1H NMR visible lipids are induced by phosphonium salts and 5-fluorouracil in human breast cancer cells



Cationic lipophilic phosphonium salts (CLPS) selectively accumulate in the mitochondria of neoplastic cells and inhibit mitochondrial function. The effects of the CLPS p-(triphenylphosphoniummethyl) benzaldehyde chloride (drug A), and [4-(hydrazinocarboxy)-1-butyl] tris-(4-dimethylaminophenyl) phosphonium chloride (drug B), on human breast cells of differing biological properties were assessed using growth inhibition assays and 1H NMR. Drug A and, to a lesser extent, drug B demonstrated selective growth inhibition of the highly tumorigenic DU4475 breast carcinoma cell line compared to the transformed HBL-100 human breast cell line. However, in contrast to previous studies using other cell lines, no synergistic activity was found when the drugs were used in combination. 1H NMR demonstrated significant increases in mobile lipid acyl chain resonances in both cell lines treated with cytotoxic doses (IC50, 48 h) of the drugs used either alone or in combination. Two-dimensional NMR revealed accompanying decreases in phosphocholine/Lys levels in HBL-100 cells treated with A, B, or a 1:1 combination A+B at the IC50, and in DU4475 cells treated with drug A (IC50). This was accompanied by significant increases in cho/Lys ratios with IC50 A or combination A+B treatment. Similar spectra were observed in cells treated with 5-fluorouracil but not methotrexate, indicating that mobile lipid accumulation is a general but not universal response to cytotoxic insult. Magn Reson Med 45:1001–1010, 2001. © 2001 Wiley-Liss, Inc.

Cationic lipophilic chemotherapeutic agents have become the focus of a number of studies over the last decade due to their novel mechanism of selectivity and toxicity for tumor cells (1–7). These compounds contain a cation and a lipophilic ring structure, and include such agents as rhodamines (e.g., Rh123 and Rh6G), MKT-077, DECA (dequalinium chloride), safranin O, Victoria Blue BO, and a class of compounds called cationic lipophilic phosphonium salts (CLPS) (2, 3). The CLPS are promising as potential antineoplastic agents compared to other lipophilic cationic compounds because a large variety of structures with a broad range of substitution patterns can be easily synthesized (2). The basic structure of CLPS consists of a triarylalkylphosphonium salt with various substituted groups on the phosphonium cation and side chains of the phenyl rings. Tetraphenylphosphonium chloride (TPP) is the prototype compound; other synthesized CLPS include p-(triphenylphosphoniummethyl) benzaldehyde chloride (drug A) and the cationic acylhydrazine, [4-(hydrazinocarboxy)-1-butyl] tris-(4-dimethylaminophenyl) phosphonium chloride (drug B). These two compounds are of particular interest because they have demonstrated synergism when used in combination, which was attributed to their potential ability to combine irreversibly under physiological conditions and form a more potent hydrazone product (see Fig. 1) (2, 3). The combination hydrazone product from a phosphonium acylhydrazine and benzaldehyde has been demonstrated in situ using mass spectroscopy (8). Moreover, a series of structurally related CLPS, including drugs A and B and the corresponding hydrazone C, have demonstrated selective in vitro cytotoxicity against breast, colon, pancreas, bladder, and hypopharynx carcinoma cells relative to two untransformed cell lines: CV-1 monkey kidney epithelial cells, and IEC-18 rat ileal epithelial cells (2–4).

Figure 1.

Chemical structure of cationic lipophilic phosphonium compounds p-(triphenylphosphoniummethyl) benzaldehyde chloride (drug A); [4-(hydrazinocarboxy)-1-butyl] tris-(4-dimethylaminophenyl) phosphonium chloride (drug B); and their combination product, C.

Cationic lipophilic compounds have been shown to selectively accumulate in the mitochondria of carcinoma cells compared to normal cells (5) due to the characteristically elevated plasma membrane potentials of neoplastic cells (1, 2, 9). Consistent associations between raised membrane potentials, drug accumulation, and cytotoxicity imply that the selective cytotoxicity of cationic lipophilic drugs is a consequence of differences in drug uptake between normal and tumorigenic cells (2, 4, 9, 10). Numerous experiments have demonstrated the selective cytotoxicity of cationic lipophilic compounds against cancer cells compared to normal cells in both tissue culture and animal studies (2–4, 6, 11–13). In addition, phase I clinical trials with the lipophilic cation MKT-077 have recently begun (5).

We have previously investigated the effects of the prototype CLPS, TPP, on the 1H NMR spectra of two human breast cancer cell lines of different biological properties (14, 15). TPP demonstrated enhanced toxicity against the poorly differentiated breast ductal carcinoma cell line, DU4475, compared to the transformed HBL-100 cell line. Both cell lines demonstrated a remarkably similar dose-dependent increase in lipid-derived resonances and changes in phospholipid metabolites. We hypothesized that these changes may be fundamentally related to structural features of CLPS that cause cytotoxicity and sought to test this hypothesis using CLPS congeners with varying toxicities. In this work, we report the effects of two cationic lipophilic phosphonium congeners of varying toxicity, alone and in combination, on cytotoxicity, NMR-visible lipids, and lipid metabolites in two human breast cancer cell lines. We compare these results with two anticancer agents in current clinical use, 5-fluorouracil (5-FU) and methotrexate (MTX).


