Effect of molecular structure on the selective ...

Effect of molecular structure on the selective phototoxicity of triarylmethane dyes towards tumor cells Irawati K. Kandela, Jeremy A. Bartlett and Guilherme L. Indig * University of Wisconsin, School of Pharmacy, Madison, WI 53705-2222, USA. E-mail: glindig@facstaff.wisc.edu Received 19th November 2001, Accepted 19th March 2002 First published as an Advance Article on the web 9th April 2002

In response to transmembrane potentials which are negative on the inner side of both the plasma and mitochondrial membranes, cationic dyes displaying appropriate structural features naturally accumulate in the cytosol and inside the mitochondria. Because enhanced mitochondrial membrane potential is a prevalent tumor cell phenotype, a number of cationic dyes preferentially accrue and are retained for longer periods in the mitochondria of tumor cells as compared to normal cells. The opportunities brought about by this phenomenon in chemo- and photochemotherapy of neoplastic diseases is highlighted by the observation that the phototoxic effects associated with some of the cationic photosensitizers known to accumulate in cell mitochondria are much more pronounced in tumor cells than in normal cells. However, the structural determinants of selective phototoxicity towards tumor cells are not well understood, and the lack of a robust model to describe the relationship between molecular structure and tumor selectivity has prevented mitochondrial targeting from becoming a more dependable therapeutic strategy. In this report we describe how the lipophilic/hydrophilic character of a series of cationic triarylmethane dyes affects the selectivity with which these photosensitizers mediate the destruction of tumor cells. Our results indicated that only the more hydrophilic triarylmethanes show tumor selectivity, presumably because these are the only dyes capable of staining energized mitochondria with a high degree of specificity. The partition of the more lipophilic dyes into a variety of extra-mitochondrial subcellular compartments occurs with comparable efficiencies in tumor and in normal cells, and this less specific subcellular localization precludes tumor selectivity from taking place.

Introduction The conceptual basis for the development of mitochondrial targeting as a novel therapeutic strategy for both chemo- and photochemotherapy of neoplastic diseases rests on the observation that enhanced mitochondrial membrane potential is a prevalent tumor cell phenotype.1 Out of more than 200 cell lines/types of tumor and normal cells originally investigated by Chen and co-workers, only approximately 2% do not show this dominant characteristic.1,2 Because transmembrane potentials are negative on the inner side of both the plasma and mitochondrial membranes, extensively conjugated cationic molecules (dyes) displaying appropriate structural features are electrophoretically driven through these membranes and tend to accumulate in the cytosol and inside energized mitochondria.1–6 The higher mitochondrial membrane potential typical of tumor cells thus opens a window for the selective destruction of these cells via mitochondrial targeting. The potential of mitochondrial targeting in photochemotherapy is highlighted by the observation that a number of cationic dyes accumulate in larger amounts and are retained for longer periods in the mitochondria of tumor cells as compared to normal cells.1–3 Besides, the phototoxic effects associated with some of the cationic photosensitizers known to accumulate in the mitochondria are much more pronounced in tumor cells than in normal cells.3–10 While several cationic dyes naturally accrue in the mitochondria of living cells, only a small subset of these “mitochondrial” dyes induce the photochemical destruction of tumor cells with desirable selectivity.7,10 The lack of information on the relationship between the molecular structure of cationic “mitochondrial” photosensitizers and their selective toxicity to tumor cells has prevented mitochondrial targeting from becoming a more dependable therapeutic strategy. In this report, we describe a systematic study of how the molecular structure of a series of closely related cationic triarylmethane (TAM⫹) DOI: 10.1039/b110572h

photosensitizers modulates their selective phototoxicity towards tumor cells. TAM⫹ dyes accumulate in energized mitochondria, and their phototoxic effects are known to occur at least in part at the mitochondrial level.5,11–14 Our results indicate that the selective phototoxicity of TAM⫹ dyes towards tumor cells is primarily controlled by the lipophilic/hydrophilic character of the photosensitizer. While the more hydrophilic TAM⫹ dyes show a high degree of tumor cell selectivity, those showing higher lipophilic character do not. Comparison of octan-1-ol–water partition coefficients indicates that only those dyes with a lipophilic/ hydrophilic character close to that of the mitochondria-specific fluorescent marker Rhodamine-123 1,2 are tumor selective. This observation suggests that only the dyes that stain the mitochondria with a high degree of specificity can promote the photoinduced destruction of tumor cells with a high degree of selectivity. When a variety of subcellular compartments (other than cell mitochondria) are simultaneously stained, differences in photosensitizer uptake and retention between tumor and normal cells are minimal. The less specific subcellular damage induced by these more lipophilic photosensitizers precludes tumor selectivity from taking place. In these cases, cell death is no longer primarily controlled by mitochondrial damage, but rather by a more complex combination of subcellular phototoxic events.

