fluorescence in the cytoplasm (Fig. 1 a). Instead, the drug was found to be accumulated in a perinuclear region and within punctate compartments throughout the cytoplasm.

Since Adriamycin is a weak base, it is predicted to accumulate inside acidic compartments. To determine if it accumulated in the lysosomes, the most acidified cellular compartment, cells were sequentially labeled with LysoSensor Blue DND-167 and Adriamycin. LysoSensor Blue is a membrane-permeable pH probe whose fluorescence emission is reduced significantly at pH >5.8 (see Materials and Methods). Thus, it selectively fluoresces only in the most highly acidic compartments of living cells such as lysosomes. Discrete punctate fluorescence in the cytoplasm was observed in cells incubated with LysoSensor Blue (Fig. 2,b). These lysosomes appeared to be distributed somewhat evenly through the cytoplasm. The same cells were then incubated with Adriamycin and excited at 488 nm, and the emission was collected >550 nm (Fig. 2,c). LysoSensor Blue is not excited at 488 nm; therefore, it did not interfere with the Adriamycin signal (data not shown). Most lysosomes (Fig. 2,b, arrows) were labeled with Adriamycin (Fig. 2,c, arrows). In addition to the lysosomal compartment, there was an additional cytoplasmic region adjacent to the nucleus that was heavily labeled with Adriamycin (Fig. 2 c).

In many cell types, the TGN and the recycling endosome compartment are found adjacent to the nucleus. These two compartments are also known to be acidic (40– 42). Therefore, specific fluorescent probes were used to determine if the perinuclear Adriamycin accumulation in the MCF-7/ADR cells colocalized with these compartments. Due to the broad excitation and emission spectrum of Adriamycin, it was not possible to simultaneously label a cell with Adriamycin and available fluorescent markers for endosomes and the TGN. Therefore, colocalization was done by sequential labeling with Adriamycin and the organelle markers.

To examine whether Adriamycin colocalized with the recycling endosome compartment, cells were first labeled with BODIPY-transferrin by receptor-mediated endocytosis. In MCF-7/ADR cells, transferrin labeling revealed that the recycling endosomal compartment has a perinuclear localization (Fig. 2,e). The transferrin label was subsequently chased out (half time = 5 min) with unlabeled transferrin. Next, the cells were incubated with Adriamycin for 20 min. In each cell, the region that labeled with BODIPY-transferrin was also labeled with Adriamycin (Fig. 2 f). Thus, the Adriamycin accumulation colocalized with the recycling endosomes.

In some cell types, the TGN is in close proximity with the recycling endosome compartment (43, 44). To test if Adriamycin colocalized with the TGN in the MCF-7/ ADR cells, the cells were sequentially labeled with Adriamycin and NBD-ceramide, a vital stain for the TGN (45). A field of cells was labeled with Adriamycin (10 μM) (Fig. 2,i). The cells were then washed with Adriamycin-free media until they no longer showed any Adriamycin labeling. Next, the cells were labeled with NBD-ceramide (Fig. 2 h). Again, the region labeled with NBD-ceramide was also labeled with Adriamycin. Thus, Adriamycin accumulation is colocalized with the acidic organelles of the drug-resistant MCF-7/ADR cells—the recycling endosome compartment, the TGN, and the lysosomes.

There was little organellar labeling of Adriamycin in drug-sensitive MCF-7 cells (Fig. 1,b). Therefore, the presence and distribution of each of these organelles in these cells were tested to determine if the organelles were either missing or present but failing to accumulate chemotherapeutic drugs. The distribution of the recycling endosome compartment was probed with BODIPY-transferrin, and the TGN was probed with NBD-ceramide (45). In the drug-sensitive MCF-7 cells, both the recycling endosome compartment (Fig. 2,d) and the TGN (Fig. 2,g) were distributed throughout the cytoplasm. Their distribution was very different from the MCF-7/ADR cells, where both compartments were distinctly perinuclear in location and polarized to one side of the nucleus. Furthermore, unlike the MCF-7/ADR cells, in the MCF-7 cells, Adriamycin was not concentrated in any of these compartments (Fig. 1 b).