Cell Culture

The two human breast cell lines used in this study were HBL-100, a well-differentiated, transformed, nonmalignant cell line originally derived from normal breast epithelial cells; and DU4475, a poorly-differentiated breast ductal carcinoma cell line originally derived from a metastatic nodule. Growth characteristics of the cell lines have been described elsewhere (14, 15). Both cell lines were grown in RPMI-1640 culture medium supplemented with 10% (v/v) fetal bovine serum (batch number 81012053; Trace Biosciences), 2 mM L-glutamine, 40,000 U of gentamicin (1 mL at 40 mg/mL), and 250 units/L of human insulin, and buffered with 26.2 mM sodium bicarbonate using standard culture conditions of 37°C and 5% CO2 in air. In contrast with previous studies (14, 15), 26.2 mM NaHCO3 was used to buffer the culture medium in place of half the bicarbonate and 20 mM HEPES. HEPES has several 1H NMR resonances in the choline/creatine region of the spectra, which affected observation of these peaks.

Growth Inhibition Assays

Cells were seeded into 24-well plates, treated with 20 μL of phosphate-buffered saline (PBS) (controls) or various drug concentrations, and the number of viable cells was determined by their ability to exclude trypan blue after 48-h drug treatment. Growth inhibitory curves were plotted as the percentage of viable drug-treated cells compared to controls, as a function of drug concentration. Each point on the growth inhibition curves represents the average of four to six determinations, and assays were performed on cells at several different passage numbers. IC50 (or IC10) values were determined by the drug concentration required to decrease the number of viable drug-treated cells by 50% (or 10%) relative to controls. Drug A (p-(triphenylphosphoniummethyl) benzaldehyde chloride; MW = 417) and drug B ([4-(hydrazinocarboxy)-1-butyl] tris-4-dimethylaminophenyl) phosphonium chloride; MW = 542) were prepared in PBS immediately prior to each experiment from 10-mM stock solutions also prepared in PBS and stored at –70°C. Frozen drug stocks, as measured by viability assays, were stable for at least 2 weeks.


The NMR methods used in this work have been described in detail elsewhere (14, 15). Briefly, cells were seeded at 2.5 × 105 cells/mL in four to six 175-cm2 filter-topped tissue culture flasks and incubated under standard culture conditions. Five hours after seeding, drugs were added (or equal volumes of PBS for controls) in various concentrations. Cells were harvested by trypsinisation 48 h later for NMR spectroscopic analysis. Cells were counted using trypan blue exclusion, washed three times in 1 mL PBS in D2O and placed into a 5-mm NMR tube in a total volume of 400–500 μL PBS/D2O. An average of 5.4 × 107 HBL-100 and 2.4 × 107 DU4475 cells were used for each experiment. A 2-mm coaxial capillary containing 60 μL of 10 mM p-aminobenzoic acid (PABA) was used as a concentration and chemical shift reference.

One-dimensional 1H NMR spectra were obtained at 360 MHz on a Bruker AM360 spectrometer (128 transients, 10 ppm spectral width, 2-s relaxation delay with presaturation of residual water by selective irradiation during the last 1 s, 90° flip angle, 3.14-s repetition time). The sample temperature was maintained at 37°C. Changes in lipid resonance intensities were quantified by measuring peak heights and peak areas (obtained by integration) following the application of a Lorentzian line-broadening of 3 Hz and Fourier transformation. The results are reported as ratios relative to the average of the two aromatic proton resonances (6.8 and 7.8 ppm) from the PABA reference capillary. All ratios relative to PABA were scaled to account for small variations in sample volume and the number of cells in the sample tube.

Magnitude-mode 2D 1H-1H correlated spectroscopy (COSY) spectra of cell samples (200 free-induction decays (FIDs) of 32 transients, acquired in <3 h) were obtained as previously described (15) using a spectral width in the F2 domain of 2874 Hz (8 ppm), 2K time-domain points in t2, and an acquisition time of 356 ms. During the 1-s relaxation delay, continuous-wave irradiation (23 dB below 0.2 W) was used to presaturate the water resonance. One-dimensional spectra were acquired before and after the COSY spectrum to check on sample stability; no significant changes in the lipid CH2 resonances were observed relative to either the internal CH3 resonance or to the resonances from the external standard, PABA, during this time period. Two-dimensional spectra were zero-filled to 1K complex points in t1, and 2D Fourier transformation was performed with a sine-bell window function in t1 and a Lorentzian-Gaussian window (LB = –30, GB = 0.12) in t2.

Cross-peak volumes were measured on each side of the diagonal of unsymmetrized spectra and expressed as a ratio relative to the lysine δCH2-ϵCH2 cross peak at 1.7, 3.0 ppm or the PABA aromatic cross peak on the same side of the diagonal. The position of the lysine cross peak, distant from the diagonal and other cross peaks and its apparent stability with drug treatment and adverse conditions (e.g., low pH, high cell density in culture) (15, 16) makes it a useful internal reference for cross-peak volume ratios. Furthermore, the ratio of the lysine cross peak relative to PABA cross peak was monitored and found to exhibit no significant variations for treated DU4475 cells. In HBL-100 cells, this ratio changed significantly only following treatment with 10 μM A + 10 μM B (data not shown).

The frequency ranges over which each lipid and amino acid cross-peak volume was measured were standardized in all spectra except for the choline metabolites, where the frequency ranges for volume integration were individually optimized due to the proximity of these cross peaks to the diagonal and to each other. Results are given as the average of at least three independent experiments, and error bars represent ± one standard deviation. Statistical significance for the 1D and 2D ratios was determined using a two-sided unpaired Student's t-test from the software package StatView 512+ for the Macintosh.