Materials and methods Reagents and cell culture media Chlorine salts of ethyl violet (EV⫹), victoria blue R (VBR⫹), victoria pure blue BO (VPBBO⫹) from Aldrich (Milwaukee, WI), and crystal violet (CV⫹, also referred to as N,N,N⬘,N⬘, N⬙,N⬙-hexamethylpararosaniline) from Sigma Chemical Co. (St. Louis, MO) were recrystallized from methanol and dried Photochem. Photobiol. Sci., 2002, 1, 309–314

This journal is © The Royal Society of Chemistry and Owner Societies 2002

309

under vacuum. The purity of recrystallized triarylmethanes was assessed by thin-layer chromatography (silica gel, methanol– acetic acid 95 : 5, v/v). Pararosaniline (PA⫹, >98% pure), methanol, n-butyl alcohol, acetic acid, acetonitrile, hydrochloric acid, perchloric acid, silica gel (230–400 mesh Merck grade 9385) from Aldrich, horseradish peroxidase (HRP, type VI), sodium chloride, potassium chloride, sodium phosphate (monobasic and dibasic salts), potassium phosphate (monobasic), and trypan blue from Sigma, and ethanol from Pharmaco Products, Inc. (Brookfield, CT), were all of high purity grade and used as supplied. RPMI 1640 cell culture media, Dulbecco’s phosphate-buffered saline (DPBS), fungizone (amphotericin B), penicillin, streptomycin, ethylenediaminetetraacetic acid (EDTA; tetrasodium salt) and trypsin (porcine) were obtained from Gibco (Rockville, MD). Fetal bovine serum (FBS) was purchased from Atlanta Biologica (Norcross, GA). Water was distilled, deionized, and filtered prior to use (Millipore Milli-Q system; resistivity, 18 MΩ cm). Instruments and methods Spectrophotometric studies were performed with a Shimadzu UV-2101PC spectrophotometer. High-performance liquid chromatography (HPLC) analyses were carried out on an HPLC system from Waters (Milford, MA) model 501 equipped with a tunable UV–Vis detector model 484. 1H-NMR spectra were obtained on a 300 MHz NMR spectrometer Bruker (Westmont, IL) model Aspect 3000. Mass spectra data were obtained either on a Micromass AutoSpec (Beverly, MA) Mass Spectrometer using fast atom bombardment (FAB) as the ionization technique or on a Applied Biosystems (Foster City, CA) Mass Spectrometer model MDS Sciex API 365 using electrospray (ESI) as the ionization technique. The cells were irradiated using a lab-built bank of four parallel 40 W tubular cold fluorescent light bulbs from Sylvania (Dania Beach, FL) model F40 CWP. The emission profile of this light source was characterized on a Timemaster Strobemaster fluorometer from Photon Technology International, Inc. (South Brunswick, NJ), and its fluence rate (18 W m⫺2) measured with a solidstate joulemeter model PM30VI from Molectron (Portland, USA). Octan-1-ol–water partition coefficients were determined at 25 ⬚C using equal volumes of these two solvents preequilibrated with each other.15 Typically, five distinct dye solutions (∼ 10 µM) were prepared in water and subsequently equilibrated with octan-1-ol. After the equilibrium was reached, the aqueous and/or organic phases were equilibrated again with new aliquots of the second solvent of the respective octan-1-ol–water solvent pair and the final dye concentration in the organic and aqueous phases determined by absorption spectroscopy. This re-partitioning procedure was repeated as needed to provide a constant value of the partition coefficient in two subsequent octan-1-ol–water equilibria. All partition coefficients were measured at neutral pH. Preparation, purification, and characterization of CVⴙ derivatives N,N,N⬘,N⬘,N⬙-Pentamethylpararosaniline, N,N,N⬘,N⬙-tetramethylpararosaniline, and N,N⬘,N⬙-trimethylpararosaniline were prepared through the sequential oxidative N-dealkylation of CV⫹ catalyzed by HRP.16 One of the major advantages of using this strategy is that, unlike chemical methods, the HRPcatalyzed reaction generates only selected isomers as reaction products. That is, while the reaction efficiently generates N,N,N⬘,N⬙-tetramethylpararosaniline and N,N⬘,N⬙-trimethylpararosaniline, it does not generate N,N,N⬘,N⬘-tetramethylpararosaniline and N,N⬘,N⬘-trimethylpararosaniline in any significant (detectable) amount. The absence of isomers in the final reaction media facilitates the purification of the reaction products to the level required for a more detailed analysis of 310