The distribution of lysosomes in MCF-7 and MCF-7/ ADR cells was compared using immunofluorescence for the lysosomal integral membrane protein LAMP-1 (46). LAMP-1 was chosen as an indicator rather than LysoSensor Blue because LysoSensor Blue requires acidified lysosomes for labeling. Both MCF-7/ADR cells (Fig. 3,a) and MCF-7 cells (Fig. 3 b) had punctate LAMP-1–labeled compartments in the cytoplasm. There was little qualitative or quantitative difference in the distribution of LAMP-1 labeling between the two cell types. In some MCF-7 and MCF-7/ADR cells, there was also a perinuclear staining for LAMP-1 which could be indicative of lysosomes or TGN since this membrane protein is sorted to the lysosomes from the TGN. Together, the results demonstrate that the recycling endosomes, TGN, and lysosomes are present in the MCF-7 cells but do not accumulate Adriamycin.

Many of the chemotherapeutic drugs, such as Adriamycin, vincristine, vinblastine, daunomycin, and mitoxantrone, are heterocyclic amines (see Fig. 4) (15, 16, 47). With pK_a values at or just above physiological pH, they are weak bases and are membrane permeable only in the noncharged form. This has been tested empirically in single-membrane model systems of liposomes (48) and anucleate red blood cells (18, 19). In liposomes in which the lumenal compartments are more acidic relative to the external medium, there is a net lumenal accumulation of Adriamycin. The magnitude of this pH-dependent lumenal accumulation closely approximates theoretical calculations. For example, Adriamycin with pK_a of 8.3 accumulates ∼100-fold in a liposome with a lumenal pH of 6 and an external pH of 8 (48). As shown above (Fig. 2, b, c, e, f, h, and i), Adriamycin accumulation colocalized with each of the acidic organelles of the cell. However, Adriamycin did not accumulate within these same organelles in the drug-sensitive MCF-7 cells. Therefore, the pH was examined in the cytosol and organelles of MCF-7 and MCF-7/ADR cells to determine whether the failure to concentrate Adriamycin in the organelles of MCF-7 cells was the consequence of a failure to acidify.

Acridine orange is a fluorescent weak base that is used frequently as a probe for acidification of organelles (49, 50). When accumulated in high concentrations in acidic compartments, its emission shifts from green to red (50). MCF-7 and MCF-7/ADR cells were incubated with acridine orange. In drug-resistant MCF-7/ADR cells, there was a red-orange fluorescence in discrete cytoplasmic organelles (Fig. 5,a), indicating acridine orange accumulation and thus an acidified compartment. This fluorescent pattern is similar to that observed in various nontransformed cells, such as the MCF-10F cells (Fig. 5 c), parietal cells (51), paramecium (52), pituitary cells (53), and Xenopus oocytes (54). The MCF-10F cells originated from a female patient with normal nonmalignant breast tissue. These cells have a normal or near-normal karyotype (55). The cytosol and nucleoplasm of MCF-10F cells show a diffuse green fluorescence with discrete punctate red-orange organelles distributed throughout the cytoplasm.

In contrast, acridine orange in the MCF-7 cells had significantly fewer red-orange fluorescent compartments, indicating many fewer acidic vesicles (Fig. 5 b). The fluorescence of acridine orange does not give any information as to the identity of the acidic compartments or the absolute value of the pH within these compartments. Therefore, subsequent experiments used ratiometric pH probes targeted to specific organelles and whose pH could be calibrated in situ. A similar qualitative difference in the distribution of acridine orange has also been observed in K-562 drug-sensitive and drug-resistant human erythroblastoma cells (26).

To measure the pH within the recycling endosome compartment, the cells were loaded with FITC-transferrin. FITC is a fluorophore sensitive to pH within the range 5.0–7.0 (56). It was found that the drug-resistant MCF-7/ADR cells had an average recycling endosome compartment pH of 6.1 ± 0.1, whereas the drug-sensitive MCF-7 cells had an average recycling endosome compartment pH of 6.6 ± 0.1 (Table 1). A similar quantitative difference in endosome pH was seen in the drug-resistant and drug-sensitive MDA-231 human breast tumor cell lines (unpublished observations).