Growth Inhibition Assays

Growth inhibition curves were used to monitor the percentage of viable drug-treated cells compared to untreated control cells as a function of drug concentration (Fig. 2). IC50s were determined for each drug or drug combination (Table 1). The selectivity of each drug or drug combination was defined as the IC50 (HBL-100 cells)/IC50 (DU4475 cells) (Table 1) (2). Drugs A and TPP inhibit 50% of DU4475 cell growth at a concentration 13.6 and 42.3 times lower, respectively, than the concentration required to inhibit HBL-100 cells, thus demonstrating significant selectivity. While the selectivity ratio of compound B is somewhat lower (3.6), it is still greater than those of the anticancer drugs 5-FU and MTX (2.5 and 0.75, respectively).

Figure 2.

Representative growth inhibition curves of HBL-100 and DU4475 cells treated with (a) drug A and (b) drug B for 48 h, as measured by trypan blue exclusion. Each data point represents the mean of four to six separate experiments ± sem.

Table 1. Drug Concentration Required for 50% Inhibition of Cell Growth Following Two Days of Drug Exposure
Drug or drug ratioIC50 (μM)Selectivity
  • Values represent averages of 4–6 determinations ± one standard deviation by trypan blue exclusion. IC50s for A, B mixtures are expressed as total drug concentration. Selectivity is calculated as the ratio of the IC50s for HBL-100 cells relative to DU4475 cells.

  • a

    IC50 values for treatment with TPP under identical conditions as reported elsewhere (14, 15).

A303 ± 9022.2 ± 9.913.6
B13 ± 33.6 ± 1.53.6
1A:1B85 ± 237.6 ± 2.511.2
3A:1B129 ± 373.8 ± 2.333.9
6A:1B166 ± 4224.0 ± 4.46.9
12A:1B224 ± 4014.0 ± 9.016.0
5-FU66 ± 2926 ± 102.5
MTX0.006 ± 0.0040.008 ± 0.0030.75

Synergism/Antagonism of Phosphonium Salt Combinations

Dose-response assays for drug A and B cotreatments were performed for several different ratios and were analyzed using isobolograms (Fig. 3). An isoeffect reference line for x% inhibition was plotted by connecting the ICx concentrations for each drug alone with a straight line, and represents the amount of each drug required in combination to produce a constant purely additive effect. Antagonistic or synergistic effects are implied when the actual data points lie above or below, respectively, the reference line (2). Combinations of compounds A and B that achieve 50% or 70% inhibition do not exhibit significant synergism against either cell line. For HBL-100 cells, drugs A and B acted antagonistically at the IC50 but were additive at the IC70 (Fig. 3). For DU4475 cells, the effect of the two drugs was generally antagonistic but close to additive (data not shown).

Figure 3.

Isobolograms for (a) 50% and (b) 70% inhibition of the growth of HBL-100 cells (trypan blue exclusion) by combinations of cationic lipophilic phosphonium salts (CLPS) A and B. Each data point represents an average of four to six IC50 or IC70 values, as shown in Table 1 and Fig. 2.

1D 1H NMR Spectra of Drug-Treated Cells

HBL-100 cells were treated with drug A, drug B, or a 1:1 combination of A and B, at concentrations necessary to inhibit cell growth by 50% (IC50) or 10% (IC10). In separate experiments, HBL-100 cells were treated with 5-FU or MTX at the IC45. DU4475 cells were treated with A, B, and A+B at the IC50, and also with drug B at the IC10. Cells were then harvested and examined by 1D and 2D NMR. The 1D NMR spectra of HBL-100 or DU4475 cells treated with high drug concentrations (IC50) were characterized by a high-resolution lipid spectrum (Fig. 4a, b, d, and e), evident in the distinct and relatively narrow methylene ((-CH2)n) resonance at 1.3 ppm and other prominent lipid resonances at 0.9 ppm (methyl, -CH3), 1.6 ppm (fatty acyl chain β-methylene, -OOC-CH2-CH2-), 2.2 ppm (fatty acyl chain α-methylene, -OOC-CH2-), 5.35 ppm (olefinic unsaturated fatty acyl chains, -CH= CH-). The spectra of untreated HBL-100 or DU4475 cells in log-phase growth showed very few lipid-derived resonances and contained predominantly contributions from amino acids, lactate, creatine, phosphocreatine, choline, and choline metabolites (Fig. 4c and f). Extensive assignment of the resonances in the 1D spectra to intracellular chemical species has been performed and reported elsewhere (14, 15).

Figure 4.

360 MHz 1H NMR spectra of HBL-100 cells treated with (a) 40 μM A + 40 μM B (IC50), (b) 300 μM drug A (IC50), and (c) PBS (control cells), and DU4475 cells treated with (d) 3.5 μM A + 3.5 μM B (IC50), (e) 20 μM drug A (IC50), and (f) PBS (control cells) for 48 h.

Peak height measurements of the lipid methylene (1.3 ppm) resonance (which may also include signals from the methyl resonances of lactate or threonine) relative to the average of the two resonances from the external standard PABA showed significant increases in both HBL-100 and DU4475 cells treated with drug A, B, or the A+B (1:1) combination at the IC50 (P < 0.01) (Fig. 5). Similarly, the methyl resonance at 0.9 ppm, which includes signals from the methyl groups of both lipid and the neutral amino acids valine, leucine, and isoleucine, increased significantly relative to peaks from the PABA standard for all cells treated with high CLPS concentrations (P < 0.05), except for HBL-100 cells treated with 40 μM A + 40 μM B. Results obtained as ratios of peak integrals were consistent with measurements of peak heights in all experiments (data not shown).