Photochem. Photobiol. Sci., 2002, 1, 309–314

structure–activity relationships in in vitro assays. Typically, HRP (0.085 µM) and hydrogen peroxide (20.4 mg, 600 µmol) were sequentially added to a solution of crystal violet (60 mg, 147.1 µmol) in 500 mL of 10 mM sodium phosphate buffer pH 7.3. The reaction mixture was kept in the dark under continuous magnetic stirring and the reaction progress monitored by HPLC (reversed phase C18 column; 90% acetonitrile–10% 50 mM aqueous perchloric acid as eluent, UV–Vis analysis at 550 and 280 nm). After the reaction was complete, the reaction products were extracted from the aqueous media with 250 mL of n-butyl alcohol, the solvent was removed by roto-evaporation, and the dyes of interest isolated by reduced pressure chromatography using columns packed with pretreated silica gel. The chromatography procedure was performed as follows: 100 g of 230–400 mesh Merck grade 9385 SiO2 were washed with 250 ml of 1 M NaCl solution and subsequently filtered through a glass Buchner funnel under reduced vacuum. This pre-treated, water-deactivated, NaClequilibrated SiO2 was then made into a slurry with propan-2-ol and packed into 1.5 cm diameter columns under reduced pressure (ca. 5–8 psi). The dye mixtures were dissolved in propan2-ol and the columns run under reduced pressure using this solvent as the mobile phase. The dye fractions (concentrated under vacuum) were subsequently analyzed by HPLC, and those containing highly purified dyes (>98% pure) were combined and dried under vacuum before storage under nitrogen in the dark. As the amount of H2O2 available in the enzymatic reaction media is increased, the fourth, fifth, and sixth members of the CV⫹ sequential N-dealkylation series are also produced in the HRP-catalyzed reaction (e.g. N,N⬘-dimethylpararosaniline, N-methylpararosaniline, and pararosaniline). However, we have found that it is substantially more difficult to control this enzymatic process to generate these last three products of sequential CV⫹ dealkylation in high yields and high purity levels than it is to generate the first three dealkylation products. For this reason, we alternatively prepared N,N⬘-dimethylpararosaniline and N-methylpararosaniline through the sequential N-methylation of commercially available pararosaniline using iodomethane as the alkylating agent. Typically, 10 ml of iodomethane (160 mmol) were slowly added to 40 ml of a stirred methanol solution of pararosaniline (0.813 g, 2.51 mmol), the reaction mixture was kept under reflux and the reaction progress followed by HPLC (reversed phase C18 column; 70% acetonitrile–30% 20 mM aqueous NH4Cl as eluent, UV–Vis analysis at 550 and 280 nm). The final reaction products were subsequently dried through roto-evaporation and separated using open column chromatography. In this case, activated SiO2 (230–400 mesh) columns were slurry packed in 70% CHCl3–30% propan-2-ol with 20 mM HCl. These columns were run under gravity flow using the same solvent system used for their packing. Mass spectrometry and 1H-NMR data for the new compounds are the following: N,N,N⬘,N⬘,N⬙-pentamethylpararosaniline, FAB-MS (m/z) for C24H28N3, calcd, 358.5, obsd, 358.2; 1H-NMR (DMSO-d6) δ (ppm) 8.28 (m, 1H), 7.26 (m, 6H), 6.99 (d, 4H, J = 9.0 Hz), 6.86 (d, 2H, J = 8.8 Hz), 3.21 (s, 12H), 2.93 (d, 3H, J = 4.8 Hz). N,N,N⬘,N⬙-Tetramethylpararosaniline, FAB-MS (m/z) for C23H26N3: calcd, 344.5; obsd, 344.1; 1H-NMR (DMSO-d6) δ (ppm) 8.17 (m, 2H), 7.25 (m, 6H), 6.98 (d, 2H, J = 7.9 Hz), 6.84 (d, 4H, J = 8.43 Hz), 3.20 (s, 6H), 2.92 (d, 6H, J = 3.87 Hz). N,N⬘,N⬙-Trimethylpararosaniline, FAB-MS (m/z) C22H24N3, calcd, 330.5; obsd, 330.2; 1H-NMR (DMSO-d6) δ (ppm) 8.19 (br s, 3H), 7.22 (d, 6H, J = 8.8 Hz), 6.84 (d, 6H, J = 8.8), 2.90 (br s, 9H). N,N⬘Dimethylpararosaniline, FAB-MS (m/z) for C21H22N3, calcd, 316.4; obsd, 316.2; 1H-NMR (DMSO-d6) δ (ppm) 8.23 (br s, 2H), 7.49 (br s, 2H), 7.20 (m, 6H), 6.82 (m, 6H), 2.90 (s, 6H). N⬘-Methylpararosaniline, ESI-MS (m/z) for C20H20N3: calcd, 302.4; obsd, 302.4.