The pH of the lysosomes in MCF-7 and MCF-7/ADR cells was measured with LysoSensor Blue. The fluorescence of this probe is strongest at pH <5.1 and disappears at pH >5.8. In most MCF-7 cells, there were few acidic compartments (10.6 ± 5.5 per cell, n = 78), and in some cells, there was virtually no fluorescence above background (Fig. 2,a), suggesting there are few compartments with pH values <5.8. In contrast, there were significantly more punctate fluorescent organelles (43.0 ± 10, n = 15) when the MCF-7/ADR cells were labeled (Fig. 2,b). The pH of the lysosomes in MCF-7/ADR cells was measured independently with FITC conjugated to dextran, a fluid phase probe that accumulates in lysosomes (28). The average pH obtained for the lysosomes using this method was pH 5.1 ± 0.1 (Table 1). This method could not be used to measure the pH of lysosomes in MCF-7 cells because of slow delivery of the fluid phase marker to the lysosome (unpublished observations). However, the lack of acidic lysosomes as monitored by LysoSensor Blue in MCF-7 cells correlated with the lack of punctate Adriamycin fluorescence (Fig. 1,b) as well as with the lack of punctate red fluorescence from acridine orange (Fig. 5,b) within the cytoplasm of these cells. While LysoSensor Blue failed to label lysosomes of MCF-7 cells (Fig. 2,a), immunolocalization indicates that these cells have numbers of lysosomes comparable to MCF-7/ADR cells (Fig. 3). This indicates that the lysosomes are present in these drug-sensitive cells but fail to acidify.

Previous work showed that many drug-resistant cell lines have higher total cellular pH than their drug-sensitive counterparts (2, 22– 25, 57). In addition, changing the cellular pH by shifting the pCO₂ of the media results in a corresponding shift in cytoplasmic Adriamycin concentration (25). When Adriamycin is incubated with cells, the first intracellular compartment into which it diffuses is the cytosol. The cytosolic pH determines the proportion of Adriamycin that will be protonated and unable to diffuse back across the plasma membrane.

These previous cellular pH measurements were made with cell-permeant pH probes (e.g., 2′7′-bis(2-carboxyethyl)- 5(6) carboxy fluorescein acetoxymethylester, and SNARF-acetoxymethylester) that are deesterified in the cytosol. However, there are several reasons why these are not good probes for cytosolic pH. First, some of the probes are internalized by endocytosis. Thus, they would be reporting a weighted sum of cytosolic and endosomal pH. Second, they accumulate in intracellular organelles which contain esterases and organic anion transporters that have been proposed to transport these probes (58). Therefore, their fluorescence is a weighted average of the cytosolic and organellar pH. Third, these probes have been suggested as substrates for Pgp (59). This would complicate considerably comparisons of results between the MCF-7 and MCF-7/ ADR cells (which express Pgp).

To specifically measure the cytosolic and nucleoplasmic pH, an ideal probe would be large, membrane impermeable, and rapidly and selectively introduced into the cytosol. SNARF, conjugated to a dextran of 10 or 70 kD, was loaded into the cytosol by scraping adherent cells rapidly off the surface of polystyrene (34). The scraping causes a temporary shearing of the plasma membrane, which allows the impermeant macromolecules to diffuse into the cytosol. Since the pH probe SNARF is conjugated to a dextran, once introduced into the cytosol it does not cross cellular membranes. To measure the pH within the cytosol, the probe was conjugated to a 70-kD dextran which is too large to pass through the nuclear pores (Fig. 6,a). To measure the pH within the nucleoplasm, the probe was conjugated to a 10-kD dextran, which is too large to cross membranes into organelles but still small enough to pass through the nuclear pores (Fig. 6 b).

These measurements show that MCF-7 cells have an average cytosolic pH of 6.75 ± 0.3, which is 0.4 U lower than the cytosolic pH of MCF-7/ADR cells (pH 7.15 ± 0.1) when the extracellular medium is buffered at pH 7.3 (Table 1). The pH of the nucleoplasm in both cell types was 0.1–0.3 pH U more alkaline than the cytosol pH (Table 1). This is consistent with another recent report, which finds a more alkaline nucleoplasmic pH (60). The higher cytosolic pH of MCF-7/ADR cells results in a lower plasma membrane ΔpH and contributes to a greater organellar ΔpH. This is consistent with the hypothesis that the distribution of the weak base Adriamycin (Fig. 1) is affected by pH gradients across membranes.