Figure 5.

NMR-visible lipid changes in HBL-100 and DU4475 cells treated with compounds A, B, or A+B (1:1) at the IC10 and IC50 for 48 h. Proton NMR spectral peak height ratios of methylene (CH2) at 1.3 ppm and methyl resonances (CH3) at 0.9 ppm are presented relative to the average of the external standard PABA at 6.8 and 7.8 ppm. Data represents the mean ± sd of at least three independent experiments. Statistical significance relative to control (PBS-treated) cells is denoted by ** (P < 0.01) and * (P < 0.05, Student's t-test).

The emergence of the mobile lipid spectrum in drug-treated cells also leads to significant increases in the CH2/CH3 ratio (data not shown). A proton NMR spectrum of pure lipid would exhibit a relatively constant CH2/CH3 ratio of around 7, based on an estimation of the contribution of common fatty acids to the resonances at these chemical shifts. In practice, the neutral amino acids always contribute significantly to the methyl resonance, such that the CH2/CH3 ratio in proton spectra of cells approaches, but rarely exceeds, a value of 4.8.

Low concentrations of the drugs A or B (IC10) used alone to treat HBL-100 cells did not induce significant changes in the resonances at 1.3 and 0.9 ppm. However, HBL-100 cells treated with IC10 concentrations of the drug combination A+B (10 μM A + 10 μM B) showed a strong increase in the methylene resonance at 1.3 ppm and in the methyl resonance at 0.9 ppm (P < 0.01; Fig. 5). Overall, the increase in methylene and methyl resonances was much greater in cells treated with 1:1 A+B combination at the IC10 compared to the IC50, consistent with the antagonism shown by these drugs for each other (Fig. 3a).

HBL-100 cells were also treated with the anticancer agents 5-FU (50 μM) and MTX (0.005 μM) at the IC45 (Fig. 6). The 1D spectra obtained showed significant increases in the lipid methyl and methylene resonances following 5-FU treatment (P < 0.01; Fig. 5), whereas a slight but significant decrease in lipid resonances was observed following treatment with MTX (Fig. 5).

Figure 6.

360 MHz 1H NMR spectra of HBL-100 cells treated with (a) 0.005 μM methotrexate and (b) 50 μM 5-FU.

2D COSY NMR Spectra of Drug-Treated Cells

Representative unsymmetrized 2D COSY spectra of control and treated (20 μM A, representing the IC50 concentration) DU4475 cells, with major cross peaks labeled, are shown in Fig. 7. Extensive assignments for the cross peaks observable in DU4475 and HBL-100 cells have been reported previously (14, 15). The cross peaks that increase in CLPS-treated cells arise from proton couplings in lipid fatty acyl chains (A–F), including those involving unsaturated bonds (C and D). The geminal cross peak from the C-1 and C-3 methylenes of the glycerol backbone (G′ = G1′ and G3′ at 4.07 and 4.27 ppm) and the two vicinal cross peaks expected for the C-1,2 and C-2,3 methylene-methine couplings (GA at 4.07, 5.20 ppm; and GB at 4.27, 5.20 ppm) appear to have increased with treatment. Other cross peaks observed in the 2D COSY of HBL-100 cells include those attributable to the CH2-CH2 moieties from choline (cho, 3.50, 4.07 ppm), phosphocholine (PC, 3.69, 4.38 ppm), glycerophosphocholine (GPC, 3.61, 4.25 ppm), and phosphoethanolamine (PE, 3.24, 4.02 ppm).

Figure 7.

Unsymmetrized 360 MHz 2D COSY spectra of DU4475 cells treated with (a) PBS (control cells) and (b) 20 μM drug A (IC50) for 48 h. Labelled cross peaks include those from fatty acyl chains: A: -(CH2)n-CH2-CH3; B: -CH = CH-CH2-CH2-; C: -CH2-CH2-CH= CH-; D: = CH-CH2-CH= CH-; E: -O-C(O)-CH2-CH2-CH2; F: -O-C(O)-CH2-CH2; and from the glycerol backbone of mobile lipids: G: (vicinal, JAB, JAC) RO-CHAHB-CHC(OR)-CHAHB-OR; G′: (geminal, JAB) RO-CHAHB-CH(OR)-CHAHB-OR. Other labelled cross peaks are described in the text.

Lipid Acyl Chains

Drug-induced cellular changes observed by 2D spectroscopy were quantified by measuring cross-peak volumes relative to standard resonances, both internal (lysine δCH2-ϵCH2; 1.7, 3.0 ppm) and external (PABA CH-CH; 6.8, 7.8 ppm) to the sample. Changes in cross-peak volumes for the fatty acyl chains confirmed the increase in NMR-visible lipid acyl chains detected by 1D spectra. The volumes for cross peaks B, C, D, F, and G (head group) increased significantly (P < 0.05) in both HBL-100 and DU4475 cells treated with drug A, B, or A+B at the IC50 (Table 2) compared to untreated controls. Treatment of HBL-100 cells with 5-FU (50 μM) showed similar increases relative to controls (P < 0.01). In contrast, spectra from MTX-treated (0.005 μM) HBL-100 cells showed no significant increases in lipid cross-peak volumes. Some cross-peak volumes, e.g., G′ and E, were not measured due to their close proximity to the diagonal, which interferes with accurate quantification, whereas others, e.g., A and PE, are affected by t1 noise and are difficult to measure accurately.