The biological model HT-29 human colon adenocarcinoma cells (ATCC HTB-38) and CV-1 green monkey kidney cells (ATCC CCL-70) were obtained from the American Type Culture Collection (Manassas, Virginia). These cells were used as models for tumor (HT-29) and normal (CV-1) cells as previously proposed by Oseroff.4,5,17 Both cell lines were grown in RPMI 1640 media supplemented with 10% FBS, amphotericin B (0.63 µg ml⫺1), penicillin (200 unit ml⫺1) and streptomycin (200 µg ml⫺1). Small Petri dishes (35 mm in diameter) were seeded at low cell densities (9 × 105 cells per 3 ml of growth media) to permit the formation of evenly distributed monolayers and the cells were allowed to attach overnight. Before irradiation, the cells were washed twice with fresh DPBS solution and subsequently incubated for 60 minutes at 37 ⬚C in the presence of 3 ml of 5 µM TAM⫹ dye solution prepared in fresh phosphate buffer saline (PBS). The original TAM⫹ stock solutions were prepared in ethanol, and the final ethanol content in the PBS solutions used for cell incubation was always kept at the 1% level. After incubation in the presence of the TAM⫹ dye, the cells were washed twice with DPBS, re-incubated with 3 ml of PBS, and immediately subjected to irradiation. During irradiation, the cells were kept at approximately 20 cm from the bank of cold fluorescent lamps and at room temperature. After irradiation, the PBS was replaced by fresh growth media and the cells incubated for 24 hours before being fixed with methanol and stained with a 0.1% aqueous CV⫹ solution. In a series of experiments designed to explore the effect of EV⫹ concentration on tumor selectivity, after this last 24 hours’ incubation in fresh growth media, the cells were detached from the culture dishes with trypsin/EDTA in PBS and the survival ratio estimated using the trypan blue dye exclusion test as previously described.18 The imaging of the cells was carried out on a microscope from Nikon (Chicago, IL) model UFX-II equipped with a Nikon camera model FX35A.

Results and discussion We conducted a model pre-clinical study to explore how the lipophilic/hydrophilic character of structurally related “mitochondrial” cationic photosensitizers affects the efficiency with which they induce the photochemical destruction of tumor cells as compared to normal cells. To this end, we compared the phototoxicity of a series of TAM⫹ dyes towards adenocarcinoma (HT-29) and normal kidney (CV-1) cells. These cells were previously employed by Oseroff and co-workers in investigations designed to demonstrate that the differences in mitochondrial membrane potential typically observed between tumor and normal cells can be explored for phototherapeutic purposes.4,5,17 One of the major advantages of exploring the concept of mitochondrial targeting using this biological model rests on the fact that it allows rapid visual inspection of whether a particular photosensitizer displays selective phototoxicity towards tumor cells. It permits the timely and inexpensive analysis of structural determinants of tumor cell selectivity on the basis of relatively large numbers of photosensitizers. Fig. 1 shows the relative phototoxicity of 10 distinct TAM⫹ dyes towards HT-29 and CV-1 cells. In these comparative experiments, half of the dye-loaded cell population present in each distinct Petri dish was irradiated with cold fluorescent light, while the other half was protected from the incident irradiation by masking half of each Petri dish with aluminium foil and black tape.4 In Fig. 1, the left-half of each panel shows the area of each cell colony exposed to light, and the right-half, the area protected from the incident irradiation. Marked differences in cell density between the left and right sides in any one of the panels indicates that substantial phototoxicity is associated with the respective dye towards that cell line. If no significant phototoxicity is observed, the cell density in the left-