These results demonstrate a large difference in the subcellular pH profile between the drug-resistant and drug-sensitive cell types (Table 1). The PSS hypothesis predicts that these differences in acidification are causally related to the drug-resistance phenotype. As a test of this hypothesis, the subcellular pH profile of drug-resistant MCF-7/ADR cells was changed to resemble that of drug-sensitive MCF-7 cells while monitoring the distribution of the chemotherapeutic drug Adriamycin. To do this, the sodium/proton exchanger monensin, which dissipates pH gradients across all membranes (21, 61, 62), and the vacuolar proton ATPase inhibitors bafilomycin A1 and concanamycin A (62, 63) were used to block acidification. Acridine orange–labeled MCF-7/ADR cells (Fig. 7,a) when treated with monensin (Fig. 7 b) did not exhibit the red spectral shift within the vesicles, indicating the loss of acidification in these compartments. Similar results were seen with the potassium/proton exchanger nigericin (data not shown).

These experiments were repeated with bafilomycin A1 and concanamycin A, which specifically block the proton ATPase. Thus, their mechanism of inhibiting the acidification of endosomes, lysosomes, and TGN is totally different from the ionophores monensin and nigericin. The punctate red acridine orange labeling of acidic organelles in MCF-7/ ADR cells (Fig. 8, a and e) was lost after incubation with bafilomycin A1 (Fig. 8,b) or concanamycin A (Fig. 8 f). Neither of these agents have any direct effect on the fluorescent properties of acridine orange (data not shown) and, unlike monensin or nigericin, they do not affect membrane permeability.

The distribution of Adriamycin in MCF-7/ADR cells (Fig. 7,c, and Fig. 8, c and g) was monitored upon addition of either monensin (Fig. 7,d), bafilomycin A1 (Fig. 8,d), or concanamycin A (Fig. 8,h). Treatment with any of the three different agents produced a similar redistribution of Adriamycin: a decrease in the punctate cytoplasmic compartments and a diffuse increase across the cytoplasm and nucleoplasm. This distribution of Adriamycin in cells where organelle acidification is blocked is similar to that observed in the drug-sensitive MCF-7 cells (Fig. 1 b).

Most chemotherapeutic agents have sites of action in the nucleus or in the cytosol. Therefore, their toxicity depends upon their concentration in either of these two compartments. The PSS hypothesis proposes that the concentration of weak base chemotherapeutic drugs in both the cytosol and nucleoplasm is regulated by the ability of cytoplasmic organelles to sequester the drugs away from the cytosol. The PSS hypothesis is based on the assumption that chemotherapeutic drugs entering acidic organelles should become protonated, thereby sequestered from the cytosol, and secreted. We find that Adriamycin colocalizes on the light microscopic level with the acidic organelles of living drug-resistant MCF-7/ADR cells, including the lysosomes, recycling endosome compartment, and the TGN. Our result is consistent with an ultrastructural study of drug-resistant NIH 3T3 cells, which showed that the Golgi compartment and lysosomes were labeled by a photoprecipitate of Adriamycin (14).

The results presented here demonstrate two aberrations of cellular pH regulation in the MCF-7 drug-sensitive tumor cell line. First, there is a failure to acidify organelles as measured both qualitatively (Fig. 5, and reference 21) and quantitatively (Table 1). Second, we have found that the cytosol in MCF-7 cells is 0.4 pH U more acidic than the cytosol of MCF-7/ADR cells. As described below, both of these features will increase the concentration of chemotherapeutic drugs in the cytosol and nucleoplasm of drug-sensitive tumor cells relative to the concentrations in drug-resistant or nontransformed cells. We have observed similar results on organellar acidification in the SW-48 human colon cell line and the MDA-231 (estrogen receptor–negative) human breast cancer line, and qualitatively similar observations have been reported for the K-562 erythroblastoma leukemia cell line (26). Chemotherapy relies upon tumor cells being more sensitive to chemotherapeutic drugs than nontransformed cells. We propose that one factor that might contribute to this enhanced sensitivity in these drug-sensitive cells is a failure of the PSS mechanism.