Table 2. Lipid Cross Peak Volumes of CLPS-Treated HBL-100 and DU4475 Human Breast Cell Lines
HBL-100 treatmentB/LysC/LysD/LysF/LysG/Lys
Control0.43 ± 0.080.25 ± 0.080.23 ± 0.090.36 ± 0.100.20 ± 0.04
10 μM A (IC10)0.48 ± 0.160.25 ± 0.100.23 ± 0.110.37 ± 0.170.19 ± 0.07
300 μM A (IC50)2.46 ± 0.55§0.98 ± 0.10§0.78 ± 0.06§1.69 ± 0.36§0.73 ± 0.41**
2 μM B (IC10)0.37 ± 0.130.26 ± 0.130.25 ± 0.160.26 ± 0.070.24 ± 0.10
10 μM B (IC50)1.90 ± 0.28§0.63 ± 0.11§0.43 ± 0.081.41 ± 0.28§0.32 ± 0.06
10 μM A + 10 μM B (IC10)2.25 ± 0.14§0.69 ± 0.05§0.47 ± 0.04§1.63 ± 0.14§0.34 ± 0.03§
40 μM A + 40 μM B (IC50)1.60 ± 0.32§0.65 ± 0.09§0.52 ± 0.10§1.20 ± 0.18§0.41 ± 0.11
50 μM 5-FU (IC45)1.35 ± 0.29§0.55 ± 0.08§0.43 ± 0.09**0.85 ± 0.19§0.31 ± 0.05**
0.005 μM MTX (IC45)0.63 ± 0.390.31 ± 0.140.26 ± 0.110.46 ± 0.280.22 ± 0.10
DU4475 treatmentB/LysC/LysD/LysF/LysG/Lys
Control0.27 ± 0.040.28 ± 0.070.28 ± 0.080.26 ± 0.030.23 ± 0.08
20 μM A (IC50)2.12 ± 0.26§0.72 ± 0.11§0.78 ± 0.19§2.39 ± 0.30§0.51 ± 0.10§
0.4 μM B (IC10)0.19 ± 0.04*0.19 ± 0.06*0.20 ± 0.060.20 ± 0.04*0.16 ± 0.04
3 μM B (IC50)2.62 ± 0.77§0.83 ± 0.240.97 ± 0.25§2.88 ± 0.89§0.52 ± 0.13
3.5 μM A + 3.5 μM B (IC50)3.51 ± 0.95§1.09 ± 0.341.26 ± 0.413.98 ± 1.07§0.70 ± 0.16§
Control0.36 ± 0.090.21 ± 0.090.20 ± 0.100.30 ± 0.080.17 ± 0.05
10 μM A (IC10)0.58 ± 0.21*0.32 ± 0.180.29 ± 0.180.42 ± 0.130.23 ± 0.11
300 μM A (IC50)2.94 ± 0.35§1.22 ± 0.32§1.01 ± 0.39§2.02 ± 0.23§0.90 ± 0.46
2 μM B (IC10)0.37 ± 0.180.27 ± 0.190.26 ± 0.230.25 ± 0.100.24 ± 0.14
10 μM B (IC50)1.94 ± 0.50§0.65 ± 0.21§0.44 ± 0.14**1.42 ± 0.36§0.32 ± 0.07
10 μM A + 10 μM B (IC10)3.85 ± 0.64§1.18 ± 0.25§0.81 ± 0.22§2.78 ± 0.46§0.58 ± 0.09§
40 μM A + 40 μM B (IC50)1.33 ± 0.20§0.56 ± 0.200.46 ± 0.19**1.00 ± 0.15§0.37 ± 0.17**
50 μM 5-FU (IC45)1.53 ± 0.48§0.63 ± 0.16§0.49 ± 0.130.96 ± 0.28§0.35 ± 0.07
0.005 μM MTX (IC45)0.65 ± 0.400.31 ± 0.160.25 ± 0.130.46 ± 0.280.21 ± 0.10
  • Cross peak volume ratios measured from 2D 1H-H COSY spectra of HBL-100 or DU4475 human breast cell lines after 48 h of drug treatment. Data is reported relative to internal (Lys 1.3, 3.0 ppm) and external (PABA, 6.8, 7.8 ppm) cross-peak volumes.

  • *

    P < 0.05

  • **

    P < 0.01;

  • P < 0.001;

  • §

    P = 0.0001, relative to control.

Control0.40 ± 0.130.44 ± 0.210.44 ± 0.230.39 ± 0.130.36 ± 0.22
20 μM A (IC50)4.26 ± 1.26§1.47 ± 0.55**1.55 ± 0.584.84 ± 1.58§1.05 ± 0.46**
0.4 μM B (IC10)0.37 ± 0.110.38 ± 0.140.39 ± 0.150.38 ± 0.090.31 ± 0.11
3 μM B (IC50)4.68 ± 2.201.47 ± 0.69**1.70 ± 0.76**5.15 ± 2.490.92 ± 0.40*
3.5 μM A + 3.5 μM B (IC50)5.87 ± 1.86§1.80 ± 0.632.06 ± 0.706.64 ± 2.09§1.17 ± 0.33

HBL-100 cells treated with drug A or B at the IC10 and DU4475 cells treated with drug B at the IC10 showed no significant increases in 2D lipid cross peaks (except for a slight increase in B/PABA that was observed for HBL-100 cells treated with IC10 B). In fact, a small but significant decrease in cross peaks B, C, and F/Lys were observed in DU4475 cells treated with IC10 B. However, as seen in 1D spectra, HBL-100 cells treated with the 10 μM A and 10 μM B (IC10) combination did show a significant increase in all lipid acyl chain ratios relative to both the internal standard lysine and the external standard, PABA (P < 0.0001) (Table 2). Moreover, correlation of the 2D cross-peak volume ratios of F/PABA and B/PABA with the corresponding 1D 1.3 ppm/PABA peak height ratio, resulted in correlation constants of 0.91–0.98 for both the DU4475 and HBL-100 cell line (data not shown).