and right-halves of the cell colony is equivalent (visually indistinguishable). A dye exhibits tumor selectivity when the exposure to light causes significant differences between the left (exposed) and right (unexposed) sides of the HT-29 plates but leaves the left and right sides of the CV-1 plates indistinguishable. For example, CV⫹ is found to be selective for tumor cells because it efficiently mediates the photoinduced destruction of HT-29 cells but does not damage the CV-1 cells. On the other hand, EV⫹ indiscriminately destroys the light-exposed colonies of both cell types, indicating that it is not selective for tumor cells. We used relatively harsh conditions in the experiments presented in Fig. 1 to allow the straightforward visual inspection of tumor selectivity. In so doing, we effectively maximized the phototoxic effects of the TAM⫹ dyes on tumor cells such that we always observed pronounced differences in cell density between the left- and right-halves of HT-29 plates. Fig. 2 shows the absorption spectra (in water) of all the TAM⫹ dyes used in this study, along with the emission profile of the light source we used to produce the data presented in Fig. 1. Note that although the overlaps between the absorption spectrum of each TAM⫹ dye and the emission profile of the light source used to irradiate the cells are different from each other, these differences do not negatively impact on this qualitative analysis. As long as the absorption spectra of the photosensitizers of interest show substantial overlap with the emission profile of the light source, it is generally straightforward to optimize dye concentration, incubation time, and irradiation time to permit the simultaneous qualitative evaluation of a large number of photosensitizers with regard to tumor selectivity. In this study, the lipophilic character of the TAM⫹ dyes was evaluated by measuring their octan-1-ol–water partition coefficient (P).15 The higher the value of P in Fig. 1, the higher the lipophilic character of the TAM⫹ dye. The results presented in Fig. 1 show that selective phototoxicity of TAM⫹ dyes towards tumor cells is observed for CV⫹ and its N-dealkylated derivatives (compounds I to VI in Fig. 1). However, those dyes with increased lipophilic character, i.e. EV⫹, VPBBO⫹, and VBR⫹ lose this selectivity. The observation that only those TAM⫹ dyes with low values of octan-1-ol–water partition coefficient are selective for tumor cells strongly suggests that this selective tumor toxicity is largely modulated by the reduced tendency of these more hydrophilic TAM⫹ dyes to partition into lipophilic subcellular compartments.19,20 This inference is supported by previous work by Chen on the determinants of subcellular distribution of Rhodamine dyes. Chen demonstrated that the mechanism of cellular uptake and mitochondrial accumulation of Rhodamine-123 (Rh-123⫹) is primarily controlled by mitochondrial membrane potentials.1–3 Accordingly, Rh-123⫹ stains the cell mitochondria with virtually absolute selectivity, and is typically taken up in larger amounts and retained for longer periods by tumor cells as compared to normal cells.1–3 It is reasonable to infer that the mechanisms of cellular uptake and retention of CV⫹ and its N-dealkylated analogs (compounds I to VI in Fig. 1) is also primarily controlled by mitochondrial membrane potentials, given that the octan-1-ol–water partition coefficients associated with these dyes (2.2 > P > 0.21, see Fig. 1) are very close to the value associated with the highly specific mitochondrial marker Rh-123⫹ (P = 0.24). The inference that CV⫹ and its N-demethylated analogs stain energized mitochondria with a high degree of specificity is in keeping with the observed tumor selectivity associated with the phototoxic effects of these dyes, since the higher mitochondrial membrane potentials typical of tumor cells would cause a higher phototoxicity towards these cells as compared to normal cells.6 The contribution of lipophilic partitioning on the mechanism of cellular uptake, subcellular distribution, and cellular retention of the more lipophilic TAM⫹ dyes, namely EV⫹ (P = 237), VPBBO⫹ (P = 180), and VBR⫹ (P = 39) is bound to be Photochem. Photobiol. Sci., 2002, 1, 309–314