The pH gradient across the plasma membrane of MCF-7 cells is 0.55 U, whereas in drug-resistant MCF-7/ADR cells, it is 0.15 U. Therefore, at equilibrium conditions, there will be ∼2.5-fold greater accumulation of Adriamycin in the cytosol of MCF-7 cells than in that of MCF-7/ADR cells.

The pH gradients across the organelles of the MCF-7 cells were almost zero (Table 1). Thus, the organelles could not accumulate chemotherapeutic drugs. In contrast, the pH gradient between the recycling endosome compartment and the cytosol of MCF-7/ADR is calculated to be an average of 1.0 pH U, which would cause a 10-fold accumulation in this organelle. The ΔpH between the lysosome and cytosol in MCF-7/ADR cells is <2.0 pH U. Thus, the lysosomes of MCF-7/ADR cells should accumulate at least 100-fold more Adriamycin than the cytosol.

If the distribution of Adriamycin reached an equilibrium, the cytosolic concentration would be solely dependent on the ΔpH across the plasma membrane and extracellular drug concentration. Acidic organelles would accumulate high levels of drugs, but could not lower the cytosolic concentration. However, a large body of work has shown that many acidic organelles, including the TGN and recycling endosomes, continuously secrete their contents by exocytosis. This active process, if fast relative to the diffusion of extracellular drug into the cytosol, will keep cytosolic and nuclear drug levels low. Drug concentrations would not reach equilibrium distribution but would remain at a steady state due to the continuous cycling of acidic organelles. Thus, organellar acidification would lower the concentration of Adriamycin in the cytosol and nucleoplasm of drug-resistant and nontransformed cells. In fact, this mechanism may account for the difference in Adriamycin distribution we observed between MCF-7 and MCF-7/ADR cells (Fig. 1). In drug-resistant MCF-7/ADR cells, Adriamycin was sequestered within subcellular organelles, decreasing the drug concentration within the nucleoplasm and, accordingly, in the cytosol as well. (The high density of organelles throughout the cytoplasm makes it impossible to resolve Adriamycin fluorescence selectively from the cytosol. As a first approximation, we assume that the nucleoplasmic concentration reflects the free cytosolic concentration, since Adriamycin should be freely permeable through both the nuclear envelope and the nuclear pores, whose size cut-off is 25 nm [64]). In drug-sensitive cells, in the absence of an organellar mechanism for sequestration, a greater percentage of incoming Adriamycin remained in the cytosol with access to binding sites within the nucleus.

The PSS hypothesis proposes that cytosolic/nucleoplasmic drug concentrations are a function of the ΔpH and drug permeability of the plasma membrane, the ΔpH and drug permeability of the organellar membrane, and the rate of exocytosis. Hence, nuclear/cytosolic drug levels would be increased by (a) elevated plasma membrane ΔpH, which would increase cytosolic drug accumulation; (b) decreased organellar ΔpH, which would decrease sequestration; and (c) decreased rate of secretion, which would permit the drug levels to equilibrate across organelle membranes. The PSS hypothesis predicts that at a steady state, dissipation of plasma membrane ΔpH will decrease drug sensitivity, whereas dissipation of organellar ΔpH and/or decreasing the rate of secretion will increase drug sensitivity.

MDR in tumors could stem from a number of cell biological changes. The most frequently proposed mechanism for MDR in tumors is a plasma membrane–based efflux pump that uses ATP to transport chemotherapeutic drugs (4). This idea is based on studies in cell lines that express two of the proteins implicated in multidrug resistance, Pgp and MRP. The evidence includes the observations that (a) addition of azide to these cells increases nuclear accumulation of chemotherapeutic drugs; (b) both Pgp and MRP have ATP-binding domains; and (c) chemotherapeutic drugs modified with photoactive groups can be used to label Pgp. The PSS mechanism tested in this paper may be an additional mechanism for drug resistance working separately from the Pgp and MRP drug-efflux pumps.