Phospholipid Metabolites

In general, CLPS treatment did not cause an increase in the cross peak arising from the methylene-methylene coupling of GPC, as has previously been demonstrated in TPP-treated cells (14, 15), except in HBL-100 cells treated with high concentrations of the A+B combination (P < 0.05) (Table 3). In HBL-100 cells, there was a consistent and significant decrease in phosphocholine (PC) for all high-dose drug treatments (P < 0.01), except for drug B when taken relative to PABA. Cross-peak volumes for PC did not change in drug-treated DU4475 cells, with the exception of a significant decrease in PC/Lys at high concentrations of drug A. Choline cross-peak volume ratios increased in both DU4475 and HBL-100 cells treated with high doses of A (P < 0.01), and in HBL-100 cells treated with 5-FU or with combinations of A and B (P < 0.05; see Table 3). There were no observed changes in phospholipid cross peaks in HBL-100 cells treated with MTX except for a slight decrease in GPC. However, treatment with high concentrations of 5-FU resulted in some changes similar to those observed with A and B: no changes in GPC, and an increase in cho/Lys and cho/PABA (P < 0.01). In contrast to CLPS treatment, 5-FU induced increases in PC that were only significant for PC/PABA.

Table 3. Lipid Metabolite Cross Peak Volumes of CLPS-Treated HBL-100 and DU4475 Human Breast Cell Lines
HBL-100 treatmentcho/LysPC/LysGPC/Lys
Control0.21 ± 0.051.28 ± 0.060.11 ± 0.06
10 μM A (IC10)0.17 ± 0.051.00 ± 0.460.09 ± 0.04
300 μM A (IC50)0.48 ± 0.08**0.22 ± 0.12**0.20 ± 0.13
2 μM B (IC10)0.19 ± 0.041.04 ± 0.16**0.06 ± 0.02
10 μM B (IC50)0.23 ± 0.051.00 ± 0.11§0.09 ± 0.03
10 μM A + 10 μM B (IC10)0.21 ± 0.051.12 ± 0.260.07 ± 0.02
40 μM A + 40 μM B (IC50)0.48 ± 0.05§0.30 ± 0.11§0.43 ± 0.11§
50 μM 5-FU (IC45)0.36 ± 0.05§1.38 ± 0.130.10 ± 0.05
0.005 μM MTX (IC45)0.18 ± 0.060.89 ± 0.570.04 ± 0.02*
DU4475 treatmentcho/LysPC/LysGPC/Lys
Control0.13 ± 0.040.18 ± 0.070.05 ± 0.02
20 μM A (IC50)0.19 ± 0.03**0.11 ± 0.02*0.05 ± 0.02
0.4 μM B (IC10)0.13 ± 0.020.21 ± 0.030.02 ± 0.00**
3 μM B (IC50)0.17 ± 0.030.14 ± 0.030.05 ± 0.02
3.5 μM A + 3.5 μM B (IC50)0.18 ± 0.040.12 ± 0.030.07 ± 0.02
HBL-100 treatmentcho/PABAPC/PABAGPC/PABA
Control0.18 ± 0.071.08 ± 0.230.09 ± 0.04
10 μM A (IC10)0.21 ± 0.081.09 ± 0.350.11 ± 0.06
300 μM A (IC50)0.60 ± 0.19§0.24 ± 0.07§0.30 ± 0.30
2 μM B (IC10)0.18 ± 0.060.99 ± 0.110.06 ± 0.03
10 μM B (IC50)0.24 ± 0.091.03 ± 0.290.09 ± 0.05
10 μM A + 10 μM B (IC10)0.39 ± 0.18*1.98 ± 0.80**0.11 ± 0.04
40 μM A + 40 μM B (IC50)0.41 ± 0.08§0.24 ± 0.07§0.38 ± 0.17
50 μM 5-FU (IC45)0.41 ± 0.121.54 ± 0.28**0.12 ± 0.05
0.005 μM MTX (IC45)0.17 ± 0.070.72 ± 0.390.04 ± 0.02*
DU4475 treatmentcho/PABAPC/PABAGPC/PABA
  • Cross peak volume ratios measured from 2D 1H-H COSY spectra of HBL-100 or DU4475 human breast cell lines after 48 h of drug treatment. Data is reported relative to internal (Lys 1.3, 3.0 ppm) and external (PABA, 6.8, 7.8 ppm) cross-peak volumes.

  • *

    P < 0.05;

  • **

    P < 0.01;

  • P < 0.001;

  • §

    P = 0.0001, relative to control.