311

Fig. 1 Phototoxicity of TAM⫹ dyes towards tumor (HT-29) and normal (CV-1) cells. Colonies of HT-29 (left panels) and CV-1 (right panels) cells were incubated for 1 hour with 5 µM TAM⫹ dyes, washed, and irradiated for 1 hour as described in the Materials and methods section. Twenty-four hours after irradiation the cells were fixed with methanol and stained. Before irradiation, half of each Petri dish was masked with black tape. The area protected from light is represented by the right-half of each panel. Upon irradiation, only the HT-29 cells were heavily destroyed when the octan-1-ol–water partition coefficient (P) associated with the TAM⫹ photosensitizer was equal to or lower than 2.2. For the cases of TAM⫹ dyes showing a partition coefficient above that of CV⫹ no tumor selectivity was observed. No phototoxic effects were observed upon irradiation of the cells in the absence of the TAM⫹ dyes.

much more pronounced for these dyes than for CV⫹ and its more hydrophilic N-demethylated analogs.19,20 Because the more lipophilic TAM⫹ dyes are expected to partition into a variety of lipophilic subcellular compartments with equivalent efficiencies in both tumor and normal cells,19,20 their putative 312


enhanced accumulation in the mitochondria of tumor cells no longer results in selective phototoxicity. That is, in these cases, the equivalent photoinduced destruction of a variety of subcellular compartments other than the mitochondria in tumor and normal cells precludes tumor selectivity, regardless of

drugs in the mitochondria of tumor cell as compared to normal cells, but also on their close to absolute specificity towards this organelle.

Concluding remarks

Fig. 2 Absorption spectra of CV⫹ and its N-demethylated derivatives in water. From right to left, CV⫹ (N,N,N⬘,N⬘,N⬙,N⬙-hexamethylpararosaniline), N,N,N⬘,N⬘,N⬙-pentamethylpararosaniline (I), N,N,N⬘, N⬙-tetramethylpararosaniline (II), N,N⬘,N⬙-trimethylpararosaniline (III), N,N⬘-dimethylpararosaniline (IV), N-methylpararosaniline (V), and pararosaniline (VI). Inset A: absorption spectra of (from right to left): VPBBO⫹ (dashed line), VBR⫹ (solid line), EV⫹ (dotted line), and CV⫹ (solid line), all recorded in water. Inset B: emission profile of the light source used to irradiate the cells.

whether the mitochondria of the tumor cells are more efficiently destroyed upon irradiation than the mitochondria of normal cells.6 The hypothesis that a highly specific mitochondrial localization of cationic photosensitizers is a major prerequisite for tumor selectivity is supported by our previous investigations of the subcellular distribution of CV⫹ and EV⫹ in rat basophilic leukemia (RBL) cells.6 In that study, we used multi-photon laser scanning microscopy 21,22 to obtain simultaneous images of the fluorescence of NAD(P)H (an endogenous mitochondrial marker 23), Rh-123⫹ and CV⫹ or EV⫹ in RBL cells incubated in the presence of both Rh-123⫹ and one of these TAM⫹ dyes. While the comparison of the subcellular distribution of the NAD(P)H, Rh-123⫹, and TAM⫹ fluorescence clearly indicated that CV⫹ does localize in the mitochondria with a high degree of specificity, the results obtained for EV⫹ indicated otherwise. The spatial distribution associated with the fluorescence of EV⫹ in RBL cells was consistent with its heavy accumulation not only in the mitochondria but also in other subcellular compartments, presumably as a result of a higher contribution of the partition phenomena to the mechanism of subcellular distribution of EV⫹ as compared to CV⫹.6 These imaging experiments indicated that while EV⫹ is expected to mediate the photochemical damage of a variety of distinct subcellular compartments in tumor and normal cells with comparable efficiency, CV⫹ is expected to induce little or no photoinduced damage towards any subcellular compartment other than the mitochondria. In addition, to explore whether a decrease in the intracellular concentration of the most lipophilic TMA⫹ photosensitizer considered here (EV⫹) might lead to a more specific mitochondrial accumulation, and therefore provide the conditions required for tumor selectivity to take place, we have compared the phototoxicity of EV⫹ towards tumor and normal cells as a function of its concentration in the incubation media. As expected, the phototoxic effects towards both CV-1 and HT-29 cells decrease upon decreasing EV⫹ concentration (5.0, 1.0, 0.5, and 0.1 µM), but we did not observe any significant tumor selectivity even for the most dilute EV⫹ solutions. For example, upon 90 minutes’ irradiation of cells incubated for one hour with 0.1 µM EV⫹ solutions, the fraction of cells killed was found to be 32 ± 4% for the case of CV-1 cells and 30 ± 2% for the case of HT-29 cells. Therefore, our observations strongly suggest that the success of the concept of mitochondrial targeting in (photo)chemotherapy of neoplastic diseases rests not only on the preferential accumulation and retention of certain cationic