It is also possible that both Pgp and MRP contribute to the PSS mechanism by affecting organelle pH. Indeed, it has been demonstrated that transfection of cells with Pgp results in an alkaline shift of the cytosolic pH (23) and total cellular pH (24) consistent with the pH shift seen in resistant cells (Table 1). One clue to the mechanism for this pH shift may come from the structural homology between Pgp, MRP, and the ATP–binding cassette family of proteins, which includes the cystic fibrosis transmembrane conductance regulator, a chloride channel expressed in lung epithelial cells. In cells defective in cystic fibrosis transmembrane conductance regulator, there is reduced acidification of the TGN and recycling endosome compartments (65). Acidification of organelles requires a proton ATPase. However, proton pumping alone is not sufficient to acidify to levels seen in organelles such as endosomes and lysosomes. The membrane potential caused by the proton gradient blocks the activity of the proton ATPase. Thus, anion channels are necessary to dissipate the membrane potential and keep the proton ATPase pumping protons (40, 66). Both Pgp and MRP have been observed to be components of intracellular membranes (67, 68). They might either form ion channels or modify the activity of ion channels (69–73). If the MCF-7 cells were missing a counterion transport, this would limit acidification of the organelles. The presence of Pgp or MRP may facilitate a counterion transport, allowing restoration of acidification within these organelles in the MCF-7/ADR cells. However, there is still disagreement over whether Pgp forms an ion channel (69, 74).

There are differences in the pH gradients across both the plasma membrane and organellar membranes between drug-sensitive and drug-resistant cells (Table 1). Results presented here indicate that the cellular organelles are critical determinants of cellular sensitivity to chemotherapeutic drugs that are weak bases. Monensin disrupts pH gradients across all cellular membranes. Dissipation of plasma membrane ΔpH should decrease the cytosolic concentration of adriamycin in MCF-7/ADR cells, and conversely, dissipation of organellar ΔpH should increase cytosolic and nucleoplasmic adriamycin. The observation that monensin increases the nucleoplasmic concentration of Adriamycin (Fig. 7,d) and increases substantially the sensitivity of MCF-7/ ADR cells to Adriamycin (21) suggests that organellar pH gradients make a greater contribution to Adriamycin distribution than plasma membrane pH gradients. Further, both bafilomycin A1 and concanamycin A, which specifically block the vacuolar proton ATPase in endosomes and lysosomes (62, 63) and TGN of eukaryotic cells (42), increase adriamycin levels in the nucleus (Fig. 8, d and h). Thus, the drug-resistant phenotype—both sequestration of drugs into cytoplasmic organelles and the sensitivity of cells to the weak base chemotherapeutic drugs—is causally dependent upon organelle acidification. The pH gradient across the plasma membrane may make significant contributions to the sensitivity of drug-resistant cells to non-weak base chemotherapeutic drugs such as colchicine and taxol. The binding of colchicine to tubulin is pH dependent and is favored at more acidic pH (75). Similarly, an acidic pH favors the stabilization of microtubules by taxol (76). Thus, the acidic cytoplasmic pH of tumor cells increases the activity of chemotherapeutic drugs. The more neutral pH of nontransformed and MDR cells decreases their activity.

In the PSS model, acidification of organelles plays a direct role in the accumulation of weak base chemotherapy agents. Still, it remains to be determined whether the drugs are actually exocytosed from the cell, and whether acidification may subserve other functions in pathways of cellular detoxification. Further, we do not know if sequestering chemotherapeutic drugs within subcellular organelles followed by secretion is sufficient to account for all the drug resistance of MCF-7/ADR cells. However, based on the observation that disrupting organellar acidification in drug-resistant MCF-7/ADR cells (with either the protonophore monensin or H⁺-ATPase blockers bafilomycin A1 and concanamycin A) both reverses the subcellular Adriamycin distribution to that seen in drug-sensitive MCF-7 cells (Figs. 7 and 8) and increases Adriamycin toxicity (21), we propose that the organelles make a significant contribution to the drug-resistance phenotype.

LAMP

lysosome-associated membrane protein

MDR

multidrug resistance

MRP

MDR-associated protein

Pgp

P-glycoprotein

PSS

protonation, sequestration, and secretion

SNARF

seminaphthorhodafluor

TGN

trans-Golgi network