Control0.20 ± 0.090.25 ± 0.080.07 ± 0.04
20 μM A (IC50)0.38 ± 0.11**0.23 ± 0.110.10 ± 0.06
0.4 μM B (IC10)0.28 ± 0.110.43 ± 0.200.05 ± 0.01
3 μM B (IC50)0.28 ± 0.060.22 ± 0.060.08 ± 0.04
3.5 μM A + 3.5 μM B (IC50)0.31 ± 0.160.21 ± 0.110.12 ± 0.06


Selective Cytotoxicity

The data presented in this work demonstrate that the TPP derivatives drug A and, to a lesser extent, drug B selectively inhibit growth of a cell line derived from a malignant breast metastasis (DU4475) compared to a transformed cell line derived from normal human breast epithelial cells (HBL-100). Combinations of A and B also display selective growth inhibition, although in contrast to previous results the combinations do not act synergistically in this cell model. Drugs A and B, used alone at high concentrations and in combination, were shown to cause increases in mobile or NMR-visible lipid as measured by 1D 1H NMR. These increases were confirmed by 2D NMR, which also revealed changes in phospholipid metabolites similar to the actions of the parent compound TPP. The significance is that congeners of the cytotoxic compound TPP cause changes in lipid metabolism similar to that of the parent compound at comparable effective doses. Moreover, 5-FU, a structurally and functionally dissimilar cytotoxic agent in current clinical use, also causes similar changes in lipid metabolism. However, the fact that increased NMR-visible lipid does not occur with MTX treatment indicates that these lipid changes are not unique to CLPS and are not characteristic of all cytotoxic agents.

The selective cytotoxicity of A and B against DU4475 relative to HBL-100 cells by factors of 13.6 and 3.6, respectively, is consistent with our previous research showing that the parent compound TPP has a selectivity ratio of 42.3 in the same cell model (14, 15). This selectivity may reflect higher transmembrane potentials in the DU4475 breast carcinoma cell line compared to the transformed breast cell line HBL-100, which is often characteristic of neoplastic cells (1, 2, 9). The growth inhibitory effects of these compounds (and selectivity ratios) occur at doses comparable to those demonstrated in PaCa-2 human pancreatic carcinoma cells and CV-1 untransformed monkey kidney epithelial cells, and similarly show compound B to be effective at lower absolute doses than compound A, albeit only marginally for DU4475 cells (2). MTX and, to a lesser extent, 5-FU were shown to be relatively nonselective against DU4475 compared to HBL-100 cells. Although many antineoplastic agents in current clinical use are effective against proliferating cells in vivo, they do not demonstrate remarkable selectivity against tumor-derived compared to untransformed cells in vitro where both cells types are actively proliferating (2, 11).

Compounds A and B are potentially capable of covalently combining more effectively in malignant cells than normal cells in situ to form an adduct more active than the precursors, hence resulting in target-selective synergism (2, 3). Combinations of A and B have been shown to exhibit target-selective synergism against Ehrlich Lettre ascites (ELA) carcinoma cells relative to untransformed CV-1 cells in vitro (3). However, cytotoxic synergism of combinations of compounds A and B was not demonstrated in either HBL-100 or DU4475 cells, with the combinations acting antagonistically or, at best, additive. This indicates either that the two individual drugs do not act synergistically or that the two drugs do not form a more active combination product in this cell model. This could be due to a number of factors. Compounds A and B could be taken up at different rates by different cells, which can be affected by differences in drug lipophilicity or cellular plasma membrane potential (2). Moreover, the difference in mitochondrial relative to plasma membrane potential between cell types may cause the drugs to localize in different parts of the cell (2). The structural similarities of A and B may result in competitive inhibition of drug uptake or action. Alternatively, compound C may self-assemble within these cells, but its cytotoxic effects may be curtailed by rapid elimination or metabolism. It is also possible that compound C is simply not as effective as either of its precursors against these cell lines, or that treatment with a different concentration or schedule may affect the synergism of these two compounds. Since the synergistic action of drug C is based on enhanced toxicity subsequent to combination of the drug precursors, further experiments are needed to assess the cytotoxicity of compound C and determine whether or not these criteria are met in HBL-100 and DU4475 cells.

Increased NMR-Visible Lipid

Proton NMR demonstrated that the major detectable change induced by treatment of two human breast cell lines with high concentrations (IC50) of the cationic lipophilic compounds A, B, and A+B (1:1) is large increases in resonances from NMR-visible lipid. This increase was also observed in HBL-100 cells following treatment with a combination of A+B at lower concentrations (IC10). The increased signal intensity at 1.3 ppm reflects predominantly an increase in the concentration of mobile lipid acyl chains that give rise to NMR-visible lipid signals, with smaller contributions from lactate and threonine methyl groups. The contribution from lactate was minimized by washing of the cells during harvest. The observed lactate levels in 1D and 2D spectra were not consistent under any of the conditions employed. Increases in mobile lipid can be due to either an increased production of fatty acyl chains or an increased mobility of existing fatty acyl chains, or both. HBL-100 cells treated with 5-FU, but not MTX, also developed increased resonances from mobile lipid acyl chains and glycerol backbone, indicating that these changes are not unique to CLPS, nor are they general for all anticancer agents. Moreover, the observation that high levels of mobile lipids could occur at a concentration that is not cytotoxic (IC10 A:B combination) indicates that the production of mobile lipid is not directly related to the degree of toxicity of CLPS and, by implication, not directly related to cell death. Furthermore, the decrease in mobile lipids observed at higher A+B combinations does demonstrate that mobile lipid production is concentration dependent. This confirms our previous study of TPP-treated HBL-100 cells, in which the maximum in lipid acyl chain resonances in 1D spectra and lipid cross-peak volumes in 2D spectra occurred at a concentration approximately 10-fold less than the IC50 (15). Increases in NMR-visible lipid have been observed in cultured cells following treatment with a number of cytotoxic agents (17–23) in response to adverse culture conditions (16, 23) and can be modulated by such factors as the degree of drug resistance (24, 25). The relationship between these diverse stimuli and increased NMR-visible lipids is not currently understood, and it is unclear whether the similar phenotype changes arise from the same or different cellular pathways.