The selective phototoxicity of CV⫹ and its N-demethylated derivatives (structures I to VI in Fig. 1) towards tumor cells indicates that these dyes display structural features that can be used as guidelines for the development of novel mitochondriontargeted drugs for photochemotherapy of neoplastic diseases. The selectivity (or lack thereof ) with which TAM⫹ dyes mediate the destruction of tumor cells is hypothesized to depend primarily on the lipophilic/hydrophilic character of these dyes. While selective phototoxicity towards tumor cells is observed for all TAM⫹ dyes with octan-1-ol–water partition coefficients close to that of the highly specific mitochondrial marker Rhodamine-123, selectivity is lost upon increasing the lipophilic character of the photosensitizer. This observation strongly suggests that for tumor selectivity to take place, the more hydrophilic TAM⫹ dyes must localize in energized mitochondria with close to absolute specificity. The simultaneous partitioning of the more lipophilic photosensitizers in a variety of subcellular domains distinct from the mitochondria precludes tumor selectivity. This inference is based on the premise that while differences in mitochondrial potential between tumor and normal cells differentiate one from the other, the distributions of lipophilic compartments within these two cell types are intrinsically identical. Therefore, if a specific cationic photosensitizer only stains the mitochondria, this specific photosensitizer will presumably accumulate in higher quantities and be retained for longer periods in tumor cells as compared to normal cells, a phenomenon that allows selective phototoxicity towards tumor cells. On the other hand, if a photosensitizer stains a variety of subcellular compartments other than the mitochondria in tumor cells as a result of lipophilic partitioning, this photosensitizer will also stain these extramitochondrial subcellular compartments in normal cells with equivalent efficiency, since tumor and normal cells are intrinsically indistinguishable from each other with regard to the intracellular distribution of extra-mitochondrial lipophilic compartments. Therefore, the photoinduced destruction of these lipophilic regions with equivalent efficiencies in tumor and normal cells precludes tumor selectivity, regardless of whether or not the sites of subcellular accumulation of the more lipophilic photosensitizers also include energized mitochondria.

Acknowledgements This work was supported by the Robert Draper Technology Innovation Fund (UW-Madison) and PharmaLux, LLC. JAB acknowledges the Pharmaceutical Research and Manufacturers of America Foundation for an Advanced Predoctoral Fellowship in Pharmaceutics.

References 1 L. B. Chen, Mitochondrial membrane potential in living cells, Annu. Rev. Cell Biol., 1988, 4, 155–181. 2 S. Davis, M. J. Weiss, J. R. Wong, T. J. Lampidis and L. B. Chen, Mitochondrial and plasma membrane potentials cause unusual accumulation and retention of rhodamine 123 by human breast adenocarcinoma-derived MCF-7 cells, J. Biol. Chem, 1985, 260, 13844–13850. 3 S. D. Bernal, T. J. Lampidis, R. M. McIsaac and L. B. Chen, Anticarcinoma Activity in vivo of Rhodamine 123, a MitochondrialSpecific Dye, Science, 1983, 222, 169–172. 4 A. R. Oseroff, D. Ohuoha, G. Ara, R. J. McAuliffe and L. Cincotta, Intramitochondrial dyes allow selective in vitro photolysis of carcinoma cells, Proc. Natl. Acad. Sci. U. S. A., 1986, 83, 9729–9733.


313

5 J. S. Modica-Napolitano, J. L. Joyal, G. Ara, A. R. Oseroff and J. R. Aprille, Mitochondrial toxicity of cationic photosensitizers for photochemotherapy, Cancer Res., 1990, 50, 7876–7881. 6 G. L. Indig, G. S. Anderson, M. G. Nichols, J. A. Bartlett, W. S. Mellon and F. Sieber, Effect of molecular structure on the performance of triarylmethane dyes as therapeutic agents for photochemical purging of autologous bone marrow grafts from residual tumor cells, J. Pharm. Sci., 2000, 89, 88–99. 7 G. L. Indig, Mechanisms of action of cationic dyes in photodynamic therapy of tumors, Recent Res. Dev. Pure Appl. Chem., 1999, 3, 9–19. 8 C. R. Shea, N. Chen, J. Wimberly and T. Hasan, Rhodamine dyes as potential agents for photochemotherapy of cancer in human bladder carcinoma cells, Cancer Res., 1989, 49, 3961–3965. 9 J. S. Modica-Napolitano, B. T. Brunelli, K. Keizo and L. B. Chen, Photoactivation enhances the mitochondrial toxicity of the cationic Rhodacyanine MKT-077, Cancer Res., 1998, 58, 71–75. 10 J. Morgan and A. R. Oseroff, Mitochondria-based anti-cancer therapy, Adv. Drug Delivery Rev., 2001, 49, 71–86. 11 F. R. Gadelha, S. N. Moreno, W. D. Souza, F. S. Cruz and R. DoCampo, The mitochondrion of Trypanosoma cruzi is a target of crystal violet toxicity, Mol. Biochem. Parasitol., 1989, 34, 117–126. 12 J. Morgan, J. E. Whitaker and A. R. Oseroff, GRP78 induction by calcium ionophore potentiates PDT using the mitochondrial targeting dye victoria blue BO, Photochem. Photobiol., 1998, 67, 155–164. 13 M. Fiedorowicz, A. Pituch-Noworolska and M. Zembala, The effect of Victoria blue BO on peripheral blood mononuclear and leukemic cells, Photochem. Photobiol., 1997, 65, 855–861. 14 A. J. Kowaltowski, J. Turin, G. L. Indig and A. E. Vercesi, Mitochondrial effects of triarylmethane dyes, J. Bioenerg. Biomembr., 1999, 31, 579–588.

314


15 A. Leo, C. Hansch and D. Elkins, Partition coefficients and their uses, Chem. Rev., 1971, 71, 525–616. 16 F. R. Gadelha, P. M. Hanna, R. P. Mason and R. Docampo, Evidence for free radical formation during horseradish peroxidasecatalyzed N-demethylation of crystal violet, Chem. Biol. Interact., 1992, 85, 35–38. 17 A. R. Oseroff, G. Ara, D. Ohuoha, J. Aprille, J. C. Bommer, M. L. Yarmush, J. Foley and L. Cincotta, Strategies for selective cancer photochemotherapy: antibody-targeted and selective carcinoma cell photolysis, Photochem. Photobiol., 1987, 46, 83–96. 18 S. W. Perry, L. G. Epstein and H. A. Gelbart, Simultaneous in situ detection of apoptosis and necrosis in monolayer cultures by TUNEL and trypan blue staining, Biotechniques, 1997, 22, 1102– 1106. 19 R. W. Horobin and F. Rashid, Interactions of molecular probes with living cells and tissues. 1. Some general mechanistic proposals making use of a simplistic Chinese box model, Histochemistry, 1990, 94, 205–209. 20 F. Rashid and R. W. Horobin, Interaction of molecular probes with living cells and tissues. 2. A structure–activity analysis of mitochondrial staining by cationic probes, and a discussion of the synergistic nature of image-based and biochemical approaches, Histochemistry, 1990, 94, 303–308. 21 W. Denk, J. H. Strickler and W. W. Webb, Two-photon laser scanning fluorescence microscopy, Science, 1990, 248, 73–76. 22 J. M. Squirrell, D. L. Wokosin, J. G. White and B. D. Bavister, Long-term two-photon fluorescence imaging of mammalian embryos without compromising viability, Nature Biotechnol., 1999, 17, 763–767. 23 B. Chance, Fluorescent probe environment and the structural and charge changes in energy coupling of mitochondrial membranes, Proc. Natl. Acad. Sci. U. S. A., 1970, 67, 560–564.