In 2D spectra, both HBL-100 and DU4475 cell lines display a CLPS-induced increase in the G′ (G1′ and G3′) cross peak arising from the geminal protons on carbons 1 and 3 of the glycerol backbone, indicating that increased NMR-visible lipid arises from triglycerides, phosphoglycerides, or their derivatives (26–28). Spectra from drug-treated HBL-100 and DU4475 cells also showed a significant increase in the volumes for the two cross peaks GA and GB arising from the vicinal coupling between HA and HB on carbons 1 and 3, and HC on C-2 on the glyceride backbone. This suggests a contribution from triglycerides, although the intensity of these cross peaks was often weak—especially in HBL-100 cells. Restriction of head group mobility means that the vicinal cross peaks GA and GB (GA > GB) are often visible only in 2D COSY spectra of cells or lipoproteins containing large amounts of triglyceride (27, 29), and cells containing a large number of cytoplasmic lipid droplets, e.g., oleate-treated myeloma cells (30). No cross peaks specific to the glyceride backbone of phospholipids were observed in the spectra of HBL-100 and DU4475 cells, which suggests that the NMR-visible lipid acyl chains are not derived from phospholipids. The asymmetry of protons in the glycerol backbone of phospholipids causes a shift in GA and GB to 4.10, 5.20 and 4.37, 5.20 ppm (with a corresponding shift in G′ to 4.10, 4.37 ppm) and the appearance an additional cross peak at 3.95, 5.20 ppm arising from the vicinal coupling of HC on C-2 to the equivalent protons on C-3 adjacent to the phosphate group (28).

The drug-induced increase in lipid acyl chains observed using 1D and 2D NMR is consistent with changes observed in HBL-100 and DU4475 cells treated with the structurally related compound, TPP. TPP treatment also caused a concentration-dependent increase in GPC and decreases in PC at low TPP concentrations (15, 16). In HBL-100 cells, PC recovered at higher TPP concentrations, and an increase in choline was also observed. In the present study, increases in choline were observed in HBL-100 cells or DU4475 cells following high-dose treatment with drug A, and in HBL-100 cells treated with 5-FU or the drug A+B combination at the IC50. PC was observed to decrease in all high-concentration treatments in HBL-100 cells. Increases in GPC were only observed for high-dose (IC50) A+B treatment. We have previously proposed that the induction of mobile lipid and the accompanying changes in phospholipid metabolites with TPP treatment can be explained by the action of phospholipases on existing cellular phospholipids (15, 16). The increased fatty acyl chain resonances and GA and GB cross peaks observed here indicate accumulation of neutral glycerides, such as triglycerides, which could be formed from the fatty acids liberated by phospholipase A1 and A2 activity. While an increase in GPC may be expected to accompany this activity, this may not be observable in all situations. In TPP-treated HBL-100 cells we only observed significant increases in GPC at concentrations above the IC50 (15).

In proliferating tumor cells, relatively high levels of PC and/or PE are detected by both in vitro and in vivo by 31P NMR, and these metabolites generally decrease significantly after effective chemotherapy (31, 32). HBL-100 cells (but not DU4475 cells) treated with A, B, or A+B showed a decrease in phosphocholine at doses that inhibit cell proliferation by 50%. In addition, a decrease in phosphocholine has been observed in HBL-100 and DU4475 breast carcinoma cells treated with TPP (14, 15), as well as in murine cells cultured under adverse conditions (16). Since PC and PE are soluble membrane precursors, they are more likely to be freely mobile, and therefore changes in cross-peak volumes are more likely an indicator of concentration changes. Inhibition of phospholipase C and/or decreased choline kinase activity could lead to a reduced concentration of PC. Decreases in PC have been correlated with reduced membrane synthesis arising from decreased activity of choline kinase (33, 34). Some investigators have suggested it could be due to reduced phospholipid breakdown associated with signal transduction pathways that stimulate cell division (35).

Further insight into the drug-induced processes that may give rise to such alterations in lipid metabolism can be obtained from other studies investigating the toxic effects of CLPS. The consistent observation that TPP and other cationic lipophilic compounds cause substantial mitochondrial damage (4, 7, 36) suggests that this may be the source of NMR-visible lipid. The damage of mitochondria could result in the release or buildup of a number of metabolites, including fatty acids and GPC. Generally, these catabolites would be used for recycling into phospholipids, production of triglycerides, or for mitochondrial β-oxidation. However the utilization of fatty acids would be restricted by the amount of available ATP, already depleted by cationic lipophilic treatment (4, 37). ATP is required to convert fatty acids to their acyl-S-CoA derivatives for further lipid synthesis. Moreover, mitochondrial damage would inhibit the process of β-oxidation. The observed NMR-visible lipid accumulations could reflect such events, and would also depend on the drug type, cytotoxicity of the drug, dose and treatment schedule, and on the cell system examined. Lipid accumulation is involved in cellular stress and cell death following a variety of insults, including treatment with CLPS, although the actual role of this process remains unclear. Elucidation of the cytotoxicity mechanism of CLPS could identify various stages in the lipid metabolism of cells, which could be used for the design of new anticancer therapies.


The authors acknowledge the critical assessments of Dr. Thomas M. Jeitner and Professor Martin Tattersall.